Bainbridge W.S. (Ed) Berkshire Encyclopedia of Human-Computer Interaction. Vol. 1

BERKSHIRE PUBLISHING GROUP
Berkshire
Encyclopedia of
HumanComputer
Interaction When science fiction
becomes science fact
William Sims Bainbridge

Edited by
National Science Foundation
Berkshire Encyclopedia of
Human-Computer
Interaction
Berkshire Encyclopedia of
Human-Computer
Interaction
VOLUME 1
Editor
Great Barrington, Massachusetts U.S.A.

www.berkshirepublishing.com
Copyright 2004 by Berkshire Publishing Group LLC
All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, elec-
tronic or mechanical, including photocopying, recording, or by any information storage and retrieval sys-
tem, without permission in writing from the publisher.
Cover photo: Thad Starner sporting a wearable computer.

Photo courtesy of Georgia Institute of Technology.
Cover background image: Courtesy of Getty Images.
For information:
Berkshire Publishing Group LLC
314 Main Street
Great Barrington, Massachusetts 01230
www.berkshirepublishing.com
Printed in the United States of America
Library of Congress Cataloging-in-Publishing Data

Berkshire encyclopedia of human-computer interaction / William Sims Bainbridge, editor.
p. cm.
A Berkshire reference work.
Includes bibliographical references and index.
ISBN 0-9743091-2-5 (hardcover : alk. paper)
1. Human-computer interaction--Encyclopedias. I. Bainbridge, William Sims. II. Title.
QA76.9.H85B46 2004
004'.01'9--dc22
2004017920
BERKSHIRE PUBLISHING STAFF
Project Director
Karen Christensen
Project Coordinators
Courtney Linehan and George Woodward
Associate Editor
Marcy Ross
Copyeditors
Francesca Forrest, Mike Nichols, Carol Parikh, and Daniel Spinella
Information Management and Programming

Deborah Dillon and Trevor Young
Editorial Assistance
Emily Colangelo
Designer
Monica Cleveland
Production Coordinator
Janet Lowry
Composition Artists
Steve Tiano, Brad Walrod, and Linda Weidemann
Composition Assistance
Pam Glaven
Proofreaders
Mary Bagg, Sheila Bodell, Eileen Clawson, and Cassie Lynch
Production Consultant
Jeff Potter
Indexer
Peggy Holloway
CONTENTS
List of Entries, ix
Readers Guide, xv
List of Sidebars, xix
Contributors, xxiii
Introduction, xxxiii
Publishers Note, xli
About the Editor, xliii
Entries
Volume I: AL
1440
Vol II: MW
441826
Appendix 1: Glossary, 827
Appendix 2: Master Bibliography of Human-Computer Interaction, 831
HCI in Popular Culture, 893
Index, 931

Index repeated in this volume, I-1
vii
LIST OF ENTRIES
Adaptive Help Systems Animation Artificial Intelligence

Peter Brusilovsky Abdennour El Rhalibi Robert A. St. Amant
Yuanyuan Shen
Adaptive Interfaces Asian Script Input
Alfred Kobsa Anthropology and HCI William Sims Bainbridge
Allen W. Batteau Erika Bainbridge
Affective Computing
Ira Cohen Anthropometry Atanasoff-Berry Computer
Thomas S. Huang Victor L. Paquet John Gustafson
Lawrence S. Chen David Feathers
Attentive User Interface
Altair Application Use Strategies Ted Selker
William Sims Bainbridge Suresh K. Bhavnani
Augmented Cognition
Alto Arpanet Amy Kruse
William Sims Bainbridge Amy Kruse Dylan Schmorrow
Dylan Schmorrow
Allen J. Sears
ix
X BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Augmented Reality Compilers Digital Divide

Rajeev Sharma Woojin Paik Linda A. Jackson
Kuntal Sengupta
Computer-Supported Digital Government
Avatars Cooperative Work Jane E. Fountain
Jeremy Bailenson John M. Carroll Robin A. McKinnon
James J. Blascovich Mary Beth Rosson
Digital Libraries
Beta Testing Constraint Satisfaction Jose-Marie Griffiths
Gina Neff Berthe Y. Choueiry
Drawing and Design
Braille Converging Technologies Mark D. Gross
Oleg Tretiakoff William Sims Bainbridge
E-business
Brain-Computer Interfaces Cybercommunities Norhayati Zakaria
Melody M. Moore Lori Kendall
Adriane D. Davis Education in HCI
Brendan Z. Allison Cybersex Jan Stage
David L. Delmonico
Browsers Elizabeth Griffin Electronic Journals
Andy Cockburn Carol Tenopir
Cyborgs
Cathode Ray Tubes William Sims Bainbridge Electronic Paper Technology
Gregory P. Crawford Gregory P. Crawford
Data Mining
CAVE Mohammad Zaki Eliza
Thomas DeFanti William H. Sterner
Dan Sandin Data Visualization
Kwan-Liu Ma E-mail
Chatrooms Nathan Bos
Amanda B. Lenhart Deep Blue
Murray Campbell Embedded Systems
Children and the Web Ronald D. Williams
Dania Bilal Denial-of-Service Attack
Adrian Perrig ENIAC
Classrooms Abraham Yaar William Sims Bainbridge
Chris Quintana
Desktop Metaphor Ergonomics
Client-Server Architecture Jee-In Kim Ann M. Bisantz
Mark Laff
Dialog Systems Errors in Interactive Behavior
Cognitive Walkthrough Susan W. McRoy Wayne D. Gray
Marilyn Hughes Blackmon
Digital Cash Ethics
Collaboratories J. D. Tygar Helen Nissenbaum
Gary M. Olson
LIST OF ENTRIES XI
Ethnography Handwriting Recognition and Information Theory

David Hakken Retrieval Ronald R. Kline
R. Manmatha
Evolutionary Engineering V. Govindaraju Instruction Manuals
William Sims Bainbridge David K. Farkas
Haptics
Expert Systems Ralph L. Hollis InternetWorldwide Diffusion
Jay E. Aronson Barry Wellman
History of Human-Computer Phuoc Tran
Eye Tracking Interaction Wenhong Chen
Andrew T. Duchowski Jonathan Grudin
Internet in Everyday Life
Facial Expressions Hollerith Card Barry Wellman
Irfan Essa William Sims Bainbridge Bernie Hogan
Fly-by-Wire Human-Robot Interaction Iterative Design

C. M. Krishna Erika Rogers Richard Baskerville
Jan Stage
Fonts Hypertext and Hypermedia
Thomas Detrie David K. Farkas Keyboard
Arnold Holland Alan Hedge
Icons
Games Stephanie Ludi Language Generation
Abdennour El Rhalibi Regina Barzilay
Identity Authentication
Gender and Computing Ashutosh P. Deshpande Laser Printer
Linda A. Jackson Parag Sewalkar Gary Starkweather
Geographic Information Systems Impacts Law and HCI

Michael F. Goodchild Chuck Huff Sonia E. Miller
Gesture Recognition Information Filtering Law Enforcement

Francis Quek Luz M. Quiroga Roslin V. Hauck
Martha E. Crosby
Graphical User Interface Lexicon Building
David England Information Organization Charles J. Fillmore
Dagobert Soergel
Grid Computing Liquid Crystal Displays
Cavinda T. Caldera Information Overload Gregory P. Crawford
Ruth Guthrie
Groupware Literary Representations
Timothy J. Hickey Information Retrieval William Sims Bainbridge
Alexander C. Feinman Dagobert Soergel
Machine Translation
Hackers Information Spaces Katrin Kirchhoff
Douglas Thomas Fionn Murtagh
XII BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Markup Languages Online Education Programming Languages

Hong-Gee Kim Robert S. Stephenson David MacQueen
Glenn Collyer
Mobile Computing Prototyping
Dharma P. Agrawal Online Questionnaires Richard Baskerville
James Witte Jan Stage
Mosaic Roy Pargas
William Sims Bainbridge Psychology and HCI
Online Voting Judith S. Olson
Motion Capture and Recognition R. Michael Alvarez
Jezekiel Ben-Arie Thad E. Hall Recommender and Reputation
Systems
Mouse Ontology Cliff Lampe
Shumin Zhai Christopher A. Welty Paul Resnick
Movies Open Source Software Repetitive Strain Injury

William Sims Bainbridge Gregory R. Madey Jack Tigh Dennerlein
MUDs Optical Character Recognition Scenario-Based Design

Richard Allan Bartle V. Govindaraju John M. Carroll
Swapnil Khedekar
Multiagent systems Search and Rescue
Gal A. Kaminka Peer-to-Peer Architecture Howie Choset
Julita Vassileva
Multimodal Interfaces Search Engines
Rajeev Sharma Pen and Stylus Input Shannon Bradshaw
Sanshzar Kettebekov Alan Hedge
Guoray Cai Security
Personality Capture Bhavani Thuraisingham
Multiuser Interfaces William Sims Bainbridge
Prasun Dewan Semantic Web
Physiology Bhavani Thuraisingham
Musical Interaction Jennifer Allanson
Christopher S. Raphael Smart Homes
Judy A. Franklin Planning Diane J. Cook
Sven Koenig Michael Youngblood
Natural-Language Processing Michail G. Lagoudakis
James H. Martin Sociable Media
Pocket Computer Judith Donath
Navigation William Sims Bainbridge
John J. Rieser Social Informatics
Political Science and HCI Howard Rosenbaum
N-grams James N. Danziger
James H. Martin Michael J. Jensen Social Proxies
Thomas Erickson
Olfactory Interaction Privacy Wendy A. Kellogg
Ricardo Gutierrez-Osuna Jeffrey M. Stanton
LIST OF ENTRIES XIII
Social Psychology and HCI Task Analysis Jenny Preece

Susan R. Fussell Erik Hollnagel Diane Maloney-Krichmar
Sociology and HCI Telecommuting Value Sensitive Design

William Sims Bainbridge Ralph David Westfall Batya Friedman
Socio-Technical System Design Telepresence Video

Walt Scacchi John V. Draper Immanuel Freedman
Software Cultures Text Summarization Video Summarization

Vaclav Rajlich Judith L. Klavans A. Murat Tekalp
Software Engineering Theory Virtual Reality

Richard Kazman Jon May Larry F. Hodges
Benjamin C. Lok
Sonification Three-Dimensional Graphics
David M. Lane Benjamin C. Lok Viruses
Aniko Sandor J. D. Tygar
S. Camille Peres Three-Dimensional Printing
William Sims Bainbridge Visual Programming
Spamming Margaret M. Burnett
J. D. Tygar Touchscreen Joseph R. Ruthruff
Andrew L. Sears
Speech Recognition Rich Goldman Wearable Computer
Mary P. Harper Thad Starner
V. Paul Harper Ubiquitous Computing Bradley Rhodes
Olufisayo Omojokun
Speech Synthesis Prasun Dewan Website Design
Jan P.H. van Santen Barbara S. Chaparro
Unicode Michael L. Bernard
Speechreading Unicode Editorial Committee
Marcus Hennecke Work
Universal Access Christine A. Halverson
Spell Checker Gregg Vanderheiden
Woojin Paik Workforce
Usability Evaluation Brandon DuPont
Sphinx Jean Scholtz Joshua L. Rosenbloom
Rita Singh
User Modeling World Wide Web
Statistical Analysis Support Richard C. Simpson Michael Wilson
Robert A. St. Amant
User Support WYSIWYG
Supercomputers Indira R. Guzman David M. Lane
Jack Dongarra
User-Centered Design
Tablet Computer Chadia Abras
READERS GUIDE
This list is provided to assist readers in locating en- Navigation

tries on related topics. It classifies articles into ten Online Education
general categories: Applications; Approaches; Online Voting
Breakthroughs; Challenges; Components; Disciplines; Planning
Historical Development; Interfaces; Methods; and Recommender and Reputation Systems
Social Implications. Some entries appear in more Search and Rescue
than one category. Statistical Analysis Support
Supercomputers
Applications Telecommuting
Classrooms Ubiquitous Computing
Digital Government Video
Digital Libraries
E-business Approaches
Games Application Use Strategies
Geographic Information Systems Beta Testing
Grid Computing Cognitive Walkthrough
Law Enforcement Constraint Satisfaction
Mobile Computing Ethics
xv
XVI BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Ethnography Components
Evolutionary Engineering Adaptive Help Systems
Information Theory Animation
Iterative Design Braille
Ontology Cathode Ray Tubes
Open Source Software Client-Server Architecture
Prototyping Desktop Metaphor
Scenario-Based Design Electronic Paper Technology
Social Informatics Fonts
Socio-Technical System Design Keyboard
Task Analysis Laser Printer
Theory Liquid Crystal Displays
Universal Access Mouse
Usability Evaluation N-grams
User Modeling Peer-to-Peer Architecture
User-Centered Design Social Proxies
Value Sensitive Design Spell Checker
Website Design Touchscreen
Unicode
Breakthroughs WYSIWYG
Altair
Alto Disciplines
Arpanet Anthropology and HCI
Atanasoff-Berry Computer Artificial Intelligence
CAVE Ergonomics
Converging Technologies Law and HCI
Deep Blue Political Science and HCI
Eliza Psychology and HCI
ENIAC Social Psychology and HCI
Hollerith Card Sociology and HCI
Mosaic
Sphinx Historical Development
Altair
Challenges Alto
Denial-of-Service Attack ENIAC
Digital Divide History of HCI
Errors in Interactive Behavior
Hackers Interfaces
Identity Authentication Adaptive Interfaces
Information Filtering Affective Computing
Information Overload Anthropometry
Privacy Asian Script Input
Repetitive Strain Injury Attentive User Interface
Security Augmented Cognition
Spamming Augmented Reality
Viruses Brain-Computer Interfaces
READERS GUIDE XVII
Compilers Markup Languages

Data Visualization Motion Capture and Recognition
Dialog Systems Natural-Language Processing
Drawing and Design Optical Character Recognition
Eye Tracking Personality Capture
Facial Expressions Programming Languages
Fly-by-Wire Search Engines
Graphical User Interface Semantic Web
Haptics Software Engineering
Multimodal Interfaces Sonification
Multiuser Interfaces Speech Recognition
Musical Interaction Speech Synthesis
Olfactory Interaction Speechreading
Online Questionnaires Text Summarization
Pen and Stylus Input User Support
Physiology Video Summarization
Pocket Computer Visual Programming
Smart Homes World Wide Web
Tablet Computer Social Implications
Telepresence Chatrooms
Three-Dimensional Graphics Children and the Web
Three-Dimensional Printing Collaboratories
Virtual Reality Computer-Supported Cooperative Work
Wearable Computer Cybercommunities
Cybersex
Methods Cyborgs
Avatars Education in HCI
Browsers Electronic Journals
Data Mining E-mail
Digital Cash Gender and Computing
Embedded Systems Groupware
Expert Systems Human-Robot Interaction
Gesture Recognition Impacts
Handwriting Recognition and Retrieval InternetWorldwide Diffusion
Hypertext and Hypermedia Internet in Everyday Life
Icons Literary Representations
Information Organization Movies
Information Retrieval MUDs
Information Spaces Multiagent systems
Instruction Manuals Sociable Media
Language Generation Software Cultures
Lexicon Building Work
Machine Translation Workforce
LIST OF SIDEBARS
Adaptive Help Systems Chatrooms

Farewell Clippy Life Online
Adaptive Interfaces Classrooms

Keeping Disabled People in the Technology Loop History Comes Alive in Cyberspace
Learning through Multimedia
Anthropology and HCI
Digital Technology Helps Preserve Tribal Language Computer-Supported Cooperative Work
Internet Singing Lessons
Anthropology and HCI Social Context in Computer-Supported
Eastern vs. Western Cultural Values Cooperative Work
Augmented Cognition Cybercommunities

Putting Humans First in Systems Design Welcome to LamdaMOO
Braille Cybersex
Enhancing Access to Braille Instructional Materials Cybersex Addiction
xix
XX BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Digital Divide Internet in Everyday Life

HomeNetToo Tries to Bridge Digital Divide Finding Work Online
Information Technology and Competitive
Digital Libraries Academic Debate
Vannevar Bush on the Memex
Law Enforcement
Education in HCI Fighting Computer Crime
Bringing HCI Into the Real World
Literary Representations
Eliza Excerpt from Isaac Asimovs I, Robot
Talking with ELIZA Excerpt from The Sand-Man (1817) by
E. T. A. Hoffman
E-mail
The Generation Gap Machine Translation
Warren Weaver on Machine Translation
Errors in Interactive Behavior
To Err Is Technological Movies
HALs Birthday Celebration
Fonts
Our Most Memorable Nightmare MUDs
The Wide World of a MUD
Gender and Computing
Computer Girl Site Offers Support for Online Education
Young Women An Online Dig for Archeology Students
Narrowing the Gap Virtual Classes Help Rural Nurses
Geographic Information Systems Political Science and HCI

Geographic Information Systems Aid Land Washington Tales of the Internet
Conservation
Psychology and HCI
Groupware Human Factors Come into the Forefront
Away Messages Virtual Flight for White-Knuckled Travelers
The Wide World of Wikis
Repetitive Strain Injury
History of HCI The Complexities of Repetitive Strain
Highlights from My Forty Years of HCI History
Scenario-Based Design
Human-Robot Interaction The Value of a Devils Advocate
Carbo-Powered Robots
Social Psychology and HCI
Hypertext and Hypermedia Love and HCI
Ted Nelson on Hypertext and the Web
Sociology and HCI
Impacts Whos on First for the Twenty-First
Therac-25 Safety Is a System Property Century
LIST OF SIDEBARS XXI
Spell Checker Work

Check the Spell Checker Software Prescribes Break Time for Enhanced
Productivity
Task Analysis
Excerpt from Cheaper by the Dozen Workforce
Cultural Differences
Unicode Employee Resistance to Technology
History and Development of Unicode
Relationship of the Unicode Standard to World Wide Web
ISO_IEC 10646 Inventing the World Wide Web
Tim Berners-Lee on the Web as Metaphor
Usability Evaluation
Global Usability WYSIWYG
Is Usability Still a Problem? The Future of HCI
CONTRIBUTORS
Abras, Chadia Alvarez, R. Michael

Goucher College Caltech-MIT Voting Technology Project
User-Centered Design Online Voting
Agrawal, Dharma P. Aronson, Jay E.

University of Cincinnati University of Georgia
Mobile Computing Expert Systems
Allanson, Jennifer Bailenson, Jeremy

Lancaster University Stanford University
Physiology Avatars
Allison, Brendan Z. Bainbridge, Erika

Georgia State University Harvard University, Center for Hellenic Studies
Brain-Computer Interfaces Asian Script Input
xxiii
XXIV BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Bainbridge, William Sims Bilal, Dania

National Science Foundation University of Tennessee
Altair Children and the Web
Alto
Asian Script Input Bisantz, Ann M.
Converging Technologies State University of New York, Buffalo
Cyborgs Ergonomics
ENIAC
Evolutionary Engineering Blackmon, Marilyn Hughes
Hollerith Card University of Colorado, Boulder
Literary Representations Cognitive Walkthrough
Mosaic
Movies Blascovich, James J.
Personality Capture University of California, Santa Barbara
Pocket Computer Avatars
Sociology and HCI
Tablet Computer Bos, Nathan
Three-Dimensional Printing University of Michigan
E-mail
Bartle, Richard Allan
Multi-User Entertainment Limited Bradshaw, Shannon
MUDs University of Iowa
Search Engines
Barzilay, Regina
Massachusetts Institute of Technology Brusilovsky, Peter
Language Generation University of Pittsburgh
Adaptive Help Systems
Baskerville, Richard
Georgia State University Burnett, Margaret M.
Iterative Design Oregon State University
Prototyping Visual Programming
Batteau, Allen W. Cai, Guoray

Wayne State University Pennsylvania State University
Anthropology and HCI Multimodal Interfaces
Ben-Arie, Jezekiel Caldera, Cavinda T.

University of Illinois, Chicago Syracuse University
Motion Capture and Recognition Grid Computing
Bernard, Michael L. Campbell, Murray

Wichita State University IBM T.J. Watson Research Center
Website Design Deep Blue
Bhavnani, Suresh K.
University of Michigan
Application Use Strategies
CONTRIBUTORS XXV
Carroll, John M. Crosby, Martha E.

Pennsylvania State University University of Hawaii
Computer-Supported Cooperative Work Information Filtering
Scenario-Based Design
Danziger, James N.
Chaparro, Barbara S. University of California, Irvine
Wichita State University Political Science and HCI
Website Design
Davis, Adriane D.
Chen, Lawrence Georgia State University
Eastman Kodak Research Labs Brain-Computer Interfaces
Affective Computing
DeFanti, Thomas
Chen, Wenhong University of Illinois, Chicago
University of Toronto Cave
Internet Worldwide Diffusion
Delmonico, David L.
Choset, Howie Duquesne University
Carnegie Mellon University Cybersex
Search and Rescue
Dennerlien, Jack Tigh
Choueiry, Berthe Y. Harvard School of Public Health
University of Nebraska, Lincoln Repetitive Strain Injury
Constraint Satisfaction
Deshpande, Ashutosh P.
Cockburn, Andy Syracuse University
University of Canterbury Identity Authentication
Browsers
Detrie, Thomas
Cohen, Ira Arizona State University
Hewlett-Packard Research Labs, Fonts
University of Illinois, Urbana-Champaign
Affective Computing Dewan, Prasun
Microsoft Corporation
Collyer, Glenn Multiuser Interfaces
iDacta, Inc. Ubiquitous Computing
Online Education
Donath, Judith
Cook, Diane J. Massachusetts Institute of Technology
University of Texas, Arlington Sociable Media
Smart Homes
Dongarra, Jack
Crawford, Gregory P. University of Tennessee
Brown University Supercomputers
Cathode Ray Tubes
Electronic Paper Technology
Liquid Crystal Displays
XXVI BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Draper, John V. Fountain, Jane E.

Raven Research Harvard University
Telepresence Digital Government
Duchowski, Andrew T. Franklin, Judy A.

Clemson University Smith College
Eye Tracking Musical Interaction
DuPont, Brandon Freedman, Immanuel

Policy Research Institute Dr. Immanuel Freedman, Inc.
Workforce Video
El Rhalibi, Abdennour Friedman, Batya

Liverpool John Moores University University of Washington
Animation Value Sensitive Design
Games
Fussell, Susan R.
England, David Carnegie Mellon University
Liverpool John Moores University Social Psychology and HCI
Graphical User Interface
Goldman, Rich
Erickson, Thomas University of Maryland, Baltimore
IBM T. J. Watson Research Center Touchscreen
Social Proxies
Goodchild, Michael F.
Essa, Irfan University of California, Santa Barbara
Georgia Institute of Technology Geographic Information Systems
Facial Expressions
Govindaraju, V.
Farkas, David K. University at Buffalo
University of Washington Handwriting Recognition and Retrieval
Hypertext and Hypermedia Optical Character Recognition
Instruction Manuals
Gray, Wayne D.
Feathers, David Rensselaer Polytechnic Institute
State University of New York, Buffalo Errors in Interactive Behavior
Anthropometry
Griffin, Elizabeth J.
Feinman, Alexander C. Internet Behavior Consulting
Brandeis University Cybersex
Groupware
Griffiths, Jose-Marie
Fillmore, Charles J. University of Pittsburgh
International Computer Science Institute Digital Libraries
Lexicon Building
CONTRIBUTORS XXVII
Gross, Mark D. Hauck, Roslin V.

University of Washington Illinois State University
Drawing and Design Law Enforcement
Grudin, Jonathan Hedge, Alan

Microsoft Research Cornell University
Computer Science Keyboard
History of HCI Pen and Stylus Input
Gustafson, John Hennecke, Marcus

Sun Microsystems TEMIC Telefunken Microelectronic GmbH
Atanasoff-Berry Computer Speechreading
Guthrie, Ruth Hickey, Timothy J.

California Polytechnic University of Pomona Brandeis University
Information Overload Groupware
Gutierrez-Osuna, Ricardo Hodges, Larry F.

Texas A&M University University of North Carolina, Charlotte
Olfactory Interaction Virtual Reality
Guzman, Indira R. Hogan, Bernie

Syracuse University University of Toronto
User Support Internet in Everyday Life
Hakken, David Holland, Arnold

State University of New York Institute of California State University, Fullerton
Technology Fonts
Ethnography
Hollis, Ralph L.
Hall, Thad E. Carnegie Mellon University
Century Foundation Haptics
Online Voting
Hollnagel, Erik
Halverson, Christine University of Linkping
IBM T. J. Watson Research Center Task Analysis
Work
Huang, Thomas S.
Harper, Mary P. University of Illinois, Urbana-Champaign
Purdue University Affective Computing
Speech Recognition
Huff, Chuck
Harper, V. Paul Saint Olaf College
United States Patent and Trademark Office Impacts
Speech Recognition
XXVIII BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Jackson, Linda A. Klavans, Judith L.

Michigan State University Columbia University
Digital Divide Text Summarization
Gender and Computing
Kline, Ronald R.
Jensen, Michael J. Cornell University
University of California, Irvine Information Theory
Political Science and HCI
Kobsa, Alfred
Kaminka, Gal University of California, Irvine
Bar Ilan University Adaptive Interfaces
Multiagent systems
Koenig, Sven
Kazman, Richard Georgia Institute of Technology
Carnegie Mellon University Planning
Software Engineering
Krishna, C. M.
Kellogg, Wendy A. University of Massachusetts, Amherst
IBM T. J. Watson Research Center Fly-by-Wire
Social Proxies
Kruse, Amy
Kendall, Lori Strategic Analysis, Inc.
State University of New York, Purchase College Arpanet
Cybercommunities Augmented Cognition
Kettebekov, Sanshzar Laff, Mark

Oregon Health and Science University IBM T.J. Watson Research Center
Multimodal Interfaces Client-Server Architecture
Khedekar, Swapnil Lagoudakis, Michail G.

University at Buffalo Georgia Institute of Technology
Optical Character Recognition Planning
Kim, Hong-Gee Lampe, Cliff

Dankook University University of Michigan
Markup Languages Recommender and Reputation Systems
Kim, Jee-In Lane, David M.

Konkuk University Rice University
Desktop Metaphor Sonification
WYSIWYG
Kirchhoff, Katrin
University of Washington Lenhart, Amanda B.
Machine Translation Pew Internet & American Life Project
Chatrooms
CONTRIBUTORS XXIX
Lok, Benjamin C. Miller, Sonia E.

University of Florida S. E. Miller Law Firm
Three-Dimensional Graphics Law and HCI
Virtual Reality
Moore, Melody M.
Ludi, Stephanie Georgia State University
Rochester Institute of Technology Brain-Computer Interfaces
Icons
Murtagh, Fionn
Ma, Kwan-Liu Queens University, Belfast
University of California, Davis Information Spaces
Data Visualization
Neff, Gina
MacQueen, David University of California, Los Angeles
University of Chicago Beta Testing
Programming Languages
Nissenbaum, Helen
Madey, Gregory R. New York University
University of Notre Dame Ethics
Open Source Software
Olson, Gary M.
Maloney-Krichmar, Diane University of Michigan
Bowie State University Collaboratories
User-Centered Design
Olson, Judith S.
Manmatha, R. University of Michigan
University of Massachusetts, Amherst Psychology and HCI
Handwriting Recognition and Retrieval
Omojokun, Olufisayo
Martin, James H. University of North Carolina, Chapel Hill
University of Colorado, Boulder Ubiquitous Computing
Natural-Language Processing
N-grams Paik, Woojin
University of Massachusetts, Boston
May, Jon Compilers
University of Sheffield Spell Checker
Theory
Paquet, Victor L.
McKinnon, Robin A. State University of New York, Buffalo
Harvard University Anthropometry
Digital Government
Pargas, Roy
McRoy, Susan W. Clemson University
University of Wisconsin, Milwaukee Online Questionnaires
Dialog Systems
XXX BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Peres, S. Camille Rogers, Erika

Rice University California Polytechnic State University
Sonification Human-Robot Interaction
Perrig, Adrian Rosenbaum, Howard

Carnegie Mellon University Indiana University
Denial-of-Service Attack Social Informatics
Preece, Jenny Rosenbloom, Joshua L.

University of Maryland, Baltimore County University of Kansas
User-Centered Design Workforce
Quek, Francis Rosson, Mary Beth

Wright State University Pennsylvania State University
Gesture Recognition Computer-Supported Cooperative Work
Quintana, Chris Ruthruff, Joseph R.

University of Michigan Oregon State University
Classrooms Visual Programming
Quiroga, Luz M. Sandin, Dan

University of Hawaii University of Illinois, Chicago
Information Filtering CAVE
Rajlich, Vaclav Sandor, Aniko

Wayne State University Rice University
Software Cultures Sonification
Raphael, Christopher S. Scacchi, Walt

University of Massachusetts, Amherst University of California, Irvine
Musical Interaction Socio-Technical System Design
Resnick, Paul Schmorrow, Dylan

University of Michigan Defense Advanced Projects Agency
Recommender and Reputation Systems Arpanet
Augmented Cognition
Rhodes, Bradley
Ricoh Innovations Scholtz, Jean
Wearable Computer National Institute of Standards and Technology
Usability Evaluation
Rieser, John J.
Vanderbilt University Sears, Andrew L.
Navigation University of Maryland, Baltimore County
Touchscreen
CONTRIBUTORS XXXI
Sears, J. Allen Stage, Jan

Corporation for National Research Initiatives Aalborg University
Arpanet Education in HCI
Iterative Design
Selker, Ted Prototyping
Massachusetts Institute of Technology
Attentive User Interface Stanton, Jeffrey M.
Syracuse University
Sewalkar, Parag Privacy
Syracuse University
Identity Authentication Starkweather, Gary
Microsoft Corporation
Sengupta, Kuntal Laser Printer
Advanced Interfaces
Augmented Reality Starner, Thad
Georgia Institute of Technology
Sharma, Rajeev Wearable Computers
Advanced Interfaces
Augmented Reality Stephenson, Robert S.
Multimodal Interfaces Wayne State University
Online Education
Shen, Yuan Yuan
Liverpool John Moores University Sterner, William H.
Animation University of Chicago
Eliza
Simpson, Richard C.
University of Pittsburgh Tekalp, A. Murat
User Modeling University of Rochester
Video Summarization
Singh, Rita
Carnegie Mellon University Tenopir, Carol
Sphinx University of Tennessee
Electronic Journals
Soergel, Dagobert
University of Maryland Thomas, Douglas
Information Organization University of Southern California
Information Retrieval Hackers
St. Amant, Robert A. Thuraisingham, Bhavani

North Carolina State University National Science Foundation
Artificial Intelligence Security
Statistical Analysis Support Semantic Web
Tran, Phuoc
University of Toronto
Internet Worldwide Diffusion
XXXII BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Tretiakoff, Oleg Westfall, Ralph David

C.A. Technology, Inc. California State Polytechnic University, Pomona
Braille Telecommuting
Tygar, J. D. Williams, Ronald D.

University of California, Berkeley University of Virginia
Digital Cash Embedded Systems
Spamming
Viruses Wilson, Michael
CCLRC Rutherford Appleton Laboratory
Unicode Editorial Committee World Wide Web
Unicode
Witte, James
van Santen, Jan P.H. Clemson University
Oregon Health and Science University Online Questionnaires
Speech Synthesis
Yaar, Abraham
Vanderheiden, Gregg Carnegie Mellon University
University of Wisconsin, Madison Denial of Service Attack
Universal Access
Youngblood, Michael
Vassileva, Julita University of Texas, Arlington
University of Saskatchewan Smart Homes
Peer-to-Peer Architecture
Zakaria, Norhayati
Wellman, Barry Syracuse University
University of Toronto E-business
Internet - Worldwide Diffusion
Internet in Everyday Life Zaki, Mohammad
Rensselaer Polytechnic Institute
Welty, Christopher A. Data Mining
IBM T.J. Watson Research Center
Ontology Zhai, Shumin
IBM Almaden Research Center
Mouse
INTRODUCTION
By William Sims Bainbridge
In hardly more than half a century, computers have hitherto existed only in the pages of science fic-
become integral parts of everyday life, at home, tion. For a sense of the wide reach of HCI, consider
work, and play. Today, computers affect almost the following vignettes:
every aspect of modern life, in areas as diverse as
car design, filmmaking, disability services, and sex Gloria, who owns a small fitness training busi-
education. Human-computer interaction (HCI) is ness, is currently trying out a new system in which
a vital new field that examines the ways in which she and a client dance on sensor pads on the floor,
people communicate with computers, robots, in- while the computer plays rhythms and scores
formation systems, and the Internet. It draws upon how quickly they are placing their feet on the
several branches of social, behavioral, and infor- designated squares.
mation science, as well as on computer science and Elizabeth has made friends through chatrooms
electrical engineering. The traditional heart of HCI connected to French and British music groups
has been user interface design, but in recent that are not well known in the United States. She
years the field has expanded to include any science occasionally shares music files with these friends
and technology related to the ways that humans before buying CDs from foreign online distrib-
use or are affected by computing technology. utors, and she has helped one of the French bands
HCI brings to the fore social and ethical issues that translate its website into English.
xxxiii
XXXIV BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Carls work team develops drivers for new color puter recognition of pen or stylus movements on
printers far more quickly and effectively than be- tablet or pocket computers.
fore, because the team comprises expert design- All of these have been very active areas of research
ers and programmers who live in different or development since he wrote, and several are fun-
time zones around the world, from India to damental to commercial products that have already
California, collectively working 24 hours a day, appeared. For example, many companies now use
7 days a week, by means of an Internet-based col- speech recognition to automate their telephone in-
laboration system. formation services, and hundreds of thousands of
Bella is blind, but her wearable computer uses people use stylus-controlled pocket computers every
Internet and the Global Positioning System not day. Many articles in the encyclopedia describe
only to find her way through the city safely but new approaches that may be of tremendous impor-
also to find any product or service she needs at tance in the future.
the best price and to be constantly aware of her Our entire perspective on HCI has been evolving
surroundings. rapidly in recent years. In 1997, the National Research
Anderson, whose Internet moniker is Neo, dis- Councila private, nonprofit institution that pro-
covers that his entire life is an illusion, main- vides science, technology, and health policy advice
tained by a vast computer plugged directly into under a congressional charterissued a major re-
his nervous system. port, More Than Screen Deep, to evaluate and sug-
gest fruitful directions for progress in user interfaces
The first three stories are real, although the names to computing and communications systems. This
are pseudonyms, and the scenarios are duplicated high-level study, sponsored by the National Science
millions of times in the modern world of personal Foundation (NSF), concluded with three recom-
computers, office automation, and the World Wide mendations to the federal government and univer-
Web. The fourth example could be realized with to- sity researchers.
days technology, simply given a sufficient investment
1. Break away from 1960s technologies and para-
in infrastructure. Not only would it revolutionize the
digms. Major attempts should be made to find
lives of blind people like Bella, it would benefit the
new paradigms for human-machine interac-
sighted public too, so we can predict that it will in
tion that employ new modes and media for in-
fact become true over the next decade or two. The
p u t a n d o u t p u t a n d t h a t i nv o l v e n e w
story about Mr. Anderson is pure fiction, no doubt
conceptualizations of application interfaces.
recognizable to many as the premise of the 1999 film
(192)
The Matrix. It is doubtful that HCI ever could (or
2. Invest in the research required to provide the com-
should) become indistinguishable from real life.
ponent subsystems needed for every-citizen in-
terfaces. Research is needed that is aimed at both
making technological advances and gaining
Background on HCI understanding of the human and organizational
In a brief history of HCI technology published in
capabilities these advances would support. (195)
1996, the computer scientist Brad Myers noted
3. Encourage research on systems-level design and
that most computer interface technology began as
development of human-machine interfaces that
government-supported research projects in univer-
support multiperson, multimachine groups
sities and only years later was developed by corpo-
as well as individuals. (196)
rations and transformed into commercial products.
He then listed six up-and-coming research areas: In 2002, John M. Carroll looked back on the his-
natural language and speech, computer-supported tory of HCI and noted how difficult it was at first to
cooperative work, virtual and augmented reality, get computer science and engineering to pay atten-
three-dimensional graphics, multimedia, and com- tion to issues of hardware and software usability. He
INTRODUCTION XXXV
argued that HCI was born as the fusion of four fields technological revolution is likely to give computer
(software engineering, software human factors, com- technology an additional powerful boost: nanotech-
puter graphics, and cognitive science) and that it con- nology. The word comes from a unit for measuring
tinues to be an emerging area in computer science. tiny distances, the nanometer, which is one billionth
The field is expanding in both scope and importance. of a meter (one millionth of a millimeter, or one mil-
For example, HCI incorporates more and more from lionth the thickness of a U.S. dime). The very
the social sciences as computing becomes increas- largest single atoms are just under a nanometer in
ingly deeply rooted in cooperative work and human size, and much of the action in chemistry (including
communication. fundamental biological processes) occurs in the range
Many universities now have research groups between 1 nanometer and 100200 nanometers. The
and training programs in HCI. In addition to the de- smallest transistors in experimental computer
signers and engineers who create computer interfaces chips are about 50 nanometers across.
and the researchers in industry and academia who are Experts working at the interface between nano-
developing the fundamental principles for success technology and computing believe that nanoelec-
in such work, a very large number of workers in many tronics can support continued rapid improvements
industries contribute indirectly to progress in HCI. in computer speed, memory, and cost for twenty
The nature of computing is constantly changing. The to thirty years, with the possibility of further progress
first digital electronic computers, such as ENIAC (com- after then by means of integrated design approaches
pleted in 1946), were built to solve military problems, and investment in information infrastructure. Two
such as calculating ballistic trajectories. The 1950s and decades of improvement in computer chips would
1960s saw a great expansion in military uses and ex- mean that a desktop personal computer bought in
tensive application of digital computers in commerce 2024 might have eight thousand times the power
and industry. In the late 1970s, personal computers of one bought in 2004 for the same priceor could
entered the home, and in the 1980s they developed have the same power but cost only twenty cents and
more user-friendly interfaces. The 1990s saw the trans- fit inside a shirt button. Already, nanotechnology
formation of Internet into a major medium of com- is being used to create networks of sensors that can
munications, culminating in the expansion of the detect and identify chemical pollutants or biologi-
World Wide Web to reach a billion people. cal agents almost instantly. While this technology
In the first decade of the twenty-first century, will first be applied to military defense, it can be
two trends are rushing rapidly forward. One is the adapted to medical or personal uses in just a few years.
extension of networking to mobile computers and The average persons wristwatch in 2024 could
embedded devices literally everywhere. The other is be their mobile computer, telling them everything
the convergence of all mass media with computing, they might want to know about their environment
such that people listen to music, watch movies, take where the nearest Thai restaurant can be found, when
pictures, make videos, carry on telephone conversa- the next bus will arrive at the corner up the road,
tions, and conduct many kinds of business on com- whether there is anything in the air the person hap-
puters or on networks of which computers are central pens to be allergic to, and, of course, providing any
components. To people who are uncomfortable with information from the worlds entire database that
these trends, it may seem that cyberspace is swal- the person might want to know. If advances in nat-
lowing real life. To enthusiasts of the technology, it ural-language processing continue at the rate they
seems that human consciousness is expanding to en- are progressing today, then the wristwatch could also
compass everything. be a universal translator that allows the person to
The computer revolution is almost certainly speak with anyone in any language spoken on the
going to continue for decades, and specialists in face of the planet. Of course, predictions are al-
human-computer interaction will face many new ways perilous, and it may be that progress will slow
challenges in the years to come. At least one other down. Progress does not simply happen of its own
XXXVI BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
accord, and the field of human-computer interac- and data. After a while I realized I was going deaf
tion must continue to grow and flourish if comput- from the noise and took to wearing earplugs. Later,
ers are to bring the marvelous benefits to human life back at Harvard in a faculty position, I began writ-
that they have the potential to bring. ing my own statistical analysis programs for my first
personal computer, an Apple II. I remember that one
kind of analysis would take a 36 hours to run, with
My Own Experience with Computers the computer humming away in a corner as I went
Computer and information technologies have pro- about my daily life. For a decade beginning in 1983,
gressed amazingly over the past fifty years, and they I programmed educational software packages in so-
may continue to do so for the next half century. My first ciology and psychology, and after a series of com-
computer, if it deserves that word, was a Geniac I re- puter-related projects found myself running the
ceived for my sixteenth birthday in 1956. Costing only sociology program at the National Science Founda-
$20, it consisted of masonite disks, wires, light bulbs tion and representing the social and behavioral
and a vast collection of nuts, bolts, and clips. From these sciences on the major computing initiatives of
parts I could assemble six rotary switches that could be NSF and the federal government more generally.
programmed (by hardwiring them) to solve simple After eight years of that experience, I moved to the
logic problems such as playing tick-tack-toe. I devel- NSF Directorate for Computer and Information
oped a great affection for the Geniac, as I did for the Science and Engineering to run the NSFs programs
foot-long slide rule I lugged to my high school in human-computer interaction, universal access,
classes, but each was a very far cry from the pocket com- and artificial intelligence and cognitive science be-
puter or even the programmable calculator my sixteen- fore becoming deputy director of the Division of
year-old daughter carries in her backpack today. Information and Intelligent Systems, which contains
Geniac was not really an electronic computer be- these programs.
cause it lacked active componentswhich in 1956 My daughters, aged sixteen and thirteen, have
meant relays or vacuum tubes, because transistors used their considerable computer expertise to cre-
were still very new and integrated circuits had not ate the Center for Glitch Studies, a research project
yet been invented. The first real computer I saw, in to discover and analyze programming errors in com-
the early 1960s, was the massive machine used by my mercial video games. So far they have documented
fathers company, Equitable Life Insurance, to keep on their website more than 230 programming errors
its records. Only decades later did I learn that my in popular video games. The hundreds of people who
uncle, Angus McIntosh, had been part of a team in visit the website are not a passive audience, but send
World War II that seized the German computer that e-mail messages describing errors they themselves
was cracking Soviet codes, and that the secret Colossus discovered, and they link their own websites into a
computer at Bletchley Park where he worked had growing network of knowledge and virtual social
been cracking German codes. In the middle of the relationships.
twentieth century, computers were huge, rare, and
isolated from the general public, whereas at the be-
ginning of the twenty-first century they are essen- A Personal StoryNSFs FastLane
tial parts of everyday life. Computers have become vastly more important at
My first experience programming computers work over recent decades, and they have come to play
came in 1974, when I was a graduate student in the increasingly more complex roles. For example,
sociology department at Harvard University, and I NSF has created an entire online system for re-
began using the machines for statistical analysis of viewing grant proposals, called FastLane, and thou-
data. Starting the next year at the University of sands of scientists and educators have become
Washington, where I was a beginning assistant familiar with it through serving as reviewers or prin-
professor, I would sit for hours at a noisy keypunch cipal investigators.
machine, making the punch cards to enter programs
INTRODUCTION XXXVII
A researcher prepares a description of the proj- the institution, and the abstract is posted on the web
ect he or she hopes to do and assembles ancillary for anyone to see. Each year, the researcher submits
information such as a bibliography and brief biog- a report, electronically of course, and the full record
raphies of the team members. The researcher sub- of the grant accumulates in the NSF computer sys-
mits this material, along with data such as the dollar tem until the work has been completed.
requests on the different lines of the formal budget. Electronic systems connect the people
The only software required is a word processor researcher, program director, and reviewersinto a
and a web browser. As soon as the head of the in- system of information flow that is also a social sys-
stitutions grants office clicks the submit button, the tem in which each person plays a specific role. Be-
full proposal appears at NSF, with the data already cause the system was designed over a number of years
arranged in the appropriate data fields, so nobody to do a particular set of jobs, it works quite well, and
has to key it in. improvements are constantly being incorporated.
Peer review is the heart of the evaluation process. This is a prime example of Computer-Supported
As director of the HCI program, I categorize pro- Cooperative Work, one of the many HCI topics cov-
posals into review panels, then recruit panelists ered in this encyclopedia.
who were experts in the field with specializations that
matched the scope of the proposals. Each panelist re-
views certain proposals and submits a written review The Role of the Berkshire Encyclopedia
electronically.
Once the individual reviews have been submit-
of Human-Computer Interaction
Because the field of HCI is new, the Berkshire
ted, the panel meets face-to-face to discuss the
Encyclopedia of Human-Computer Interaction breaks
proposals and recommend funding for the best ones.
new ground. It offers readers up-to-date information
The panelists all have computers with Electronic
about several key aspects of the technology and its
Panel System (EPS) groupware that provides easy ac-
human dimensions, including
cess to all the proposals and reviews associated with
the particular panel. During the discussion of a par- applicationsmajor tools that serve human
ticular proposal, one panelist acts as scribe, keep- needs in particular ways, with distinctive usability
ing a summary of what was said in the EPS. Other issues.
panelists can read the summary, send written com- approachestechniques through which scien-
ments to the scribe, and may be asked to approve the tists and engineers design and evaluate HCI.
final draft online. breakthroughsparticular projects that marked
Next the NSF program officer combines all the a turning point in the history of HCI.
evaluations and writes a recommendation in the elec- challengesproblems and solutions, both tech-
tronic system, for approval by the director of the di- nical and human, especially in controversial
vision in which the program is located. More often areas.
than not, unfortunately, the decision is to decline to componentskey parts of a software or hard-
fund the proposal. In that case, the program officer ware system that are central to how people use it.
and division director processes the action quickly on disciplinesthe contributions that various sci-
their networked computers, and an electronic no- ences and academic fields make to HCI.
tification goes immediately to the principal inves- interfaceshardware or software systems that
tigator, who can access FastLane to read the reviews mediate between people and machines.
and summary of the panel discussion. methodsgeneral computer and information
In those rarer and happier situations when a science solutions to wide classes of technical
grant is awarded, the principal investigator and pro- problems.
gram officer negotiate the last details and craft an social implicationstechnological impacts on so-
abstract, describing the research. The instant the ciety and policy issues, and the potential of multi-
award is made, the money goes electronically to user HCI systems to bring about social change.
XXXVIII BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
These categories are not mutually exclusive; many of HCI. I have written occasional encyclopedia arti-
articles fit in two or more of them. For example, the cles since the early 1990s, when I was one of sev-
short article on laser printers concerns an output in- eral subject matter editors of The Encyclopedia of
terface and explains how a laser printer puts words Language and Linguistics. Often, an editor working
and pictures on paper. But this article also concerns on a specialized encyclopedia for one publisher or
a breakthrough, the actual invention of the laser another would send me an e-mail message asking
printer, and it was written by the inventor himself, if I would write a particular essay, and I would
Gary Starkweather. send it in, also by e-mail. I had a very good experi-
ence contributing to the Encyclopedia of Community,
Contributors edited by Karen Christensen and David Levinson
The 175 contributors to the encyclopedia possess the of Berkshire Publishing. I suggested to Karen that
full range and depth of expertise covered by HCI, Berkshire might want to do an encyclopedia of
and more. They include not only computer scien- human-computer interaction and that I could re-
tists and electrical engineers, but also social and cruit excellent authors for such a project. Berkshire
behavioral scientists, plus practicing engineers, sci- has extensive experience developing high-quality ref-
entists, scholars, and other experts in a wide range erence works, both in partnership with other pub-
of other fields. The oldest authors were born around lishing houses and on its own.
the time that the very first experimental digital elec- Almost all the communication to create the
tronic computer was built, and the entire history encyclopedia was carried out online. Although I know
of computing has taken place during their lives. many people in the field personally, it was a great
Among the influential and widely respected con- help to have access to the public databases placed on
tributors is Jose-Marie Griffiths, who contributed the Web by NSF, including abstracts of all grants
the article on digital libraries. As a member of the made in the past fifteen years, and to the online pub-
U.S. Presidents Information Technology Advisory lications of organizations such as the Association for
Committee, Griffiths understands the full scope and Computing Machinery and to the websites of all of
social value of this new kind of public resource. the authors, which often provide copies of their pub-
Contributors Judith S. Olson, Gary M. Olson, and lications. Berkshire created a special password-
John M. Carroll are among the very few leaders who protected website with information for authors and
have been elected to the Academy of the Special a section where I could review all the essays as they
Interest Group on Computer-Human Interaction of were submitted.
the Association for Computing Machinery (SIGCHI).
In 2003 Carroll received the organizations Life- For the Reader
time Achievement Award for his extensive accomplish- There are many challenges ahead for HCI, and many
ments, including his contributions to the Blacksburg are described in this encyclopedia. Difficult prob-
Electronic Village, the most significant experiment lems tend to have both technical and human aspects.
on community participation in computer-mediated For the benefit of the reader, the articles identify stan-
communication. Jack Dongarra, who wrote the con- dard solutions and their ramifications, both positive
tribution on supercomputers, developed the and negative, and may also cover social or political
LINPACK Benchmark, which is used to test the speed controversies surrounding the problem and its pos-
of these upper-end machines and which is the sible solutions. Many of the articles describe how a
basis of the annual list of the five hundred fastest particular scientific discipline or branch of engi-
computers in the world. neering approaches HCI, and what it contributes to
the multidisciplinary understanding of and im-
Building the Encyclopedia: provement in how computers, robots, and informa-
Computer-Supported Cooperative Work tion systems can serve human needs. Other articles
The creation of this encyclopedia is an example of focus on a particular interface, modality, or medium
computer-supported cooperative work, a main area in which people receive information and control the
INTRODUCTION XXXIX
computer or system of which it is a part. These articles Some seventy-five diverse illustrations, which range
explain the technical features of the hardware or soft- from antique photos of the ENIAC computer (c.
ware; they also explain the way humans perceive, 1940s) to cutting-edge computerized images.
learn, and behave in the particular context. Still other A bibliography of HCI books and journal
articles concern how computer and information sci- articles.
ence has developed to solve a wide class of problems, A popular culture appendix that includes more
using vivid examples to explain the philosophy of than 300 annotated entries on books, plays,
the method, paying some attention as well to the hu- movies, television shows, and songs that have
man side of the equation. connections to HCI.
Many articlessometimes as their central focus
and sometimes incidentallyexamine the social im- William Sims Bainbridge
plications of HCI, such as the impact of a particular
kind of technology, the way that the technology The views expressed are those of the author and do not
fits into societal institutions, or a social issue involving necessarily reflect the position of the National Science
computing. The technology can strengthen either Foundation
cooperation or conflict between human beings, and
the mutual relations between technological change
and social change are often quite complex. FURTHER READING
For information technology workers, this ency-
clopedia provides insight into specialties other than Asher, R. E., & Simpson, J. M. Y. (Eds.). (1994). The encyclopedia of
the one they work in and offers useful perspectives on language and linguistics. Oxford, UK: Pergamon.
the broad field. For policy makers, it provides a basis Bainbridge, W. S. (1989). Survey research: A computer-assisted intro-
duction. Belmont, CA: Wadsworth.
for thinking about the decisions we face in exploit- Bainbridge, W. S. (1992). Social research methods and statistics: A
ing technological possibilities for maximum human computer-assisted introduction. Belmont, CA: Wadsworth.
benefit. For students, this encyclopedia lays out how Carroll, J. M. (Ed.). (2002). Human-computer interaction in the new
millennium. Boston: Addison-Wesley.
to use the technology to make a better world and of- Christensen, K., & Levinson, D. (2003). Encyclopedia of community:
fers a glimpse of the rapidly changing computer-as- From the village to the virtual world. Thousand Oaks, CA: Sage.
sisted human world in which they are living their lives. Myers, B. A. (1996). A brief history of human computer interaction
To illuminate and expand on the articles them- technology. ACM Interactions, 5(2), 4454.
National Research Council. (1997). More than screen deep. Washington,
selves, the encyclopedia includes the following spe- DC: National Academy Press.
cial features: Roco, M. C., & Bainbridge, W. S. (2001). Societal implications of
nanoscience and nanotechnology. Dordrecht, Netherlands: Kluwer.
Approximately eighty sidebars with key primary Roco, M. C., & Bainbridge, W. S. (2003). Converging technologies for
text, glossary terms, quotes, and personal stories improving human performance. Dordrecht, Netherlands: Kluwer.
about how HCI has had an impact on the work
and lives of professionals in the field.
PUBLISHERS NOTE
By Karen Christensen
The Berkshire Encyclopedia of Human-Computer relationships than in binary code; but it was books
Interaction (HCI) is our first independent title. Weve and a career in publishingthat at last brought home
done many other award-winning encyclopedias but to me that computers can support and expand hu-
HCI will always have a unique place in our hearts man connections and improve our lives in myriad
and in our history. ways. Berkshire Publishing Group, based in a tiny
Even though most of our work has been in the New England town, depends on human-computer
social sciences, when William Bainbridge at the interaction to maintain working relationships, and
National Science Foundation wrote to suggest the friendships too, with many thousands of experts
topic of HCI, I knew instantly that it was the right around the world. We are convinced, in fact, that this
topic for our knowledge and technology company. topic is central to our development as a twenty-first
I grew up with the computer industry. My father, a century publishing company,
computer engineer in the Silicon Valley, tried very The Berkshire Encyclopedia of Human-Computer
hard to explain the fundamentals of computing, and Interaction takes computing into new realms, intro-
even built a machine out of plywood and blinking ducing us to topics that are intriguing both in their
lights to show my sixth-grade class that information technical complexity and because they present us
can be captured and communicated with nothing human beingswith a set of challenging questions
more than a combination of on-off switches. I was about our relationship with thinkingmachines. There
a reader, much more interested in human stories and are opportunities and risks in any new technology, and
xli
XLII BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
HCI has intrigued writers for many decades because The Berkshire Encyclopedia of Human-Computer
it leads us to a central philosophical, religious, and even Interaction provides us with an essential grounding
historical question: What does it mean to be human? in the most relevant and intimate form of technol-
Well be exploring this topic and related ones in fur- ogy, making scientific and technological research
ther works about technology and society. available to a wide audience. This topic and other as-
Bill Bainbridge was an exceptional editor: or- pects of what Bill Bainbridge likes to refer to as con-
ganized, focused, and responsive. Working with him verging technologies will continue to be a core part
has been deeply rewarding, and its no surprise of our print and online publishing program. And, as
that the hundreds of computer scientists and engi- befits a project so closely tied to electronic tech-
neers he helped us recruit to contribute to the en- n o l o g y, a n o n l i n e ve r s i o n o f t h e B e r k s h i re
cyclopedia were similarly enthusiastic and gracious. Encyclopedia of Human-Computer Interaction will be
All these expertscomputer scientists and engineers available through xrefplus. For more information,
as well as people working in other aspects of HCI visit www.xreferplus.com.
truly wanted to work with us to ensure that their
work would be accessible and understandable. Karen Christensen
To add even greater interest and richness to the CEO, Berkshire Publishing Group
work, weve added dozens of photographs, personal [email protected]
stories, glossary terms, and other sidebars. In addi-
tion to article bibliographies, there is a master bib-
liography at the end, containing all 2,590 entries in Editors Acknowledgements
the entire encyclopedia listed together for easy ref- Karen Christensen, cofounder of the Berkshire
erence. And weve added a characteristic Berkshire Publishing Group, deserves both thanks and
touch, an appendix designed to appeal to even the praise for recognizing that the time had come when
most resolute Luddite: HCI in Popular Culture, a a comprehensive reference work about human re-
database compilation listing with 300 sci-fi novels, lations with computing systems was both possible
nonfiction titles, television programs and films from and sorely needed. Courtney Linehan at Berkshire
The Six-Million Dollar Man to The Matrix (per- was both skilled and tireless in working with the au-
haps the quintessential HCI story), and even a hand- thors, editor, and copyeditors to complete a mar-
ful of plays and songs about computers and velous collection of articles that are technically
technology. accurate while communicating clearly to a broad
The encyclopedia has enabled us to develop a public. At various stages in the process of develop-
network of experts as well as a cutting-edge resource ing the encyclopedia, Marcy Ross and George
that will help us to meet the needs of students, Woodward at Berkshire made their own indispen-
professionals, and scholars in many disciplines. Many sable contributions. Among the authors, Mary
articles will be of considerable interest and value to Harper, Bhavani Thuraisingham, and Barry Wellman
librariansDigital Libraries, Information Filtering, were unstinting in their insightful advice. I would
Information Retrieval, Lexicon Building, and much particularly like to thank Michael Lesk who, as di-
moreand even to publishers. For example, we have rector of the Division of Information and Intelligent
an article on Text Summarization written by Judith Systems of the National Science Foundation, gave
Klavans, Director of Research at the Center for me the opportunity to gain invaluable experience
Advanced Study of Language, University of managing the grant programs in Universal Access
Maryland. Summarization is a technique for and Human-Computer Interaction.
identifying the key points of a document or set of
related documents, and presenting these selected William Sims Bainbridge
points as a brief, integrated independent represen- Deputy Director,
tation and is essential to electronic publishing, a key Division of Information and Intelligent Systems
aspect of publishing today and in the future. National Science Foundation
ABOUT THE EDITOR
William Sims Bainbridge is deputy director of the resented the social and behavioral sciences on five ad-
Division of Information and Intelligent Systems of vanced technology initiatives: High Performance
the National Science Foundation, after having di- Computing and Communications, Knowledge and
rected the divisions Human-Computer Interaction, Distributed Intelligence, Digital Libraries, Information
Universal Access, and Knowledge and Cognitive Technology Research, and Nanotechnology.
Systems programs. He coedited Converging Tech- Bill Bainbr idge is also the author of ten
nologies to Improve Human Performance, which books, four textbook-software packages, and some
explores the combination of nanotechnology, bio- 150 shorter publications in information science,
technology, information technology, and cognitive social science of technology, and the sociology of
science (National Science Foundation, 2002; culture. He earned his doctorate from Harvard
www.wtec.org/ConvergingTechnologies). He has rep- University.
xliv
ADAPTIVE HELP SYSTEMS
ADAPTIVE INTERFACES
AFFECTIVE COMPUTING
ALTAIR
A
ALTO
ANIMATION
ANTHROPOLOGY AND HCI
ANTHROPOMETRY
APPLICATION USE STRATEGIES
ARPANET
ARTIFICIAL INTELLIGENCE
ASIAN SCRIPT INPUT
THE ATANASOFF-BERRY COMPUTER
ATTENTIVE USER INTERFACE
AUGMENTED COGNITION
AUGMENTED REALITY
AVATARS
The first wave of research on adaptive help emerged

ADAPTIVE HELP in early 1980 when the UNIX systemdue to its low
cost and efficiencyreached many universities whose
SYSTEMS users lacked the advanced technical training (such as
knowledge of complicated commands) needed to op-
Adaptive help systems (AHSs; also called intelligent erate UNIX. Early work on adaptive and intelligent
help systems) are a specific kind of help system and a help systems focused almost exclusively on UNIX and
recognized area of research in the fields of artificial inits utilities, such as text editors and e-mail. From 1980
telligence and human-computer interaction. The goal to 1995 this research direction involved more than a
of an adaptive help system is to provide personalized hundred researchers working on at least two-dozen
help to users working with complex interfaces, from projects. The most representative projects of this gen-
operating systems (such as UNIX) to popular appli- eration were UNIX Consultant and EUROHELP. The
cations (such as Microsoft Excel). Unlike traditional widespread use of graphical user interfaces (GUIs)
static help systems that serve by request the same in- in early 1990 caused a pause in AHS research, because
formation to different users, AHSs attempt to adapt to GUIs resolved a number of the problems that the early
the knowledge and goals of individual users, offering generation of AHS sought to address. In just a few years,
the most relevant information in the most relevant however, GUIs reached the level of complexity
way. where adaptive help again became important, giving
1
2 BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION
Many AHSs use two classic approaches to model

INTERFACE Interconnections between a device, program, the user. First, they track the users actions to under-
or person that facilitate interaction. stand which commands and concepts the user
knows and which are not known, and second, they use
task models to deduce the users current goal and miss-
rise to a second wave of research on AHSs. Lumire, ing knowledge. The first technology is reasonably sim-
the most well known project of this wave, intro- ple: The system just records all used commands and
duced the idea of intelligent help to millions of users parameters, assuming that if a command is used, it
of Microsoft applications. must be known. The second is based on plan recog-
nition and advanced domain knowledge representa-
tion in such forms as a goal-plan-action tree. To identify
Active and Passive AHSs the current goal and missing pieces of knowledge, the
Adaptive help systems are traditionally divided into system first infers the users goal from an observed se-
two classes: active and passive. In a passive AHS, the
user initiates the help session by asking for help. An
active help system initiates the help session itself. Both Farewell Clippy
kinds of AHSs have to solve three challenging prob-
lems: They must build a model of user goals and knowl-
M
any PC users through the years quickly learned
edge, they must decide what to present in the next help how to turn off Clippy, the Microsoft Office
message, and they must decide how to present it. In helper who appeared out of nowhere eagerly hop-
addition, active AHSs also need to decide when to ing to offer advice to the baffled. The Microsoft press re-
intervene with adaptive help. lease below was Clippys swan song.
User Modeling REDMOND, Wash., April 11, 2001Whether

you love him or you hate him, say farewell to
To be useful, a help message has to present informa- Clippy automatically popping up on your screen.
tion that is new to the user and relevant to the users Clippy is the little paperclip with the soulful
current goal. To determine what is new and relevant, eyes and the Groucho eyebrows. The electronic
AHSs track the users goals and the users knowledge ham who politely offers hints for using Microsoft
about the interface and maintain a user model. Two Office software.
major approaches to user modeling in AHSs are ask But, after four years on-screen, Clippy will
lose his starring role when Microsoft Office XP
the user and observe the user. Most passive AHSs debuts on May 31. Clippy, the Office Assistant in-
have exploited the first of these approaches. UNIX troduced in Office 97, has been demoted in Office
Consultant demonstrates that a passive AHS can be XP. The wiry little assistant is turned off by de-
fairly advanced: It involves users in a natural-language fault in Office XP, but diehard supporters can
dialogue to discover their goals and degree of knowl- turn Clippy back on if they miss him.
edge and then provides the most relevant information. Office XP is so easy to use that Clippy is
no longer necessary, or useful, explained Lisa
In contrast, active AHSs, introduced by the com- Gurry, a Microsoft product manager. With new
puter scientist Gerhard Fischer in 1985, strive to de- features like smart tags and Task Panes, Office XP
duce a users goals by observing the user at work; they enables people to get more out of the product
then strive to identify the lack of knowledge by de- than ever before. These new simplicity and ease-
tecting errors and suboptimal behavior. EURO- of-use improvements really make Clippy obso-
HELP provides a good example of an active help system lete, she said.
Hes quite down in the dumps, Gurry joked.
capable of identifying a knowledge gap and filling it He has even started his own campaign to try
provocatively. In practical AHSs the two approaches to get his old job back, or find a new one.
often coexist: The user model is initiated through a Source: Microsoft. Retrieved March 10, 2004, from
short interview with the user and then kept updated http://www.microsoft.com/presspass/features/2001/apr01/04-11clippy.asp
through observation.
ADAPTIVE INTERFACES 3
quence of commands. It then tries to find a more ef- intelligence and HCI and has helped to establish re-
ficient (or simply correct) sequence of commands to search on intelligent interfaces and user modeling. A
achieve this goal. Next, it identifies the aspects of the treasury of knowledge accumulated by various AHS
interface that the user needs to know to build this se- projects over the last thirty years is being used now
quence. These aspects are suspected to be unknown to develop practical adaptive help and adaptive per-
and become the candidates to be presented in help formance support systems.
messages.
Peter Brusilovsky
Providing Adaptive Help:
Deciding What to Present and How See also Artificial Intelligence; Task Analysis; User
Deciding what should be the focus of the next help Modeling
message is the most challenging job of an adaptive help
system. A number of passive AHSs simply avoid this
problem, allowing the users to determine what they FURTHER READING
need and focusing on adaptive presentation only.
Classic AHSs, which use plan recognition, can deter- Brusilovsky, P., Kobsa, A., & Vassileva, J. (Eds.). (1998). Adaptive hy-
pertext and hypermedia. Dordrecht, Netherlands: Kluwer.
mine quite precisely what the user needs, but this func- Encarnao, L. M., & Stoev, S. L. (1999). An application-independent
tionality requires elaborate knowledge representation. intelligent user support system exploiting action-sequence based
To bypass the knowledge representation barrier, mod- user modeling. In J. Kay (Ed.), Proceedings of 7th International
ern practical AHSs use a range of alternative Conference on User Modeling, UM99, June 2024, 1999 (pp.
245254). Vienna: Springer.
(though less precise) technologies that are either sta- Fischer, G. (2001). User modeling in human-computer interaction.
tistically or socially based. For example, Lumire used User Modeling and User-Adapted Interaction, 11(12), 6586.
a complex probabilistic network to connect ob- Goodman, B. A., & Litman, D. J. (1992). On the interaction between
plan recognition and intelligent interfaces. User Modeling and User-
served user actions with available help interventions, Adapted Interaction, 2(1), 83115.
while the system developed by MITRE researchers Hegner, S. J., Mc Kevitt, P., Norvig, P., & Wilensky, R. L. (Eds.). (2001).
Linton and Schaefer compared the skills of individual Intelligent help systems for UNIX. Dordrecht, Netherlands: Kluwer.
users with a typical set of interface skills assembled by Horvitz, E., Breese, J., Heckerman, D., Hovel, D., & Rommelse, K.
(1998). The Lumire project: Bayesian user modeling for inferring
observing multiple users. the goals and needs of software users. In Proceedings of Fourteenth
As soon as the focus of the next help message is Conference on Uncertainty in Artificial Intelligence (pp. 256265).
determined, the AHS has to decide how to present the San Francisco: Morgan Kaufmann.
target content. While some AHSs ignore this part and Linton, F., & Schaefer, H.-P. (2000). Recommender systems for learn-
ing: Building user and expert models through long-term obser-
focus solely on the selection part, it has been shown vation of application use. User Modeling and User-Adapted
that adaptive presentation of help information can in- Interaction, 10(23), 181208.
crease the users comprehension speed and decrease Oppermann, R. (Ed.). (1994). Adaptive user support: Ergonomic de-
sign of manually and automatically adaptable software. Hillsdale,
errors. Most often the content presentation is adapted NJ: Lawrence Erlbaum Associates.
to the users knowledge, with, for example, expert users Wilensky, R., Chin, D., Luria, M., Martin, J., Mayfield, J., & Wu, D.
receiving more specific details and novice users re- (1988). The Berkeley UNIX Consultant project. Computational
ceiving more explanations. To present the adaptive Linguistics, 14(4), 3584.
Winkels, R. (1992). Explorations in intelligent tutoring systems and help.
content, classic AHSs that operated in a line-based Amsterdam: IOS Press.
UNIX interface relied mostly on a natural language
generation approach. Modern AHSs operating in the
context of Graphical User Interfaces exploit adaptive
hypermedia techniques to present the content and links
to further information that is most suitable for the ADAPTIVE INTERFACES
given user.
Research into adaptive help systems has contributed Computer interfaces are becoming ever richer in
to progress in a number of subfields within artificial functionality, software systems are becoming more
complex, and online information spaces are be- perform certain indicative actions (such as entering
coming larger in size. On the other hand, the num- certain keywords in search machines).
ber and diversity of people who use computer systems User adaptability and adaptivity recently gained
are increasing as well. The vast majority of new users strong popularity on the World Wide Web under the
are thereby not computer experts, but rather layper- notion of personalization. This popularity is due
sons such as professionals in nontechnical areas, eld- to the fact that the audiences of websites are often
erly people, and children. These users vary with even less homogeneous than the user populations of
respect not only to their computer skills, but also commercial software. Moreover, personalization has
to their fields of expertise, their tasks and goals, their been recognized as an important instrument for on-
mood and motivation, and their intellectual and line customer relationship management.
physical capabilities.
The traditional strategy for enabling heteroge-
neous user groups to master the complexity and rich- Acquiring Information about Users
ness of computers was to render computer interaction To acquire the information about users that is needed
as simple as possible and thereby to cater to the low- to cater to them, people can use several methods. A
est common denominator of all users. Increasingly, simple way is to ask users directly, usually through an
though, developers are creating computer applica- initial questionnaire. However, this questionnaire must
tions that can be manually customized to users be kept extremely short (usually to less than five ques-
needs by the users themselves or by an available ex- tions) because users are generally reluctant to spend
pert. Other applications go beyond this capability. efforts on work that is not directly related to their cur-
They are able within certain limits to recognize rent tasks, even if this work would save them time in
user needs and to cater to them automatically. the long run. In certain kinds of systems, specifically
Following the terminology of Reinhard Oppermann, tutoring systems, user interviews can be clad in the
we will use the term adaptable for the manual type form of quizzes or games. In the future, basic infor-
of application and adaptive for the automatic type. mation about users may be available on smartcards,
that is, machine-readable plastic cards that users swipe
through a reading device before the beginning of a
Adaptable and Adaptive Systems computer session or that can even be read from a dis-
Adaptable systems are abundant. Most commercial tance as users approach a computer terminal.
software allows users to modify system parameters Various methods draw assumptions about
and to indicate individual preferences. Web portals users based on their interaction behavior. These
permit users to specify the information they want to methods include simple rules that predict user
see (such as stock quotes or news types) and the form characteristics or assign users to predetermined
in which it should be displayed by their web browsers. user groups with known characteristics when cer-
Web shops can store basic information about their tain user actions are being observed (the latter
customers, such as payment and shipping data, past method is generally known as the stereotype
purchases, wish lists for future purchases, and birth- approach to user modeling). Probabilistic reason-
dates of friends and family to facilitate transac- ing methods take uncertainty and evidences from
tions online. In contrast, adaptive systems are still different sources into account. Plan recogni-
quite rare. Some shopping websites give purchase tion methods aim at linking individual actions of
recommendations to customers that take into ac- users to presumable underlying plans and goals.
count what these customers bought in the past. Machine-learning methods try to detect regu-
Commercial learning software for high school math- larities in users actions (and to use the learned
ematics adapts its teaching strategies to the presumed patterns as a basis for predicting future actions).
level of expertise of each student. Advertisements on Clique-based (collaborative) filtering methods
mobile devices are already being targeted to users in determine those users who are closest to the cur-
certain geographical locations only or to users who rent user in an n-dimensional attribute space and
ADAPTIVE INTERFACES 5
Keeping Disabled People in the Technology Loop
AUSTIN, Texas (ANS)If communications technology is been defined in recent years. While the proliferation of
fueling the economy and social culture of the 21st century, computers, home gadgets and gizmos is on the rise, its
why should 18 percent of the population be left behind? workers like Berger who make sure the disabled arent left
Stephen Berger, a specialist in retrofitting the latest out of the loop.
computer and phone technology for the disabled, is Other workers in the field, according to Berger, are
trying to make sure theyre not. coming from educational institutions. For example, Neil
From an office in Austin, Berger works to make sure Scott and Charlie Robinson, from Stanford University
that those with hearing and vision impairments or other and Louisiana Tech University respectively, are working
disabilities can benefit from the latest in Internet, cell on the things the Hollywood movies are made of.
phone and other technologies. []
As a project manager at Siemens Information and Guys like this are breaking the barrier between the
Communication Mobile, where hes responsible for stan- blind and computers, he said.(The blind) will soon have
dards and regulatory management, Berger works to un- an interface with no visual, just audio computer controls
ravel such problems as why those who use hearing aids with no touch, just head position and voice controls.
couldnt use many brands of cell phones. Other devices, like the Home RF systemsthats home
Some new cell phones make a buzz in hearing aids, radio frequencylink all the major appliances and
Berger explained. The Federal Communications electronics of the home together. That means tele-
Commission took note and said it needed to be resolved. phone, Dolby sound, Internet, entertainment electronics
But what was needed was either better technology or and other devices are all connected into one wireless net-
protocols that both the hearing impaired and the cell work with voice control for those who arent mobile.
phone companies could agree on. Berger helped deter- Its microphones implanted in wallpaper, security
mine what types of hearing aids work with certain systems by voice, household appliances that work on a
types of phones. The intelligence was passed around the vocal command, Berger said. Its what the movies are
industry, and the problem is now minimal. made of and its here today.
Berger is one of the many technology specialists in Source: Innovations keep disabled in the technology loop. American
News Services, October 12, 2000.
huge communications companies whose niche has
use them as predictors for unknown attributes of Usage data, such as selections (e.g., of webpages
the current user. Clustering methods allow one to or help texts with certain content), temporal
generalize groups of users with similar behaviors viewing behavior (particularly skipping of
or characteristics and to generate user stereotypes. webpages or streaming media), user ratings (e.g.,
regarding the usefulness of products or the rel-
evance of information), purchases and related
Types of Information about the User actions (e.g., in shopping carts, wish lists), and
Researchers have considered numerous kinds of
usage regularities (such as usage frequencies,
user-related data for personalization purposes,
high correlations between situations and spe-
including the following:
cific actions, and frequently occurring sequences
Data about the user, such as demographic data, of actions)
and information or assumptions about the Environmental data, such as data about the
users knowledge, skills, capabilities, interests, users software and hardware environments and
preferences, goals, and plans information about the users current location
(where the granularity ranges from country level dents learning (Conati et al. 2000, 404). Corbett and
to the precise coordinates) and personalization- Trask showed that a certain tutoring strategy (namely
relevant data of this location. subgoal scaffolding based on a continuous knowl-
edge trace of the user) decreases the average number
of problems required to reach cognitive mastery of
Privacy Lisp concepts. In studies reviewed by Specht and
Storing information about users for personaliza- Kobsa, students learning time and retention of learn-
tion is highly privacy relevant. Numerous con- ing material improved significantly if learners with
sumer surveys show consistently that users are low prior knowledge received strict recommen-
concerned about their privacy online, which also dations on what to study next (which amounted to
affects personalized systems on the Web. Some the blocking of all other learning material), while stu-
popular personalization methods also seem in con- dents with high prior knowledge received noncom-
flict with privacy laws that protect the data of iden- pulsory recommendations only. Strachan and
tified or identifiable individuals in more than thirty colleagues found significantly higher user ratings for
countries. Such laws usually call for parsimony, pur- the personalized version of a help system in a com-
pose specificity, and user awareness or even user con- mercial tax advisor system than for its nonperson-
sent in the collecting and processing of personal data. alized version.
The privacy laws of many countries also restrict the Personalization for e-commerce on the Web
transborder flow of personal data or even extend their has also been positively evaluated to some extent,
coverage beyond the national boundaries. Such laws both from a business and a user point of view. Jupiter
then also affect personalized websites abroad that Communications reports that personalization at
serve users in these regulated countries, even if there twenty-five consumer e-commerce sites boosted the
is no privacy law in place in the country where the number of new customers by 47 percent and revenues
websites are located. Well-designed user interaction by 52 percent in the first year. Nielsen NetRatings re-
will be needed in personalized systems to commu- ports that registered visitors to portal sites (who
nicate to users at any point the prospective benefits obtain the privilege of adapting the displayed in-
of personalization and the resulting privacy conse- formation to their interests) spend more than three
quences to enable users to make educated choices. times longer at their home portal than other users
A flexible architecture, moreover, will be needed to and view three to four times more pages. Nielsen
allow for optimal personalization within the con- NetRatings also reports that e-commerce sites offer-
straints set by users privacy preferences and the le- ing personalized services convert approximately twice
gal environment. Alternatively, anonymous yet as many visitors into buyers than do e-commerce sites
personalized interaction can be offered. that do not offer personalized services. In design stud-
ies on beneficial personalized elements in a Web-based
procurement system, participants, however,expressed
Empirical Evaluation their strong desire to have full and explicit control of
A number of empirical studies demonstrate in sev- data and interaction and to readily be able to make
eral application areas that well-designed adaptive user sense of site behavior, that is, to understand a sites
interfaces may give users considerable benefits. Boyle rationale for displaying particular content (Alpert
and Encarnacion showed that the automatic adet al. 2003, 373).
justment of the wording of a hypertext document to User-adaptable and user-adaptive interfaces have
users presumed familiarity with technical vocabu- shown their promise in several application areas. The
lary improved text comprehension and search increase in the number and variety of computer users
times significantly in comparison with static hyper- is likely to increase their promise in the future. The
text. Conati and colleagues presented evidence that observation of Browne still holds true, however:
adaptive prompts based on the student model ef- Worthwhile adaptation is system specific. It is de-
fectively elicited self-explanations that improved stu- pendent on the users of that system and requirements
AFFECTIVE COMPUTING 7
to be met by that system (Browne 1993, 69). Careful Specht, M., & Kobsa, A. (1999). Interaction of domain expertise and inter-
user studies with a focus on expected user benefits face design in adaptive educational hypermedia. Retrieved March 24,
2004, from http://wwwis.win.tue.nl/asum99/specht/specht.html
through personalization are, therefore, indispensa- Strachan, L., Anderson, J., Sneesby, M., & Evans, M. (2000). Minimalist
ble for all practical deployments. user modeling in a complex commercial software system. User
Modeling and User-Adapted Interaction, 10(23), 109146.
Alfred Kobsa Teltzrow, M., & Kobsa, A. (2004). Impacts of user privacy preferences
on personalized systemsA comparative study. In C.-M. Karat,
See also Artificial Intelligence and HCI; Privacy; User J. Blom, & J. Karat (Eds.), Designing personalized user experiences
Modeling for e-commerce (pp. 315332). Dordrecht, Netherlands: Kluwer
Academic Publishers.
FURTHER READING
Alpert, S., Karat, J., Karat, C.-M., Brodie, C., & Vergo, J. G. (2003). User AFFECTIVE COMPUTING
attitudes regarding a user-adaptive e-commerce web site. User
Modeling and User-Adapted Interaction, 13(4), 373396. Computations that machines make that relate to hu-
Boyle, C., & Encarnacion, A. O. (1994). MetaDoc: An adaptive hy-
pertext reading system. User Modeling and User-Adapted Interaction,
man emotions are called affective computations.
4(1), 119. Such computations include but are not limited to
Browne, D. (1993). Experiences from the AID Project. In M. Schneider- the recognition of human emotion, the expression
Hufschmidt, T. Khme, & U. Malinowski (Eds.), Adaptive user of emotions by machines, and direct manipulation
interfaces: Principles and practice (pp. 6978). Amsterdam: Elsevier.
Carroll, J., & Rosson, M. B. (1989). The paradox of the active user. In of the human users emotions. The motivation for
J. Carroll (Ed.), Interfacing thought: Cognitive aspects of human- the development of affective computing is derived
computer interaction (pp. 80111). Cambridge, MA: MIT Press. from evidence showing that the ability of humans
Conati, C., Gertner, A., & VanLehn, K. (2002). Using Bayesian networks
to manage uncertainty in student modeling. User Modeling and
to feel and display emotions is an integral part of hu-
User-Adapted Interaction, 12(4), 371417. man intelligence. Emotions help humans in areas
Corbett, A. T., & Trask, H. (2000). Instructional interventions in com- such as decisionmaking and human-to-human com-
puter-based tutoring: Differential impact on learning time and ac- munications. Therefore, it is argued that in order
curacy. Proceedings of ACM CHI 2000 Conference on Human Factors
in Computing Systems (pp. 97104).
to create intelligent machines that can interact ef-
Hof, R., Green, H., & Himmelstein, L. (1998, October 5). Now its YOUR fectively with humans, one must give the machines
WEB. Business Week (pp. 6875). affective capabilities.
ICONOCAST. (1999). More concentrated than the leading brand. Although humans interact mainly through
Retrieved August 29, 2003, from http://www.iconocast.com/is-
sue/1999102102.html speech, we also use body gestures to emphasize
Kobsa, A. (2002). Personalized hypermedia and international privacy. certain parts of the speech and as one way to display
Communications of the ACM, 45(5), 6467. Retrieved August 29, emotions. Scientific evidence shows that emo-
2003, from http://www.ics.uci.edu/~kobsa/papers/2002-CACM-
kobsa.pdf
tional skills are part of what is called intelligence.
Kobsa, A., Koenemann, J., & Pohl, W. (2001). Personalized hyperme- A simple example is the ability to know when some-
dia presentation techniques for improving customer relationships. thing a person says to another is annoying or
The Knowledge Engineering Review, 16(2), 111155. Retrieved August pleasing to the other, and be able to adapt accord-
29, 2003, from http://www.ics.uci.edu/~kobsa/papers/2001-KER-
kobsa.pdf
ingly. Emotional skills also help in learning to dis-
Kobsa, A., & Schreck, J. (2003). Privacy through pseudonymity in user- tinguish between important and unimportant things,
adaptive systems. ACM Transactions on Internet Technology, 3(2), an integral part of intelligent decision-making. For
149183. Retrieved August 29, 2003, from http://www.ics.uci.edu/ computers to be able to interact intelligently with
~kobsa/papers/2003-TOIT-kobsa.pdf
Oppermann, R. (Ed.). (1994). Adaptive user support: Ergonomic design humans, they will need to have such emotional skills
of manually and automatically adaptable software. Hillsdale, NJ: as the ability to display emotions (for example,
Lawrence Erlbaum. through animated agents) and the ability to recog-
Rich, E. (1979). User modeling via stereotypes. Cognitive Science, 3,
329354.
nize the users emotions. The ability to recognize
Rich, E. (1983). Users are individuals: Individualizing user models. emotions would be useful in day-to-day interaction,
International Journal of Man-Machine Studies, 18, 199214. for example, when the user is Web browsing or
lem with this approach is that the humans often feel

FUNCTIONALITY The capabilities of a given program or blended emotions. In addition, the choice of words
parts of a program. may be too restrictive or culturally dependent. Another
way to describe emotions is to have multiple dimen-
sions or scales. Instead of choosing discrete labels, emo-
searching: If the computer can recognize emotions are describe on several continuous scales, for
tions, it will know if the user is bored or dissatis- example from pleasant to unpleasant or from simple
fied with the search results. Affective skills might also to complicated. Two common scales are valence and
be used in education: A computer acting as a virtual arousal.Valence describes the pleasantness of the stim-
tutor would be more effective if it could tell by stu- uli, with positive (or pleasant) on one end and nega-
dents emotional responses that they were having dif- tive (or unpleasant) on the other. For example,
ficulties or were boredor pleased. happiness has a positive valence, while disgust has a
It would not, however, be necessary for com- negative valence. The other dimension is arousal, or
puters to recognize emotions in every application. activation, which describes the degree to which the
Airplane control and banking systems, for example, emotion stimulates the person experiencing it. For ex-
do not require any affective skills. However, in ap- ample, sadness has low arousal, whereas surprise has
plications in which computers take on a social role high arousal. The different emotional labels could be
(as a tutor, assistant, or even companion), it may en- plotted at various positions on a two-dimensional plane
hance their functionality if they can recognize users spanned by these two axes to construct a two-di-
emotions. Computer agents could learn users mensional emotion model. In 1954 the psychologist
preferences through the users emotions. Computers Harold Schlosberg suggested a three-dimensional
with affective capabilities could also help human model in which he added an axis for attention-rejec-
users monitor their stress levels. In clinical set- tion to the above two. This was reflected by facial ex-
tings, recognizing a persons inability to interpret pressions as the degree of attention given to a person
certain facial expressions may help diagnose early or object. For example, attention is expressed by
psychological disorders. In addition to recognizing wide open eyes and an open mouth. Rejection shows
emotions, the affective computer would also have contraction of eyes, lips, and nostrils.
the ability to display emotions. For example, syn- Although psychologists and others argue about
thetic speech with emotions in the voice would sound what exactly emotions are and how to describe them,
more pleasing than a monotonous voice and everyone agrees that a lack of emotions or the
would enhance communication between the user presence of emotional disorders can be so dis-
and the computer. For computers to be affective, they abling that people affected are no longer able to lead
must recognize emotions, be capable of measuring normal lives or make rational decisions.
signals that represent emotions, and be able to
synthesize emotions.
Technology for Recognizing Emotions
Technologies for recognizing human emotions be-
General Description of Emotions gan to develop in the early 1990s. Three main modal-
Human beings possess and express emotions in every- ities have been targeted as being relevant for this task:
day interactions with others. Emotions are often visual, auditory, and physiological signals. The vi-
reflected on the face, in hand and body gestures, and sual modality includes both static images and videos
in the voice. The fact that humans understand containing information such as facial expressions
emotions and know how to react to other peoples and body motion. The audio modality uses prima-
expressions greatly enriches human interaction. rily human voice signal as input, while the physio-
There is no clear definition of emotions. One way logical signals measure changes in the human body,
to handle emotions is to give them discrete labels, such such as changes in temperature, blood pressure, heart
as joy, fear, love, surprise, sadness, and so on. One prob- rate, and skin conductivity.
AFFECTIVE COMPUTING 9
Facial Expressions of the universality of how emotions are expressed

One of the most common ways for humans to display vocally, unlike the case for facial expressions. Research
emotions is through facial expressions. The best-known that began in the late 1990s concentrated on com-
study of facial expressions was done by the psychol- bining voice and video to enhance the recognition
ogist Paul Ekman and his colleagues. Since the capabilities of voice-only systems.
1970s, Ekman has argued that emotions are manifested
directly in facial expressions, and that there are six ba-
sic universal facial expressions corresponding to Multimodal Input
happiness, surprise, sadness, fear, anger, and disgust. Many researchers believe that combining different
Ekman and his collaborator, the researcher Wallace modalities enables more accurate recognition of a
Friesen, designed a model linking facial motions to ex- users emotion than relying on any single modality
pression of emotions; this model is known as the Facial alone. Combining different modalities presents both
Action Coding System (FACS). The facial action technological and conceptual challenges, however.
coding system codes facial expressions as a combina- On the technological side, the different signals
tion of facial movements known as action units. The have different sampling rates (that is, it may take
action units have some relation to facial muscular mo- longer to register signals in one modality than in an-
tion and were defined based on anatomical knowledge other), and the existence of one signal can reduce the
and by studying videotapes of how the face changes reliability of another (for example, when a person is
its appearance. speaking, facial expression recognition is not as re-
Ekmans work inspired many other researchers to liable). On the conceptual side, emotions are not al-
analyze facial expressions by means of image and video ways aligned in time for different signals. For
processing. Although the FACS is designed to be example, happiness might be evident visually before
performed by human observers viewing a video frame it became evident physiologically.
by frame, there have been attempts to automate it in
some fashion, using the notion that a change in fa-
cial appearance can be described in terms of a set of Computers Displaying Emotions
facial expressions that are linked to certain emo- For affective computing, it is as important that com-
tions. Work on automatic facial-expression recogni- puters display emotions as it is that they recognize them.
tion started in the early 1990s. In all the research, some There are a number of potential ways in which com-
method to extract measurements of the facial features puters could evince an emotion.A computer might de-
from facial images or videos was used and a classifier pend on facial expressions of animated agents or on
was constructed to categorize the facial expressions. synthesized speech, or emotion could be conveyed to
Comparison of facial expression recognition methods a user through wearable devices and text messages. The
shows that recognition rates can, on limited data sets method would be determined by the application do-
and applications, be very high. The generality of these main and preset goals. For example, interaction in an
results has yet to be determined. office environment requires emotion to be expressed
differently from the way it would be expressed during
pursuit of leisure activities, such as video games; sim-
Voice ilarly, in computer-assisted tutoring, the computers
Quantitative studies of emotions expressed in the goal is to teach the human user a concept, and the
voice have had a longer history than quantitative display of emotions facilitates this goal, while in a game
studies of facial expressions, starting in the 1930s. of poker, the computers goal is to hide its intention and
Studies of the emotional content of speech have deceive its human adversary.
examined the pitch, duration, and intensity of the A computer could also synthesize emotions in
utterance. Automatic recognition systems of emo- order to make intelligent decisions with regard to prob-
tions from voice have so far not achieved high ac- lems whose attributes cannot all be quantified ex-
curacy. In addition, there is no agreed-upon theory actly, or when the search space for the best solution is
large. By assigning valence to different choices based

on different emotional criteria, many choices in a large ALTAIR
space can be eliminated quickly, resulting in a quick
and good decision. People have called the legendary Altair the first true
personal computer. However, although it played
an important role in the development of per-
Research Directions sonal computers, we would be more correct to say
Affective computing is still in its infancy. Some com- that Altair was the last hobby computer that set
puter systems can perform limited recognition of hu- many of the ambitions for the social movement
man emotion with limited responses in limited that produced real personal computers during the
application domains. As the demand for intelligent 1970s. Thus, from the standpoint of human-com-
computing systems increases, however, so does the need puter interaction, Altair is worth remembering be-
for affective computing. cause it marked a crucial transition between two
Various moral issues have been brought up as eras of amateur computing: an experimental era
relevant in the design of affective computers. Among lasting from about 1950 until the mid-1970s, when
them are privacy issues: If a computer can recognize home computers were among the more esoteric
human emotions, a user may want assurances that projects attempted by electronics hobbyists, and
information on his or her emotional state will not be the true personal computer era, beginning with
abused. There are also issues related to computers ma- such computers as the Apple II in 1977.
nipulation of peoples emotions: Users should have as- Altair was announced to the world late in 1974
surance that computers will not physically or in the issue of Popular Electronics magazine
emotionally harm them. There are also questions re- dated January 1975. Some controversy exists about
garding who will have responsibility for computer how active a role the magazine played in launch-
actions. As affective technology advances, these issues ing Altair, but clearly Altair was actually designed
will have increasing relevance. and manufactured by the small company Micro
Instrumentation and Telemetry Systems (MITS)
Ira Cohen, Thomas S. Huang, Lawrence S. Chen in Albuquerque, New Mexico, headed by H. Edward
Ed Roberts. Altair was a kit, costing $397, that
required much skill from the builder and did not
FURTHER READING include sufficient memory or input-output devices
to perform any real tasks. The central processing
Darwin, C. (1890). The expression of the emotions in man and animals unit was the new Intel 8080 microprocessor chip.
(2nd ed.). London: John Murray. Altair came with only 256 bytes of memory, and
Ekman, P. (Ed.). (1982). Emotion in the human face (2nd ed.). New
York: Cambridge University Press.
a notoriously unreliable 4-kilobyte memory ex-
Ekman, P., & Friesen,W.V. (1978). Facial action coding system: Investigators pansion board kit cost an additional $264. After
guide. Palo Alto, CA: Consulting Psychologists Press. a while MITS offered data input and output by
James, W. (1890). The principles of psychology. New York: Henry Holt. means of an audio cassette recorder, but initially
Jenkins, J. M., Oatley, K., and Stein, N. L. (Eds.). (1998). Human emotions:
A reader. Malden, MA: Blackwell.
only a dedicated amateur was in a practical posi-
Lang, P. (1995). The emotion probe: Studies of motivation and attention. tion to add a keyboard (perhaps a used teletype
American Psychologist, 50(5), 372385. machine) or punched paper tape reader. Input-
Pantic, M., & Rothkrantz, L. J. M. (2000). Automatic analysis of facial ex- output for the original computer was accomplished
pressions: The state of the art. IEEE Transactions on Pattern Analysis
and Machine Intelligence, 22(12), 14241445. by switches and lights on the front of the cabinet.
Picard, R. W. (1997). Affective computing. Cambridge, MA: MIT Press. Popular Electronics hyped the Altair as if it were
Picard, R. W., Vyzas, E., & Healey, J. (2001). Toward machine emotional a fully developed minicomputer and suggested
intelligence: Analysis of affective physiological state. IEEE Transactions
Pattern Analysis and Machine Intelligence, 23(10), 11751191.
some excessively demanding applications: an au-
Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review, topilot for airplanes or boats, a high-speed in-
61(2), 8188. put-output device for a mainframe computer, a
ALTAIR 11
brain for a robot, an automatic controller for an

air-conditioning system, and a text-to-Braille con- Personal computers are notorious for having a half-life
verter to allow blind people to read ordinary of about two years. In scientific terms, this means that
printed matter. Altair rescued MITS from the verge two years after you buy the computer, half of your friends
of bankruptcy, but the company could never will sneer at you for having an outdated machine.
fully deliver on the promise of the computer and Peter H. Lewis
was absorbed by another company in 1977.
Hundreds of amateurs built Altairs on the way to
careers in the future personal computer industry,
its subcomponent interface bus became the widely soldering each wire, capacitor, and resistor into
used S100 standard, and the computer contributed place manually. Any of these parts might burn out
greatly to the revolution in human-computer in- in use, so repair shops flourished, and companies
teraction that occurred during its decade. Notably, such as Allied Radio and Lafayette Electronics sold
the mighty Microsoft corporation began life as a individual parts to hobbyists and to anyone else
tiny partner of MITS, producing a BASIC inter- who was willing to buy. For the novice, these dis-
preter for programming the Altair. tributors sold kits that provided all the parts needed
Altair was far from being the first hobby to build a project, and more advanced amateurs
computer, however. That honor probably be- followed instructions in a number of magazines to
longs to Edmund C. Berkeleys Simon relay- build projects from parts they bought separately
based computer produced in 1950 and publicized from distributors.
among hobbyists in the pages of Radio-Electronics In purely financial terms building a stereo sys-
magazine. The most widely owned hobby digital tem from a kit during the 1960s, as tens of thou-
computer before Altair was probably Berkeleys sands of people did, made little sense, but the result
GENIAC (Genius Almost-Automatic Computer), was fully as good as the best ready-made system
which cost less than twenty dollars in 1955. Lacking that could be bought in stores, and in some cases
vacuum tubes, relays, or transistors, this assem- the designs were identical. The introduction of in-
bly of Masonite board, rotary switches, lights, and tegrated circuits gradually reduced the role of re-
wires instructed students in the rudiments of logic pairpersons, and by the dawn of the twenty-first
programming (programming the steps of logical century much electronic equipment really could
deductions). Immediately prior to Altair, two less not be repaired effectively and was simply replaced
influential hobby computers were also based on when it broke down. Already by the late 1970s
Intel chips: the Scelbi 8H and the Titus Mark-8. the electronics hobby was in decline, and the home-
The difference is that Altair was expandable and built computer craze during that decade was prac-
intended to evolve into a full-featured personal tically a last fling.
computer. For a decade after the introduction of Altair,
The 1970s marked a turning point in the his- a vibrant software hobbyist subculture prevailed
tory of hobby electronics, and innovative proj- as people manually copied programs from a host
ects such as Altair could be seen as desperate of amateur computer magazines, and many peo-
measures in the attempt to keep the field alive. ple brewed their own personally designed word
Today some enthusiasts build electronic equipment processors and fantasy games. This subculture de-
from kits or from scratch, just as others build their clined after the introduction of complicated graph-
own harpsichords, but they no longer have the same ical user interface operating systems by Apple and
relationship to the electronics industry that they Microsoft, but it revived during the mid-1990s as
enjoyed during the middle decades of the twenti- vast numbers of people created their own web-
eth century. Prior to the development of integrated sites in the initially simple HTML (hypertext
circuits, factories constructed radios, televisions, markup language). During its heyday this sub-
and audio amplifiers largely by hand, laboriously culture was a great training ground of personnel
for the electronics and computer industries because Data Systems (SDS), and it established Xerox PARC
amateurs worked with the same technology that pro- near Stanford University in the area that would soon
fessionals worked with. Altair was a watershed per- be nicknamed Silicon Valley. Xerox proclaimed the
sonal computer in the sense that amateurs assembled grand goal of developing the general architecture of
it personally and that it transformed them person- information rather than merely producing a num-
ally into computer professionals. ber of unconnected, small-scale inventions.
The Alto was part of a larger system of software
William Sims Bainbridge and hardware incorporating such innovations as ob-
ject-oriented programming, which assembles pro-
See also Alto grams from many separately created, reusable
objects, the ethernet LAN, and laser printers. At the
time computers were large and expensive, and a com-
FURTHER READING mon framework for human-computer interaction
was time sharing: Several users would log onto a
Freiberger, P., & Swaine, M. (1999). Fire in the valley: The making of the mainframe or minicomputer simultaneously from
personal computer (2nd ed.). New York: McGraw-Hill.
Roberts, H. E., & Yates, W. (1975). Altair minicomputer. Popular
dumb terminals, and it would juggle the work from
Electronics, 7(1), 3338. all of the users simultaneously. Time sharing was
Roberts, H. E., & Yates, W. (1975). Altair minicomputer. Popular an innovation because it allowed users to interact
Electronics, 7(2), 5658. with the computer in real time; however, because the
Mims, F. M. (1985, January). The tenth anniversary of the Altair 8800.
Computers & Electronics, 23(1), 5860, 8182. computer was handling many users it could not de-
vote resources to the HCI experience of each user. In
contrast, Alto emphasized the interface between the
user and the machine, giving each user his or her own
ALTO computer.
In April 1973 the first test demonstration of an
The Alto computer, developed at the Xerox Alto showed how different using it would be from us-
Corporations Palo Alto Research Center (Xerox ing the text-only computer terminals that people were
PARC) in the 1970s, was the prototype of the late used to when it began by painting on its screen a pic-
twentieth-century personal computer. Input was by ture of the Cookie Monster from the television
means of both keyboard and mouse; the display program Sesame Street. The Altos display em-
screen integrated text and graphics in a system of win- ployed bitmapping (controlling each pixel on the
dows, and each computer could communicate with screen separately) to draw any kind of diagram,
others over a local area network (LAN). The Alto was picture, or text font, including animation and pull-
significant for human-computer interaction (HCI) down menus. This capability was a great leap forward
in at least three ways. First, it established a new dom- for displaying information to human beings, but it
inant framework for how humans would interact with required substantial hardware resources, both in terms
computers. Second, it underscored the importance of memory size and processing speed, as well as rad-
of theory and research in HCI. Third, the failure of ically new software approaches.
Xerox to exploit Alto technology by gaining a dom- During the 1970s the typical computer display
inant position in the personal computer industry is consisted of letters, numbers, and common punctu-
a classic case study of the relationship between in- ation marks in a single crude font displayed on a black
novators and the technology they create. background in one color: white or green or amber.
During the late 1960s the Xerox Corporation was In contrast, the default Alto display was black on
aware that it might gradually lose its dominant posi- white, like printed paper. As originally designed,
tion in the office copier business, so it sought ways of the screen was 606 pixels wide by 808 pixels high, and
expanding into computers. In 1969 it paid $920 mil- each of those 489,648 pixels could be separately con-
lion to buy a computer company named Scientific trolled. The Xerox PARC researchers developed sys-
ANIMATION 13
tems for managing many font sizes and styles si- computing. In contrast, the model that flourished
multaneously and for ensuring that the display screen during the 1980s was autonomous personal com-
and a paper document printed from it could look puting based on stand-alone computers such as the
the same. All this performance placed a heavy bur- Apple II and original IBM PC, with networking de-
den on the computers electronics, so an Alto often veloping fully only later. The slow speed and lim-
ran painfully slow and, had it been commercialized, ited capacity of the Alto-like Lisa and original
would have cost on the order of $15,000 each. 128-kilobyte Macintosh computers introduced by
People have described the Alto as a time ma- Apple in 1983 and 1984 suggest that Alto would re-
chine, a computer that transported the user into ally not have been commercially viable until 1985,
the office of the future, but it might have been too a dozen years after it was first built.
costly or too slow to be a viable personal com- One lesson that we can draw from Altos history
puter for the average office or home user of the is that corporate-funded research can play a deci-
period in which it was developed. Human-computer sive role in technological progress but that it can-
interaction research of the early twenty-first cen- not effectively look very far into the future. That role
tury sometimes studies users who are living in the may better be played by university-based labora-
future. This means going to great effort to create an tories that get their primary funding from govern-
innovation, such as a computer system or an envi- ment agencies free from the need to show immediate
ronment such as a smart home (a computer-con- profits. On the other hand, Xerox PARC was so spec-
trolled living environment) or a multimedia tacularly innovative that we can draw the opposite
classroom, that would not be practical outside the lessonthat revolutions in human-computer in-
laboratory. The innovation then becomes a test bed teraction can indeed occur inside the research lab-
for developing future systems that will be practical, oratories of huge corporations, given the right
either because the research itself will overcome some personnel and historical circumstances.
of the technical hurdles or because the inexorable
progress in microelectronics will bring the costs William Sims Bainbridge
down substantially in just a few years.
Alto was a remarkable case study in HCI with See also Altair; Graphical User Interface
respect to not only its potential users but also its cre-
ators. For example, the object-oriented program-
ming pioneered at Xerox PARC on the Alto and FURTHER READING
other projects changed significantly the work of pro-
grammers. Such programming facilitated the sepa- Hiltzik, M. (1999). Dealers of lightning: Xerox PARC and the dawn of
the computer age. New York: HarperBusiness.
ration between two professions: software engineering Lavendel, G. (1980). A decade of research: Xerox Palo Alto Research
(which designs the large-scale structure and func- Center. New York: Bowker.
tioning of software) and programming (which writes Smith, D. C., & Alexander, R. C. (1988). Fumbling the future: How
the detailed code), and it increased the feasibility of Xerox invented, then ignored the first personal computer. New
York: William Morrow.
dividing the work of creating complex software Waldrop, M. M. (2001). The dream machine: J. C. R. Licklider and the
among many individuals and teams. revolution that made computing personal. New York: Viking.
People often have presented Alto as a case study
of how short sighted management of a major cor-
poration can fail to develop valuable new technol-
ogy. On the other hand, Alto may have been both
too premature and too ambitious. When Xerox fi- ANIMATION
nally marketed the Alto-based Star in 1981, it was a
system of many small but expensive computers, con- Animation, the creation of simulated images in mo-
nected to each other and to shared resources such tion, is commonly linked with the creation of car-
as laser printersa model of distributed personal toons, where drawn characters are brought into play
to entertain. More recently, it has also become a side and a cycle of still images on the inside that
significant addition to the rich multimedia mate- show an image in graduating stages of motion.
rial that is found in modern software applications Whenever the barrel spins rapidly, the dark frames
such as the Web, computer games, and electronic of the still pictures disappear and the picture appears
encyclopedias. to move. Another, even simpler example is the flipbook,
a tablet of paper with a single drawing on each page.
When the book is flicked through rapidly, the draw-
Brief History ings appear to move.
Animations are formed by showing a series of still pic- Once the basic principles of animation were dis-
tures rapidly (at least twelve images per second) so that covered, a large number of applications and techniques
the eye is tricked into viewing them as a continuous emerged. The invention of these simple animation de-
motion. The sequence of still images is perceived as vices had a significant influence on the development
motion because of two phenomena, one optical of films, cartoons, computer-generated motion graph-
(persistence of vision) and one psychological (phi prin- ics and pictures, and more recently, of multimedia.
ciple). Persistence of vision can be explained as the pre-
disposition of the brain and eye to keep on seeing a
picture even after it has moved out of the field of vi- Walt Disney and Traditional
sion. In 1824 British scientist, physician, and lexicog-
rapher Peter Mark Roget (17791869) explained this
Animation Techniques
During the early to mid-1930s, animators at
phenomenon as the ability of the retina to retain the
Walt Disney Studios created the twelve animation
image of an object for 1/20 to 1/5 second after its re-
principles that became the basics of hand-drawn car-
moval; it was demonstrated two years later using a
toon character animation. While some of these prin-
thaumatrope, which is a disk with images drawn on
ciples are limited to the hand-drawn cartoon animation
both sides that, when twirled rapidly, gives the illusion
genre, many can be adapted for computer animation
that the two images are combined together to form
production techniques. Here are the twelve principles:
one image.
The other principle is the phi phenomenon or stro- 1. Squash and stretchUse shape distortion to em-
boscopic effect. It was first studied by German psy- phasize movement.
chologist Max Wertheimer (18801943) and 2. AnticipationApply reverse movement to pre-
German-American psycho-physiologist Hugo pare for and bring out a forward movement.
Munsterberg (18631916) during the period from 1912 3. StagingUse the camera viewpoint that best
to 1916. They demonstrated that film or animation shows an action.
watchers form a mental connection that completes the 4. Straight-ahead vs. pose-to-pose actionApply
action frame-to-frame, allowing them to perceive a se- the right procedure.
quence of motionless images as an uninterrupted 5. Follow-through and overlapping actionAvoid
movement. This mental bridging means that even if stopping movement abruptly.
there are small discontinuities in the series of 6. Slow-in and slow-outAllow smooth starts and
frames, the brain is able to interpolate the missing de- stops by spacing frames appropriately.
tails and thus allow a viewer to see a steady movement. 7. ArcsAllow curved motion in paths of action.
In the nineteenth century, many animation de- 8. Secondary actionsAnimate secondary actions
vices, such as the zoetrope invented by William George to bring out even more life.
Horner (17861837), the phenakistiscope (1832), the 9. TimingApply time relations within actions
praxinoscope (1877), the flipbook, and the thaumatrope to create the illusion of movement.
were direct applications of the persistence of vision. 10. ExaggerationApply caricature to actions and
For example, the zoetrope is a cylindrical device timing.
through which one can see an image in action. The ro- 11. Solid drawingLearn and use good drawing
tating barrel has evenly spaced peepholes on the out- techniques.
ANIMATION 15
12. AppealCreate and animate appealing or a running character. In sprite-based animation, a

characters. single image or sequence of images can be at-
tached to a sprite. The sprite can animate in one place
Traditional animation techniques use cel anima- or move along a path. Many techniquesfor exam-
tion in which images are painted on clear acetate sheets ple, tiling, scrolling, and parallax have been devel-
called cels. Animation cels commonly use a layering oped to process the background layer more efficiently
technique to produce a particular animation frame. and to animate it as well. Sometimes sprite-based
The frame background layer is drawn in a separate animation is called path-based animation. In path-
cel, and there is a cel for each character or object that based animation, a sprite is affixed to a curve drawn
moves separately over the background. Layering through the positions of the sprite in consecutive
enables the animator to isolate and redraw only the frames, called a motion path. The sprite follows
parts of the image that change between consecutive this curve during the course of the animation. The
frames. There is usually a chief animator who sprite can be a single rigid bitmap (an array of pix-
draws the key-frames, the ultimate moments in the els, in a data file or structure, which correspond bit
series of images, while in-between frames are drawn for bit with an image) that does not change or a se-
by others, the in-betweeners. Many of the processes ries of bitmaps that form an animation loop. The
and lingo of traditional cel-based animation, such as animation techniques used by computers can be
layering, key-frames, and tweening (generating im- frame-by-frame, where each frame is individually cre-
mediate frames between two images to give the ap- ated, or real-time, where the animator produces
pearance that the first image evolves smoothly into the key-frames and the computer generates the
the next), have carried over into two-dimensional frames in between when the animation is dis-
and three-dimensional computer animation. played at run time.
Two-dimensional computer animation tech-
Two-Dimensional niques are widely used in modern software and
can be seen in arcade games, on the Web, and even
Computer Animation in word processors. The software used to design two-
In recent years, computer programs have been de- dimensional animations are animation studios
veloped to automate the drawing of individual that allow animators to draw and paint cels, provide
frames, the process of tweening frames between key- key-frames with moving backgrounds, use multiple
frames, and also the animation of a series of frames. layers for layering, support linking to fast video disk
Some animation techniques commonly used in two- recorders for storage and playback, and allow scans
dimensional (2D) computer animation are either to be imported directly. Examples of this software
frame-based or sprite-based. include Adobe Photoshop (to create animated GIFs),
Frame-based animation is the simplest type of Macromedia Director (multimedia authoring tool
animation. It is based on the same principle as the that includes sophisticated functions for animation),
flipbook, where a collection of graphic files, each con- Macromedia Flash (vector-based authoring tool to
taining a single image, is displayed in sequence and produce real-time animation for the Web). Even some
performs like a flipbook. Here again, to produce programming languages such as Java are used to pro-
the illusion of motion, graphic images, with each im- duce good quality animation (frame-by-frame and
age slightly different from the one before it in the se- real-time) for the Web.
quence, are displayed at a high frame-rate (the number
of frames of an animation displayed every second). Three-Dimensional
Sprite-based animation uses a technique that is
similar to the traditional animation technique in Computer Animation
which an object is animated on top of a static graphic Three-dimensional computer animations are based
background. A sprite is any element of an animation on a three-dimensional (3D) coordinate system,
that moves independently, such as a bouncing ball which is a mathematical system for describing
three-dimensional space. Space is measured along Skeletal Structure Animation

three coordinates, the X direction, Y direction, and Skeletal structures are bones-based. Widely used to
Z direction. These coordinates correspond to the control three-dimensional creatures, they appear in
width, length, and depth of objects or space. The practically all modern three-dimensional model-
X, Y, Z coordinates of points in space are used to de- ing software studios. They enable the artist to preset
fine polygons, and collections of polygons make and control the rotation points of a three-dimen-
up the definition of three-dimensional objects. sional creature, facilitating its animation. The ani-
The process of 3D animation involves at least the mator can then model a geometric skin (representing
following stages: modeling, rendering, and anima- how the creature would appear) and link it to the
tion. Modeling is the process of creating 3D ob- bones structure. Skeletal structures software with
jects from simple 2D objects by lofting (the process graphical and powerful interfaces provide rich en-
of transforming a two-dimensional cross section ob- vironments in which artists can control the complex
ject into a complete three-dimensional object) or algorithms involved in creating animated three-
from other simple 3D objects called primitives dimensional creatures (human, animal, or imaginary).
(spheres, cubes, cylinders, and so on). Primitives can The availability of a skeletal animation environment
be combined using a variety of Boolean operations characteristically brings another advantagethe
(union, subtraction, intersection, and so on). They exploitation of inverse kinematics (IK) to bring a char-
can also be distorted in different ways. The resulting acter to life.
model is called a mesh, which is a collection of
faces that represent an object. Rendering is used to Inverse Kinematics
create an image from data that represents objects IK is a common technique for positioning multi-
as meshes, and to apply colors, shading, textures, and linked objects, such as virtual creatures. When using
lights to them. In its simplest form, the process of an animation system capable of IK, a designer can
three-dimensional computer animation is very sim- position a hand in space by grabbing the hand and
ilar to the two-dimensional process of key-frames leading it to a position in that space. The con-
and tweening. The main differences are that three- nected joints rotate and remain connected so that,
dimensional animations are always vector-based and for example, the body parts all stay connected. IK
real-time. provides a goal-directed method for animating a 3D
creature. It allows the animator to control a three-
Spline-Based Animation dimensional creatures limbs by treating them as a
Motion paths are more believable if they are curved, kinematics chains. The points of control are attached
so animation programs enable designers to create to the ends of these chains and provide a single han-
spline-based motion paths. (Splines are algebraic dle that can be used to control a complete chain.
representations of a family of curves.) To define IK enables the animator to design a skeleton system
spline-based curves, a series of control points is de- that can also be controlled from data sets gener-
fined and then the spline is passed through the con- ated by a motion capture application.
trol points. The control points define the beginning
and end points of different parts of the curve. Each Motion Capture
point has control handles that enable designers to Motion capture is the digital recording of a creatures
change the shape of the curve between two control movement for immediate or postponed analysis and
points. The curves and the control points are de- playback. Motion capture for computer character an-
fined in 3D space. Most computer animation sys- imation involves the mapping of human action onto
tems enable users to change the rate of motion along the motion of a computer character. The digital data
a path. Some systems also provide very sophisti- recorded can be as simple as the position and ori-
cated control of the velocity of an object along entation of the body in space, or as intricate as the
paths. deformations of the expression of the visage.
ANTHROPOLOGY AND HCI 17
Advances in work in animation lies in physic-based modeling in

which objects or natural phenomena are animated
Three-Dimensional Animation according to their real physical properties, in real-
With the support of powerful computers, three-
time motion capture, and in goal-orientated ani-
dimensional animation allows the production and
mation. Considering the numerous applications of
rendering of a photo-realistic animated virtual world.
animation, from multimedia to archeology and
Three-dimensional scenes are complex virtual envi-
chemistry, the future possibilities seem endless.
ronments composed of many elements and effects,
such as cameras, lights, textures, shading, and envi-
Abdennour El Rhalibi and Yuanyuan Shen
ronment effects, and all these elements can be ani-
mated. Although cel animation is traditionally
See also Data Visualization; Graphic Display;
two-dimensional, advances in three-dimensional
Graphical User Interface
rendering techniques and in camera animation have
made it possible to apply three-dimensional tech-
niques to make two-dimensional painted images ap-
FURTHER READINGS
pear visually three-dimensional. The 3D animation
techniques described in this section are supported CoCo, D. (1995). Real-time 3D games take off. Computer Graphics
by modern 3D animation studios that are software World, 8(12), 2233.
programs such as Maya (alias|wavefront), Softimage Corra, W. T., Jensen, R. J., Thayer, C. E., & Finkelstein, A. (1998).
Texture mapping for cel animation. In Proceedings of SIGGRAPH 98,
(Softimage), 3D Studio Max (Discreet), or Rhino3D Computer Graphics Proceedings, Annual Conference Series (pp. 435446).
(Robert McNeel & Associates). Kerlow, I. V. (2000). The art of 3-D computer animation and imaging
Examples of environment effects include rain, (2nd ed.). New York: Wiley.
fire, fog, or dying stars. A technique widely used in Lassiter, J. (1987). Principles of traditional animation applied to 3D
computer animation. SIGGRAPH 87 (pp. 3544).
real-time applications involving an environmental Maltin, L. (1987). Of mice and magicA history of American animated
effect is called a particle system. A particle system is cartoons. New York: Penguin Books.
a method of graphically producing the appearance ORourke, M. (1995). Principles of three-dimensional computer ani-
mation. New York: W. W. Norton.
of amorphous substances, such as clouds, smoke,
Parent, R. (2001). Computer animation: Algorithms and techniques.
fire, or sparkles. The substance is described as a col- San Francisco: Morgan-Kaufmann.
lection of particles that can be manipulated dy- Potter, C. D. (1995). Anatomy of an animation. Computer Graphics
namically for animation effects. Some even more World, 18(3). 3643.
Solomon, C. (1994). The history of animation: Enchanted drawings.
recent techniques include physics-based behavior such New York: Wings Books.
as a realistic animation of cloth, hair, or grass affected Thomas, F., & Johnson, O. (1981). The illusion of life. New York:
by the wind. Abbeville Press.
Watt, A., & Policarpo, F. (2001). 3D gamesreal-time rendering and
software technology. New York: Addison-Wesley.
Watt, A. H., & Watt, M. (1992). Advanced animation and rendering.
Endless Possibilities New York: Addison-Wesley.
Williams, R. (2001). The animators survival kit. New York: Faber &
Animation has become an ubiquitous component Faber.
of human-computer interfaces. It has evolved
from prehistoric paintings in Altamira caves to re-
alistic virtual worlds in sophisticated multimedia
computers. The technologies supporting animation ANTHROPOLOGY AND HCI
are still emerging and will soon support even more
complex worlds, more realistic character animation, As a social science that brings together social anthro-
considerably easier 3D animation development, bet- pology, linguistics, archaeology, and human biology,
ter quality animations on the Web, and better inter- anthropology clearly has a major contribution to
actions with virtual reality interfaces. The current make to the study of human-computer interaction
(HCI). However, bringing that contribution into fo- A third concept, actually a method, employed by
cus is at times a challenge, not only because of the anthropologists in the study of HCI is actor-network
extreme interdisciplinarity but also because of collab- theory. This theory views artifacts and social roles as
orations between anthropologists and computer sci- coevolving nodes in a common network. Insofar as
entists and the sometimes-blurred boundaries each node encodes information about the entire net-
between anthropology and related disciplines, includ- work (for example, in any country, electrical appli-
ing sociology and psychology. Despite these challenges, ances are tailored to the specific power system of the
anthropology has created distinctive methods and a country and the expectations of the users) and is ca-
distinctive epistemology, and has offered new insights pable of state changes based on network inputs, both
for understanding human-computer interaction. artifacts and social roles can be considered to have
Anthropology also poses three profound questions. agency within the network. This concept, originally
developed by the sociologist Michel Callon, in his study
of the French governments involvement in techno-
Methods logical projects, and elaborated by the sociologist John
Anthropologys development of ethnographic meth- Law in a study of Portuguese sailing vessels in the
ods is a notable contribution to research in HCI. More sixteenth century, is very pertinent to rapidly chang-
than simple naturalistic observation, ethnography is ing technologies such as computers. Indeed, observ-
a structured process informed by theoretical models ing the shifting topology of the Internet and Internet
through which researchers attempt to elucidate the co- computing makes it clear that user roles are anticipated
herence of a context. For example, anthropologist and complemented by machine behavior (for instance,
Bonnie Nardi, in her study of end-user computing used collaborative filtering), and machine states enable or
concepts of formalisms and communication to inter- constrain users agency within the network (for ex-
pret how users developed their own programs; an- ample, the structures of search engines). Although sil-
thropologist Lucy Suchman used a mechanistic concept icon and carbon units are distinct, for now, the
of cognition as a foil to understand how users inter- image of the cyborg (cybernetic organism), and the
acted with an expert-system-based help facility em- emergence of integrated biological/computational sys-
bedded in a copying machine. In both these cases tems, suggests other possibilities.
researchers combined intensive naturalistic observa- This hints at the final, and perhaps most impor-
tion with conceptual insights to develop new HCI tant anthropological contribution to HCI, the evolu-
models. tionary perspective. All branches of anthropology have
A frequently employed variation on ethnographic been concerned with the evolution of human societies,
methods is called ethnomethodology. As originally languages, and even genotypes. Although there is room
developed by sociologist Harold Garfinkel, eth- for debate over the telos or chaos of evolutionary
nomethodology stipulates that individuals make processes, understanding humans and their artifacts
sense out of a context in an ad hoc, almost indeter- as goal-seeking objects who learn is fundamental to
minate manner. In place of social order, the actors any anthropological viewpoint. Using the archaeo-
in a given context are synthesizing what appears to logical record and anthropological knowledge of so-
be order, accepting or rejecting information as it fits cieties with simpler toolkits, the anthropologist David
with their synthesis. The mutual intelligibility of Hakken has questioned the extent to which the
an interaction is thus an ongoing achievement be- widespread use of computers in society justifies being
tween the actors, a result rather than a starting point. called a revolution; he concludes that due to their
Thus, two users can construct two quite different failure to transform the character of labor, computers
meanings out of similar interactions with comput- are just one more technology in the implementation
ers, depending on the experiences they bring to the of an automated, massified Fordist model of pro-
interaction. This suggests some obvious limita- ductiona model inspired by Henry Ford in which
tions on the abilities of computers to constrain or large quantities of products are produced through the
reproduce human actions. repetitive motions of unskilled workers.
Epistemology of passage for most anthropologists. When re-

What distinguishes anthropology from other disciplines searchers have lived for an extended period of time
such as psychology and sociology that use similar in an unfamiliar village, cut off from their normal
methods is in many ways a matter of epistemology social moorings, when cultural disorientation be-
that is, the stance it takes toward the subject mat- comes embedded in their daily routine, they acquire
ter. Central to this stance is the orthogonal view, that a profound conviction that all social forms are
is, the ability to analyze a situation from a fresh conventional, that otherness is not alien, and that
and original yet plausible perspective. It is the or- belonging and familiarity are rare and fragile flow-
thogonal view that enables anthropologist Constance ers. It is this experience and this conviction more
Perin to see office automation as a panopticon, than any methodological or conceptual apparatus
that suggests to linguistic anthropologist Charlotte that defines anthropology and that enables the or-
Linde that failed communication can improve per- thogonal view.
formance, or that led Edwin Hutchins, a cognitive It is this implicitly critical stance that has con-
anthropologist, to understand a cockpit as a cogni- strained anthropologys contribution to the study
tive device. Orthogonal viewpoints originate from of automation human factors. Human factors is
the experience of fieldwork, or rather, field immer- an engineering discipline using engineering meth-
sion, preferably in a remote setting, which is the rite ods of analytic decomposition to solve engineering
A Personal StoryEastern vs. Western Cultural Values
My understanding of human communication using mediated technologies is primarily based on cultural assumptions.
Cultural values could influence the way a human chooses its medium of communication. On the other hand, with the
advancement of computer-mediated communication (CMC) technologies (e.g., e-mail, e-commerce sites, weblogs, bul-
letin boards, newsgroups) people could also change their communication patterns to suit the different forms of a medium.
Whichever way, apparently, people will not adopt CMC unless and until it fits with their cultural values. Based on my in-
terviews with a number of informants from different cultural backgrounds, I have observed some disparate yet interesting
views on communication patterns and preferences, i.e., why and when people use CMC. Let me briefly illustrate one case
of contrasting communication preferences and patterns.
When I asked the informants from Eastern cultures why they would use CMC, one of the key responses was that they
can express themselves better over mediated technologies than to voice their opinions in face-to-face. Public self-expres-
sion is avoided due to the value of saving face. Also, using asynchronous medium such as e-mail, does not require
spontaneous response. People could first think, reflect, and then express. On the contrary, the informants from Western
cultures felt that using e-mail is best for complex and detailed information, as they require very explicit forms of instruc-
tions. Additionally, people send messages via CMC in order to get quick response so that tasks can get completed. Also,
based on a written format, the text becomes an evidence or proof of say for a job accomplished. Getting a job or as-
signment done is perceived as a priority and building a relationship is thus secondary.
Cultural values could present a new lens to understand why and how certain a communication medium offers differ-
ent functions or purposes. What is more important is the uniqueness of human beings with a set of cultural assump-
tions and values, and not the technological features. Anthropologist Edward T. Hall postulates that communication is
culture and culture is communication. Hence, organizations need to understand fully the myriad cultural preferences be-
fore making a substantial investment in CMC technology. Without such understanding, technology will simply be another
gadget that gets rusty and dusty!
Norhayati Zakaria
Digital Technology Helps Preserve Tribal Language
(ANS)The American Indian language of Comanche of American Indian Studies at the University of South
was once taught through conversationa vocabulary Dakota in Vermillion. Students at that university, located
passed on and polished as it moved from one genera- in the midst of a large Sioux community, are increasingly
tion to the next. But as fluency among Comanches de- interested in learning indigenous languages, he said.
clines, the tribe has turned to cutting-edge technology to Under a federal policy of discouraging use of American
preserve this indigenous language. By next winter, mem- Indian languages by allowing only English to be spoken
bers hope to produce an interactive CD-ROM that will by American Indian children at schools run by the Bureau
create a digital record of the language and help tribe mem- of Indian Affairs, Comanche began faltering about 50
bers learn it. years ago.
You cant say youre Comanche without knowing Without preservation efforts, researchers predict that
your own language. Thats the way I feel, said Billie Kreger 90 percent of the worlds languages, including those of
of Cache, Okla., vice president of the Comanche Language the 554 American Indian tribes, will disappear in the next
and Cultural Preservation Committee. Kreger, 47, did- century, said Peg Thomas, executive director of The Grotto
nt learn much Comanche as a child but has begun study- Foundation, a nonprofit organization in St. Paul,
ing it in the past few years. Minn., that provides funding to American Indian organ-
Of the 10,000 Comanches that still remain in the izations. Each year about five languages fall into ex-
United States, roughly 200 are fluent, according to Karen tinction, meaning that they have no youthful speakers,
Buller, president and chief executive officer of the Santa she said.
Fe, N.M.-based organization that is paying for the CD- According to some estimates, between 300 and 400
ROM project, the first of its kind in the United States. American Indian languages have become extinct since
Tribe members are anxious to record the language European settlers first arrived in North America.
while the fluent speakers, who are in their 70s and 80s, The point of preserving the languages is partly to
are still living, she said. maintain a connection to the past and learn the history
Bullers group, the National Indian Telecom- of a culture, said Buller. Students of the Comanche lan-
munications Institute, is paying for the project with guage discover, for instance, that the words for food prepa-
$15,000 in grant money from the Fund for the Four ration are based on the root word for meatbecause
Directions. meat was a key part of the Comanche diet. She and oth-
The CD-ROM will teach about 1,500 vocabulary ers say that American Indian children who learn indige-
words. Students will see Comanche elders pronouncing nous languages in addition to English appear to perform
the words and hear the words used in conversations. better in school.
Bullers group is recording conversations on videotape. But language programs are targeting adults, too.
Other indigenous language revitalization efforts are Kreger, of the Comanche Language and Cultural
under way around the country, too, including language Preservation Committee, says she is looking forward to
immersion programs in Alaskan and Hawaiian schools. using the CD-ROM for her own language studies. I
The institute provided teacher training for those projects. can hardly wait, she said.
All the tribes are saying, Weve got to save the lan- Nicole Cusano
guage, said Leonard Bruguier, who heads the Institute Source: Digital technology helps preserve tribal language. American
News Service, June 15, 2000.
problemsin other words, the improved perform- that is, constructthe context of information, a
ance of artifacts according to some preestablished structure of irreducible complexity. The context is
set of specifications. Anthropology, by contrast, far more than simply a compilation of information.
would begin by questioning the specifications, adopt- Computers and other information technologies,
ing a holistic point of view toward the entire proj- by contrast, focus on the processing of information,
ect. Holism is the intellectual strategy of grasping stripping information of its contextual properties
the entire configuration rather than breaking it down and thus of the attributes that humans use to turn
into separate elements. From an anthropological information into (warranted, usable, and meaning-
viewpoint, specifications are not a given, but open ful) knowledge.
to interrogation. A holistic viewpoint requires John Seely Brown, the former director of Xerox
that the researcher adopt multiple disciplinary tools, Palo Alto Research Center, and researcher Paul
including (but certainly not limited to) direct ob- Duguid, for example, describe the importance of
servation, interviewing, conversation analysis, en- context for using information. The news, for in-
gineering description, survey research, documentary stance, is not simply unfiltered information from a
study, and focus groups. For many, anthropology is distant place; it is information that has been selected,
highly interdisciplinary, assembling research tools aggregated, evaluated, interpreted, and warranted by
as the contextualized problem requires. human journalists, trained in face-to-face classrooms
How far the anthropologist is permitted to go or mentored by over-the-shoulder coaches.
with this approach is one of the dilemmas of an- Physicality is an important component of these re-
thropologists working in software design. The emerg- lationships: Although people can learn technical skills
ing fields of design ethnography and user-centered online, they learn integrity and morality only in-
design have employed ethnographers to better terpersonally. Making a convincing case for the crit-
understand users requirements, and to elicit expert icality of context for human users, Brown and
knowledge in the construction of expert systems. Duguid describe six of the context-stripping
However, these efforts are at times compromised by mechanisms that are supposedly inherent in infor-
a substantial disconnect between the anthropolo- mation technologies: demassification, decentraliza-
gists understanding of requirements and knowledge, tion, denationalization, despacialization,
and the eng ineers understanding of them. disintermediation, and disaggregation. These are
Anthropologists see human needs (that is, require- said to represent forces that, unleashed by infor-
ments) as emergent rather than given, and knowl- mation technology, will break society down into its
edge (even expert knowledge) as embedded in a fundamental constituents, primarily individuals and
culturally contingent body of assumptions called information (Brown and Duguid 2000, 22). The
common sense. Many systems designers, as the late sum of their argument is that such 6D thinking
medical anthropologist Diana Forsythe put it, view is both unrealized and unrealizable. Information
common sense as unproblematic and universal. This technology does not so much eliminate the social
assumption and others will be discussed below. context of information, for this is either pointless or
impossible, as it displaces and decomposes that con-
text, thus posing new difficulties for users who need
Insights to turn information into knowledge.
The most important anthropological insight to HCI Contexts can be high (rich, detailed, and full of
is the emphasis on context for understanding hu- social cues) or low (impoverished and monochro-
man behavior, including human interaction with cy- matic), they can be familiar or unfamiliar, and they
bernetic devices. The human organism is unique can include information channels that are broadband
in its ability to integrate information from a vari- (a face-to-face conversation) or narrowband (read-
ety of sensory inputs and to formulate an infinite ar- ing tea leaves, for example). From a human perspec-
ray of potential behavioral responses to these inputs. tive, all computer interaction, even the most
These arrays of inputs and responses constitute multimedia-rich, is narrowband: Sitting where I am,
my computer screen and keyboard occupy no more ble human action, according to Suchman, it must not
than 25 percent of my field of vision, until I turn my attempt to anticipate every user state and response (for
head. Looking around, the percentage shrinks to unit cannot). Alternatively, a strategy of real-time user
der 5 percent. The other 95 percent is filled with other modeling that incorporates (a) continually updated
work and information storage devices (bookshelves models of user behavior, (b) detection (and adapta-
and filing cabinets), task aids (charts on the wall), tion to) diagnostic inconsistencies, (c) sensitivity to
and reminders of relationships: a social context. As a local conditions, and (d) learning from fault states
low-context device, the computer must be supple- (such as false alarms and misleading instructions) sug-
mented by these other more social artifacts if it is gests a better approximation of situated action than
to have human usefulnessthat is, if it is to be preconceived user models.
used for knowledge work rather than mere infor- Suchmans findings are based on the concept
mation processing. of distributed cognition originally developed by
Applying the concept of context to a specific tech- Edwin Hutchins. Instead of understanding cogni-
nological problem, the design of intelligent systems, tion as information processing (searching, aggre-
Suchman developed a concept of situated action as gating, parsing, and so on), Hutchins saw mental
an alternative explanation for the rationality of activity as contextually emergent, using contextual
human action. In place of seeing activity as the ex- resources (including language and artifacts) as
ecution of a plan (or program), or inversely, seeing part of an interactive process.
a plan as a retrospective rationalization of activity, These insights are derived from efforts to use an-
Suchmans concept of situated action sees plans as thropological methods in the development of expert
only one of several resources for making sense out systems and other artificial intelligence devices.
of the ongoing flow of activity. Expert systems hold out the hope that in class-
Human action, or more accurately interaction (for room instruction, in routine bureaucratic problem
all action is by definition social, even if only one ac- solving, in medical diagnosis, and in other fields, cer-
tor is physically present), is an ongoing flow of mes- tain low-level mental tasks could be accomplished
sage input and output. Traditionally social studies have by computers, in much the same manner as repeti-
assumed that actors have a scheme or mental program tive manual tasks have been automated. Building
which they are enacting: a plan. In contrast to this, these systems requires a process of knowledge ac-
Suchman demonstrates that the rationality of an acquisition that is viewed as linear and unproblem-
tion is an ongoing construction among those involved atic. An alternative view, suggested by anthropologist
in the action. The default state of this rationality is a Jean Lave and computer scientist Etienne Wenger, is
transparent spontaneity in which the participants act that learning is embedded in (and a byproduct of)
rather than think. Only when the ongoing flow breaks social relationships and identity formation, and that
down does it become necessary to construct a rep- people learn by becoming a member of a commu-
resentation (that is, a plan or image) of what is hap- nity of practice.
pening. (Breakdowns, while frequent, are usually easily The concept of community of practice is fur-
repaired.) Language, due to its ability to classify, is a ther developed by Wenger to describe how experts
powerful resource for constructing such representa- acquire, share, and use their expertise. Communities
tions, although it is only one of several channels of practice are groups that share relationships, mean-
that humans use for communication. Using language, ing, and identity around the performance of some
the participants in an action understand what they set of tasks, whether processing insurance claims
are doing. Rationality (understanding what they are or delivering emergency medicine. The knowledge
doing) is the achievement rather than the configu- that they share is embedded in these relationships
ration state of interaction. and identities, not something that can be abstracted
The implications of this for constructing intelli- and stored in a database (or knowledge base).
gent devices (such as expert systems) are profound. In Anthropologist Marietta Baba has applied these
order for an intelligent device to reproduce intelligi- concepts along with the concept of sociotechnical
systems developed by the Tavistock Institute to ex- are indicative of a propensity to create normative
amine the response of work groups to the introduc- closure within any ongoing collectivity.
tion of office automation and engineering systems. At Both these concepts, of work group cultures and
major corporations she found that efforts to intro- online communities, point up the importance of
duce new automated systems frequently failed because culture for computing. As anthropologys signature
they were disruptive of the work processes, social concept, culture has an important (if sometimes un-
relationships, identities, and values of the work group, stated) place in anthropological thinking about hu-
considered as a community of practice. man-computer interaction.
Understanding cognitive activity as distributed
among multiple agents is closely related to the is-
sue of man/machine boundaries, an issue clearly Culture
of interest to anthropologists. Cyborg anthropol- For anthropologists, culture is more profound
ogy has been an ongoing professional interest at than simply the attitudes and values shared by a popu-
least since the 1991 publication of anthropologist lation. As a system of shared understandings, culture
Donna Haraways Simians, Cyborgs, and Women. represents the accumulated learning of a people (or a
Although most cyborg anthropology has focused on group), rooted in their history, their identity, and their
medical technology (such as imaging systems and relationship with other groups. Cultures evolve as
artificial organs) rather than on computational tech- shared projects with other groups. Although they are
nology, the basic conceptof human bodies and invented and imagined, cultures cannot be conjured
lives becoming increasingly embedded within auto- up at will, as much of the recent management lit-
mated information (control) circuitswill have in- erature on corporate culture seems to suggest.
creasing relevance for understanding the adaptation This is significant, because much of computing
of humans to advanced information technology: As use is in a corporate or organizational context (even
more and more human faculties, such as memory, if the organization is virtual). From an anthropo-
skilled manipulation, and interpersonal sensitivity, logical perspective, it is highly important to note that
are minimalized, disaggregated, and shifted away much of human-computer interaction is influenced
from the individual organism to automated devices, either directly, by the regimes of instrumental ra-
the dependence of carbon-based humans on their tionality in which it takes place, or indirectly, by
artifactual prostheses will increase. the fact that it follows protocols established by in-
Communities also form around technologies. fluential corporations. Several ethnographies of high-
Technology writer Howard Rheingold has described tech companies suggest that computerization and
participation in a San Francisco-based usenet as a the high-tech expectations associated with it are cre-
form of community building. Hakken describes the ating new corporate cultures: sociologist Gideon
influence of class on the experiences of users with Kunda and anthropologist Kathleen Gregory-
computing in Sheffield, England. Sociolologist Sherry Huddleston have described the working atmosphere
Turkle describes the identity experimentation con- of two high-tech corporations, noting that despite a
ducted by users of multiuser domains. Anthro- technological aura and emancipatory rhetoric, their
pologist Jon Anderson has examined how Middle corporate cultures are still mechanisms of control.
Eastern countries have used and adapted the Internet It should be noted that high tech is less an engineering
with unique methods for unique social goals. These concept for explaining functionality or performance
include the maintenance of diaspora relationships than it is an aesthetic conceit for creating auras of
with countrymen scattered around the globe. Online power and authority.
communities quickly evolve (actually adapt from Others have taken note of the fact that com-
surrounding norms) distinctive norms, including puters create new forms of culture and identity
styles of communication and categories of iden- and have described numerous microcultures that
tity. Although such collections of norms and val- have sprung up around such systems as textual data-
ues fall short of full-fledged human cultures, they banks, engineering design, and online instruction.
The culture of systems developers, as described by mation society; (2) family, work groups, and personal
Diana Forsythe is particularly notable. Insofar as de- relationships; (3) public institutions and private cor-
velopers and users have separate and distinctive cul- porations; (4) communities, both virtual and real;
tural outlooks, there will be a mismatch between (5) public policy and decision-making; (6) the chang-
their tacit understandings of system functionality ing shapes of knowledge and culture; and (7) the glob-
and system performance. The frequent experience alization of the information infrastructure (NSF
of systems not living up to expectations when de- 1996). In many ways this workshop both captured
ployed in the field is less a consequence of poor and projected forward the anthropological research
engineering than of the fundamental cultural rela- agenda for understanding the changing social face of
tionships (or disconnects) between developers and advanced information technology.
users.
Finally, anthropologys original interest in the re-
mote and exotic has often taken its attention away Questions
from the laboratories and highly engineered envi- Anthropologys orthogonal viewpoint proposes sev-
ronments in which the most advanced informa- eral unique questions. Perhaps the first of these is
tion technologies are found. In 2001 Allen Batteau, the question of control versus freedom. On the
an industrial anthropologist, observed that many one hand, cybernetic devices exist to create and in-
factories and field installations usually lack the re- tegrate hierarchies of control, and the fifty-year his-
liable infrastructure of universities or development tor y of the development of automation has
laboratories. As a consequence, computationally in- demonstrated the effectiveness of this strategy. On
tensive applications that work so well in the labo- the other hand, this poses the question of the proper
ratory (or in the movies) crash and burn in the field. role of a unique node in the control loop, the human
This lack, however, is not simply a matter of these user: How many degrees of freedom should the user
production environments needing to catch up to the be allowed? The designers answer, No more than
laboratories: Moores Law for nearly forty years necessary, can be unsatisfying: Systems that con-
has accurately predicted a doubling of computational strain the behavior of all their elements limit the
capability every eighteen months, a geometric growth users learning potential.
that outstrips the arithmetic pace of technological The related concepts of system learning and evo-
diffusion. The dark side of Moores Law is that the lution raise the second outstanding question, which
gap between the technological capabilities of the has to do with the nature of life. Should systems that
most advanced regions and those of the remote cor- can evolve, learn from, and reproduce themselves
ners of the human community will continue to grow. within changing environments be considered liv-
In 1995 Conrad Kottak, an anthropologist, observed ing systems? Studies of artificial life suggest that
that High technology has the capacity to tear all they should. The possibility of a self-organizing sys-
of us apart, as it brings some of us closer together tem that can replicate itself within a changing en-
(NSF 1996, 29). vironment has been demonstrated by anthropologist
Many of these observations grew out of a work- Chris Langston, enlarging our perspective beyond
shop organized by the American Anthropological the carbon-based navet that saw only biological or-
Association and the Computing Research Association ganisms as living.
called Culture, Society, and Advanced Information The final question that this raises, which is the ulti-
Technology. Held (serendipitously) at the time of the mate anthropological question, is about the nature
first deployment of graphical Web browsers (an event or meaning of humanity. Etymologically, anthro-
that as much as any could mark the beginning of the pology is the science of man, a collective term that
popular information revolution), this workshop iden- embraces both genders, and possibly more. Anthro-
tified seven areas of interest for social research in ad- pologists always anchor their inquiries on the ques-
vanced information technology: (1) the nature of tion of What does it mean to be human? Otherwise,
privacy, identity, and social roles in the new infor- their endeavors are difficult to distinguish from com-
parative psychology, or comparative linguistics, or Deal, T., & Kennedy, A. (1999). The new corporate cultures. Reading,
comparative sociology. However, the rise of infor- MA: Perseus Books.
Emery, F., & Trist, E. (1965). The causal texture of organizational
mation technology has fundamentally challenged some environments. Human Relations, 18, 2131.
received answers to the question of what it means to Forsythe, D. (2001). Studying those who study us: An anthropologist in the
be human. What are the human capabilities that world of artificial intelligence. Stanford, CA: Stanford University Press.
computers will never mimic? As Pulitzer-prize-win- Garfinkel, H. (1967). Studies in ethnomethodology. Englewood
Cliffs, NJ: Prentice-Hall.
ning writer Tracy Kidder asked, Do computers have Gregory-Huddleston, K. (1994). Culture conflict with growth:
souls? Will there ever be a computer that meets the Cases from Silicon Valley. In T. Hamada & W. Sibley (Eds.),
Turing testthat is, a computer that is indistin- Anthropological Perspectives on Organizational Culture. Washington,
DC: University Press of America.
guishable from a fully social human individual? More Hakken, D. (1999). Cyborgs@Cyberspace: An ethnographer looks to the
specifically, how many generations are required to future. New York: Routledge.
evolve a cluster of computers that will (unaided by Haraway, D. (1991). Simians, cyborgs, and womenThe reinvention of
human tenders) form alliances, reproduce, wor- nature. London: Free Association Books.
Hutchins, E. (1994). How a cockpit remembers its speeds. Cognitive
ship a deity, create great works of art, fall into petty Science, 19, 265288.
bickering, and threaten to destroy the planet? As the Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press.
abilities of silicon-based artifacts to think, feel, learn, Kidder, T. (1981). The soul of a new machine. Boston: Little, Brown.
adapt, and reproduce themselves continue to de- Kunda, G. (1992). Engineering culture: Control and commitment in a
high-tech corporation. Philadelphia: Temple University Press.
velop, the question of the meaning of humanity will Langston, C. G. (Ed.). (1989). Artificial life (Santa Fe Institute Studies
probably become the most challenging scientific and in the Sciences of Complexity, Proceedings, Volume 6).
philosophical question of the information age. Redwood City, CA: Addison-Wesley.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral
participation. Cambridge, UK: Cambridge University Press.
Allen W. Batteau Law, J. (1987). Technology and heterogeneous engineering: The case
of Portuguese expansion. In W. E. Bijker, T. P. Hughes, &
See also Ethnography; Sociology and HCI; Social T. Pinch (Eds.), The social construction of technological systems.
Psychology and HCI (pp. 111134). Cambridge, MA: MIT Press.
Linde, C. (1988). The quantitative study of communicative success:
Politeness and accidents in aviation discourse. Language in Society,
17, 375399.
Moore, G. (1965, April 19). Cramming more components onto in-
FURTHER READING tegrated circuits. Electronics.
Nardi, B. A. (1993). A small matter of programming: Perspectives on end
user computing. Cambridge, MA: MIT Press.
Anderson, J. (1998). Arabizing the Internet. Emirates Occasional Papers
National Science Foundation. (1996). Culture, society, and advanced
# 30. Abu Dhabi, United Arab Emirates: Emirates Center for
information technology (Report of a workshop held on June 12,
Strategic Studies and Research.
1995). Washington, DC: U. S. Government Printing Office.
Anderson, J., & Eikelman, D. (Eds.). (2003). New media in the Muslim
Perin, C. (1991). Electronic social fields in bureaucracies.
world: The emerging public sphere (Indiana Series in Middle East
Communications of the ACM, 34(12), 7482.
Studies). Bloomington: Indiana University Press.
Rheingold, H. (1993). The virtual community: Homesteading on the
Baba, M. L. (1995). The cultural ecology of the corporation: Explaining
electronic frontier. Reading, MA: Addison-Wesley.
diversity in work group responses to organizational transforma-
Star, S. L. (Ed.). (1995). The cultures of computing. Oxford, UK:
tion. (1995). Journal of Applied Behavioral Science, 31(2), 202233.
Blackwell Publishers.
Baba, M. L. (1999). Dangerous liaisons: Trust, distrust, and infor-
Stone, A. R. (1995). The war of desire and technology at the close of
mation technology in American work organizations. Human
the mechanical age. Cambridge, MA: MIT Press.
Organization, 58(3), 331346.
Suchman, L. (1987). Plans and situated actions: The problem of human-
Batteau, A. (2000). Negations and ambiguities in the cultures of or-
machine communication. Cambridge, UK: Cambridge University
ganization. American Anthropologist, 102(4), 726740.
Press.
Batteau, A. (2001). A report from the Internet2 Sociotechnical Summit.
Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
Social Science Computing Review, 19(1), 100105.
New York: Simon & Schuster.
Borofsky, R. (1994). Introduction. In R. Borofsky (Ed.), Assessing cul-
Wenger, E. (1998). Communities of practice: Learning, meaning, and
tural anthropology. New York: McGraw-Hill.
identity. Cambridge, UK: Cambridge University Press.
Brown, J. S., & Duguid, P. (2000). The social life of information. Boston:
Harvard Business School Press.
Callon, M. (1980). The state and technical innovation: A case study
of the electrical vehicle in France. Research Policy, 9, 358376.
German scholar Rodolpho von Ihering. These pro-

ANTHROPOMETRY posals called upon German anatomists and anthro-
pologists to reinvestigate craniometric (relating to
The word anthropometry, which means the meas- measurement of the skull) and anthropometric meas-
urement of the physical characteristics and physical urement methods. The German Anthropological
abilities of people, is derived from the Greek words Society convened in Munich and Berlin during the
anthropo meaning human being and metry mean- 1870s and early 1880s to establish what anthro-
ing measure. Physical characteristics, also called pometrist J. G. Garson and others have called the
structural dimensions, include such aspects as Frankfort Agreement of 1882. This agreement in-
heights, widths, depths, and body segment circum- troduced craniometric methods distinct from the
ferences. Physical abilities, also called functional di- predominant French methods and established a new
mensions, include such aspects as grip, push and nomenclature and measurement methods. The ex-
pull strength, reaching capabilities, fields of vision, istence of the French and German schools only
and functional task performance. further cemented the belief that international con-
Anthropologists, clinicians, and engineers use sensus on methods, nomenclature, and measure-
anthropometric information in a variety of ways. ments was needed.
For engineers, in particular, anthropometry provides During the early twentieth century people at-
information that can be used for the design of oc- tempted to develop an international consensus on
cupational, pubic, and residential environments. The the nomenclature of body dimensions and meas-
information can also be used for the design of tools, urement methods. In 1906, at the Thirteenth
protective head gear, clothing, and workstation equip- International Congress of Prehistoric Anthropology
ment. Doorway widths, tool handle lengths and cir- and Archaeology in Monaco, an international agree-
cumferences, ranges of clothing sizes, and the location ment of anthropometry took form. This congress
of displays and controls on workstations are some and the Fourteenth International Congress in
of the design applications. Anthropometry also pro- Geneva, Switzerland, in 1912 began to formalize the
vides information about body segment center of mass body of anthropometric work. The foundations of
and joint center of rotation characteristics that is a normative framework and a standardization of an-
used for biomechanical modeling (the study of joint thropometric measurement had been laid and trans-
forces and torques on the body). lated into French, German, and English by 1912. This
framework standardized anthropometric measure-
ments on both skeletal and living human subjects.
A Brief History Since 1912 several works by Hrdlicka, Rudolf
Although anthropometry was applied when Greek Marting, and James Gaven have increased the aware-
and Egyptian artists created standards (canons) ness of anthropometry and its uses and added to
for the human form centuries ago, not until the its scientific rigor.
nineteenth century were thought and dialogue on After the initial congresses, people attempted
anthropometry organized. Early work in anthro- to establish consensus throughout the twentieth cen-
pometry focused on the human anatomy, racial char- tury. Congresses meeting under the name of Hrdlicka
acteristics, skeletal remains, and human growth. convened on the topic of anthropometry and meas-
Among the noteworthy work documented by phys- urement methods. Other congresses aimed to create
ical anthropologist Ales Hrdlicka was that of French standards and databases for general use. During the
anthropologist Paul Pierre Broca and the Belgian sci- late twentieth century authors such as Bruce
entist Adolphe Quetelet. During the mid-eighteenth Bradtmiller and K. H. E. Kroemer chronicled these
century Quetelet used statistics to describe anthro- congresses and offered unique ways to manage
pometric information. Shortly after the Franco- anthropometric data. During recent years the
Prussian War of 1870, a growing emphasis on International Standardization Organization (ISO)
individualism was evident in the proposals of the technical committee on ergonomics published ISO
ANTHROPOMETRY 27
7250: Basic Human Body Measurements for Technical circumferences, and tape measures are used to meas-
Design (1996) to standardize the language and measure other circumferences such as the distance
urement methods used in anthropometry and ISO around the waist. Scales are used to measure body
15535: General Requirements for Establishing an weight. Photographs and video are used to meas-
Anthropometric Database (2003) to standardize the ure body dimensions in two dimensions. One
variables and reporting methods of anthropometric method uses grids that are attached behind and
studies. to the side of the person measured. Photographs
are then taken perpendicular to the grids, and the
space covered by the person in front of the grids
Structural Anthropometric can be used to estimate body segment heights,
widths, and depths. A variant of this method uses
Measurement Methods digital photography for which an anthropometric
Structural anthropometric measurement methods measurement is obtained by comparing the num-
require a person to be measured while standing or ber of pixels (small discrete elements that together
sitting. Anatomical landmarksobservable body constitute an image, as in a television or com-
features such as the tip of the finger, the corner of puter screen) for a dimension to the number of pix-
the eye, or the bony protrusion of the shoulder els of a reference object also located in the digital
known as the acromion processstandardize photograph.
the locations on the body from which measurements Attempts to develop three-dimensional com-
are made. The desire to achieve consistent measure- puter human models with conventional anthropo-
ments has led to the use of standardized measure- metric data reveal that limitations exist, such as the
ment postures held by people who are being uncertainty about three-dimensional definition of
measured. The anthropometric standing posture re- key points on the body surface, locations of cir-
quires the person to hold the ankles close together, cumferences, and posture. These limitations have re-
standing erect, arms relaxed and palms facing me- sulted in the development of more sophisticated
dially (lying or extending toward the median axis of three-dimensional anthropometric measurement
the body) or anteriorly (situated before or toward methods.
the front), the head erect and the corners of the eyes Digital anthropometry is the use of digital and
aligned horizontally with the ears. The anthropo- computerized technology in the collection of infor-
metric seated posture requires the person to be seated mation about body size and physical ability. In this
erect on a standard seating surface. The elbows use, computers are responsible for the actual collec-
and knees are flexed 90 degrees. The palms face me- tion of anthropometric data and are not relegated
dially with the thumb superior (situated above or solely to data analysis or storage. Digital anthro-
anterior or dorsal to another and especially a cor- pometry varies greatly from conventional anthro-
responding part) to the other digits. pometry. This variation has changed the nature of
Structural dimensions include the distances be- anthropometry itself for both the anthropometrist
tween anatomical landmarks, the vertical distance and the experimental context in which measurements
from a body landmark to the floor, and the cir- are taken. Human factors engineer Matthew Reed
cumferences of body segments and are measured and colleagues have identified some of the potential
with a variety of instruments. Among the most com- benefits of digital anthropometry:
mon instruments is the anthropometer, which is
a rod and sliding perpendicular arm used to meas- The capacity to assemble more accurate mod-
ure heights, widths, and depths. A spreading caliper els of human form, dimensions, and postures
having two curved arms that are hinged together is The capacity to evaluate multiple body dimen-
sometimes used to measure segment widths and sions simultaneously
depths defined by the distance between the tips of The capacity to measure the human and the en-
the arms. Graduated cones are used to measure grip vironment together
The improved description of joint centers of ro- those that employ photography and potentiometry
tation and movement in three dimensions as described above, or methods that require an in-
The capacity to make corrections to dimensions dividual to mark with a hand-held pen or pencil the
or create new dimensions after measurements maximum or comfortable reach locations on a
have been recorded vertical or horizontal grid.
Electromagnetic and video-based motion analy-
Laser scanning is often used in digital anthro-
sis systems provide new measures of physical abili-
pometry because it allows excellent resolution of
ties related to the way people move (kinematics) and
the morphological (relating to the form and structure
can be used with other types of instrumentation,
of an organism or any of its parts) features of the hu-
such as force plates (hardware that measure the force
man body and can be completed rapidly. Laser
applied to it), to provide biomechanical (the me-
scanning produces accurate three-dimensional rep-
chanics of biological and especially muscular ac-
resentations of the complex body surfaces, and
tivity) information or measures of balance. These
most protocols (detailed plans of a scientific or med-
systems allow positions of body landmarks to be
ical experiment, treatment or procedure) require
tracked over time during a physical activity. The data
the placement of surface markers on the body to
can be evaluated statistically or can serve as an ex-
ensure the proper location of bony protrusions that
ample of a human task simulation. Such methods of
are used as measurement landmarks beneath the sur-
data collection allow more lifelike dynamic digital
face of the skin. Other protocols using laser scans have
human models that can be used to evaluate human
morphological extraction algorithms (procedures for
performance in virtual environments. However, use
solving a mathematical problem in a finite number of
of these methods is expensive and time consuming.
steps that frequently involve repetition of an opera-
tion) to estimate landmark locations based on mor-
phological features.
Potentiometry can also be used to collect digital Measurement Consistency
anthropometric measurements. Electromechanical and Variation
potentiometric systems allow the measurer to man-
Anthropometric measurements are recordings of
ually digitize points in three-dimensional space. The
body dimensions and physical abilities that are sub-
measurer guides a probe tip manually to render dis-
ject to variability. No correct measurement exists
crete points or body surface contours.
because a measurement is simply an observation
or recording of an attribute that is the cumulative
contribution of many factors.
Functional Performance Anthropometric studies have investigated the
topic of measurement consistency in relation to
Measurements intrinsic qualities of variability within a given meas-
Conventional functional performance measurements urement. J. A. Gavan (1950) graded anthropome-
include grip, push, and pull strength, and reaching try dimensions in terms of consistencies seen through
abilities. For grip strength measurement, an indi- expert anthropometrists and concluded that con-
vidual is required to squeeze for several seconds at sistency increased as: the number of technicians de-
maximum effort a hand dynamometer (a force meas- creased, the amount of subcutaneous [under the
urement device) set at one or more grip circumfer- skin] tissue decreased, the experience of the tech-
ences. For the measurement of push and pull nician increased, and as the landmarks were more
strength, an individual usually holds a static (un- clearly defined (Gavan 1950, 425). Claire C. Gordon
changing) posture while either pushing on or pulling and Bruce Bradtmiller (1992), Charles Clauser and
against a force gauge at a maximum effort over associates (1998), Gordon and associates (1989), and
several seconds. An individuals reaching abilities can others have also studied intra- and interobserver er-
be evaluated with a number of methods, including ror contributions in anthropometric measurements,
ANTHROPOMETRY 29
including the contributions of different measure- modeling methods. The correlation between two di-
ment instruments and the effects of breathing cy- mensions provides a measure of how strongly two
c l e s . O t h e r re s e a rch e r s , s u ch a s Ka t h e r i n e dimensions covary linearly. When two measurements
Brooke-Wavell and colleagues (1994), have evalu- are highly correlated the values of one measurement
ated the reliability of digital anthropometric meas- can be used to predict the values of another in a
urement systems. These evaluations have brought regression analysis, therefore reducing the total num-
about an awareness of anthropometric reliability and ber of measurements needed to construct a com-
error as well as acceptable levels of reliability. prehensive set of anthropometric tables and human
Anthropometric data typically are collected for models based on partially extrapolated data. When
large samples of populations to capture distribu- combinations of anthropometric dimensions are
tional characteristics of a dimension so that it is rep- considered simultaneously in the evaluation of a
resentative of a target population. Many sources of product or environment, mockups and task trialing
anthropometric variability exist within populations. involving people or simulation approaches using dig-
Men and women differ greatly in terms of structural ital human modeling of people are required.
and functional anthropometric dimensions.
Additionally, the anthropometric dimensions of peo-
ple have changed systematically through time. Todays Important Data Sources
people are generally taller and heavier than those The most comprehensive anthropometric studies
of previous generations, perhaps because of improved have focused on military personnel, at least in part
availability and nutrition of food in developed coun- due to the need for the military to have information
tries. Of course, a persons body size also changes to provide well-designed uniforms, equipment, land
through time, even throughout the course of vehicles, and aircraft. Perhaps one of the most
adulthood. As a person ages, for example, his or comprehensive studies was the 1988 U.S. Army
her height decreases. Other sources of anthropo- Anthropometric Survey (ANSUR), which summa-
metric variability include ethnicity, geography, and rized 132 dimensions of approximately nine thou-
occupational status. sand army personnel.
The distribution characteristics of an anthropo- One of the most inclusive sources of civilian an-
metric dimension are often reported for different thropometric data is a U.S. National Aeronautics and
categories of age and gender, and sometime for Space Administration (NASA) technical report pro-
different ethnicities or countries. Because the vari- duced by the staff of the Anthropology Research
ability of anthropometric dimensional values within Project in 1978. This report contains anthropo-
such subgroups often takes the shape of a Gausian metric data across a variety of civilian and military
(bell-shaped) distribution, the mean deviation and populations for a large number of anthropometric
standard deviation of the sample data are often used variables, including information about the mass dis-
to describe the distributional characteristics of a di- tribution of body segments.
mension. The percentile valuethe value of a di- More recently, the Civilian American and
mension that is greater than or equal to a certain European Surface Anthropometry Resource (CAE-
percentage of a distributionalso provides useful SAR) project used laser scanning to collect the body
information. For example, the fifth and ninety-fifth surface contours and sizes of approximately twenty-
percentiles of a dimensional value define the outer four hundred North American and two thousand
boundaries of the 90 percent midrange of a popu- European civilians from 1998 to 2000. Measurements
lation distribution that might enable a designer to were recorded with people in standing, standardized
develop an adjustable consumer product or envi- seated, and relaxed seated postures. Thousands of
ronment feature that can accommodate 90 percent points that define the location of the bodys sur-
or more of the target population. face were collected with each scan, providing ex-
Multivariate data analysis includes the use of cor- tremely accurate three-dimensional representations
relation and regression analyses, as well as human of the body surface contours for individual human
models that can be used to evaluate the fit of a prod- that apply anthropometric data to the development
uct or of the person in an environment. Because of design guidelines. These publications include ISO
markers are also placed over key body landmarks, 14738 Safety of MachineryAnthropometric
conventional descriptive analysis of dimensions has Requirements for the Design of Workstations at
also been performed. CAESAR is the largest and most Machinery (2002), ISO 15534 Ergonomic Design for
valuable anthropometric data source of its kind. the Safety of Machinery (2000), and ISO 9241
Documents on the Ergonomic Requirements for Office
Work with Visual Display Terminals (19922001).
Using Anthropometric Data in Design The latter publications were developed to improve
Conventional use of anthropometric data in de- the fit between people and their computers at work.
sign requires determining (1) the population for
which a design in intended, known as the target
population, (2) the critical dimension or dimen- Future Research
sions of the design, (3) appropriate anthropometric A major challenge of future research is how to sum-
data source, (4) the percentage of the population marize and interpret the information-rich but com-
to be accommodated by the design, (5) the portion plex three-dimensional data that accompany the new
of the distribution that will be excluded, usually the methods of measurement described here. New meth-
largest and/or smallest values of the distribution, and ods of three-dimensional measurement of body
(6) the appropriate design values through the use of dimensions such as whole-body scanning provide
univariate or bivariate statistical methods. new opportunities to move conventional univariate
Conventional application of anthropometric anthropometric applications to complete three-di-
data, however, is not able to address the design prob- mensional static human models that can be used
lems that require the evaluation of many design char- to evaluate design in new ways. Motion analysis
acteristics simultaneously. Multivariate analysis using methods in dynamic human modeling also pro-
mockups and task trialing requires recruiting peo- vide a powerful tool to improve our understanding
ple with the desired range of body size and ability of the functional abilities of people. The reliability,
and assessing human performance during the sim- accuracy, and applications of many of these anthro-
ulation, such as judging whether people can reach pometric measurement methods, however, have yet
a control or easily see a display for a particular de- to be fully explored.
sign. Static and dynamic digital human modeling Perhaps what is most needed is simply more in-
approaches require manipulating models of various formation about the physical dimensions and abili-
sizes in virtual environments to assess the person- t i e s i n m o r e d i ve r s e u s e r g r o u p s . L a c k o f
design fit. Analysis methods for dynamic digital hu- anthropometric information severely limits the use
man modeling approaches are still in their infancy of anthropometry in the design of living and work-
due to the limited amount of studies recording the ing spaces that can be used by diverse populations.
needed information and the complicated nature of U.S. government agencies, particularly the U.S.
the data. Architectur al and Tr anspor tation Bar r iers
A variety of fields uses anthropometric data, in- Compliance Board (Access Board) and the U.S.
cluding anthropology, comparative morphology, hu- Department of Educations National Institute on
man factors engineering and ergonomics, medicine, Disability and Rehabilitation Research (NIDRR), re-
and architectural design. Additionally, digital an- cently have started to address the information gap
thropometry has been used outside of scientific and by studying the physical abilities of people with dis-
research endeavors, as seen in the application of a abilities, such as people who use wheelchairs.
new suit-making technology for Brooks Brothers However, much work remains to be done. In par-
(known as digital tailoring). ticular, the need for anthropometric data to in-
The International Organization for Stan- form the design of occupational, public, and
dardization has published numerous publications residential environments of the elderly is expected
ANTHROPOMETRY 31
to increase substantially as the proportion of the eld- display terminals (VDTs), (ISO Standard 9241). Geneva, Switzerland:
erly in the population continues to increase dra- International Organization for Standardization.
International Organization for Standardization. (Ed.). (1996). Basic
matically during the years to come. human body measurements for technical design (ISO Standard 7250).
G e n e v a , Sw i t ze r l a n d : In te r n a t i o n a l O r g a n i z a t i o n f o r
Victor Paquet and David Feathers Standardization.
International Organization for Standardization. (Ed.). (2000).
Ergonomic design for the safety of machinery (ISO Standard 15534).
See also Motion Capture G e n e v a , Sw i t ze r l a n d : In te r n a t i o n a l O r g a n i z a t i o n f o r
Standardization.
International Organization for Standardization. (Ed.). (2002).
Safety of machineryAnthropometric requirements for the design
FURTHER READING of workstations at machinery (ISO Standard 14738). Geneva,
Switzerland: International Organization for Standardization.
International Organization for Standardization. (Ed.). (2003). General
Annis, J. F. (1989). An automated device used to develop a new 3-D
requirements for establishing an anthropometric database (ISO
database for head and face anthropometry. In A. Mital (Ed.),
Standard 15535). Geneva, Switzerland: International Organization
Advances in industrial ergonomics and safety (pp. 181188). London:
for Standardization.
Taylor & Francis.
Kroemer, K. H. E., Kroemer, H. J., & Kroemer-Elbert, K. E. (1997).
Annis, J. F., Case, H. W., Clauser, C. E., & Bradtmiller, B. (1991).
Engineering anthropometry. In K. H. E. Kroemer (Ed.), Engineering
Anthropometry of an aging work force. Experimental Aging
physiology (pp. 160). New York: Van Nostrand Reinhold.
Research, 17, 157176.
Marras, W., & Kim, J. (1993). Anthropometry of industrial popula-
Brooke-Wavell, K., Jones, P. R. M., & West, G. M. (1994). Reliability
tions. Ergonomics, 36(4), 371377.
and repeatability of 3-D body scanner (LASS) measurements com-
Molenbroek, J. (1987) Anthropometry of elderly people in the
pared to anthropometry. Annals of Human Biology, 21, 571577.
Netherlands: Research and applications. Applied Ergonomics, 18,
Clauser, C., Tebbetts, I., Bradtmiller, B., McConville, J., & Gordon,
187194.
C. (1998). Measurers handbook: U.S. Army anthropometric sur-
Paquet, V. (Ed.). (2004). Anthropometry and disability [Special issue].
vey (Technical Report No. TR-88/043). Natick, MA: U.S. Army
International Journal of Industrial Ergonomics, 33(3).
Natick Research, Development and Engineering Center.
Paquette, S., Case, H., Annis, J., Mayfield, T., Kristensen, S., & Mountjoy,
Damon, A., & Stout, H. (1963). The functional anthropometry of old
D. N. (1999). The effects of multilayered military clothing ensembles
men. Human Factors, 5, 485491.
on body size: A pilot study. Natick, MA: U.S. Soldier and Biological
Dempster, W. T., Gabel, W. C., & Felts, W. J. L. (1959). The anthro-
Chemical Command Soldier Systems Center.
pometry of the manual work space for the seated subject. American
Reed, M., Manary, M., Flannagan, C., & Schneider, L. (2000). Effects
Journal of Physical Anthropometry, 17, 289317.
of vehicle interior geometry and anthropometric variables on auto-
Eastman Kodak Company. (2003). Ergonomic design for people at work
mobile driving posture. Human Factors, 42, 541552.
(2nd ed.). New York: Wiley.
Reed, M., Manary, M., & Schneider, L. (1999). Methods for measur-
Garson, J. (1885). The Frankfort Craniometric Agreement, with
ing and representing automobile occupant posture (SAE Technical
critical remarks thereon. Journal of the Anthropological Institute of
Paper No. 1999-01-0959). Warrendale, PA: Society of Automotive
Great Britain and Ireland, 14, 6483.
Engineers.
Gavan, J. (1950). The consistency of anthropometric measure-
Robinette, K. (1998). Multivariate methods in engineering anthro-
ments. American Journal of Physical Anthropometry, 8, 417426.
pometry. In Proceedings of the Human Factors and Ergonomics
Gordon, C., & Bradtmiller, B. (1992). Interobserver error in a large
Society 42nd annual meeting (pp. 719721). Santa Monica, CA:
scale anthropometric survey. American Journal of Human Biology,
Human Factors and Ergonomics Society.
4, 253263.
Robinette, K. (2000). CAESAR measures up. Ergonomics in Design,
Gordon, C., Bradtmiller, B., Clauser, C., Churchill, T., McConville,
8(3), 1723.
J., Tebbetts, I., & Walker, R. (1989). 19871988 anthropometric sur-
Roebuck, J., Kroemer, K. H. E., & Thomson, W. (1975). Engineering
vey of U.S. Army personnel: Methods and summary statistics
anthropometry methods. New York: Wiley.
(Technical Report No. TR-89/027). Natick, MA: U.S. Army Natick
Steenbekkers, L., & Molenbroek, J. (1990). Anthropometric data of
Research, Development and Engineering Center.
children for non-specialized users. Ergonomics, 33(4), 421429.
Haddon, A. (1934). The history of anthropology. London: Watts & Co.
Ulijaszek, S., & Mascie-Taylor, C. G. N. (Eds.). (1994). Anthropometry:
Hobson, D., & Molenbroek, J. (1990). Anthropometry and design for
The individual and the population. Cambridge, UK: Cambridge
the disabled: Experiences with seating design for cerebral palsy
University Press.
population. Applied Ergonomics, 21(1), 4354.
Hoekstra, P. (1997). On postures, percentiles and 3D surface an-
thropometry. Contemporary Ergonomics (pp. 130135). London:
Taylor & Francis.
Hrdlicka, A. (1918). Physical anthropology; its scope and aims, etc.
American Journal of Physical Anthropometry, 1, 323.
I n t e r n a t i o n a l O r g a n i z a t i o n f o r S t a n d a rd i z a t i o n . ( E d . ) .
(19922003). Ergonomics requirements for office work with visual
manipulate strategy is also goal-directed because it

APPLICATION USE can be used to complete the task of drawing three
arched windows.
STRATEGIES The definition of a strategy given above subsumes
more limited strategy definitions used in fields as di-
Strategies for using complex computer applications verse as business management and cognitive psy-
such as word-processing programs and computer- chology. These definitions may be stated in terms of
aided drafting (CAD) systems are general and goal- time (they may define strategy as a long-term plan
directed methods for performing tasks. These for achieving a goal), the existence of alternate meth-
strategies are important to identify and learn be- ods (they may consider a strateg y to be any
cause they can make users more efficient and effec- method that is nonobligatory), or performance out-
tive in completing their tasks, and they are often comes (they may define a strategy as a method that
difficult to acquire just by knowing commands on results in a competitive advantage). However, ex-
an interface. cluding these particulars (time, existence of alter-
To understand strategies for using computer ap- nate methods, and performance outcomes) from the
plications, consider the task of drawing three iden- definition of strategy enables us to describe strate-
tical arched windows in a CAD system. As shown in gies in a more encompassing way, irrespective of
Figure 1A, one way to perform the task is to draw all whether they are short term or long term, unique or
the arcs across the windows, followed by drawing all one of many, or efficient or inefficient.
the vertical lines, followed by drawing all the hori-
zontal lines. Another way to perform the same task
(Figure 1B) is to draw all the elements of the first The Costs and Benets
window, group the elements and then make three
copies of the grouped elements. of Using Strategies
The first method is called sequence-by-operation Although the two strategies shown in Figure 1 achieve
because it organizes the drawing task by performing the same goal, different costs and benefits are asso-
one set of identical operations (in this case draw arc), ciated with each ones use. By drawing all the arcs
followed by performing the next set of similar op- before the lines, the sequence-by-operation strategy
erations (in this case draw line). The second method reduces the cost of switching between the draw arc,
is called detail-aggregate-manipulate because it or- and the draw line commands. Furthermore, the strat-
ganizes the task by first detailing all the elements egy uses simple commands that are useful for per-
of the first object (in this case drawing the parts of forming a large set of tasks. Therefore, the short-term
the first window), aggregating the elements of the learning cost of using this strategy is small. However,
first object (in this case grouping all the parts of because the user is constructing every element in the
the first window), and then manipulating that ag- drawing, the performance cost (measured in terms
gregate (in this case making two copies of the of time and effort) can become large when draw-
grouped elements of the first window). Both the ing repeated elements across many tasks, especially
methods are strategies because they are general in the long term. In contrast, the detail-aggregate-
and goal-directed. For example, the detail-aggregate- manipulate strategy requires the user to draw the el-
manipulate strategy is general because it can be used ements of only one window, and makes the computer
to create multiple copies of sets of objects in a construct the rest of the windows using the group,
wide range of applications. The above example and copy commands. For a novice CAD user, the
was for a CAD application, but the same strategy short-term learning cost for the detail-aggregate-
could be used to create many identical paragraphs manipulate strategy involves learning the group and
for address labels in a word-processing appli- copy commands and how to sequence them.
cation, such as Microsoft Word. The detail-aggregate- However, as is common in the use of any new tool,
APPLICATION USE STRATEGIES 33
this short-term learning cost is amortized over the A. Sequence-by-Operation Strategy

long term because of the efficiency gained over many
invocations of the strategy. This amortization there-
fore lowers the overall performance cost.
Research has shown that strategies like detail-ag- 1.Draw arcs. 2. Draw vertical 3. Draw horizontal
gregate-manipulate can save users between 40 per- lines. lines.
cent and 70 percent of the time to perform typical
drawing tasks, in addition to reducing errors. B. Detail-Aggregate-Manipulate Strategy
Furthermore, with properly designed strategy-based
training, such strategies can be taught to novice com-
puter users in a short amount of time. For users who
care about saving time and producing accurate draw- 1.Draw arc. 2.Draw lines. 3. Group lines. 4. Copy group.
ings, learning such strategies can therefore make them detail aggregate manipulate
more efficient (save time) and more effective (reduce
errors) with relatively short training. Source: Bhavnani, S. K., John, B. E. (1996). Exploring the unrealized potential of computer-aided drafting.
Proceedings of CHI96, 337. Copyright 1996 ACM, Inc. Reprinted by permission.
FIGURE 1. Two strategies to perform the 3-window

drawing task.
A Framework That Organizes
Strategies for Using Complex
broad in scope because the powers they exploit are
Computer Applications offered by a large range of computer applications
Given the important role that strategies can play in such as authoring and information retrieval appli-
improving overall productivity, researchers have at- cations. Other strategies are narrower in scope and
tempted to identify and organize strategies for com- applicable to a smaller range of computer applica-
puter application use, such as authoring and tions such as only to word processors.
information retrieval applications. Frameworks to
organize strategies have suggested the design of: [Large-Scope Strategies
(1) training that teaches the strategies in a system- Given the ubiquity of graphical user interfaces (GUIs)
atic way, (2) new systems that provide effective and across computer applications, most useful computer
efficient strategies to users with little experience, and applications require some interaction with a visual
(3) evaluation methods to ensure that designers con- interface. Such computer applications offer the power
sistently offer the commands for using efficient and of visualization, that is, the power to selectively view
effective strategies. information on the screen. For example, a common
One useful way to organize strategies is based on word-processing task is to compare information from
the general capabilities of computer applications that one part of a document with information in another
the strategies exploit. For example, the detail-aggre- part of the document. When these two parts of the
gate-manipulate strategy described in Figure 1 ex- document cannot fit simultaneously on the screen,
ploits the iterative power of computers; it makes the the user can perform the comparison task in several
computer (instead of the user) perform the repeti- ways. One way is to scroll back and forth between
tious task of copying the elements multiple times. the relevant parts of the document. This method is
Strategies have also been identified that exploit other time-consuming and error-prone because it requires
powers of computers, such as the powers of propa- the user to remember the information that is not vis-
gation, organization, and visualization. ible. Another way to perform the same compari-
Another way to organize strategies is by the scope son task is to first bring together on the computer
of their use. For example, some strategies are screen the two relevant parts of the document, before
Large Scope Computer Applications

Visualization Strategies
Other computer
Medium Scope Authoring Applications
applications (e.g.
Iteration Strategies information retrieval )
Propagation Strategies
Word Processors Spreadsheets Drawing Systems Other authoring

applications (e.g.
Small Scope Small Scope Small Scope Web authoring)
Text Transformation Formula Decomposition Graphic Precision
Strategies Strategies Strategies
F I G U R E 2 . Strategies have been identified to exploit different powers of computers at different scopes levels. Large
scope strategies are useful to many classes of computer applications, such as authoring and information retrieval
applications. Medium scope strategies apply to a single class of computer applications, such as authoring appli-
cations. Small scope strategies apply to a single sub-class of applications, such as only to word processors. The
dotted lines represent how future strategies can be included in the framework.
comparing them. The information can be brought Second, there are strategies that exploit the power
together on the screen by different commands, such of propagation provided by authoring applications.
as by opening two windows of the same document The power of propagation enables users to set up de-
scrolled to the relevant parts of the document, or by pendencies between objects, such that modifications
using the split window command in Microsoft Word automatically ripple through to the dependent ob-
to view two parts of the document simultaneously. jects. For example, often users have to change the
In addition to being useful for word-process- font and size of headings in a document to conform
ing tasks, this visualization strategy is also useful to different publication requirements. One way to
when one is drawing a complex building in a CAD perform this task is to make the changes manually.
system, or when one is comparing information from This is time-consuming, especially when the docu-
two different webpages when retrieving information ment is long, and error-prone, because certain head-
on the Web. Hence strategies that exploit the ings may be missed or incorrectly modified. A more
power of visualization have wide scope, spanning efficient and effective method of performing the same
many different classes of computer applications. task is to first make the headings in a document
dependent on a style definition in Microsoft Word.
Medium-Scope Strategies When this style definition is modified, all depend-
While visualization strategies have the widest use ent headings are automatically changed. This strat-
across classes of computer applications, there are egy is useful across such applications as spreadsheets
three sets of strategies that are limited in scope to (where different results can be generated by altering
only one class of computer applications: a variable such as an interest rate), and CAD systems
First, there are strategies that exploit the itera- (where it can be used to generate variations on a re-
tive power of computers, such as the detail-aggre- peated window design in a building faade).
gate-manipulate strategy discussed earlier. These are Third, there are strategies that exploit the power
useful mainly for authoring applications such as of organization provided by authoring applications.
drawing systems and word processors. The power of organization enables users to explic-
APPLICATION USE STRATEGIES 35
itly structure information in representations (such ate and manipulate precise graphic objects. For ex-
as in a table). These explicit representations enable ample, a common precision drawing task is to cre-
users to make rapid changes to the content with- ate a line that is precisely tangent and touching the
out having to manually update the structure of the end of an arc (as shown in the arched windows in
representation. For example, one way to represent Figure 1). One way to perform this task is to visually
tabular information in a word-processing applica- locate, and then click the end of the arc when draw-
tion is by using tabs between the words or numbers. ing the line. This is error-prone because the user re-
However, because tabs do not convey to the com- lies on visual feedback to detect the precise location
puter an explicit tabular representation consisting of the end of the arc. Another way is to use the snap-
of rows and columns, the tabular structure may to-object command, which enables the user to click
not be maintained when changes are made to the a point that is only approximately at the end of the
content. A more efficient and effective way to per- arc. The computer responds by automatically locat-
form this task is to first make the table explicit to the ing the precise end of the arc, and therefore en-
computer by using the command insert table, and ables the user to draw a line that is precisely tangent
then to add content to the table. Because the com- to the end of the arc.
puter has an internal data structure for representing Similar small-scope strategies have been iden-
a table, the tabular representation will be maintained tified for word-processing applications (such as those
during modifications (such as adding more content that assist in transforming text to generate summaries
to a cell in the table). Organization strategies are also or translations) and for spreadsheets (such as
useful in other authoring applications. For example, those that decompose formulas into subformulas to
information can be stored using a set-subset repre- enable quick debugging).
sentation in a spreadsheet (as when different
sheets are used to organize sets of numbers) and in Future Extensions of the
a CAD system (as when different layers are used to Strategy Framework
organize different types of graphic information). The strategy framework described above focuses on
As discussed above, strategies that exploit the authoring applications. However, the framework can
powers of iteration, propagation, and organization also be extended to organize the large number of
are useful mainly for authoring applications. search strategies that have been identified for use
However, it is important to note that the powers of with information retrieval applications such as gen-
iteration, propagation, and organization can also be eral-purpose search engines like Google. In contrast
offered by other classes of computer applications, to computer powers that are useful in organizing
such as information retrieval applications. For ex- strategies for use with authoring applications, strate-
ample, many Web browsers offer users ways to or- gies for use with information retrieval systems ap-
ganize the addresses of different retrieved webpages. pear to be driven by attributes of how information
(The organizing features provided by the favorites sources are structured. For example, a large portion
command in Internet Explorer is one example.) of the Web comprises densely connected webpages
However, while powers provided by authoring ap- referred to as the core of the Web. The densely
plications can be provided in other classes of com- connected structure of information sources in the
puter applications, the strategies that they exploit core suggests the importance of using a variety of
will tend to be the same. browsing strategies (that rely on using hyperlinks to
move from one page to another) to locate relevant
Small-Scope Strategies sources. There is also a large portion of the Web that
Small-scope strategies exploit powers provided by consists of new pages that are not linked to many
particular subclasses of applications. For example, other pages. Strategies to find these pages therefore
the power of graphic precision is offered mainly by require the use of different query-based search en-
drawing systems, such as CAD systems. Strategies gines, given that no single search engine indexes all
that exploit graphic precision enable users to cre- webpages.
While there has been much research on edge across applications than did students who were
strategies for finding relevant sources of informa- taught only commands.
tion, one set of strategies works by selecting and or-
dering relevant sources of information based on New Search Systems
the way information is distributed across sources. The identification of search strategies to deal with
For example, health care information is typically the scatter of information across the Web has led
scattered across different health care portals. In this to the design of a new kind of domain portal
situation a useful strategy is to visit specific kinds called a Strategy Hub. This type of domain portal
of portals in a particular order to enable compre- implements the heuristic search strategy of visiting
hensive accumulation of the relevant information. sources of information in a particular order. Recent
Such strategies become critical when incomplete studies show that such a system enables users to find
information can have dangerous consequences (as more comprehensive information on specific topics
is the case with incomplete information on health when compared to the information retrieved by users
issues). of other search systems.
An important difference between strategies for
using authoring applications and strategies for us- An Analysis Method To Ensure Consistency in
ing information retrieval systems is that search strate- Capabilities across Applications
gies are fundamentally heuristicthat is, they are To enable the widest use of strategies across com-
rules of thumb that do not guarantee successful task puter applications, designers must provide a con-
completion. This is in part because users evaluation sistent set of commands. Therefore, a method called
of what is relevant changes based on what is being designs conducive to the use of efficient strategies
learned during the search process. (Design-CUES) has been developed that enables de-
signers to systematically check if their designs pro-
vide the commands necessary for users to implement
efficient and effective strategies.
How the Identication of
Strategies Can Improve
Looking Forward
Human-Computer Interaction Many years of research has shown that merely learn-
The identification and analysis of application use ing commands does not make for the best use of
strategies suggests three practical developments: strat- complex computer applications. The effective and
egy-based instruction, new search systems, and an efficient use of computer applications often requires
analysis method to ensure consistency in capabili- the use of strategies in addition to commands. An
ties across applications. important research goal has therefore been to
identify strategies for using a wide range of computer
Strategy-Based Instruction applications. The strategies that have been identified
Strategies for using authoring applications have to date have benefited users through strategy-
led to the design of strategy-based instruction. based instruction, new forms of search systems, and
Strategy-based instruction teaches commands in new design methods. As research on strategy iden-
combination with the authoring strategies that make tification continues, we can expect more develop-
use of authoring applications powers of iteration, ments along those lines, all with the ultimate goal of
propagation, and organization. Research has shown making users more effective and efficient in the
that students who took the strategy-based training use of complex computer applications.
acquired more efficient and effective strategies and
demonstrated a greater ability to transfer that knowl- Suresh K. Bhavnani
ARPANET 37
Advanced Research Projects Agency (DARPA),

FURTHER READING funded some of the most important research of
the twentieth century.
Bates, M. (1979). Information search tactics. Journal of the American
Society for Information Science 30(4), 205214.
Bates, M. J. (1998). Indexing and access for digital libraries and the
Internet: Human, database, and domain factors. Journal of the The Arpanet Concept
American Society for Information Science, 49(13), 11851205.
Belkin, N., Cool, C., Stein, A., & Thiel, U. (1995). Cases, scripts, and in-
The Arpanet long-distance computer network was
formation-seeking strategies: On the design of interactive informa- a collection of ideas, breakthroughs, and people.
tion retrieval systems. Expert Systems with Applications, 9(3), 379395. The roots of the Arpanet can be traced to one of
Bhavnani, S. K. (2002). Domain-specific search strategies for the ef- ARPAs most famous managers, J. C. R. Licklider.
fective retrieval of healthcare and shopping information. In
Proceedings of CHI02 (pp. 610611). New York: ACM Press.
In 1962 Licklider was recruited to work at ARPA,
Bhavnani, S. K. (in press). The distribution of online healthcare infor- then housed in the Pentagon, to start a behavioral
mation: A case study on melanoma. Proceedings of AMIA 03. sciences program. Although a psychologist by train-
Bhavnani, S. K., Bichakjian, C. K., Johnson, T. M., Little, R. J., Peck, ing, Licklider had a passion for the emergent field of
F. A., Schwartz, J. L., et al. (2003). Strategy hubs: Next-generation
domain portals with search procedures. In Proceedings of CHI computers and was adamant that the future of com-
03, (pp. 393400). New York: ACM Press. puting resided in the interactions between humans
Bhavnani, S. K., & John, B. E. (2000). The strategic use of complex and computers. In his seminal work, a paper enti-
computer systems. Human-Computer Interaction, 15(23), 107137.
Bhavnani, S. K., Reif, F., & John, B. E. (2001). Beyond command knowl-
tled Man-Computer Symbiosis written in 1960,
edge: Identifying and teaching strategic knowledge for using com- Licklider predicted that computers would not be
plex computer applications. In Proceedings of CHI 01 (pp. 229236). merely tools for people to use but also extensions of
New York: ACM Press. people, forming a symbiotic relationship that would
Drabenstott, K. (2000). Web search strategies. In W. J. Wheeler
(Ed.), Saving the users time through subject access innovation: Papers
revolutionize the way people interact with the world.
in honor of Pauline Atherton Cochrane (pp. 114161). Champaign: Through ARPA Licklider began to interact with
University of Illinois Press. the brightest minds in computingscientists at
Mayer, R. E. (1988). From novice to expert. In M. Helander (Ed.), Stanford, Berkeley, UCLA, MIT, and a handful of com-
Handbook of human-computer interaction (pp. 781796).
Amsterdam: Elsevier Science. panies that made up what Licklider considered to be
ODay, V., & Jeffries, R. (1993). Orienteering in an information his intergalactic computer network. Of course, this
landscape: How information seekers get from here to there. In network existed only in theory because people had
Proceedings of CHI 93 (pp. 438445). New York: ACM Press.
Shute, S., & Smith, P. (1993). Knowledge-based search tactics. Infor-
no way to bring these resources together other than
mation Processing & Management, 29(1), 2945. telephone or face-to-face meetings. However, Licklider
Siegler, R. S., & Jenkins, E. (1989). How children discover new strate- had the vision of gathering these people and resources,
gies. Hillsdale, NJ: Lawrence Erlbaum Associates. making the intergalactic network a physical network
Singley, M., & Anderson, J. (1989). The transfer of cognitive skill.
Cambridge, MA: Harvard University Press.
through an integrated network of computers.
Although originally brought on board to work on be-
havioral science issues in command-and-control
systems, Licklider was directly responsible for trans-
ARPANET forming his command-and-control research office
into the Information Processing Techniques Office
The Arpanet, the forerunner of the Internet, was devel- (IPTO), which would be responsible for critical ad-
oped by the U.S. Department of Defenses Advanced vanced computing achievements for decades to come.
Research Projects Agency (ARPA) in the early 1960s. Although Licklider left ARPA in 1964, he had a last-
ARPA was created in 1958 by President Dwight D. ing effect on the field of computing and the devel-
Eisenhower to serve as a quick-response research and opment of the Arpanet.
development agency for the Department of Defense, In 1966 another computer visionary, Bob Taylor,
specifically in response to the launch of the Soviet became director of IPTO and immediately began
satellite Sputnik. The agency, now the Defense to address the computer networking problem. The
ple locations), packet switching, dynamic routing al-

We were just rank amateurs, and we were expecting that gorithms (computational means of directing data
some authority would finally come along and say,Heres flows), and network survivability/redundancy. Packet
how we are going to do it. And nobody ever came along. switching would be a critical element of network de-
Vint Cerf on the design of Arpanet sign because it would allow information to be bro-
ken down into pieces or packets that would be sent
over the network and reassembled at their final des-
tination. This design was a much more efficient mes-
computing field at that time suffered from duplica- saging design, particularly when contrasted to
tion of research efforts, no electronic links between analogue phone lines. Additionally, the distributed
computers, little opportunity for advanced graph- network design would be more efficient and robust.
ics development, and a lack of sharing of valuable Without central nodes (locations that contain all the
computing resources. Taylor asked the director of resources and then distribute them to the rest of the
ARPA, Charles Herzfeld, to fund a program to create system), the system could survive a loss of one or
a test network of computers to solve these problems. more nodes and still route data traffic. This design
Herzfeld granted Taylors request, and Taylors office also would allow more efficient data trafficking when
received more than one million dollars to address the coupled with an adaptive networking algorithm ca-
problems. Thus, the Arpanet project was born. pable of determining the most efficient path for any
Taylor needed a program manager for the packet to travel.
Arpanet project. He recruited Larry Roberts from Researchers addressed these issues prior to the
MITs Lincoln Labs. Roberts, twenty-nine years old, Arpanet project. RANDs Paul Baran recommended a
arrived at the Pentagon in 1966 and was ready to ad- distributed switching network to the U.S. Air Force in
dress head on the problem of communications be- 1965 for the communications network of the Strategic
tween computers. Air Command, but the network was not developed. In
the United Kingdom Don Davies was working on
packet switching and adaptive networking for the
Fundamental Issues in Networking Ministry of Defense. The two men independently came
Several fundamental issues existed in the network- up with many of the same answers that would even-
ing of computers. Networking had been conceived tually be incorporated into the Arpanet.
of to solve the problem of resource sharing between
computers. During the 1960s computers were ex-
tremely large, expensive, and time consuming to op- The Arpanet Experiment
erate. ARPA had already invested in computing Larry Roberts arrived at ARPA in 1966 with the
resources at several computing centers across the charge to solve the computer networking problem.
country, but these centers had no way to communi- At an ARPA investigators meeting in Ann Arbor,
cate among one another or to share resources. At the Michigan, Roberts proposed a networking experi-
same time, Cold War concerns were causing U.S. sci- ment that would become the Arpanet. He proposed
entists to take a hard look at military communica- that all of the ARPA time-sharing computers at var-
tions networks across the country and to evaluate ious sites across the country be connected over dial-
the networks survivability in case of a nuclear strike. up telephone lines. The time-sharing (or host)
In the United Kingdom scientists were looking at computers would serve double dutyboth as re-
networks for purely communications use and were sources and routers. Meeting participants met
evaluating digital communication methods to work Robertss proposal with a great deal of skepticism.
around the inefficiency of the analogue telephone Why would people want to spend valuable comput-
system. Both U.S. and United Kingdom scientists ing resources to communicate between computers
were researching distributed networks (digital data when people already had all the computing they
communication networks that extend across multi- needed at their site? At the time, sharing between
ARPANET 39
computing centers was a goal of ARPA and not nec- Lyon, assistant to the president of the University of
essarily of the scientific community itself. In addi- Texas, unveil the Sputnik-era beginnings of the
tion, researchers would be reluctant to give up Internet, the groundbreaking scientific work that cre-
valuable computing power just so they could share ated it, and the often eccentric, brilliant scientists and
with other researchers. However, a researcher at engineers responsible. The team, led by Frank
the meeting, Wes Clark, struck upon a solution Heart, was dedicated to building the Arpanet on time
that would allow the experiment to be carried out. and to specifications and had only nine months to de-
Clark recommended keeping the host computers out liver the first IMP. Despite hardware setbacks, the team
of the networking duties. Instead, he suggested us- delivered the first IMP to UCLA early. UCLA was also
ing a subnetwork of intermediary computers to han- the site of the network management center, the test
dle packet switching and data trafficking. This track for the Arpanet. The team was charged with
subnetwork would reduce the computing demand testing the networks limits and exposing bugs, flaws,
on the host computers, and the use of a subnetwork and oddities. The initial Arpanet experiment consisted
of specialized computers would provide uniformity of four nodes, with an IMP at UCLA, Stanford
and control. This suggestion solved many problems, Research Institute (SRI), University of Utah, and
both technical and administrative, and would allow University of California at Santa Barbara. BBN also
ARPA to control the subnetwork. The computers was responsible for two critical elements: the IMPs
used at the subnetwork level were called interface themselves (including IMP-to-IMP communications)
message processors (IMPs). In addition to design- and the specifications for the IMP-to-host commu-
ing IMPs, researchers would have to develop pro- nications. The specifications for the IMP-to-host com-
tocols for how the IMPs would communicate with munications were drafted by Bob Kahn, who became
host computers and create the network. the intermediary between the Arpanet research com-
ARPA issued a request for proposals (RFP) in munity and BBN. Graduate students of the host in-
1968, because the specifications for the network had stitutions digested those specifications and developed
become so detailed. These specifications included: the code that would serve as the interface between host
and IMP. They formed the Network Working
Transfer of digital bits from source to specified
Group to hammer out the details of protocols, shared
location should be reliable.
resources, and data transfer. They created file trans-
Transit time through the subnetwork should
fer protocols (which layout the rules for how all com-
be one-half second or less.
puters handle the transfer of files) that became the
The subnetwork had to operate autonomously.
backbone of the Arpanet and made it functional. This
The subnetwork had to function even when IMP
experiment was so successful that the Arpanet was ex-
nodes went down.
panded to include other research sites across the coun-
The ARPA RFP was issued to determine which try until it grew to twenty-nine nodes. In 1972 the
company could build the Arpanet to these specifi- Arpanet made its public debut at the International
cations. After much debate, the contract was awarded Conference on Computer Communication. It was an
in 1969 to the Bolt, Baranek, and Newman company unequivocal hit, and the computer networking con-
(BBN), which had assembled an amazing team of cept was validated in the public arena.
scientists to transform this vision into reality. The
choice of BBN was a surprise to many people be-
cause BBN was considered to be a consulting firm, The Arpanet Evolves
not a computing heavy hitter. However, its proposal As members of a user community, the researchers
was so detailed and exacting that it could begin work involved in the Arpanet were always adding, creat-
immediately upon awarding of the contract. BBN ing, experimenting. The Arpanet became a bargain-
had only twelve months to do the work. ing tool in the recruiting of computer science faculty
In their 1996 book, Where Wizards Stay Up Late, and an impromptu communication tool for net-
Katie Hafner, co-author of Cyberpunk, and Matthew work mail or electronic mail (e-mail). In 1973 an
ARPA study showed that 75 percent of all traffic

on the Arpanet was e-mail. Researchers eventually FURTHER READING
wrote dedicated software to handle this side use of
the Arpanet. Adam, J. (1996, November). Geek gods: How cybergeniuses Bob Kahn
and Vint Cerf turned a Pentagon project into the Internet and con-
In 1972 Bob Kahn left BBN and went to work at nected the world. Washingtonian Magazine, 66.
ARPA with Larry Roberts. Kahn was now in charge Baranek, B., & Newman. (1981, April). A history of the ARPANET: The
of the network that he had helped create. He formed first decade. NTIS No. AD A 115440). Retrieved March 23, 2004,
a fruitful collaboration with Vint Cerf of Stanford from http://www.ntis.gov
Evenson, L. (1997, March 16). Present at the creation of the
( w h o w a s a g r a d u a te s t u d e n t o n t h e U C L A Internet: Now that were all linked up and sitting quietly, Vint Cerf,
Arpanet project) that led to the next evolution of one of its architects, describes how the Internet came into being.
networking. Together they tackled the problem of San Francisco Chronicle (p. 3ff).
Hafner, K., & Lyon, M. (1996). Where wizards stay up late: The origins
packet switching in internetworking, which would of the Internet. New York: Simon & Schuster.
eventually become the Internet. In 1975 Vint Cerf Hughes, T. J. (1998). Rescuing Prometheus. New York: Pantheon Books.
went to DARPA to take charge of all of the ARPA Norberg, A., & ONeill, J. (1997). Transforming computer technology.
Internet programs, and the Arpanet itself was trans- Ann Arbor: Scholarly Publishing Office, University of Michigan
Library.
ferred to the Defense Communication Agency, a Salus, P. (1995). Casting the Net. Reading, MA: Addison-Wesley.
transfer that upset some people in the non-bu-
reaucratic computing research community. The
Internet was created by the merging of the Arpanet,
SATNET (Atlantic Packet Satellite Network), and a
packet radio networkall based on the transmis- ARTIFICIAL
sion-control protocol/Internet protocol (TCP/IP)
standard that Cerf and Kahn createdand then more INTELLIGENCE
and more networks were created and connected un-
til the Internet was born. The Arpanet eventually Most research in mainstream artificial intelligence
burgeoned to 113 nodes before it adopted the new (AI) is directed toward understanding how people
TCP/IP standard and was split into MILNET and (or even animals or societies) can solve problems ef-
Arpanet in 1983. In 1989 the Arpanet was officially fectively. These problems are much more general
powered down, and all of the original nodes than mathematical or logical puzzles; AI researchers
were transferred to the Internet. are interested in how artificial systems can perceive
and reason about the world, plan and act to meet
goals, communicate, learn, and apply knowledge such
The Internet and Beyond that they can behave intelligently.
The creation of the Arpanetand then the Internet In the context of human-computer interaction
was the work of many researchers. Only with diffi- (HCI), research in AI has focused on three general
culty can we imagine our modern society without the questions:
interconnectedness that we now share. The
How can the process of designing and imple-
Arpanet was a testament to the ingenuity of the hu-
menting interactive systems be improved?
man mind and peoples perhaps evolutionary de-
How can an interactive system decide which
sire to be connected to one another. The Arpanet not
problems need to be solved and how they should
only brought us closer together but also brought us
be solved?
one step closer to J. C. R. Lickliders vision of human-
How can an interactive system communicate
computer interaction more than four decades ago.
most effectively with the user about the prob-
lems that need to be solved?
Amy Kruse, Dylan Schmorrow, and J. Allen Sears
The first question deals with the development
See also InternetWorldwide Diffusion process in HCI, the others with user interaction,
ARTIFICIAL INTELLIGENCE 41
specifically the issues of control and communica- acting as automated tax advisors, automobile repair
tion. These questions have been a central concern in advisors, or medical consultants, search oppor-
HCI for the past thirty years and remain critical tunistically for combinations of if-then rules that de-
today. AI has been able to provide useful insights into rive plausible conclusions from input data and
how these questions can be answered. In sum, existing knowledge. Machine learning systems, in-
what AI brings to HCI development is the possi- cluding neutral networks, incrementally refine an
bility of a more systematic exploration and evalua- internal representation of their environment, in a
tion of interface designs, based on automated search for improved performance on given tasks.
reasoning about a given application domain, the Natural language understanding systems search
characteristics of human problem solving, and gen- for correct interpretations through a space of am-
eral interaction principles. The AI approach can ben- biguous word meanings, grammatical constructs,
efit end users because it encourages tailoring the and pragmatic goals. These brief descriptions are
behavior of an interactive system more closely to only approximate, but they help us understand how
users needs. a system can represent and deal with some of the
problems that arise in interacting with users or in-
terface developers in an intelligent way.
The Concept of Search
Almost all techniques for problem solving in AI
are based on the fundamental concept of search. One ARTIFICIAL INTELLIGENCE (AI) The subfield of computer
way to understand search is by analogy to naviga- science that is concerned with symbolic reasoning and
tion on the World Wide Web. Imagine that my goal problem solving.
is to reach a specific webpage starting from my home-
page, and that I have no access to automated facili-
ties such as search engines. I proceed by clicking
on the navigation links on my current page. For each AI and the Development
new page that comes up, I decide whether I have
reached my goal. If not, then I evaluate the new page,
of User Interfaces
comparing it with the other pages that I have en- Considerable attention in AI has focused on the
countered, to see whether I am moving closer to my process of developing user interfaces. Experienced
goal or farther away. Based on my evaluation, I developers generally have a working knowledge of
may continue forward or go back to an earlier, more software engineering practice, interface architectures,
promising point to take a different path. graphic design, and related areas, plus information
An automated search process works in the same about the purpose for which the interface is to be
way. Pages correspond to states in a search space, used. If this knowledge can be captured in compu-
or relevant information about the environment; nav- tational form, an intelligent development environ-
igation actions are operators, which transform one ment can aid developers by testing and validating
state into another; an evaluation function assesses design specifications, by producing alternative de-
information about the state to guide the selection of signs for a given specification, by generating po-
operators for further transformations. tential improvements to a design, and by automating
A large number of AI techniques have been de- some of the more common implementation tasks.
veloped to address specific classes of search prob- The motivation for a search-based approach can
lems, representing the problems in different ways. be seen most clearly in the problem of layout design.
For example, planning systems search for sequences If an experienced designer were asked to organize
of interdependent operators to reach a set of goals; ten loosely related items of information (represented
these systems can deal with complex tasks ranging in text, pictures, and buttons) on a companys top-
from planning space missions to helping robots nav- level webpage, the final product might be the re-
igate over unfamiliar terrain. Expert systems, whether sult of comparing several alternatives, perhaps a few
dozen at most. The number of all possible layouts of the product in real-world situations). Thanks to AI
ten items, however, runs into the millions and higher; research, however, it is becoming possible to build ar-
this is much more than a designer can humanly con- tificial software agents that can stand in for real users.
sider. Most of these layouts will be unacceptable (for It is common to think about user interaction with
example, all possible orderings of items diagonally the software in problem-solving terms, as goal-oriented
across the page), but there may be many effective de- behavior. For example, if my goal is to send an e-mail
signs that are missed simply because the number message, I divide this into subgoals: entering the re-
of possibilities is so enormous. A system that can cipient information and subject line information, writ-
search through different spatial relationships and ing a short paragraph of text, and attaching a picture.
evaluate the results, even without perfect accuracy, My paragraph subgoal breaks down further into writ-
can give designers a more comprehensive view of the ing individual sentences, with the decomposition
problem and its solutions. continuing to the point of mouse movements and key
Automated layout design is just one aspect of in- presses. In AI terms, these decompositions can be rep-
terface design. Research in the general area of model- resented by plans to be constructed and executed au-
based interface design aims to support developers in tomatically. The PATHS system, a system designed to
all stages of the design process. In MOBI-D and help automate the testing of graphical user interfaces,
Mastermind, which are user interface generation tools, lets developers specify a beginning state, an end state,
developers build and evaluate abstract models of com- and a set of goals to be accomplished using the inter-
puter applications (such as word processing appli- face. PATHS then creates a comprehensive set of plans
cations, spreadsheet applications, or photographic to achieve the goals. For example, given the goal of mod-
design applications), interaction tasks and actions, ifying a document, the planner will generate sequences
presentations, even users and workplaces. The goal of actions for opening the document, adding and delet-
is to give developers decision-making tools that al- ing text, and saving the results, accounting for all the dif-
low them to apply their design skills but do not overly ferent ways that each action can be carried out. If a given
restrict their choices. These tools test constraints, eval- sequence is found not to be supported when it should
uate design implications, present suggestions, track be, PATHS will record this as an error in the application.
changes, and so forth, facilitating the eventual con- Similar work is carried out in the related field of
struction of the actual interface. For example, if a de- cognitive modeling, which shares many concepts
veloper specifies that the user must enter a number with AI. Cognitive modelers build computational
at some point, MOBI-D can present different inter- models of human cognitive processingperception,
face alternatives, such as a slider (the software equiv- attention, memory, motor action, and so forthin
alent of a linear volume control) or a text box that order to gain insight into human behavior. To make
the user can type into directly, for the developer to valid comparisons between a models performance
choose from. In Mastermind, the developer can switch and human performance, a common experimental
between a number of visual formats, avoiding ones ground is needed. User interfaces provide that com-
that are cumbersome. Current research in this area mon ground. Cognitive models comparable to plan-
is helping to improve webpage design and build in- ning systems have been developed for evaluating user
terfaces that meet the constraints of the next gener- interfaces, and they have the added benefit of giving
ation of interactive devices, including cell phones and developers information about the human side of in-
handheld computers. teraction as well as the application side.
AI research is also helping software companies
with product evaluation. Partially automated testing
of noninteractive software is now commonplace, but Interaction
conventional techniques are not well suited to test- The metaphor of tool use has come to dominate the
ing user interfaces. Software companies usually rely way we understand human interaction with com-
on limited user studies in the laboratory, plus a large puters, especially with regard to graphical user inter-
population of alpha and beta testers (people who test faces. Just as a carpenter keeps specialized sets of tools
ARTIFICIAL INTELLEGENCE 43
A Personal StoryPutting Humans First in Systems Design
The field of augmented cognition is pushing the integration of human systems and information technology to the fore-
front, while also attempting to maximize human potential. My current (and anticipated future) experience with using an
ever-increasing number of technologies during my everyday life compels me (propels me!) to help design a new class of
systems for the user to interact with. Practitioners of traditional human-systems integration research and design have stead-
fastly urged that the human must be considered when designing systems for human use.
An emerging concept is that not only are human beings the weak link in current human-systems relationships, but also
that the number of systems that a single human interacts with is growing so rapidly that the human is no longer capable
of using these technologies in truly meaningful ways. This specifically motivates me to develop augmented cognition tech-
nologies at the Defense Advanced Research Projects Agency (where I am a program manager). I want to decrease the num-
ber of system interfaces that we need to interact with, and increase the number of advanced systems that individuals are
capable of using simultaneously. On any given day, I typically wear (carry) five computers: my wristwatch, cell phone, two-
way pager with e-mailing capability, a personal digital assistant, and a laptop. I find these systems intrusive and the asso-
ciated demands on my time to be unacceptable. My home is inundated with appliances that are evolving into computer
devicesthese systems have advanced features that require significant attention in order to use them optimally. Even
with the worlds greatest human factors interface, I would never have time to interact with all of these systems that I use
on a daily basis.
Having said all of this, I need the systems that support me to exhibit some intelligence; I need them to be able to per-
ceive and understand what is going on around and inside of me. I do not have time to overtly direct them. Ideally they
will support me by sensing my limitations (and my capabilities) and determining how best to communicate with me if
absolutely necessary. Augmented cognition technology will imbue into these systems the ability to interact with me. Indeed,
augmented cognition is about maximizing human potential. If we humans are the weak link, it is because our current
advanced computer systems are actually limiting our performance. In the future, we must have transparent technologies
addressing our needs, or we will be overwhelmed by meaningless interactions.
Dylan Schmorrow
for framing a house or building fine furniture, an expe- The principles of direct manipulation provide a
rienced computer user has a variety of software tools foundation for tool-based environments. Direct-ma-
for word processing, analyzing data with spread- nipulation interfaces, as defined by Ben Shneiderman,
sheets, or creating graphics and illustrations. User inter- the founding director of the Human-Computer
faces are often thought of as tool-using environments, Interaction Laboratory at the University of Maryland,
which has important implications for the involve- provide a visual representation of objects, allow rapid
ment of AI in user interaction. Let us extend the car- operations with visible feedback, and rely mainly on
penter analogy. If I am intent on hammering a physical actions (such as selecting and dragging or
nail, I am not constantly reconsidering and recali- pressing buttons) to initiate actions. Modern graph-
brating the relationship between the hammer and ical user interfaces can trace much of their power to
my hand, or the head of the hammer and the nail. direct-manipulation principles. Nevertheless, as pow-
Instead, after an initial adjustment, the hammer erful as direct-manipulation interfaces can be, they
effectively becomes an extension of my arm, so are not appropriate in all situations. For example,
that I can use it without thinking about it. Similarly, sometimes in using a piece of software I know what
for a tool-based software environment, selecting in- needs to be doneI can even describe in words what
dividual tools should be intuitive, and applying a tool I would like to dobut I do not know exactly how to
should quickly become second nature. accomplish my task given the tools at hand.
These potential limitations, among others, have Other approaches to building intelligent envi-
led AI researchers to consider alternatives to a strict ronments, such as programming by example (PBE),
tool-based approach. First, it is possible to build in- involve more significant changes to user interaction.
telligent environments that take a more active role in PBE systems watch the user perform a procedure a
assisting the userfor example, by automatically number of times and then automatically general-
adapting their behavior to the users goals. Second, ize from these examples to create a fully functional
intelligent behavior can be encapsulated within a soft- program that can execute the repetitive actions so
ware agent that can take responsibility for different the user does not have to.
tasks in the environment, reducing the burden on the The SMARTedit system is an example of a ma-
user. Third, these agents and environments can com- chine-learning approach to PBE, in the context of
municate with the user, rather than passively being a text-editing application. Suppose that the user
acted upon by the user, as tools are. moves the cursor to the beginning of the word ap-
ple, erases the lowercase a, and types an uppercase
Intelligent Environments A. There are several ways that those actions could be
Some intelligent environments work by integrating interpreted. Perhaps, for example, the user wanted
AI search into an otherwise conventional interface. to move the cursor forward n characters and replace
One recently developed technique, human-guided the arbitrary character at that location with A, or
simple search, is intended to solve computationally perhaps the user wanted to move to the next oc-
intensive problems such as the traveling salesman currence of the letter a and capitalize it, or to correct
problem. This problem involves a salesman who must the capitalization of the first word in a sentence, or
visit a number of cities while keeping the distance some other possibility. Each of these interpretations
traveled as small as possible. Finding the optimal is a different hypothesis maintained by SMARTedit
route for even a small number of locations is beyond about the users intentions. As the user takes further
what can be done with pencil and paper; for ten actions, repeating similar sequences on different text,
locations there are over three million possible routes. ambiguity is reduced. Some hypotheses become more
Large problems are challenging even for the most plausible while others are pruned away because they
sophisticated computer programs. predict actions inconsistent with the users behav-
The user works with the human-guided search ior. At any point, the user can direct SMARTedit to
(HUGSS) tool kit through a graphical display of take over the editing process and watch the system
routes that the system has found. By pressing a apply its most hig hly ranked hy pothesis. If
button, the user activates a search process that com- SMARTedit carries out a sequence incorrectly, the
putes the best route it can find within a fixed period user can interrupt and correct the mistake, with
of time. The user examines the solution and modi- the system learning from the feedback.
fies it by selecting parts of the route that need fur- Adaptive user interfaces are another type of in-
ther refinement or identifying those parts that already telligent environment. Their development is moti-
have a reasonable solution. The user brings human vated by the observation that while the ideal software
perception and reasoning to bear on the problem by system is tailored to an individual user, for economic
constraining the space that the search process con- reasons a single system must be designed and released
siders (for example, by temporarily focusing the to thousands or even millions of users, who differ
search on routes between five specific locations, rather widely from one another in expertise, interests, needs,
than the entire set). Problem-solving responsibility and so forth. The solution is a system that can
is explicitly shared between the user and the system, adapt to its users when in use. A simple example is
with the amount and timing of the systems effort adaptive menus. A system can record how often the
always under the users control. HUGSS works faster user selects different menu options, and modify the
than the best fully automated systems currently in menu structure so that more frequently chosen op-
use, and it produces results of equal quality. tions can be reached more efficiently. This basic idea
ARTIFICIAL INTELLEGENCE 45
also works in more sophisticated adaptive systems, Information retrieval is just one area in which
many of which compile detailed models of users and agents have become popular. Agents have also ap-
their particular tasks and adapt accordingly. Adaptive peared in help systems, planning and scheduling aids,
systems have become especially relevant in efforts to scripting systems, intelligent tutoring systems, col-
personalize the World Wide Web as well as in research laborative filtering applications, matchmaking ap-
on intelligent tutoring systems and other applications plications, and electronic auctions. Work on agents
of AI to education. is one of the fastest-growing areas of AI.
An important topic within research on agents is
how to make agents interact most effectively with users.
Intelligent Agents Who should take the initiativethe user or the agent?
The engineer Michael Huhns and the computer sci- And when? Should one ever interrupt the other? These
entist Munindar Singh define intelligent agents as are questions of mixed-initiative interaction. Some
active, persistent (software) components that per- work on these questions is carried out in the area of
ceive, reason, act, and communicate (Huhns and rational decision making, wherein rationality is in-
Singh 1997, 1). For our purposes, the most impor- terpreted in an economic sense. If an agent has knowl-
tant characteristic of an agent is its autonomyits edge of the users preferences and can reason about
ability to carry out activities without the constant, the users goals, then it can, for example, determine
direct supervision of a human being. Agents in use that the value of the information it can contribute
at present include animated characters or believ- at some point will offset the cost of the user having to
able agents, autonomous agents such as softbots deal with an interruption. A different direction is taken
(software agents that perform tasks on the Internet) by projects that are influenced by the ways that peo-
and physical robots, and mobile agents whose pro- ple interact with one another, especially in dialogue.
cessing is not limited to a single computer platform. TRIPS (The Rochester Interactive Planning System)
Agents are also used in multi-agent systems, which is a mixed-initiative planning and scheduling assis-
may involve mixed teams of humans and agents. tant that collaborates with a human user to solve prob-
Most relevant to HCI are interface agents, which act lems in crisis situations, such as planning and
as intelligent assistants within a user interface, some- managing an evacuation. The COLLAGEN (from
times carrying out tasks on their own but also able COLLaborative AGENt) system is a collaboration sys-
to take instructions and guidance from the user. tem that can be incorporated into agents to give them
Letizia is an interface agent that assists users in sophisticated collaboration capabilities across a range
browsing the World Wide Web. Letizia operates in of application domains. TRIPS and COLLAGEN
conjunction with a standard Web browser, main- agents can interact with users via everyday natural
taining two open windows for its own use. As the language as well as through multimedia presentations,
user navigates through the Web, Letizia records the which leads to the topic of communication.
information on each page that the user visits and
performs an independent search of nearby pages that Communication
the user may not have seen. Letizias evaluation func- Some agents communicate by conventional means
tion compares the information on the pages that it in a graphical user interface, for example by raising
visits with the information that the user has seen up dialog windows and accepting typed input and but-
to the current point. In this way Letizia can make ton presses for responses. A common and reason-
suggestions about what the user might be interested able expectation, however, is that if a system is
in seeing next. As Letizia visits pages, it displays the intelligent, we should be able to talk with it as we
most promising ones for a short time in one win- would with other people, using natural language.
dow and the overall winner it has encountered in the (Natural language refers to the languages that peo-
other window. The user can watch what Letizia is ple commonly use, such as English or French, in con-
doing and take control at will. trast to programming languages.) Unfortunately,
even a brief treatment of natural-language under- hance its spoken conversation. Users can communi-
standing and generation, not to mention voice recog- cate via speech or gesture, even by simply looking at
nition and speech output, is beyond the scope of this particular objects, nonverbal behavior that is sensed
article. An example, however, may give some idea of by cameras. Systems like REA aim to make the com-
the issues involved. Consider three remarks from the puter side of face-to-face human-computer commu-
users side of a dialogue with a natural-language sys- nication as rich and nuanced as the human side.
tem (the bracketed text is not spoken by the user):
User (1): Show me document.txt.
User (2): Whats the last modification date [on Future Directions
the file document.txt]? This article has introduced the reader to AI approaches
User (3): Okay, print it [i.e., document.txt]. to HCI rather than give a taxonomy of AI systems;
To respond correctly, the system must be able many of the systems touched upon are much broader
to reason that modification dates are associated with in scope than can be conveyed through a category
files and that files rather than dates are usually printed assignment and a few sentences. Developments that
(it could grammatically refer to either.) Reading do not fit neatly within the categories discussed are
this dialogue, English-speaking humans make these listed below.
inferences automatically, without effort or even aware-
ness. It is only recently that computer systems have Smart Rooms and Intelligent Classrooms
been able to match even a fraction of our abilities. Much of what makes a software environment intelli-
The QuickSet communication system combines gent can be generalized to the physical domain. Smart
natural language and other methods of interaction rooms and intelligent classrooms rely on the same
for use in military scenarios. Shown a map on a tablet kind of technology as an embodied conversational
PC, the user can say, Jeep 23, follow this evacuation agent; they register users gestures and spoken com-
route, while drawing a path on the display. The sys- mands and adjust thermostats, change lighting, run
tem responds with the requested action. This in- presentations and the like, accordingly.
teraction is striking for its efficiency: the user has two
simultaneous modes of input, voice and pen-aided Games and Virtual Environments
gesture, and the ambiguities in one channel (in Intelligent agents have begun to enrich games and
this example, the interpretation of the phrase this virtual environments, acting as teammates or oppo-
route) are compensated for by information in the nents. Extending this line of research, the Mimesis
other channel (the drawn path). In general, voice and system imposes a nonscripted, dynamic narrative
natural language can support a more engaging, nat- structure onto a virtual gaming environment, so that
ural style of interaction with the interface than ap- external goals (for example, education on a histor-
proaches that use a single vector of communication. ical period) can be met without compromising the
Embodied conversational agents take work in nat- users direct control over the environment.
ural language a step further. When people speak with
one another, communication is not limited to the Human-Robot Interaction
words that are spoken. Gestures, expressions, and other Robots are appearing outside the laboratory, in our
factors can modify or even contradict the literal mean- workplaces and homes. Human-robot interaction
ing of spoken words. Embodied conversational agents examines issues of interaction with physical agents
attempt to recognize and produce these broader cues in real-world environments, even in social situations.
in communication. REA, a simulated real estate agent Robots can be used to explore otherwise inaccessi-
research prototype developed at the Massachusetts ble environments and in search-and-rescue missions.
Institute of Technology, is represented by a full body It should be clear from this discussion that the
figure on a large-scale display. REA shows users around most interesting problems in HCI are no longer
a house, making appropriate use of eye gaze, body found in software technology, at the level of the
posture, hand gestures, and facial expressions to en- visible components of the interface. Effective AI
ASIAN SCRIPT INPUT 47
approaches to HCI focus on issues at deeper levels, Shneiderman, B. (1998). Designing the user interface: Strategies for
probing the structure of problems that need to be effective human-computer interaction. Boston: Addison-Wesley.
Shneiderman, B., & Maes, P. (1997). Debate: Direct manipulation
solved, the capabilities and requirements of users, vs. interface agents. Interactions, 4(6), 4261.
and new ways of integrating human reasoning with Sullivan, J. W., & Tyler, S. W. (Eds.). (1991). Intelligent user inter-
automated processing. faces. New York: ACM Press.
St. Amant, R., & Healey, C. G. (2001). Usability guidelines for inter-
Robert St. Amant active search in direct manipulation systems. In Proceedings of the
International Joint Conference on Artificial Intelligence (IJCAI) (pp.
11791184). San Francisco: Morgan Kaufman.
Szekely, P., Sukaviriya, P., Castells, P., Muthukumarasamy, J., & Salcher,
FURTHER READING E. (1996). Declarative interface models for user interface construc-
tion tools: The Mastermind approach. In L. Bass & C. Unger (Eds.),
Anderson, D., Anderson, E., Lesh, N., Marks, J., Mirtich, B., Ratajczak,
Engineering for human-computer interaction (pp. 120150). London
D., et al. (2000). Human-guided simple search. In Proceedings of
and New York: Chapman & Hall.
the National Conference on Artificial Intelligence (AAAI) (pp.
Wolfman, S. A., Lau, T. A., Domingos, P., & Weld, D. S. (2001).
209216). Cambridge, MA: MIT Press.
Mixed initiative interfaces for learning tasks: SMARTedit talks back.
Cassell, J. (Ed.). (2000). Embodied conversational agents. Cambridge,
In Proceedings of the International Conference on Intelligent User
MA: MIT Press.
Interfaces (pp. 67174). New York: ACM Press.
Cassell, J., Bickmore, T., Billinghurst, M., Campbell, L., Chang, K.,
Vilhjlmsson, H., et al. (1999). Embodiment in conversational in-
terfaces: REA. In Proceedings of the ACM Conference on Human
Factors in Computing Systems (CHI) (pp. 520527). New York:
ACM Press.
Cypher, A., (Ed.). (1993). Watch what I do: Programming by demon-
stration. Cambridge, MA: MIT Press. ASIAN SCRIPT INPUT
Huhns, M. N., & Singh, M. P. (Eds.). (1997). Readings in agents. San
Francisco: Morgan Kaufmann. The Asian languages that employ the Chinese alphabet
Kobsa, A. (Ed.). (2001). Ten year anniversary issue. User Modeling and
User-Adapted Interaction, 11(12).
in their writing systems present difficult challenges
Lester, J. (Ed.). (1999). Special issue on intelligent user interfaces. AI for entering text into computers and word proces-
Magazine, 22(4). sors. Many Asian languages, such as Korean and Thai,
Lieberman, H. (1995). Letizia: An agent that assists Web browsing. In have their own alphabets, and the Devanagari al-
Proceedings of the International Joint Conference on Artificial
Intelligence (IJCAI) (pp. 924929). San Francisco: Morgan Kaufmann.
phabet is used to write Sanskrit, Hindi, and some
Lieberman, H. (Ed.). (2001). Your wish is my command. San Francisco: other languages of India. Designing keyboards and
Morgan Kaufmann. fonts for alphabets of languagessuch as Hebrew,
Lok, S., & Feiner, S. (2001). A survey of automated layout tech- Greek, Russian, and Arabicthat do not employ the
niques for information presentations. In Proceedings of the First
International Symposium on Smart Graphics (pp. 6168). New York: Roman alphabet used by English and other west-
ACM Press. ern European languages is relatively simple. The chal-
Maybury, M. T., & Wahlster, W. (Eds.). (1998). Readings in intelli- lenge with Chinese, simply put, is that a standard
gent user interfaces. San Francisco: Morgan Kaufmann.
Memon, A. M., Pollack, M. E., Soffa, M. L. (2001). Hierarchical GUI
national database contains 6,763 symbols (called
test case generation using automated planning. IEEE Transactions characters rather than letters), and a keyboard
on Software Engineering, 27(2), 144155. with so many keys would be completely unwieldy.
Newell, A., & Simon, H. (1972). Human problem solving. Englewood As was the case with ancient Egyptian hiero-
Cliffs, NJ: Prentice-Hall.
Oviatt, S. L., Cohen, P. R., Wu, L., Vergo, J., Duncan, L., Suhm, B., et
glyphics and Mesopotamian cuneiform, Chinese writ-
al. (2002). Designing the user interface for multimodal speech and ing began as pictographs that represented particular
gesture applications: State-of-the-art systems and research di- things. Evolving through time and modified for grace-
rections. In J. Carroll (Ed.), Human-computer interaction in the ful drawing with an ink brush, these pictographs be-
new millennium (pp. 419456). Reading, MA: Addison-Wesley.
Puerta, A.R. (1997). A model-based interface development environ- came the current system of characters representing
ment. IEEE Software, 14(4), 4147. concepts and sounds in a complex interplay of func-
Ritter, F. E., & Young, R. M. (Eds.). (2001). Special issue on cognitive tions. A person fully literate in Chinese today uses
modeling for human-computer interaction. International Journal
of Human-Computer Studies, 55(1).
3,000 to 4,000 characters; newspapers have 6,000 to
Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern ap- 7,000 available, but some dictionaries list as many as
proach. Englewood Cliffs, NJ: Prentice-Hall. 50,000. In 1958 a standardized phonetic system based
on the Roman alphabet and called pinyin was in- play the equivalent hiragana characters instead of
troduced, but it has not replaced the traditional Western letters on the screen.
system of writing. Japanese employs two phonetic al-
phabets called kana, as well as Chinese characters
called kanji. In 1983 the Japan Industrial Standard Many Meanings
listed 2,963 commonly used characters plus an- The writer probably does not want the hiragana but
other 3,384 that appear only rarely. Korean also makes rather the kanji, but many Japanese words can be ro-
some use of Chinese characters, but the chief form manized kannsou. Asian languages have many
of writing is with an alphabet historically based on homonyms (words that sound similar but have dif-
Chinese but phonetically representing the sounds of ferent meanings), and Chinese characters must rep-
spoken Korean. resent the one intended meaning. The standard way
Because Japan has been a leader in developing com- in which word processors handle this awkward fact, in
puter technology for decades, its language is the best Chinese as well as Japanese, is to open a selection win-
example. Around 1915 Japan began experimenting dow containing the alternatives. For example, lets say
with typewriters, but they were cumbersome and rare. the user typed kannsou, then hit the spacebar (which
Typewriters could be made simply for the kana, a cen- is not otherwise used in ordinary Japanese) to open
turies-old phonetic system for writing Japanese syl- the selection window with the first choice highlighted.
lables, either in the traditional hiragana form or in the The user can select the second choice, which is the cor-
equivalent katakana form used for writing foreign rect Chinese characters for the Japanese word mean-
words or telegrams. Occasionally reformers have sug- ing a comment (ones thoughts and impressions
gested that Chinese characters should be abandoned about something). If the user wanted kannsou to mean
in favor of the kana or the Roman alphabet, but this not comment, but rather dry, he or she would se-
reform has not happened. Thus, newspapers employed lect the third choice. The fourth through ninth choices
vast collections of Chinese type, and careful hand- mean welcome and farewell,a musical interlude,
writing was used in business, schools, and forms of completion, as of a race,meditate,hay(dry grass),
printing such as photocopying that could duplicate and telling peoples fortunes by examining their faces.
handwriting. During the 1980s word processors Good Asian-language word processing software
were introduced that were capable of producing the presents the choices in descending order of likelihood,
traditional mixture of kanji, hiragana, and katakana, and if a person selects a particular choice repeatedly it
along with occasional words in Roman script and other will appear on the top of the list. The word processor
Western symbols. The Macintosh, which was the can be set so that the first kanji choice, instead of the
first commercially successful computer with bitmapped hiragana, appears in the text being written. Pressing
(relating to a digital image for which an array of bi- the spacebar once would transform it to the second
nary data specifies the value of each pixel) screen choice, and pressing again could select the next choice
and printing, became popular in Japan because it could and open the selection window. The choices may in-
handle the language, but all Windows-based comput- clude a katakana choice as well. Many choices exist,
ers can now as well, as, of course, can indigenous and some Chinese word processors often fill the se-
Japanese word processors. lection window four times over. Thus, research on the
Kana computer keyboards exist in Japan, but the frequency of usage of various Chinese words is im-
most common input method for Chinese characters portant in establishing their most efficient ordering in
in both China and Japan requires the user to enter text the selection window. Human-computer interaction
into a Western keyboard, romanizing the words. (HCI) research has explored other ways of making the
Suppose that someone is using Microsoft Word in word selection, including eye tracking to select the al-
Japanese and wants to type the word meaning com- ternative that the users eyes focus upon.
ment. The writer would press the Western keys that The chief substitutes for keyboard text input are
phonetically spell the Japanese word kannsou. If the speech recognition and handwriting recognition.
word processor is set to do so, it will automatically dis- Speech recognition systems developed for English are
ASIAN SCRIPT INPUT 49
unsuitable for Asian languages. Notably, spoken Modern word processors may change the balance
Chinese is a tonal language in which each syllable of forces working for or against change in the tradi-
has a characteristic pitch pattern, an important feature tional Asian scripts. They may degrade peoples Chinese
absent from English. Experts have done a good deal of character handwriting skills, but they may simultane-
research on computer recognition of Japanese and ously help people employ more obscure characters. In
Chinese, but speech input introduces errors while the psychology of memory people have the ability to
requiring the same selection among choices, as does recognize things they would not have spontaneously
keyboard input. Handwriting recognition avoids the produced. Chinese-language and Japanese-language
problem of alternative ways of writing homonyms, but word processors often include character palettes (com-
despite much research it remains excessively error parable ranges, qualities, or uses of available elements),
prone. Three approaches are being tried with Chinese: allowing users to select even obscure characters with
recognizing (1) the whole word, (2) the individual a single click of the mouse, thereby perhaps encour-
characters, or (3) parts of characters, called radicals, aging them to do so. Computer and information sci-
that may appear in many characters. All three ap- entists and engineers are rapidly producing search
proaches have high error rates because many charac- engines and a whole host of other tools that are giving
ters are graphically complex, and people vary the ancient Asian scripts a new life on the Internet and
considerably in how they draw them. Thus, key- the World Wide Web.
board input remains by far the most popular method.
William Sims Bainbridge and Erika Bainbridge
East and West See also Handwriting Recognition and Retrieval;

China, Japan, and Korea have from time to time Keyboard
considered abandoning the traditional Chinese
characters, with Korea coming the closest to actually
doing so. A phonetic writing system is easier to FURTHER READING
learn, thus giving students more time to study other
things. The traditional Chinese system supported an Apple Computer Company. (1993). Macintosh Japanese input method
guide. Cupertino, CA: Apple.
entrenched intellectual elite, who feared that a sim- Asher, R. E., & Simpson, J. M. Y. (Eds.). (1994). The encyclopedia of lan-
ple alphabet might democratize writing. On the other guage and linguistics. Oxford, UK: Pergamon.
hand, one advantage of the traditional system is that Fujii, H., & Croft, W. B. (1993). A comparison of indexing techniques
a vast region of the world speaking many dialects for Japanese text retrieval. In Proceedings of the 16th annual ACM
SIGIR Conference on Research and Development in Information
and languages could be united by a single writing sys- Retrieval (pp. 237246). New York: ACM Press.
tem, and even today a Chinese person can commu- Ho, F.-C. (2002). An analysis of reading errors in Chinese language. In
nicate to some extent with a Japanese personeven L. Jeffrey (Comp.), AARE 2002 conference papers (n.p.).
Melbourne, Australia: Australian Association for Research in
though neither knows the others spoken lan- Education.
guageby drawing the characters. Fluent bilingual Li,Y., Ding, X., & Tan, C. L. (2002). Combining character-based bigrams
readers of an Asian language and a Western language with word-based bigrams in contextual postprocessing for Chinese
sometimes say they can read Chinese characters more script recognition, ACM Transactions on Asian Language Information
Processing, 1(4), 297309.
quickly because the characters directly represent con- Shi, D., Damper, R. I., & Gunn, S. R. (2003). Offline handwritten Chinese
cepts, whereas Western letters represent sounds and character recognition by radical decomposition. ACM Transactions
thus only indirectly relate to concepts. Some writers on Asian Language Information Processing, 2(1), 2748.
have conjectured that dyslexia should be rare in Wang, J. (2003). Human-computer interaction research and practice
in China. ACM Interactions, 10(2), 8896.
Chinese, if difficulties in learning to read are an in- Wang, J., Zhai, S., & Su, H. (2001). Chinese input with keyboard and eye-
ability to connect letters with sounds. However, dyslexia tracking. In Proceedings of the SIGCHI Conference on Human Factors
seems to exist in every language, although its causes in Computing Systems (pp. 349356). New York: ACM Press.
and characteristics might be somewhat different in
Asian languages than in English.
like that of current scientific computers: curve-fit-

THE ATANASOFF-BERRY ting, circuit analysis, structural analysis, quantum
physics, and problems in mechanics and astronomy.
COMPUTER The desktop calculators of the era were not up to the
equation-solving task, and Atanasoff identified their
The Atanasoff-Berry Computer (ABC) was the limits as a common bottleneck in scientific research.
first electronic digital computer and the inspiration His conception of a high-speed solution made sev-
for the better-publicized 1946 ENIAC. It was con- eral unprecedented leaps: binary internal arithmetic
ceived in late 1938, prototyped in 1939 at Iowa State (with automatic binary-decimal conversion), all-
College (now Iowa State University) in Ames, electronic operation using logic gates, dynamically-
Iowa, and made usable for production computing refreshed memory separated from the arithmetic
by 1941. John Atanasoff, a professor of mathemat- units, parallel operation of up to thirty simultane-
ics and physics, collaborated with Clifford Berry, a ous arithmetic units, and a synchronous system clock.
graduate student, to develop the system. The ABC achieved practical success at the curve-
fitting application. Atanasoff collaborated with a stat-
istician colleague at Iowa State, George Snedecor,
Physical Description who supplied a steady stream of small linear-system
In contrast to the computers that followed in the problems to the ABC. Snedecors secretary was given
1940s, the ABC was compact, movable, and easily the task of checking the results by desk calculation,
operated by a single user. The original system no which was simpler than solving the equations and
longer exists except for a logic module and a mem- could be performed manually.
ory drum, but a functioning replica was constructed
in the late 1990s.
Human Interface
The Atanasoff-Berry Computer Compared to modern interfaces, the ABC interface
The ABC weighed about 750 pounds. It had the resembled that of an industrial manufacturing
weight and maneuverability of an upright piano and machine. The user controlled the system with throw
could roll on four heavy-duty casters. The total power switches and card readers (decimal for input and bi-
it drew was less than a kilowatt, and the heat gen- nary for intermediate results). The user was also
erated by its vacuum tubes was low enough to dis- responsible for moving a jumper from one pair of
sipate without requiring fan-forced air. The ABC contacts to another to indicate a particular vari-
used ordinary 117-volt line power. An electric mo- able in the system of equations. The ABC commu-
tor synchronized to standard 60-hertz line voltage nicated to the user through a few incandescent lamp
served as the system clock. The electromechanical indicators, an ohmmeter to indicate correct work-
parts of the ABC, like those of a modern coming voltages, a binary punch card output, and a cylin-
puter, were for purposes other than calculation; drical readout for decimal numbers that resembled
the computing itself was completely electronic. The a car odometer.
arithmetic modules were identical and could easily The inventors clearly designed the machine for
be interchanged, removed, and repaired. operation by themselves, not general users. None of
the switches or lamps was labeled; it was up to the
user to remember what each switch did and what
Intended Applications each lamp meant. One switch instructed the ABC to
read a base-10 punch card, convert it to binary,
and Production Use and store it in the dynamic memory, for example.
The ABC was intended to solve dense systems of Furthermore, the open design of the ABC pro-
up to thirty simultaneous linear equations with vided far less protection from electric shock than a
15-decimal precision. Atanasoff targeted a workload modern appliance does. Exposed surfaces only a few
ATTENTIVE USER INTERFACE 51
centimeters apart could deliver a 120-volt shock to

the unwary. FURTHER READING
A user entered the coefficients of the equations on
standard punch cards, using an IBM card punch. Each Atanasoff, J. V. (1984). Advent of electronic digital computing. Annals
coefficient required up to fifteen decimals and a of the History of Computing, 6(3), 229282.
Burks, A. R. (2003). Who invented the computer? The legal battle that
sign, so five numbers fit onto a single eighty-column changed computing history. Amherst, NY: Prometheus Books.
card. It was in the users best interest to scale up the Burks, A. R., & Burks, A. W. (1989). The first electronic computer: The
values to use all fifteen decimals, since the arithmetic Atanasoff story. Ann Arbor: University of Michigan Press.
Gustafson, J. (2000). Reconstruction of the Atanasoff-Berry com-
was fixed-point and accumulated rounding error. puter. In R. Rojas & U. Hashagen (Eds.), The first computers:
Because the ABC could hold only two rows of History and architectures (91106). Cambridge, MA: MIT
coefficients in its memory at once, it relied on a mass Press.
storage medium to record scratch results for later use. Mackintosh, A. R. (1988, August). Dr. Atanasoff s computer. Scientific
American (pp. 7278).
(The solution of two equations in two unknowns Mollenhoff, C. R. (1988). Atanasoff: Forgotten father of the computer.
did not require scratch memory.) Since magnetic stor- Ames: Iowa State University Press.
age was still in its infancy, Atanasoff and Berry devel- Randell, R. (Ed.). (1982). The or ig ins of dig ital computers
oped a method of writing binary numbers using (pp. 305325). New York: Springer-Verlag.
Reconstruction of the Atanasoff-Berr y Computer. (n.d.).
high-voltage arcs through a paper card. The pres- Retrieved on January 27, 2004, from http://www.scl.ameslab
ence of a hole, representing a 1, was then readable with .gov/ABC
lower voltage electrodes. Both reading and writing took Sendov, B. (2003). John Atanasoff: The electronic Prometheus. Sofia,
Bulgaria: St. Kliment Ohridski University Press.
place at 1,500 bits per second, which was a remarkable Silag, W. (1984). The invention of the electronic digital computer at
speed for input/output in 1940. However, the reliabil- Iowa State College, 19301942. The Palimpsest, 65(5), 150177.
ity of this system was such that a 1-bit error would oc-
cur every 10,000 to 100,000 bits, and this hindered the
ability to use the ABC for production computing be-
yond five equations in five unknowns.
To obtain human-readable results, the ABC con-
verted the 50-bit binary values stored in the memory
to decimals on the odometer readout. The total process ATTENTIVE USER
of converting a single 15-decimal number and mov-
ing the output dials could take anywhere from 1 sec- INTERFACE
ond to 150 seconds depending on the value of the
number. Atanasoff envisioned automating the man- An attentive user interface is a context-aware hu-
ual steps needed for operation, but enhancement of man-computer interface that uses a persons at-
the ABC was interrupted by World War II and never tention as its primary input to determine and act
resumed. upon a persons intent. Although we can read a
The ABC was a landmark in human-computer in- persons attention in her every word and action
teraction by virtue of being the first electronic com- (even the way a person moves a cursor on a com-
puter. Its use of punch cards for the input of puter interface shows what she is attending to), we
high-accuracy decimal data, binary internal represen- usually read attention in what and how people look
tation, operator console, and the management of mass at things.
storage and volatile storage were major advancements Visual attentive user interfaces concentrate on
for the late 1930s when Atanasoff and Berry conceived the autonomic (involuntary) and social responses
and developed it. that eyes communicate and read such eye move-
ments as a lingering stare, a roving gaze, and a nerv-
John Gustafson ous blink in a language of ocular attention. Such
interfaces also monitor the order in which people
See also ENIAC visually scan objects.
Eye Tracking
Eye tracking is a technique that monitors a persons
eye movements to determine where she is looking.
Eye tracking has long held promise as the ultimate hu-
man-computer interface, although eye tracking prod-
ucts have not been a commercial success. Original eye
tracking approaches used mechanical/optical instru-
ments that tracked mirrored contact lens reflections
or even instruments that measured eye muscle ten-
sion. Newer approaches illuminate the eye with in-
frared light and watch reflections with a camera.
Researchers can indirectly determine where a persons
eye is focusing by noting that an electroencephalo-
gram (EEG) signal is dominated by an ocular stim- Researcher Mike Li demonstrates the technology used
ulus. Four or five video strobe rates on different parts in the Invision eye-tracking experiment. The balls on
of a display can be distinguished in an EEG. When a the screen have names of companies that move around
person attends to one of them, his EEG pulses at the as he looks at them. The object under the screen is
video strobe rate. Codings of attention on a screen the eye tracker. Photo courtesy of Ted Selker.
can be identified with an EEG frequency counter.
eye tracker enables the users gaze to roughly posi-

Attention Can Be Detected tion the cursor, which the mouse can then manip-
Whereas the advertising and psychology fields ulate. If the user wants to change the application
have long used eye movement to understand what a window he is working with, he stares at the applica-
person is looking at, the human-computer interface tion window that he wants to work in; this stare
field has struggled to use the eye as a controller. How- warps the cursor to that application window.
ever, the breakthrough in visual attentive user inter- MAGIC pointing speeds up context changes on
faces is in observing what the eye does, not in giving the screen.
it a tracking task. Interest Tracker is a system that
monitors the time that a person spends gazing
over a title area instead of the time that the person The Path of Attention Can
spends gazing at a specific character to determine se-
lection. For example, the title of an article is pre- Demonstrate Intention
sented at the bottom of a computer screen. A user During the late 1960s it was shown that the way that
might glance down to read the title; if his glance plays a persons eyes move while scanning a picture de-
over the title for more than .3 seconds a window scribes aspects of what she is thinking. When re-
opens on the computer screen with the full article. searchers asked viewers seven questions about a
That .3 seconds of dwell time is less than the typi- painting entitled The Unexpected Visitor, seven iden-
cal 1 second that is required for a computer user to tifiable eye-scan patterns were recognizable. The or-
select something on a screen by using a pointing de- der in which a person looks at things also is a key to
vice. Interest Tracker registers whether a person is what that person is thinking. Research on an ex-
paying attention to, for example, news feeds, stock periment called Invision uses this fact in a user in-
prices, or help information and learns what titles terface to prioritize activities. Invisions grouping of
to audition at the bottom of the screen. things by the way a person looks improves eye track-
MAGIC (Manual and Gaze Input Cascaded) ing and uses gaze to group things of interest.
pointing is a technique that lets a computer mouse Knowing that an eye moves between staring fixa-
manipulate what a users eyes look at on a screen. An tions can help find those fixations. By analyzing the
ATTENTIVE USER INTERFACE 53
eye-travel vectors between fixation vertices, Invision a sensor can recognize many aspects of attention.
gains a much more accurate idea of what a person EyeaRe consists of a Microchip PIC microprocessor
is trying to look at than by analyzing that persons that records and runs the system, an LED and a
dwell time on a particular item. photo diode looking at the eye, and another
Attending to the order in which people look at LED/photo diode pair that measures whether it is
things provides a powerful interface tool. Invision in front of other EyeaRe devices and communicates
demonstrates that an attentive user interface can be information. An IR channel communicates to a
driven from insights about where people look. video base station or a pair of glasses.
Scenarios are created in which the attentive pattern If an EyeaRe user is staring, the IR reflection off
of the eye gaze can be understood by a com- his eye does not change. Staring at a video base sta-
puter. By watching the vertices of a persons eye mov- tion starts a video; glancing away stops it. The video
ing through a visual field of company names, the image can detect whether a user is paying attention
system notices which ones interest the person. to it; if the user doesnt like it and blinks her eyes
The company names aggregate themselves into clus- in frustration, the system puts up a more pleasing
ters on the screen based on the persons scanning image.
patterns. When two people stare at each other, EyeaRe
A similar approach uses an ecological interface uses the IR communication channel to exchange in-
that is an image of a kitchen with several problems. formation. When one person stares at another per-
On the counter is a dish with some food on it; the son, the person being stared at receives the contact
oven door is slightly ajar, as are the dishwasher information of the person who is staring.
and refrigerator doors. The manner in which a per- People tend to move their eyes until they have
sons eyes move around the kitchen image allows the to look 15 degrees to the side; EyeaRe has an 18-
interface to understand whether the person is hun- degree horizontal field of view. Thus, gaze and blink
gry, thinking of taking care of problems, or think- detection occurs when a person looks at the EyeaRe
ing about something else in the kitchen. The base station or glasses. EyeaRe demonstrates that a
interface uses the order in which the person views system that doesnt even track the eye can under-
things in the image to bring up a menu and so forth. stand the intentions of attention.
This approach aggregates eye motions into a
story of what the person wants to do. The attention
model drives the interface. The vertices of change A Simple Attentive
in direction of eye movements easily give focus lo-
cations that have eluded most eye tracking research. Eye-Gesture Language
To take eye communication one step further, the Eye
Bed interface uses an eye-gesture language to perform
Ocular Attention tasks that are helpful to a person lying in bed. The Eye
Bed demonstrates that computers can be attentive to
without Eye Tracking peoples need to be horizontal eight hours a day. The
EyeaRe is an ocular attention system that is based Eye Bed interface uses eye tracking housed in a con-
on the fact that many of the social cues that are made verted lamp hanging over the head of the person in
by an eye do not depend on where the eye is look- bed. This interface easily distinguishes between star-
ing. In fact, EyeaRe has no eye tracking system. It ing at an object on the ceiling and glancing around in-
simply measures reflected infrared (IR) from the differently. A language of attentional eye gestures drives
sclera (the opaque white outer coat enclosing the the scenario. Glancing around shows lack of attention,
eyeball except the part covered by the cornea) and whereas staring demonstrates attention. Blinking a
pupil to a photo diode. The system uses this reflected long wink-like blink means selection. Blinking rapidly
infrared to determine whether the eye is open, closed, means dislike. Closing the eyes could mean that the
blinking, winking, or staring. Without a camera such user is going to sleep; thus, a sunset and a nighttime
scenario begin. Opening the eyes makes a morning their attentioneven encouraging their ocular focus
and wakeup scenario begin. Intelligent systems ana- to be on what they want to do. Attentive user inter-
lyze a persons reactions to media on music and video faces allow peoples attention to make things happen.
jukeboxes. The media offerings are auditioned to de-
tect the attention shown them. Blinking when one Ted Selker
doesnt like the media makes the system know that
it should choose other music or video to show the per- See also Eye Tracking
son. Winking or closing the eyes turns off the system.
The reading of eye gestures becomes an attentive user
interface. FURTHER READING
Understanding attention requires a model of what
eye movement means. Researchers can make a com- Bolt, R. A. (1985). Conversing with computers. Technology Review, 88(2),
plexity of interfaces from some simple observations 3443.
of eye behavior. As an output device the eye is a Gregory, R. L. (1997). Eye and brain: The psychology of seeing. Oxford,
UK: Oxford University Press.
simpler user interface tool than is normally described. Guo, X. (1999). Eye contactTalking about non-verbal communica-
The eye can easily be used with a language of closing, tion: A corpus study. Retrieved April 29, 2004, from http://www.lan-
opening, blinking, winking, making nervous move- guagemagazine.com/internetedition/ma99/sprpt35.html
ments, glancing around, and staring. This language Maglio, P. P., Barrett, R., Campbell, C. S., & Selker, T. (2000). SUITOR:
An attentive information system. New York: ACM Press.
can be sensed with eye-tracking cameras or with a Morimoto, D., & Flickner, M. (2000). Pupil detection using multiple light
simple reflected LED, as the EyeaRe system demon- sources. Image and Vision Computing, 18, 331335.
strates. Nervous TV newscasters blink more. (1999). Retrieved April 29, 2004,
from http://www.doctorbob.com/news/7_24nervous.html
Rice, R., & Love, G. (1987). Electronic emotion: Socioemotional content
in a computer-mediated communication. Communication Research,
Promises of the Future 14(1), 85108.
Attentive user interfaces hold great promise. People Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern approach.
Upper Saddle River, NJ: Prentice Hall.
are now in a position to implement and extend such Selker, T., & Burleson, W. (2000). Context-aware design and interaction
interfaces. The hardware to create and test them is eas- in computer systems. IBM Systems Journal, 39(34), 880891.
ily accessible. With the use of the eye as a secondary Shepard, R. N. (1967). Recognition memory for words, sentences and
indicator of intention, researchers can make robust pictures. Journal of Verbal Learning and Verbal Behavior, 6, 156163.
and computationally simple visual interfaces.

Models of human intention and attention are be-
coming part of all human-computer interfaces. The
context of where we are and what we are doing can AUGMENTED
accomplish more than automatically opening the gro-
cery store door. Many interfaces can be driven com- COGNITION
pletely by noticing a persons attention.
Sensors in a given context can detect many things Augmented cognition is a field of research that seeks
about human attention. For example, a sensor pad in to extend a computer users abilities via technologies
front of an office door can detect if a person has ar- that address information-processing bottlenecks in-
rived to visit. Many biometrics (relating to the sta- herent in human-computer interaction (HCI). These
tistical analysis of biological observations and bottlenecks include limitations in attention, mem-
phenomena) such as EEG changes, sweat responses, ory, learning, comprehension, visualization abilities,
and heart rate variability are candidates for attentive and decision-making. Limitations in human cogni-
user interfaces. tion (the act or process of knowing) are due to in-
People want to focus on what they are doing trinsic restrictions in the number of mental tasks
and on the people they are with. Attentive user in- that a person can execute at one time, and these re-
terfaces can detect peoples intentions without taking strictions may fluctuate from moment to moment
AUGMENTED COGNITION 55
depending on a host of factors, including mental fa- Early Investments in Related Work
tigue, novelty, boredom, and stress. Augmented cognition does not draw from just one
As computational interfaces have become more scientific fieldit draws from fields such as neuro-
prevalent in society and increasingly complex with science, biopsychology, cognitive psychology, human
regard to the volume and type of information pre- factors, information technology, and computer
sented, researchers have investigated novel ways to science. Each of these fields has itself undergone a
detect these bottlenecks and have devised strategies substantial revolution during the past forty years that
to aid users and improve their performance via tech- has allowed the challenges raised by researchers to
nologies that assess users cognitive status in real time. begin to be investigated. Although many individ-
A computational interaction monitors the state of a ual research projects contributed to the general de-
user through behavioral, psychophysiological, and/or velopment and direction of augmented cognition,
neurophysiological data and adapts or augments the several multimillion-dollar projects helped shape the
computational interface to significantly improve users foundation on which the field is built.
performance on the task at hand. Since the invention of the electronic computer,
scientists and engineers have speculated about the
unique relationship between humans and comput-
Emergence of ers. Unlike mechanized tools, which are primarily
devices for extending human force and action, the
Augmented Cognition computer became an entity with which humans
The cognitive science and HCI communities have forged an interactive relationship, particularly as
researched augmented cognition for several decades. computers came to permeate everyday life. In 1960
Scientific papers in this field increased markedly dur- one of the great visionaries of intelligent comput-
ing the late 1990s and addressed efforts to build and ing, J. C. R. Licklider, wrote a paper entitled Man-
use models of attention in information display and Computer Symbiosis. Licklider was director of the
notification systems. However, the phrase aug- Information Processing Techniques Office (IPTO)
mented cognition associated with this research did at the Defense Departments Advanced Research
not find widespread use until the year 2000, when Projects Agency (ARPA) during the 1960s. In his pa-
a U.S. Defense Department Defense Advanced per he stated, The hope is that, in not too many
Research Project Agency (DARPA) Information years, human brains and computing machines will
Science and Technology (ISAT) group study and a be coupled together very tightly, and that the resulting
workshop on the field at the National Academy of partnership will think as no human brain has ever
Sciences were held. During the year 2002 the num- thought and process data in a way not approached
ber of papers about augmented cognition increased by the information-handling machines we know to-
again. This increase was due, in part, to the start of day (Licklider 1960, 4). Almost prophetic, this de-
a DARPA research program in augmented cognition scription of the symbiotic relationship between
in 2001 with a focus on challenges and opportuni- humans and computers is one of the first descrip-
ties with the real-time monitoring of cognitive states tions of what could be considered an augmented cog-
with physiological sensors. This substantial invest- nition computational system. Although research on
ment in these developing technologies helped bring this topic was not conducted during his tenure at
together a research community and stimulated a set ARPA during the 1960s, Licklider championed the
of thematically related projects on addressing cog- research that developed into the now-burgeoning
nitive bottlenecks via the monitoring of cognitive field of computer science, including creation of the
states. By 2003 the augmented cognition field ex- Arpanet computer network (forerunner of the
tended well beyond the boundaries of those specific Internet). His research, vision, and direction had a
Defense Department research projects, but that ini- significant impact on both computer science and in-
tial investment provided impetus for the infant field formation technology and set the stage for the
to begin to mature. field of augmented cognition.
During the early 1960s researchers speculated corporate any physiological monitoring. Cognitive
that electrical signals emanating from a human brain modeling was the cornerstone of the pilot-vehicle in-
in the form of electroencephalographic (EEG) terface (PVI), which had the critical task of manag-
recordings could be used as indicators of specific ing all pilot interactions with the system by inferring
events in human cognitive processing. Several the pilots intentions and communicating these in-
Department of Defense investigations into detect- tentions to the other components of the PA system.
ing these signals and other measurements oc- The PVI was also responsible for modeling pilot work-
curred through the biocybernetics and learning load to adapt and configure the information displays
strategies programs sponsored by ARPA during in the cockpit, conveying workload information to
the 1970s and 1980s. The earliest program was the other subsystems, and compensating for pilot be-
biocybernetics, which tested the hypothesis that EEG havior that might result in an error. An example of
activity might be able to control military devices and this work was a PA program at NASA-Ames Research
serve as indicators of user performance. In this pro- Center that explored the use of probabilistic mod-
gram biocybernetics was defined as a real-time con- els of a pilots goals and workload over time, based
nection between the operator and computational on multiple inputs and the use of models to con-
system via physiological signals recorded during spe- trol the content and complexity of displays. Such
cific tasks. Both the biocybernetics and learning models did not employ physiological measures of a
strategies programs centered around the creation of pilots cognitive status.
closed-loop feedback systems (the relationship be- Other research occurred in the academic and pri-
tween user and computational system, where changes vate sectors, including the attentional user interface
in the computational interface are driven by detected (AUI) project at Microsoft Research during the
changes in the users physiological status, which in late 1990s, which provided conceptual support to ef-
turn change as a result of the new format of the in- forts in augmented cognition. Researchers developed
terface) between operator and computer for the methods for building statistical models of attention
selection and training of personnel, display/con- and workload from data. Researchers built archi-
trol design, and online monitoring of operator sta- tectures to demonstrate how cognitive models could
tus (although with slightly different military be integrated with real-time information from mul-
application domains between the two programs). tiple sensors (including acoustical sensing, gaze and
In both programs researchers saw the real-time head tracking, and events representing interaction
identification of cognitive events as critical to un- with computing systems) to control the timing
derstanding the best methods for aiding military and communication medium of incoming notifica-
users in a rapid and contextually appropriate way. tions. AUI work that included psychological studies
However, when this research was begun, both complemented the systems and architectures work.
computational systems and neuroscience were in
their infancy, and the results of this research were
not incorporated into production military sys- Foundations of Augmented Cognition
tems. Augmented cognition can be viewed as a de- In light of these earlier research efforts, the logical
scendant of these early programs. question arises: What sets augmented cognition apart
Another investigation in this field was the Pilots from what has already been done? As mentioned,
Associate (PA) program sponsored by DARPA dur- augmented cognition relies on many fields whose
ing the 1980s and early 1990s. Pilots Associate was maturity is critical for its success. Although programs
an integrated system of five components that in- such as biocybernetics during the 1970s had similar
corporated AI (artificial intelligence) techniques and goals, they did not have access to the advanced com-
cognitive modeling to aid pilots in carrying out their putational power necessary to process brain sig-
missions with increased situational awareness and nals in real time, nor did researchers know enough
enhanced decision-making. Unlike biocybernetics, about those signals to use them to control displays
PA utilized cognitive modeling alone and did not in- or machines. Likewise, the Pilots Associate program
AUGMENTED COGNITION 57
during the 1980s shared many aspirations of todays

augmented cognition, namely to develop adaptive INTELLIGENT AGENT Software program that actively lo-
interfaces to reduce pilot workload. However, PA cates information for you based on parameters you set.
could assess the status of a pilot from inferences and Unlike a search engine or information filter, it actively
models based only on the pilots overt behavior and seeks specific information while you are doing
the status of the aircraft. What distinguishes aug- other things.
mented cognition is its capitalization on advances
in two fields: behavioral/neural science and com-
puter science. ical sensors, but the field of augmented cognition is
At the start of the twenty-first century researchers broadened even further by their inclusion.
have an unparalleled understanding of human brain As a result of the Decade of the Brain, research-
functioning. The depth of this understanding is due to ers have an increased knowledge of the cognitive lim-
the development of neuroscientific techniques funded itations that humans face. The HCI field focuses on
by the U.S. National Institutes of Health (NIH) and the design, implementation, and evaluation of in-
other agencies during the 1990s, a period now referred teractive systems in the context of a users work.
to as the Decade of the Brain. The billion-dollar fund- However, researchers in this field can work only
ing of the fields of neuroscience, cognitive science, and with the data and observations easily accessible to
biopsychology resulted in some of the greatest advances them, that is, how people overtly behave while using
in our understanding of the human biological system interfaces. Through efforts in neuroscience, biopsy-
in the twentieth century. For example, using techniques chology, and cognitive neuroscience we can locate and
such as functional magnetic resonance imaging (fMRI), measure activity from the brain regions that are ac-
scientists were able to identify discrete three-dimen- tively involved in day-to-day information-processing
sional regions of the human brain active during spe- tasks. Researchers will have a greater understanding
cific mental tasks. This identification opened up the of the cognitive resources that humans possess and
field of cognitive psychology substantially (into the how many of these resources are available during a
new field of cognitive neuroscience) and enabled re- computationally based task, whether or not the com-
searchers to test their theories of the human mind and putational systems will include advanced sensors. After
associate previously observed human thoughts and these cognitive resources are identified and their ac-
behaviors with neural activity in specific brain regions. tivity (or load) measured, designers of computational
Additional investment from the Department of Defense interfaces can begin to account for these limitations
and other agencies during the twenty-first century has (and perhaps adapt to their status) in the design of
allowed researchers to develop even more advanced new HCI systems.
sensors that will eventually be used in augmented cog- Finally, without advances in computer science
nition systems. Novel types of neurophysiological sig- and engineering, none of the neuroscientific devel-
nals that are measurable noninvasively include electrical opments listed here would be possible, and the field
signalsusing electroencephalography and event-re- of augmented cognition would certainly not be fea-
lated potentials (identifiable patterns of activity within sible. During the past forty years society has experi-
the EEG that occur either before specific behaviors are enced leaps in computational prowess and the
carried out, or after specific stimuli are encountered) sophistication of mathematical algorithms. These
and local cortical changes in blood oxygenation leaps have been due in part to the miniaturization of
(BOLD), blood volume, and changes in the scattering transistors and other silicon-based components so
of light directly due to neuronal firing (using near that more computational power is available per square
infrared [NIR] light). Most of these signals, unlike inch of hardware. This miniaturization has allowed
fMRI, can be collected from portable measurement computers to shrink in size until they have perme-
systems in real time, making them potentially avail- ated the very fabrics that people wear and even
able for everyday use. All augmented cognition systems their environments. Computer code itself has become
do not necessarily contain advanced neurophysiolog- smaller and more flexible, with the emergence of
agent-based computing (the instantiation of active, tention, memory, learning, comprehension, sensory
persistent software components that perceive, rea- processing, visualization abilities, qualitative judg-
son, act, and communicate in software code), JAVA, ments, serial processing, and decision-making. For an
and Internet services. Thus, augmented cognition has augmented cognition system to be successful it
benefited from two computing advancesimprove- must identify at least one of these bottlenecks in real
ments in raw computational resources (CPUs, phys- time and alleviate it through a performance-enhanc-
ical memory) and improvements in the languages ing mitigation strategy. Such mitigation strategies are
and algorithms that make adaptive interfaces possi- conveyed to the user through the adaptive interface
ble. Many other fields have benefited from these and might involve modality switching (between
advances as well and in turn have fed into the aug- visual, auditory, and haptic [touch]), intelligent in-
mented cognition community. These fields include terruption, task negotiation and scheduling, and as-
user modeling, speech recognition, computer vision, sisted context retrieval via book marking. When a
graphical user interfaces, multimodal interfaces, and user state is correctly sensed, an appropriate strat-
computer learning/artificial intelligence. egy is chosen to alleviate the bottleneck, the inter-
face is adapted to carry out the strategy, and the
Components of an resulting sensor information indicates that the aiding
has workedonly then has a system closed the loop
Augmented Cognition System and successfully augmented the users cognition.
At the most general level, augmented cognition har-
nesses computation and knowledge about human
limitations to open bottlenecks and address the bi- Applications of Augmented Cognition
ases and deficits in human cognition. It seeks to The applications of augmented cognition are nu-
accomplish these goals through continual back- merous, and although initial investments in systems
ground sensing, learning, and inferences to under- that monitor cognitive state have been sponsored by
stand trends, patterns, and situations relevant to a military and defense agencies, the commercial sec-
users context and goals. At its most general level, an tor has shown interest in developing augmented cog-
augmented cognition system should contain at least nition systems for nonmilitary applications. As
four componentssensors for determining user mentioned, closely related work on methods and ar-
state, an inference engine or classifier to evaluate in- chitectures for detecting and reasoning about a users
coming sensor information, an adaptive user in- workload (based on such information as activity with
terface, and an underlying computational architecture computing systems and gaze) have been studied
to integrate the other three components. In reality a for nonmilitary applications such as commercial no-
fully functioning system would have many more tification systems and communication. Agencies such
components, but these are the most critical. as NASA also have shown interest in the use of meth-
Independently, each of these components is fairly ods to limit workload and manage information over-
straightforward. Much augmented cognition research load. Hardware and software manufacturers are
focuses on integrating these components to close always eager to include technologies that make their
the loop and create computational systems that systems easier to use, and augmented cognition sys-
adapt to their users. tems would likely result in an increase in worker pro-
Thus, the primary challenge with augmented cog- ductivity with a savings of both time and money
nition systems is not the sensors component (although to companies that purchased these systems. In more
researchers are using increasingly complex sensors). specific cases, stressful jobs that involve constant in-
The primary challenge is accurately predicting/as- formation overload from computational sources,
sessing, from incoming sensor information, the cor- such as air traffic control, would also benefit from
rect state of the user and having the computer select such technology. Finally, the fields of education and
an appropriate strategy to assist the user at that training are the next likely targets for augmented
time. As discussed, humans have limitations in at- cognition technology after it reaches commercial vi-
AUGMENTED REALITY 59
ability. Education and training are moving toward Wilson, G. F. (2001). Real-time adaptive aiding using psychophysio-
an increasingly computational medium. With dis- logical operator state assessment. In D. Harris (Ed.), Engineering
psychology and cognitive ergonomics (pp. 175182). Aldershot, UK:
tance learning in high demand, educational systems Ashgate.
will need to adapt to this new nonhuman teaching Wilson, R. A., & Keil, F. C. (Eds.). (2001). The MIT encyclopedia of the
interaction while ensuring quality of education. cognitive sciences (MITECS). Cambridge, MA: MIT Press.
Augmented cognition technologies could be applied
to educational settings and guarantee students a
teaching strategy that is adapted to their style of learn-
ing. This application of augmented cognition could AUGMENTED REALITY
have the biggest impact on society at large.
Augmented reality is a new field of research that con-
Dylan Schmorrow and Amy Kruse centrates on integrating virtual objects into the
real world. These virtual objects are computer graph-
See also Augmented Reality; Brain-Computer ics displayed so that they merge with the real world.
Interfaces; Information Overload Although in its infancy, augmented reality holds out
the promise of enhancing peoples ability to perform
certain tasks. As sensing and computing technolo-
FURTHER READING gies advance, augmented reality is likely to come to
play a significant role in peoples daily lives.
Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical
review of 275 PET and fMRI studies. Journal of Cognitive
Neuroscience, 12(1), 147.
Dix, A., Finlay, J., Abowd, G., & Beale, R. (1998). Human computer in-
teraction (2nd ed.). London, New York: Prentice Hall.
Augmented Reality
Donchin, E. (1989). The learning strategies project. Acta Psychologica,
71(13), 115.
and Virtual Reality
Freeman, F. G., Mikulka, P. J., Prinzel, L. J., & Scerbo, M. W. (1999). An augmented-reality system merges the real scene
Evaluation of an adaptive automation system using three EEG in- viewed by the user with computer-generated virtual
dices with a visual tracking task. Biological Psychology, 50(1), 6176.
Gevins, A., Leong, H., Du, R., Smith, M. E., Le, J., DuRousseau, D.,
objects to generate a composite view for the user.
Zhang, J., & Libove, J. (1995). Towards measurement of brain func- The virtual objects supplement the real scene with
tion in operational environments. Biological Psychology, 40, 169186. additional and useful information. Sounds may be
Gomer, F. (1980). Biocybernetic applications for military systems. added through the use of special headphones that
Chicago: McDonnell Douglas.
Gray, W. D., & Altmann, E. M. (2001). Cognitive modeling and hu- allow the user to hear both real sounds and syn-
man-computer interaction. In W. Karwowski (Ed.), International thesized sounds. There are also special gloves that
encyclopedia of ergonomics and human factors (pp. 387391). New a user can wear that provide tactile sensation such
York: Taylor & Francis.
Horvitz, E., Pavel, M., & Schmorrow, D. D. (2001). Foundations of aug-
as hardness or smoothness. A user wearing such
mented cognition. Washington, DC: National Academy of Sciences. gloves could feel virtual furniture in a real room.
Humphrey, D. G., & Kramer, A. F. (1994). Toward a psychophysio- In an augmented-reality system, users can walk
logical assessment of dynamic changes in mental workload. Human around a real room, hear the echo of their footsteps,
Factors, 36(1), 326.
Licklider, J. C. R. (1960). Man-computer symbiosis: IRE transac-
and feel the breeze from an air conditioning unit,
tions on human factors in electronics. HFE-1 (pp. 411). while at the same time they can see computer-gen-
Lizza, C., & Banks, S. (1991). Pilots Associate: A cooperative, erated images of furniture or paintings.
knowledge-based system application. IEEE Intelligent Systems, 6(3), One of the requirements of an augmented-real-
1829.
Mikulka, P. J., Scerbo, M. W., & Freeman, F. G. (2002). Effects of a bio- ity system is that it needs to be interactive in real
cybernetic system on vigilance performance. Human Factors, 44, time. Animation, sound, and textures are added in
654664. real time so that what the user sees, hears, and feels
Prinzel, L. J., Freeman, F. G., Scerbo, M. W., Mikulka, P. J., & Pope,
A. T. (2000). A closed-loop system for examining psychophysio-
reflects the true status of the real world. The most
logical measures for adaptive task allocation. International Journal important characteristic of augmented reality is the
of Aviation Psychology, 10, 393410. ability to render objects in three-dimensional space,
which makes them much more realistic in the eyes This reduces the safety risk, since the user can see the
of the user. Virtual objects are drawn in relationship real world in real time. If there is a power failure, the
to the real objects around them, both in terms of po- user will still be able to see as well as he or she would
sition and size. If a virtual object is situated partially when wearing dark sunglasses. If there is some kind
behind a real object (or vice versa) then the user of hazard moving through the areaa forklift, for
should not see part of the obscured object. Occlusion examplethe wearer does not have to wait for the sys-
of objects is the largest contributor to human depth tem to process the image of the forklift and display
perception. it; the wearer simply sees the forklift as he or she
The major difference between augmented reality would when not wearing the HMD. One disadvan-
and virtual reality is that in virtual reality everything tage is that the virtual objects may appear to lag
that is sensed by the user is computer generated. behind the real objects; this happens because the
Therefore the virtual objects must be rendered as virtual objects must be processed, whereas real objects
photorealistically as possible in order to achieve the do not need to be. In addition, some users are reluc-
feeling of immersion. Augmented reality uses both tant to wear the equipment for fear of harming their
real and synthetic sights, sounds, and touches to con- vision, although there is no actual risk, and other users
vey the desired scene, so virtual objects do not bear dislike the equipments cumbersome nature. A new
the entire burden of persuading the user that the scene version of the see-through HMD is being developed
is real, and therefore they do not need to be so pho- to resemble a pair of eyeglasses, which would make
torealistic. Augmented reality lies in the middle of the it less cumbersome.
continuum between absolute reality (in which every- Closed-view HMDs cannot be seen through.
thing sensed is real) and virtual reality (in which They typically comprise an opaque screen in front
everything that is sensed is created). of the wearers eyes that totally blocks all sight of the
real world. This mechanism is also used for tradi-
tional virtual reality. A camera takes an image of the
Different Types of Displays real world, merges it with virtual objects, and
presents a composite image to the user. The advan-
for Augmented Reality tage the closed-view has over the see-through ver-
Most people depend on vision as their primary sen- sion of the HMD is that there is no lag time for
sory input, so here we will discuss several types of the virtual objects; they are merged with the real
visual displays that can be used with augmented scene before being presented to the user. The dis-
reality, each with its own advantages and disadvan- advantage is that there is a lag in the view of the real
tages. Visual displays include head-mounted displays world because the composite image must be
(HMDs), monitor-based displays, projected images, processed before being displayed. There are two
and heads-up displays (HUDs). safety hazards associated with closed-view HMD.
First, if the power supply is interrupted, the user is
Head-Mounted Displays essentially blind to the world around him. Second,
HMDs are headsets that a user wears. HMDs can the user does not have a current view of the real
either be see-through or closed view. The see-through world. Users have the same concerns and inhibitions
HMD works as its name implies: The user looks regarding closed-view HMD as they do regarding
through lenses to see the real world, but the lenses see-through HMD.
are actually display screens that can have graphics
projected onto them. The biggest advantage of the Monitor-Based Displays
see-through HMD mechanism is that it is simple Monitor-based displays present information to the
to implement because the real world does not have user for configuring an augmented-reality system
to be processed and manipulated; the mechanisms this way. First, because a monitor is a separate dis-
only task is to integrate the visual augmentations. play device, more information can be presented to
the user. Second, the user does not have to wear involves placing the virtual objects in the proper lo-
(or carry around) heavy equipment. Third, graphi- cations in the real world. This is an important ele-
cal lag time can be eliminated because the real world ment of augmented reality and includes sensing,
and virtual objects are merged in the same way they calibration, and tracking. Appearance concerns what
are for closed-view HMDs. The safety risk is avoided the virtual objects look like. In order to achieve seam-
because the user can see the real world in true real less merging of real and virtual objects, the virtual
time. objects must be created with realistic color and
There are also some drawbacks to using moni- texture.
tor-based displays instead of HMDs. First, the user In virtual-reality systems, tracking the relative
must frequently look away from the workspace in position and motion of the user is an important
order to look at the display. This can cause a slow- research topic. Active sensors are widely used to track
down in productivity. Another problem is that the position and orientation of points in space. The
user can see both the real world andon the tracking information thus obtained is fed into the
monitorthe lagging images of the real world. In computer graphics system for appropriate render-
a worse case situation in which things in the scene ing. In virtual reality, small errors in tracking can be
are moving rapidly, the user could potentially see a tolerated, as the user can easily overlook those errors
virtual object attached to a real object that is no in the entirely computer-generated scene. In aug-
longer in the scene. mented-reality systems, by contrast, the registration
is performed in the visual field of the user. The
Projected-Image Displays type of display used in the system usually determines
Projected-image displays project the graphics and the accuracy needed for registration.
annotations of the augmented-reality system onto One popular registration technique is vision-
the workspace. This method eliminates the need for based tracking. Many times, there are fiducials
extra equipment and also prevents the user from hav- (reference marks) marked out in the scene in which
ing to look away from the work area to check the the virtual objects need to be placed. The system rec-
monitor-based display. The biggest disadvantage is ognizes these fiducials automatically and determines
that the user can easily occlude the graphics and an- the pose of the virtual object with respect to the scene
notations by moving between the projector and before it is merged. There are also techniques that
the workspace. Users also can put their hands and use more sophisticated vision algorithms to deter-
arms through the projected display, reducing their mine the pose without the use of fiducials. The mo-
sense of the reality of the display. tion of the user and the structure of the scene are
computed using projective-geometry formulation.
Heads-Up Displays (Projective geometry is the branch of geometry that
Heads-up displays are very similar to see-through deals with projecting a geometric figure from one
HMDs. They do not require the user to wear special plane onto another plane; the ability to project points
headgear, but instead display the data on a see-through from one plane to another is essentially what is
screen in front of the user. As in seethrough HMDs, needed to track motion through space.)
these systems are easy to implement, however, there For a seamless augmented-reality system, it is
may be a lag time in rendering the virtual object. important to determine the geometry of the vir-
tual object with respect to the real scene, so that
occlusion can be rendered appropriately. Stereo-based
Challenges in Augmented Reality depth estimation and the z-buffer algorithm (an al-
A majority of the challenges facing augmented real- gorithm that makes possible the representation of
ity concern the virtual objects that are added to the objects that occlude each other) can be used for
real world. These challenges can be divided into two blending real and virtual objects. Also, using research
areas: registration and appearance. Registration results in radiosity (a technique for realistically
simulating how light reflects off objects), it is possi- is placed on top of it on the television video to make
ble to shade the virtual object appropriately so that it easier for those watching the game on television
it blends properly with the background scene. to follow the rapid motion of the puck. Augmented
reality could also make possible a type of virtual set,
very similar to the blue-screen sets that are used
Applications today to film special effects. Augmented-reality
Augmented reality has applications in many fields. sets would be interactive, would take up less space,
In medicine, augmented reality is being researched and would potentially be simpler to build than tra-
as a tool that can project the output of magnetic res- ditional sets. This would decrease the overall cost of
onance imaging (MRI), computed tomography (CT) production. Another example, already developed is
scans, and ultrasound imaging onto a patient to the game AR2 Hockey, in which the paddles and field
aid in diagnosis and planning of surgical operations. (a table, as in air hockey) are real but the puck is vir-
Augmented reality can be used to predict more actual. The computer provides visual tracking of the
curately where to perform a biopsy for a tiny tumor: virtual puck and generates appropriate sound effects
All the information gathered from traditional meth- when the paddles connect with the puck or when the
ods such as MRIs can be projected onto the patient puck hits the table bumpers.
to reveal the exact location of the tumor. This en- One military application is to use the technol-
ables a surgeon to make precise incisions, reducing ogy to aim weapons based on the movement of the
the stress of the surgery and decreasing the trauma pilots head. Graphics of targets can be superimposed
to the patient. on a heads-up display to improve weapons accuracy
In architecture and urban planning, annotation by rendering a clearer picture of the target, which
and visualization techniques can be used to show will be hard to miss.
how the addition of a building will affect the sur- Many examples of assembly augmented-real-
rounding landscape. Actually seeing the future build- ity systems have been developed since the 1990s.
ing life sized, in the location it will occupy, gives a One of the best known is the Boeing wire-bundling
more accurate sense of the project than can be project, which was started in 1990. Although well
conveyed from a model. Augmented-reality simula- known, this project has not yet been implemented
tions also make it easier to recognize potential prob- in a factory as part of everyday use. The goal is
lems, such as insufficient natural lighting for a relatively straightforward: Use augmented reality to
building. aid in the assembly of wire bundles used in Boeings
Augmented reality also has the potential to let 747 aircraft.
developers, utility companies, and home owners see For this project, the designers decided to use a
where water pipes, gas lines, and electrical wires see-through HMD with a wearable PC to allow work-
are run through walls, which is an aid when it comes ers the freedom of movement needed to assemble
to maintenance or construction work. In order for the bundles, which were up to 19 meters long. The
this technique to be implemented, the data must subjects in the pilot study were both computer sci-
be stored in a format the augmented-reality system ence graduate students who volunteered and Boeing
can use. Simply having a system that can project the employees who were asked to participate.
images of electrical wiring on a wall would not be The developers ran into both permanent and
sufficient; the system first must know where all the temporary problems. One temporary problem, for
wires are located. example, was that the workers who participated in
Augmented reality has the potential to make a the pilot program were typically tired because
big impact on the entertainment industry. A sim- the factory was running the pilot study at one of
ple example is the glowing puck that is now used the busier times in its production cycle. Workers
in many televised hockey games. In this application, first completed their normal shift before working
the hockey puck is tracked and a brightly colored dot on the pilot project. Another temporary problem
was the curiosity factor: Employees who were tial calibration that must be performed as part of the
not involved with the project often came over to start-up process. The calibration is then per-
chat and check out what was going on and how the formed periodically when the system becomes con-
equipment worked. More permanent problems fused or the error rate increases past a certain
were the employees difficulties in tracing the wires threshold. Users seemed to have difficulty keeping
across complex subassemblies and their hesi- their heads still enough for the sensitive calibra-
tance to wear the headsets because of fear of the tion process, so a headrest had to be built. Another
lasers located close to their eyes and dislike of the problem was that the magnetic tracking devices did
helmet head effect that came from wearing the not work well because there were so many metal parts
equipment. in the assembly. In addition, the speech recogni-
One of the strongest success points for this pilot tion part of the system turned out to be too sensi-
study was that the bundles created using the aug- tive to background noise, so it was turned off.
mented-reality system met Boeings quality assur- The pilot study for this project was used as a
ance standards. Another good thing was that the demonstration at a trade show in Germany in 1998.
general background noise level of the factory did not The program ran for one week without difficulty.
interfere with the acoustic tracker. In the pilot study, Due to time considerations, the system was not cal-
augmented reality offered no improvement in pro- ibrated for each user, so some people were not as
ductivity and the only cost savings came from no impressed as the developers had hoped. Also,
longer needing to store the various assembly boards. even with the headrest, some users never stayed still
(This can be, however, a significant savings.) The de- long enough for a proper calibration to be per-
velopers concluded that the reason there was no sig- formed. Their reactions showed researchers that av-
nificant improvement in assembly time was because erage users require some degree of training if
they still had some difficulty using the systems in- they are to use this sort of equipment success-
terface to find specific wires. The developers are work- fully. Despite setbacks, the developers considered
ing on a new interface that should help to solve the pilot a success because it brought the technol-
this problem. ogy to a new group of potential users and it gener-
Augmented reality has also been used in BMW ated several possible follow-up ideas relating to the
automobile manufacture. The application was de- door lock assembly.
signed to demonstrate the assembly of a door lock
for a car door, and the system was used as a feasi-
bility study. The annotations and graphics were taken The Future
from a CAD (computer-aided design) system that Augmented reality promises to help humans in
was used to construct the actual physical parts for many of their tasks by displaying the right infor-
the lock and the door. In this case, the augmented- mation at the right time and place. There are many
reality system uses a see-through HMD and a voice- technical challenges to be overcome before such
activated computerin part because the assembly interfaces are widely deployed, but driven by com-
process requires that the user have both hands free pelling potential applications in surgery, the mili-
for the assembly process. Because this augmented- tary, manufacturing, and entertainment, progress
reality system mimicked an existing virtual-reality continues to be made in this promising form of
version of assembly planning for the door lock as- human-computer interaction.
sembly, much of the required data was already avail-
able in an easily retrievable format, which simplified Rajeev Sharma and Kuntal Sengupta
the development of the augmented-reality system.
The developers had to overcome certain prob- See also Augmented Cognition; Virtual Reality
lems with the system in order to make the pilot work.
The first was the issue of calibration. There is an ini-
FURTHER READING AVATARS

Aliaga, D. G. (1997). Virtual objects in the real world. Communi- Avatar derives from the Sanskrit word avatarah,
cations of the ACM, 40(3), 4954.
Azuma, R. (1997). A survey of augmented reality. Presence:
meaning descent and refers to the incarnation
Teleoperators and Virtual Environments, 6(3), 355385. the descent into this worldof a Hindu god. A Hindu
Bajura, M., Fuchs, H., & Ohbuchi, R. (1992). Merging virtual objects deity embodied its spiritual being when interacting
with the real world: Seeing Ultrasound imagery within the with humans by appearing in either human or ani-
patient. Computer Graphics (Proceedings of SIGGRAPH92),
26(2), 203210. mal form. In the late twentieth century, the term
Das, H. (Ed.). (1994). Proceedings of the SPIEThe International avatar was adopted as a label for digital represen-
Society for Optical Engineering. Bellingham, WA: International tations of humans in online or virtual environments.
Society for Optical Engineering.
Elvins, T. T. (1998, February). Augmented reality: The futures so
Although many credit Neal Stephenson with being
bright I gotta wear (see-through) shades. Computer Graphics, the first to use avatar in this new sense in his semi-
32(1), 1113. nal science fiction novel Snow Crash (1992), the term
Ikeuchi, K., Sato, T., Nishino, K., & Sato, I. (1999). Appearance mod- and concept actually appeared as early as 1984 in on-
eling for mixed reality: photometric aspects. In Proceedings of the
1999 IEEE International Conference on Systems, Man, and
line multiuser dungeons, or MUDs (role-playing en-
Cybernetics (SMC99) (pp. 3641). Piscataway, NJ: IEEE. vironments), and the concept, though not the term,
Milgram, P., & Kishino, F. (1994, December). A taxonomy of mixed appeared in works of fiction dating back to the mid-
reality visual displays. IEICE Transactions on Information 1970s. This entry explores concepts, research, and
Systems, E77-D(12), 13211329.
Neumann, U., & Majoros, A. (1998). Cognitive, performance, and
systems issues for augmented reality applications in manufacturing
and maintenance. In Proceedings of the IEEE 1998 Virtual Reality
Annual International Symposium (pp. 411). Los Alamitos, CA:
IEEE. Embodied Avatar
Ohshima, T., Sato, K., Yamamoto, H., & Tamura, H. (1998). AR2 Agent
hockey: A case study of collaborative augmented reality. In
Proceedings of the IEEE 1998 Virtual Reality Annual International
Symposium (pp. 268275) Los Alamitos, CA: IEEE.
Ong, K. C., Teh, H. C., & Tan, T. S. (1998). Resolving occlusion in im- Digital
age sequence made easy. Visual Computer, 14(4), 153165. Representation
Raghavan, V., Molineros, J., & Sharma, R. (1999). Interactive evalua-
tion of assembly sequences using augmented reality. IEEE
Transactions on Robotics and Automation, 15(3), 435449.
Rosenblum, L. (2000, JanuaryFebruary). Virtual and augmented re-
ality 2020. IEEE Computer Graphics and Applications, 20(1), 3839.
State, A., Chen, D. T., Tector, C., Brandt, A., Chen, H., Ohbuchi, R.,
et al. (1994). Observing a volume rendered fetus within a pregnant
patient. In Proceedings of IEEE Visualization 94 (pp. 364368). Los
Agent Live Human
Alamitos, CA: IEEE. Being
Stauder, J. (1999, June). Augmented reality with automatic illumi-
nation control incorporating ellipsoidal models. IEEE Transactions
on Multimedia, 1(2), 136143.
Tatham, E. W. (1999). Getting the best of both real and virtual worlds.
Communications of the ACM, 42(9), 9698.
Tatham, E. W., Banissi, E., Khosrowshahi, F., Sarfraz, M., Tatham, E., F I G U R E 1 . A representational schematic of avatars and
& Ursyn, A. (1999). Optical occlusion and shadows in a see- embodied agents. When a given digital representation
through augmented reality display. In Proceedings of the 1999 IEEE
International Conference on Information Visualization (pp. 128131). is controlled by a human, it is an avatar, and when it
Los Alamitos, CA: IEEE. is controlled by a computational algorithm it is an em-
Yamamoto, H. (1999). Case studies of producing mixed reality worlds. bodied agent. Central to the current definition is the
In Proceedings of the 1999 IEEE International Conference on Systems,
Man, and Cybernetics (SMC99) (pp. 4247). Piscataway, NJ: IEEE.
ability for real-time behavior, in that the digital rep-
resentation exhibits behaviors by the agent or hu-
man as they are performed.
AVATAARS 65
ethical issues related to avatars as digital human rep- gestures and an agent controls more mundane au-
resentations. (We restrict our discussion to digital tomatic behaviors.
avatars, excluding physical avatars such as puppets One should also distinguish avatars from online
and robots. Currently, the majority of digital avatars identities. Online identities are the distributed dig-
are visual or auditory information though there is ital representations of a person. Humans are known
no reason to restrict the definition as such.) to each other via e-mail, chat rooms, homepages, and
other information on the World Wide Web.
Consequently, many people have an online identity,
Agents and Avatars constituted by the distributed representation of all
Within the context of human-computer interaction, relevant information, though they may not have
an avatar is a perceptible digital representation whose an avatar.
behaviors reflect those executed, typically in real time,
by a specific human being. An embodied agent, by
contrast, is a perceptible digital representation whose Realism
behaviors reflect a computational algorithm designed Avatars can resemble their human counterparts along
to accomplish a specific goal or set of goals. Hence, a number of dimensions, but the two that have re-
humans control avatar behavior, while algorithms ceived the most attention in the literature are be-
control embodied agent behavior. Both agents and havioral realism (reflected in the number of a given
avatars exhibit behavior in real time in accordance humans behaviors the avatar exhibits) and photo-
with the controlling algorithm or human actions. graphic realism (reflected in how many of a given hu-
Figure 1 illustrates the fact that the actual digi- mans static visual features the avatar possesses).
tal form the digital representation takes has no bear- Behavioral realism is governed by the capability
ing on whether it is classified as an agent or avatar: of the implementation system to track and render
An algorithm or person can drive the same repre- behavior in real time. Currently, real-time behavioral
sentation. Hence, an avatar can look nonhuman de- tracking technology, while improving steadily, does
spite being controlled by a human, and an agent can not meet expectations driven by popular culture; for
look human despite being controlled by an algorithm. example, online representations of the character Neo
Not surprisingly, the fuzzy distinction between in The Matrix (1999), Hiro from Snow Crash (1992),
agents and avatars blurs for various reasons. or Case from Neuromancer (1984). In those fictional
Complete rendering of all aspects of a humans ac- accounts, the movements and gestures of avatars and
tions (down to every muscle movement, sound, and the represented humans are generally perceptually
scent) is currently technologically unrealistic. Only indistinguishable. However, in actual practice,
actions that can be tracked practically can be ren- complete real-time behavior tracking is extremely
dered analogously via an avatar; the remainder are difficult. Although gesture tracking through vari-
rendered algorithmically (for example, bleeding) or ous mechanical, optical, and other devices has im-
not at all (minute facial expressions, for instance). proved, the gap between actual movements and avatar
In some cases avatar behaviors are under non- movements remains large, reducing behavioral real-
analog human control; for example, pressing a ism at least in situations requiring real-time tracking
button and not the act of smiling may be the way and rendering, such as online social interaction
one produces an avatar smile. In such a case, the be- (for example, collaborative virtual work groups).
haviors are at least slightly nonanalogous; the Fewer barriers exist for photographic realism.
smile rendered by the button-triggered computer al- Three-dimensional scanners and photogrammetric
gorithm may be noticeably different from the actual software allow for the photographically realistic recre-
humans smile. Technically, then, a human repre- ation of static, digital human heads and faces that
sentation can be and often is a hybrid of an avatar cannot be easily distinguished from photographs and
and an embodied agent, wherein the human con- videos of the underlying faces. Nonetheless, the
trols the consciously generated verbal and nonverbal key challenge to avatar designers is creating faces and
Three views of a digital avatar modeled after a human head and face. This avatar is built by creating a three-
dimensional mesh and wrapping a photographic texture around it. Photo courtesy of James J. Blascovich.
bodies in sufficient detail to allow for the realistic users interact with one another using either a key-
rendering of behavior, which brings us back to be- board or a joystick, typing messages back and
havioral realism. In summary, static avatars currently forth and viewing one anothers avatars as they move
can look quite a bit like their human controllers but around the digital world. Typically, these are avatars
can only perform a small subset of a dynamic hu- in the minimal sense of the word; behavioral and
mans actions in real time. photographic realism is usually quite low. In the case
of online role-playing games, users typically navi-
gate the online world using stock avatars with lim-
Current Use of Avatars ited behavioral capabilities.
Depending on how loosely one defines digital rep-
resentation, the argument can be made that
avatars are quite pervasive in society. For example, Avatar Research
sound is transformed into digital information as it Computer scientists and others have directed much
travels over fiber-optic cables and cellular networks. effort towards developing systems capable of pro-
Consequently, the audio representation we perceive ducing functional and effective avatars. They have
over phone lines is actually an avatar of the speaker. striven to develop graphics, logic, and the tracking
This example may seem trivial at first, but be- capabilities to render actual movements by humans
comes less trivial when preset algorithms are applied on digital avatars with accuracy, and to augment
to the audio stream to cause subtle changes in the those movements by employing control algo-
avatar, for example, to clean and amplify the sig- rithms that supply missing tracking data or infor-
nal. This can only be done effectively because the mation about static visual features.
voice is translated into digital information. Furthermore, behavioral scientists are examin-
More often, however, when people refer to ing how humans interact with one another via
avatars, they are referring to visual representations. avatars. These researchers strive to understand so-
Currently, millions of people employ avatars in cial presence, or copresence, a term referring to the
online role-playing games as well as in chat rooms degree to which individuals respond socially towards
used for virtual conferencing. In these environments, others during interaction among their avatars, com-
AVATARS 67
pared with the degree to which they respond to ac- to have systems operators do this for them) within
tual humans. virtual environments by amplifying or suppressing
The behavioral scientist Jim Blascovich and his communication signals.
colleagues have created a theoretical model for so- TSI algorithms can impact interactants abilities
cial influence within immersive virtual environments to influence interaction partners. For example, sys-
that provides specific predictions for how the inter- tem operators can tailor the nonverbal behaviors
play of avatars photographic and behavioral realism of online teachers lecturing to more than one stu-
will affect peoples sense of the relevance of the avatar- dent simultaneously within an immersive virtual
mediated encounter. They suggest that the inclusion classroom in ways specific to each student inde-
of certain visual features is necessary if the avatar pendently and simultaneously. Student A might
is to perform important, socially relevant behavioral respond well to a teacher who smiles, and Student B
actions. For example, an avatar needs to have rec- might respond well to a teacher with a neutral ex-
ognizable eyebrows in order to lower them in a frown. pression. Via an avatar that is rendered separately for
Other data emphasize the importance of behav- each student, the teacher can be represented si-
ioral realism. In 2001 Jeremy Bailenson and his multaneously by different avatars to different stu-
colleagues demonstrated that making a digital rep- dents, thereby communicating with each student in
resentation more photographically realistic does not the way that is optimal for that student. The psy-
increase its social presence in comparison with an chologist Andrew Beall and his colleagues have used
agent that is more cartoon-like as long as both types avatars to employ such a strategy using eye con-
of agents demonstrate realistic gaze behaviors. In tact; they demonstrated that students paid greater
findings presented in 2003, Maia Garau and her col- attention to the teacher using TSI.
leagues failed to demonstrate an overall advantage However, there are ethical problems associated
for more photographically realistic avatars; more- with TSIs. One can imagine a dismal picture of the fu-
over, these researchers demonstrated that increasing ture of online interaction, one in which nobody is who
the photographic realism of an avatar can actually they seem to be and avatars are distorted so much from
cause a decrease in social presence if behavioral re- the humans they represent that the basis for judging
alism is not also increased. the honesty of the communication underlying social
In sum, though research on avatars currently is interactions is lost. Early research has demonstrated
largely in its infancy, investigators are furthering our that TSIs involving avatars are often difficult to detect.
understanding of computer-mediated human in- It is the challenge to researchers to determine the
teraction. As avatars become more commonplace, best way to manage this issue as the use of avatars
research geared towards understanding these appli- becomes more prevalent.
cations should increase.
State of the Art

Ethical Issues Currently, there are many examples of humans in-
Interacting via avatars allows for deceptive interac- teracting with one another via avatars. For the most
tions. In 2003 Bailenson and his colleagues intro- part, these avatars are simplistic and behaviorally
duced the notion of transformed social interactions and photographically unrealistic. The exception oc-
(TSIs). Using an avatar to interact with another per- curs in research laboratories, in which scientists
son is qualitatively different from other forms of are beginning to develop and test avatars that are
communication, including face-to-face interaction, similar in appearance and behavior to their human
standard telephone conversations, and videocon- counterpart. As avatars become more ubiquitous,
ferencing. An avatar that is constantly rerendered in it is possible that we may see qualitative changes in
real time makes it possible for interactants to sys- social interaction due to the decoupling and trans-
tematically filter their appearance and behaviors (or formation of behavior from human to avatar. While
there are ethical dangers in transforming behav- Blascovich, J. (2001). Social influences within immersive virtual
iors as they pass from physical actions to digital rep- environments. In R. Schroeder (Ed.), The social life of avatars. Berlin,
Germany: Springer-Verlag.
resentations, there are also positive opportunities Blascovich, J., Loomis, J., Beall, A. C., Swinth, K. R., Hoyt, C. L., &
both for users of online systems and for research- Bailenson, J. N. (2001). Immersive virtual environment technol-
ers in human-computer interaction. ogy as a methodological tool for social psychology. Psychological
Inquiry, 13, 146149.
Brunner, J. (1975). Shockwaver rider. New York: Ballantine Books.
Jeremy N. Bailenson and James J. Blascovich Cassell, J., & Vilhjlmsson, H. (1999). Fully embodied conversa-
tional avatars: Making communicative behaviors autonomous.
See also Animation; Telepresence; Virtual Reality Autonomous Agents and Multi-Agent Systems, 2(1), 4564.
Garau, M., Slater, M.,Vinayagamoorhty,V., Brogni, A., Steed, A., & Sasse,
M. A. (2003). The impact of avatar realism and eye gaze control on
perceived quality of communication in a shared immersive virtual
environment. In Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems (pp. 529536). New York: ACM Press.
FURTHER READING Gibson, W. (1984). Neuromancer. New York: Ace Books.
Morningstar, C., & Farmer, F.R. (1991). The lessons of Lucasfilms
Badler, N., Phillips, C., & Webber, B. (1993). Simulating humans: habitat. In M. Benedikt (Ed.), Cyberspace: First steps. Cambridge,
Computer graphics, animation, and control. Oxford, UK: Oxford MA: MIT Press.
University Press. Slater, M., Howell, J., Steed, A., Pertaub, D., Garau, M., & Springel,
Bailenson, J. N., Beall, A. C., Blascovich, J., & Rex, C. (in press). S. (2000). Acting in virtual reality. ACM Collaborative Virtual
Examining virtual busts: Are photogrammetrically generated head Environments, CVE2000, 103110.
models effective for person identification? PRESENCE: Teleoperators Slater, M., Sadagic, A., Usoh, M., & Schroeder, R. (2000). Small group
and Virtual Environments. behaviour in a virtual and real environment: A comparative study.
Bailenson, J. N., Beall, A. C., Loomis, J., Blascovich, J., & Turk, M. (in PRESENCE: Teleoperators and Virtual Environments, 9, 3751.
press). Transformed social interaction: Decoupling representation Stephenson, N. (1993). Snow crash. New York: Bantam Books.
from behavior and form in collaborative virtual environments. Thalmann, M. N, & Thalmann D. (Eds). (1999). Computer Animation
PRESENCE: Teleoperators and Virtual Environments. and Simulation 99. Vienna, Austria: Springer-Verlag.
Bailenson, J. N., Blascovich, J., Beall, A. C., & Loomis, J. M. (2001). Turk, M., & Kolsch, M. (in press). Perceptual Interfaces. In G. Medioni
Equilibrium revisited: Mutual gaze and personal space in virtual & S. B. Kang (Eds.), Emerging topics in computer vision. Upper
environments. PRESENCE: Teleoperators and Virtual Environments, Saddle River, NJ: Prentice-Hall.
10, 583598. Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
Beall, A. C., Bailenson, J. N., Loomis, J., Blascovich, J., & Rex, C. (2003). New York: Simon & Schuster.
Non-zero-sum mutual gaze in immersive virtual environments. Yee, N. (2002). Befriending ogres and wood elves: Understanding rela-
In Proceedings of HCI International 2003 (pp. 11081112). New tionship formation in MMORPGs. Retrieved January 16, 2004, from
York: ACM Press. http://www.nickyee.com/hub/relationships/home.html
BETA TESTING
BRAILLE
BRAIN-COMPUTER INTERFACES
BROWSERS
will be discussed below, beta tests make it possible for

BETA TESTING endusers to contribute to the design and develop-
ment of a product and may represent a shift in the
Beta testing, a stage in the design and development organization of the production process.
process of computer software and hardware, uses
people outside a company, called beta testers, to be
sure that products function properly for typical end- Denitions of Beta Testing
users outside the firm. Does a piece of software work A beta test is an early (preshipping or prelaunch), un-
under normal operating conditions? Can users nav- official release of hardware or software that has already
igate important features? Are there any critical been tested within the company for major flaws. In
programming flaws? These are the questions beta theory, beta versions are very close to the final prod-
tests answer. uct, but in practice beta testing is often simply one way
The widespread use of beta tests warrants the ex- for a firm to get users to try new software under real
amination of the process. Because the trade literature conditions. Beta tests expose software and hardware
in computer programming focuses on the mechan- to real-world configurations of computing platforms,
ics of designing, conducting, and interpreting beta operating systems, hardware, and users. For exam-
tests, less has been written on the social implications ple, a beta test of a website is the time period just
of the growing use of beta testing. For example, as before a sites official launch when a fully operational
69
product is used under normal operating conditions to ogists Gina Neff and David Stark, establishing a
identify any programming bugs or interface issues cycle of testing, feedback, and innovation that facil-
(Grossnickle and Raskin 2001, 351). David Hilbert de- itates negotiations about what is made can make it
scribes beta testing as a popular technique for evalu- possible to incorporate broader participation into
ating the fit between application design and use. the design of products and organizations.
The term beta testing emerged from the practice However, in practice, beta tests may be poorly
of testing the unit, module, or components of a sys- designed to incorporate user feedback. Advice in the
tem first. This test was called alpha, whereas beta re- trade literature suggests that beta tests may not be
ferred to the initial test of the complete system. Alpha constructed to provide more than bug squashing
and beta, derived from earlier nomenclature of hard- and usability testing (Grossnickle and Raskin
ware testing, were reportedly first used in the n.d., 1). Beta tests also present firms with a chance
1960s at IBM. Now alpha typically refers to tests con- to conduct research on their users and on how their
ducted within the firm and beta refers to tests con- products are used. Ideally, beta testers are statistically
ducted externally. representative of typical product users. However, em-
There is ample evidence that beta testing has in- pirical research suggests that beta testers may not ac-
creased in various forms over the last decade. James curately reflect end-users because testers tend to have
Daly, a technology business reporter and founder of more technical training and hold more technical jobs
the magazine Business 2.0, reports that by 1994, 50 than typical office workers.
percent of Fortune 1000 companies in the United States
had participated in beta testing and 20 percent of those
companies had used beta testing widely. However, the Critical Views of Beta Testing
implementationand the purposesof beta testing The shift from total quality management to a test-
vary by company. An online market-research hand- ing-driven model of development means that the
book suggests that for most ventures, standard generation and detection of error plays a renewed
beta-testing technique involves e-mailing friends, fam- and desired role in the production cycle (Cole 2002,
ily, and colleagues with the URL of a new site 1052). With the rise of the acceptance of beta ver-
(Grossnickle and Raskin 2001, 351), which clearly sions, companies and users alike may be more will-
would not produce a statistically representative sam- ing to tolerate flaws in widely circulated products,
ple of end users. A meta study of beta-test evaluations and end-users (including beta testers) may bear an
done more than a decade ago found that most beta increased burden for the number of errors that com-
testing was actually driven by convenience or tradi- panies allow in these products. Some criticism has
tion rather than recognition of the costs and benefits emerged that companies are releasing products for
involved (Dolan and Matthews 1993, 318). beta testing that are clearly not ready for the mar-
In addition to determining whether or not a ket and are exploiting free labor by using beta testers
product works, a beta test can be used to increase a as unpaid consultants to find the bugs in their prod-
firms knowledge about the user base for its products (Garman 1996, 6).
ucts, to support its marketing and sales goals, and to Users may also be frustrated by the continually
improve product support. More importantly, beta updated products that beta testing can enable. The
testers suggestions may be incorporated into the de- distribution of software in non-shrink-wrapped ver-
sign of the product or used to develop subsequent sions means that products are not clean end-versions
generations of the product. b u t d e s t a b i l i ze d a n d con s t a n t l y ch a n g i n g .
Technological advances in distribution, such as
User Participation in online distribution of software products, makes it
possible to distribute products that are continually
Product Development updateable and almost infinitely customizable
Beta testing allows users to become involved in the products that, in effect, never leave a type of beta
product-development process. According to sociol- phase (Neff and Stark 2003, 177).
BETA TESTING 71
Benets to Beta Testers tion of the product (Dolan and Matthews 1993, 20),
Because they are willing to risk bugs that could po- beta tests present crucial opportunities to incorpo-
tentially crash their computers, beta testers accrue rate user suggestions into the design of a product.
benefits such as getting a chance to look at new
features and products before other users and con- Gina Neff
tributing to a product by detecting software bugs or
minor flaws in programming. More than 2 million See also Prototyping; User-Centered Design
people volunteered to be one of the twenty thousand
beta testers for a new version of Napster. There is also
an increase of beta retail productsearly and of- FURTHER READING
ten cheaper versions of software that are more ad-
Cole, R. E. (2002). From continuous improvement to continuous
vanced than a traditional beta version but not yet a innovation. Total Quality Management, 13(8), 10511056.
fully viable commercial release. Although Apples Daly, J. (1994, December). For beta or worse. Forbes ASAP, 3640.
public beta release of OS X, its first completely Dolan, R. J., & Matthews, J. M. (1993). Maximizing the utility of con-
new operating system since 1984, cost $29.95, thou- sumer product testing: Beta test design and management. Journal
of Product Innovation Management, 10, 318330.
sands downloaded it despite reports that it still Hove, D. (Ed.). The Free online dictionary of computing. Retrieved
had many bugs and little compatible software was March 10, 2004, from http://www.foldoc.org
available. These beta users saw the long-awaited new Garman, N. (1996). Caught in the middle: Online professionals and
operating system six months before its first com- beta testing. Online, 20(1), 6.
Garud, R., Sanjay, J., & Phelps, C. (n.d.). Unpacking Internet time in-
mercial release, and Apple fans and the press pro- novation. Unpublished manuscript, New York University, New
vided invaluable buzz about OS X as they tested it. York.
Many scholars suggest that the Internet has com- Grossnickle, J., & Raskin, O. (2001). Handbook of online marketing
research. New York: McGraw Hill.
pressed the product-development cycles, especially Grossnickle, J., & Raskin, O. (n.d.). Supercharged beta test. Webmonkey:
in software, often to the extent that one generation Design. Retrieved January 8, 2004, from http://hotwired.lycos.com/
of product software is hard to distinguish from the webmonkey
next. Netscape, for example, released thirty-nine dis- Hilbert, D. M. (1999). Large-scale collection of application usage data
and user feedback to inform interactive software development.
tinct versions between the beta stage of Navigator Unpublished doctoral dissertation, University of California, Irvine.
1.0 and the release of Communicator 4.0. Kogut, B., & Metiu, A. (2001). Open source software development and
distributed innovation. Oxford Review of Economic Policy, 17(2),
248264.
Krull, R. (2000). Is more beta better? Proceedings of the IEEE Professional
Future Developments Communication Society, 301308.
Production is an increasingly dense and differenti- Metiu, A., & Kogut, B. (2001). Distributed knowledege and the global
ated layering of people, activities and things, each organization of software development. Unpublished manu-
script, Wharton School of Business, University of Pennsylvania,
operating within a limited sphere of knowing and Philadelphia.
acting that includes variously crude or sophisticated Neff, G., & Stark, D. (2003). Permanently beta: Responsive organi-
conceptualizations of the other (Suchman 2003, zation in the Internet era. In P. Howard and S. Jones (Eds.), Society
62). Given this complexity, beta testing has been wel- Online. Thousand Oaks, CA: Sage.
O'Mahony, S. (2002). The Emergence of a new commercial actor: Com-
comed as a way in which people who create prod- munity managed software projects. Unpublished doctoral disser-
ucts can inter act w ith those who use them. tation, Stanford University, Palo Alto, CA. Retrieved on January 8,
Internet communication facilitates this communi- 2004, from http://opensource.mit.edu/
cation, making the distribution of products in ear- Raymond, E. (1999). The Cathedral and the bazaar: Musings on Linux
and open source from an accidental revolutionary. Sebastapol, CA:
lier stages of the product cycle both easier and O'Reilly and Associates.
cheaper; it also facilitates the incorporation of user Ross, R. (2002). Born-again Napster takes baby steps. Toronto Star,
feedback into the design process. E04.
Suchman, L. (2002). Located accountabilities in technology production.
While it is true that most design-change ideas Retrieved on January 8, 2004, from http://www.comp.lancs.ac.
surfaced by a beta test are passed onto product de- uk/sociology/soc039ls.html. Centre for Science Studies, Lancaster
velopment for incorporation into the next genera- University.
Techweb (n.d.). Beta testing. Retrieved on January 8, 2004, from and too expensive to give its users unlimited and
http://www.techweb.com/encyclopedia quick access to an increasing amount of printed ma-
Terranova, T. (2000). Free labor: Producing culture for the digital econ-
omy. Social Text 18(2), 3358.
terial: books, newspapers, leaflets, and so forth.
The invention of the transistor in 1947 by three U.S.
physicists and of integrated circuits in the late 1960s
provided the solution: electromechanical tactile dis-
BRAILLE plays. After many attempts, documented by numer-
ous patents, electronic Braille was developed
Access to printed information was denied to blind simultaneously during the early 1970s by Klaus-Peter
people until the late 1700s, when Valentin Hay, hav- Schnherr in Germany and Oleg Tretiakoff in France.
ing funded an institution for blind children in Paris,
embossed letters in relief on paper so that his pupils
could read them. Thus, two hundred and fifty First Electronic Braille Devices
years after the invention of the printing press by the In electronic Braille, Braille codesand therefore
German inventor Johannes Gutenberg, blind people Braille booksare stored in numerical binary for-
were able to read but not to write. mat on standard mass storage media: magnetic tapes,
magnetic disks, and so forth. In this format the bulk
and cost of Braille books are reduced by several orders
Historical Background of magnitude. To be accessible to blind users, elec-
In 1819 a French army officer, Charles Barbier, in- tronically stored Braille codes must be converted into
vented a tactile reading system, using twelve-dot raised-dot patterns by a device called an electro-
codes embossed on paper, intended for nighttime mechanical Braille display. An electromechanical
military communications. Louis Braille, who had Braille display is a flat reading surface that has holes
just entered the school for the blind in Paris, learned arranged in a Braille cell pattern. The hemispherical
of the invention and five years later, at age fifteen, tip of a cylindrical pin can either be raised above the
developed a much easier-to-read six-dot code, reading surface to show a Braille dot or lowered under
providing sixty-three dot patterns. Thanks to his in- the reading surface to hide the corresponding Braille
vention, blind people could not only read much dot. The Braille dot vertical motion must be con-
faster, but also write by using the slate, a simple hand trolled by some kind of electromechanical actuator.
tool made of two metal plates hinged together be- Two such displays were almost simultaneously put
tween which a sheet of paper could be inserted onto the market during the mid-1970s.
and embossed through cell-size windows cut in The Schnherr Braille calculator had eight Braille
the front plate. Six pits were cut in the bottom cells of six dots each, driven by electromagnetic ac-
plate to guide a hand-held embossing stylus inside tuators and a typical calculator keyboard. The dot
each window. spacing had to be increased to about 3 millimeters
In spite of its immediate acceptance by his fel- instead of the standard 2.5 millimeters to provide
low students, Brailles idea was officially accepted enough space for the actuators.
only thirty years later, two years after his death in The Tretiakoff Braille notebook carried twelve
1852. Eighty more years passed before English-speak- Braille standard cells of six dots each, driven by piezo-
ing countries adapted the Braille system in 1932, and electric (relating to electricity or electric polarity due
more than thirty years passed before development to pressure, especially in a crystalline substance) reeds,
of the Nemeth code, a Braille system of scientific no- a keyboard especially designed for blind users, a cas-
tation, in 1965. Braille notation was also adopted by sette tape digital recorder for Braille codes storage,
an increasing number of countries. and a communication port to transfer data between
In spite of its immense benefits for blind people, the Braille notebook and other electronic devices. Both
the Braille system embossed on paper was too bulky devices were portable and operated on replaceable or
BRAILLE 73
Enhancing Access to Braille Instructional Materials
(ANS)Most blind and visually impaired children at- have misunderstood each other's business, he said, which
tend regular school classes these days, but they are often led to frustration on both sides.
left waiting for Braille and large-print versions of class Most blind children are mainstreamed into public
texts to arrive while the other children already have the school classrooms and receive additional help from a cadre
books. of special teachers of the blind. Technology is also giv-
There are 93,000 students in kindergarten through ing blind students more options. Scanning devices now
12th grade who are blind or have limited vision. Because download texts into Braille and read text aloud. Closed
this group represents a small minority of all school- circuit television systems can enlarge materials for low-
children, little attention has been paid to updating the vision students.
cumbersome process of translating books into Braille, ad- These kids have very individual problems, noted
vocates said. Kris Kiley, the mother of a 15-year-old who has limited
Traditionally, publishers have given electronic copies vision.It's not one size fits all. But if you don't teach them
of their books to transcribers, who often need to com- to read you've lost part of their potential.
pletely reformat them for Braille. Lack of a single techno- New tools also bring with them new problems. For
logical standard and little communication between example, the new multimedia texts, which are available
publishing houses and transcribers led to delays in blind to students on CD-ROM, are completely inaccessible to
students receiving instructional materials, experts said. blind students. And because graphics now dominate many
The solution, said Mary Ann Siller, a national pro- books, lots of information, especially in math, does not
gram associate for the American Foundation for the Blind reach those with limited vision.
who heads its Textbook and Instructional Materials Simply recognizing the challenges faced by the
Solutions Forum, is to create a single electronic file for- blind would go a long way toward solving the problem,
mat and a national repository for textbooks that would said Cara Yates. Yates, who recently graduated from law
simplify and shorten the production process. And that's school, lost her sight at age 5 to eye cancer. She recalls one
exactly what is happening. of her college professors who organized a series of tutors
In October, the American Printing House for the Blind to help her see star charts when she took astrophysics.
in Louisville, Ky., took the first step in creating a reposi- A lot of it isn't that hard, she said.It just takes some
tory by listing 140,000 of its own titles on the Internet. thought and prior planning. The biggest problem for the
The group is now working to get publishers to deposit blind is they can't get enough information. There's no ex-
their text files, which transcribers could readily access. cuse for it. It's all available.
Everyone is excited about it, said Christine Anderson, Siller said the foundation also hoped to raise aware-
director of resource services for the Kentucky organiza- ness about educational assessment; the importance of
tion. By having a central database with information about parental participation; better preparation for teachers; a
the files for all books available in Braille, large print, sound core curriculum for blind students in addition to the
recording or computer files, costly duplications can be sighted curriculum; and better Braille skills and a reduced
eliminated, she said. caseload for teachers who often travel long distances to
Pearce McNulty, director of publishing technology at assist their students.
Houghton Mifflin Co. in Boston, which is a partner in the Mieke H. Bomann
campaign, said he is hopeful the repository will help solve Source: Campaign seeks to end blind students' wait for Braille
textbooks. American News Service, December 16, 1999.
the problem. Publishers and Braille producers historically
rechargeable batteries. The Tretiakoff Braille note- cells used in Tretiakoff 's extremely portable Braille
book, called Digicassette, measured about 20 by notebook, the P-Touch. In these vertical cells each
25 by 5 centimeters. piezoelectric actuator was located underneath the
A read-only version of the Digicassette was man- corresponding tactile dot, allowing tactile dots to be
ufactured for the U.S. National Library Services for arranged in arrays of regularly spaced rows and
the Blind of the Library of Congress. columns for the electronic display of graphics. These
vertical cells were about twice as high as conventional
horizontal cells and no less expensive. Multiline or
Personal Braille Printers graphic displays were thus made technically feasible
Braille books consist of strong paper pages embossed but remained practically unaffordable at about $12
with a Braille dot pattern by high-speed machines per dot for the end user as early as 1985.
and then bound together much like ordinary books.
A typical Braille page can carry up to twenty-five lines
of forty Braille characters each and can be explored Active versus Passive Reading
rapidly from left to right and from top to bottom by Since Louis Braille, blind people have performed tac-
a blind reader. Electronic Braille displays consist gentile reading by moving the tip of one to three fingers
erally of a single line comprising usually from eight- across a Braille page or along a Braille line while ap-
een to forty Braille characters to keep the displays plying a small vertical pressure on the dot pattern in
portable and affordable for individual users. The shift a direction and at a speed fully controlled by the
from a full page to a single line delayed the accept- reader, hence the name active reading.
ance of Braille displays in spite of their ability to pro- Louis Braille used his judgment to choose tactile
vide easy and high-speed access to electronically dot height and spacing; research performed during
stored information. the last thirty years has shown that his choices
Personal Braille printers, also made possible by were right on the mark. Objective experiments, in
the development of integrated circuits, appeared which the electrical response of finger mechanore-
soon after the first personal computers to fill the gap ceptors (neural end organs that respond to a me-
between industrially produced Braille books and sin- chanical stimulus, such as a change in pressure) is
gle-line Braille displays. Similar in concept to dot- measured from an afferent (conveying impulses
matrix ink printers, personal Braille printers allowed toward the central nervous system) nerve fiber, have
a blind user to emboss on a sheet of strong paper a shown that strokingthe horizontal motion of
few lines of Braille characters per minute from Braille the fingerplays an essential role in touch resolu-
codes received from an external source. tion, the ability to recognize closely spaced dots.
Conversely, if a blind reader keeps the tip of one
or more fingers still on an array of tactile dots that
Tactile Graphics is moved in various patterns up or down under the
Although the first personal Braille printers were de- fingertips, this is called passive reading. Passive read-
signed to print only a regularly spaced Braille pat- ing has been suggested as a way to reduce the num-
ternat .6 centimeter spacing between characters ber of dots, and therefore the cost of tactile displays,
some were outfitted with print heads capable of by simulating the motion of a finger across a wide ar-
printing regularly spaced dots, in both the hori- ray of dots by proper control of vertical dot motion
zontal and the vertical directions, allowing the under a still finger. The best-known example of this
production of embossed tactile graphics. approach is the Optacon (Optical to Tactile Converter),
Although the first electronic Braille displays were invented during the mid-1970s by John Linvill to give
built with horizontally stacked piezoelectric reeds, blind people immediate and direct access to printed
whose lengthabout 5 centimetersprevented the material. The Optacon generated a vibrating tactile
juxtaposition of more than two Braille lines, the mid- image of a small area of an object viewed by its
1980s brought the first vertical piezoelectric Braille camera placed and moved against its surface.
BRAIN-COMPUTER INTERFACES 75
Research has shown that touch resolution and

reading speed are significantly impaired by passive BRAIN-COMPUTER
reading, both for raised ordinary character shapes
and for raised-dot patterns. INTERFACES
A brain-computer interface (BCI), also known as a
Current and Future direct brain interface (DBI) or a brain-machine in-
terface (BMI), is a system that provides a means
Electronic Tactile Displays for people to control computers and other devices
At the beginning of the twenty-first century, sev- directly with brain signals. BCIs fall into the cate-
eral companies make electromechanical tactile cells, gory of biometric devices, which are devices that de-
which convert electrical energy into mechanical en- tect and measure biological properties as their basis
ergy and vice versa, but the dominant actuator tech- of operation. Research on brain-computer interfaces
nology is still the piezoelectric (relating to electricity spans many disciplines, including computer science,
or electric polarity due to pressure, especially in a neuroscience, psychology, and engineering. BCIs
crystalline substance) bimorph reed, which keeps were originally conceived in the 1960s, and since the
the price per tactile dot high and the displays bulky late 1970s have been studied as a means of provid-
and heavy. The majority of electronic tactile dis- ing a communication channel for people with very
plays are single-line, stand-alone displays carry- severe physical disabilities. While assistive technol-
ing up to eighty characters or Braille computers ogy is still the major impetus for BCI research, there
carrying from eighteen to forty characters on a sin- is considerable interest in mainstream applications
gle line. Their costs range from $3,000 to more than as well, to provide a hands-free control channel that
$10,000. A small number of graphic tactile mod- does not rely on muscle movement.
ules carrying up to sixteen by sixteen tactile dots Despite characterizations in popular fiction, BCI
are also available from manufacturers such as KGS systems are not able to directly interpret thoughts or
in Japan. perform mind reading. Instead, BCI systems mon-
Several research-and-development projects, us- itor and measure specific aspects of a users brain
ing new actuator technologies and designs, are un- signals, looking for small but detectable differences
der way to develop low-cost g raphic tactile that signal the intent of the user. Most existing BCI
displays that could replace or complement visual dis- systems depend on a person learning to control an
plays in highly portable electronic communication aspect of brain signals that can be detected and meas-
devices and computers. ured. Other BCI systems perform control tasks, such
as selecting letters from an alphabet, by detecting
Oleg Tretiakoff brain-signal reactions to external stimuli.
Although BCIs can provide a communications
See also Sonification; Universal Access channel, the information transmission rate is low
compared with other methods of control, such as
keyboard or mouse. The best reported user per-
FURTHER READING formance with current BCI systems is an informa-
tion transfer rate of sixty-eight bits per minute, which
American Council of the Blind. (2001). Braille: History and Use of roughly translates to selecting eight characters per
Braille. Retrieved May 10, 2004, from http://www.acb.org/re-
sources/braille.html
minute from an alphabet. BCI studies to date have
Blindness Resource Center. (2002). Braille on the Internet. been conducted largely in controlled laboratory set-
Retrieved May 10, 2004, from http://www.nyise.org/braille.html tings, although the field is beginning to target real-
world environments for BCI use.
A driving motivation behind BCI research has been
the desire to help people with severe physical disabil-
ities such as locked-in syndrome, a condition caused Sensorimotor Cortex Rhythms

by disease, stroke, or injury in which a person remains Cortical rhythms represent the synchronized activ-
cognitively intact but is completely paralyzed and unity of large numbers of brain cells in the cortex
able to speak. Traditional assistive technologies for that create waves of electrical activity over the brain.
computer access depend on small muscle movements, These rhythms are characterized by their frequency
typically using the limbs, eyes, mouth, or tongue to ac- of occurrence; for example, a rhythm occurring
tivate switches. People with locked-in syndrome between eight and twelve times a second is denoted as
have such severely limited mobility that system in- mu, and one occurring between eighteen and twenty-
put through physical movement is infeasible or un- six times a second is referred to as beta. When
reliable. A BCI system detects tiny electrophysiological recorded over the motor cortex, these rhythms are
changes in brain signals to produce control instruc- affected by movement or intent to move. Studies have
tions for a computer, thereby making it unnecessary shown that people can learn via operant-condition-
for a user to have reliable muscle movement. ing methods to increase and decrease the voltage
Researchers have created applications for non- of these cortical rhythms (in tens of microvolts) to
disabled users as well, including gaming systems and control a computer or other device. BCIs based on
systems that allow hands-free, heads-up control of processing sensorimotor rhythms have been used to
devices, including landing an aircraft. Brain signal operate a binary spelling program and two-dimen-
interfaces have been used in psychotherapy to mon- sional cursor movement.
itor relaxation responses and to teach meditation,
although these are biofeedback rather than control
interfaces. Slow Cortical Potentials
Slow cortical potentials (SCPs) are low-frequency
shifts of cortical voltage that people can learn to con-
Brain Signal Characteristics trol with practice. SCP shifts can occur in durations
Brain signals are recorded using two general ap- of a hundred milliseconds up to several seconds. SCP
proaches. The most ubiquitous approach is the elec- signals are based over the frontal and central cor-
troencephalogram (EEG), a recording of signals tex area, and are typically influenced by emotional
representing activity over the entire surface of the or mental imagery, as well as imagined movement.
brain or a large region of the brain, often incorpo- SCPs are recorded both from electrodes on the scalp
rating the activity of millions of neurons. An EEG using an operant conditioning approach and from
can be recorded noninvasively (without surgery) positive reinforcement to train users to alter their
from electrodes placed on the scalp, or invasively (re- SCPs. Both nondisabled and locked-in subjects have
quiring surgery) from electrodes implanted inside been able to learn to affect their SCP amplitude, shift-
the skull or on the surface of the brain. Brain signals ing it in either an electrically positive or negative di-
can also be recorded from tiny electrodes placed direction. Locked-in subjects have used SCPs to
rectly inside the brain cortex, allowing researchers communicate, operating a spelling program to write
to obtain signals from individual neurons or small letters and even surfing with a simple web browser.
numbers of colocated neurons.
Several categories of brain signals have been ex-
plored for BCIs, including rhythms from the senso- Evoked Potentials
rimotor cortex, slow cortical potentials, evoked The brain's responses to stimuli can also be detected
potentials, and action potentials of single neurons. and used for BCI control. The P300 response occurs
A BCI system achieves control by detecting changes when a subject is presented with something famil-
in the voltage of a brain signal, the frequency of a iar, such as a photo of a loved one, or of interest, such
signal, and responses to stimuli. The type of brain as a letter selected from an alphabet. The P300 re-
signal processed has implications for the nature of sponse can be evoked by almost any stimulus, but
the users interaction with the system. most BCI systems employ either visual or auditory
stimuli. Screening for the P300 is accomplished by learning to raise or lower some aspect of his or
through an oddball paradigm, where the subject her brain signals, usually amplitude or frequency.
views a series of images or hears a series of tones, at- Continuous transducers have enabled users to per-
tending to the one that is different from the rest. If form selections by raising or lowering a cursor to hit
there is a spike in the signal power over the parietal a target on a screen. A continuous transducer is anal-
region of the brain approximately 300 millisec- ogous to a continuous device, such as a mouse or joy-
onds after the oddball or different stimulus, then stick, that always reports its current position.
the subject has a good P300 response. One practical A discrete transducer is analogous to a switch de-
application that has been demonstrated with P300 vice that sends a signal when activated. Discrete trans-
control is a spelling device. The device works by flash- ducers produce a single value upon activation. A user
ing rows and columns of an alphabet grid and aver- typically activates a discrete transducer by learning
aging the P300 responses to determine which letter to cause an event in the brain that can be detected
the subject is focusing on. P300 responses have by a BCI system. Discrete transducers have been used
also been used to enable a subject to interact with to make decisions, such as whether to turn in navi-
a virtual world by concentrating on flashing vir- gating a maze. Continuous transducers can emulate
tual objects until the desired one is activated. discrete transducers by introducing a threshold that
the user must cross to activate the switch.
Direct-spatial-positioning transducers produce a
Action Potentials of Single Neurons direct selection out of a range of selection choices.
Another approach to BCI control is to record from These transducers are typically associated with evoked
individual neural cells via an implanted electrode. responses, such as P300, that occur naturally and do
In one study, a tiny hollow glass electrode was im- not have to be learned. Direct transducers have been
planted in the motor cortices of three locked-in sub- used to implement spelling, by flashing letters arranged
jects, enabling neural firings to be captured and in a grid repeatedly and averaging the brain signal re-
recorded. Subjects attempted to control this form of sponse in order to determine which letter the user was
BCI by increasing or decreasing the frequency of neu- focusing on. A direct spatial positioning transducer is
ral firings, typically by imagining motions of para- analogous to a touch screen.
lyzed limbs. This BCI was tested for controlling BCI system architectures have many common
two-dimensional cursor movement in communica- functional aspects. Figure 1 shows a simplified model
tions programs such as virtual keyboards. Other ap- of a general BCI system design as described by Mason
proaches utilizing electrode arrays or bundles of and Birch (2003).
microwires are being researched in animal studies. Brain signals are captured from the user by an
acquisition method, such as scalp electrodes or im-
planted electrodes. The signals are then processed by
Interaction Styles With BCIs an acquisition component called a feature extractor
How best to map signals from the brain to the that identifies signal changes that could signify in-
control systems of devices is a relatively new area tent. A signal translator then maps the extracted sig-
of study. A BCI transducer is a system component nals to device controls, which in turn send signals to
that takes a brain signal as input and outputs a con- a control interface for a device, such as a cursor, a
trol signal. BCI transducers fall into three general television, or a wheelchair. A display may return feed-
categories: continuous, discrete, and direct spatial back information to the user.
positioning. Feedback is traditionally provided to BCI users
Continuous transducers produce a stream of through both auditory and visual cues, but some test-
values within a specified range. These values can be ing methods allow for haptic (touch) feedback and
mapped to cursor position on a screen, or they can di- electrical stimulation. Which feedback mechanisms
rectly change the size or shape of an object (such as a are most effective usually depends on the abilities
progress bar). A user activates a continuous transducer and disabilities of the user; many severely disabled
FIGURE 1. BCI system architecture
users have problems with vision that can be com- Neural Prosthetics
pensated for by adding auditory cues to BCI tasks. A BCI application with significant implications is
Some research teams have embraced usability test- neural prostheses, which are orthoses or muscle stim-
ing to determine what forms of feedback are most ulators controlled by brain signals. In effect, a neu-
effective; this research is under way. ral prosthesis could reconnect the brain to paralyzed
limbs, essentially creating an artificial nervous sys-
tem. BCI controls could be used to stimulate mus-
Applications for BCIs cles in paralyzed arms and legs to enable a subject to
As the BCI field matures, considerable interest has learn to move them again. Preliminary work on a
arisen in applying BCI techniques to real-world prob- neurally controlled virtual hand was reported in 2000
lems. The principal goal has been to provide a com- with implanted electrodes: a noninvasive BCU has
munication channel for people with severe motor been demonstrated to control a hand-grasp ortho-
disabilities, but other applications may also be sis for a person whose hand was paralyzed. An
possible. Researchers are focusing on applications SSVEP-based BCI has also been used to control a
for BCI technologies in several critical areas: functional electrical stimulator to activate paralyzed
muscles for knee extension.
Communication
Making communication possible for a locked-in per- Mobility
son is a critical and very difficult task. Much of the Restoring mobility to people with severe disabilities
work in BCI technology centers around communi- is another area of research. A neurally controlled
cation, generally in the form of virtual keyboards or wheelchair could provide a degree of freedom and
iconic selection systems. greatly improve the quality of life for locked-in peo-
ple. Researchers are exploring virtual navigation tasks,
Environmental Control such as virtual driving and a virtual apartment, as
The ability to control the physical environment is well as maze navigation. A noninvasive BCI was used
also an important quality-of-life issue. Devices to direct a remote-control vehicle, with the aim of
that permit environmental control make it possi- eventually transferring driving skills to a power
ble for locked-in people to turn a TV to a desired wheelchair.
channel and to turn lights on and off, as well as con-
trolling other physical objects in their world.
Issues and Challenges for BCI
Internet Access There are many obstacles to overcome before BCIs
The Internet has the potential to enhance the lives can be used in real-world scenarios. The minute elec-
of locked-in people significantly. Access to the trophysiological changes that characterize BCI con-
Internet can provide shopping, entertainment, edu- trols are subject to interference from both electrical
cation, and sometimes even employment opportu- and cognitive sources. Brain-signal complexity and
nities to people with severe disabilities. Efforts are variability make detecting and interpreting changes
under way to develop paradigms for BCI interaction very difficult except under controlled circumstances.
with Web browsers. Especially with severely disabled users, the effects of
medications, blood sugar levels, and stimulants such

as caffeine can all be significant. Cognitive distrac- FURTHER READING
tions such as ambient environmental noise can af-
fect a persons ability to control a BCI in addition to Bayliss, J. D., & Ballard, D. H. (2000). Recognizing evoked potentials
in a virtual environment. Advances in Neural Information Processing
increasing the cognitive load the person bears. Systems, 12, 39.
Artifacts such as eye blinks or other muscle move- Birbaumer, N., Kubler, A., Ghanayim, N., Hinterberger, T., Perelmouter,
ments can mask control signals. J. Kaiser, J., et al. (2000). The thought translation device (TTD) for
BCIs and other biometric devices are also plagued completely paralyzed patients. IEEE Transactions on Rehabilitation
Engineering, 8(2), 190193.
by what is termed the Midas touch problem: How Birch, G. E., & Mason, S. G. (2000). Brain-computer interface research
does the user signal intent to control when the brain at the Neil Squire Foundation. IEEE Transactions on Rehabilitation
is active constantly? Hybrid discrete/continuous Engineering, 8(2), 193195.
Chapin, J., & Nicolelis, M. (2002). Closed-loop brain-machine in-
transducers may be the answer to this problem, terfaces. In J. R. Wolpaw & T. Vaughan (Eds.), Proceedings of Brain-
but it is still a major issue for BCIs in the real world. Computer Interfaces for Communication and Control: Vol. 2. Moving
Another important issue currently is that BCI Beyond Demonstration, Program and Abstracts (p. 38). Rensse-
systems require expert assistance to operate. As BCI laerville, NY.
Donchin, E., Spencer, K., & Wijesinghe, R. (2000). The mental
systems mature, the expectation is that more of the prosthesis: Assessing the speed of a P300-based brain-computer
precise tuning and calibration of these systems interface. IEEE Transactions on Rehabilitation Engineering, 8(2),
may be performed automatically. 174179.
Although BCIs have been studied since the mid- Kandel, E., Schwartz, J., & Jessell, T. (2000). Principles of neural sci-
ence (4th ed.). New York: McGraw-Hill Health Professions Division.
1980s, researchers are just beginning to explore their Kennedy, P. R., Bakay, R. A. E., Moore, M. M., Adams, K., & Goldwaithe,
enormous potential. Understanding brain signals and J. (2000). Direct control of a computer from the human central
patterns is a difficult task, but only through such an nervous system. IEEE Transactions on Rehabilitation Engineering,
8(2), 198202.
understanding will BCIs become feasible. Currently Lauer, R. T., Peckham, P. H., Kilgore, K. L., & Heetderks, W. J.
there is a lively debate on the best approach to acquir- (2000). Applications of cortical signals to neuroprosthetic control:
ing brain signals. Invasive techniques, such as im- A critical review. IEEE Transactions on Rehabilitation Engineering,
planted electrodes, could provide better control 8(2), 205207.
Levine, S. P., Huggins, J. E., BeMent, S. L., Kushwaha, R. K., Schuh,
through clearer, more distinct signal acquisition. Non- L. A., Rohde, M. M., et al. (2000). A direct-brain interface based
invasive techniques, such as scalp electrodes, could on event-related potentials. IEEE Transactions on Rehabilitation
be improved by reducing noise and incorporating so- Engineering, 8(2), 180185.
phisticated filters. Although research to date has fo- Mankoff, J., Dey, A., Moore, M., & Batra, U. (2002). Web accessibility
for low bandwidth input. In Proceedings of ASSETS 2002 (pp. 8996).
cused mainly on controlling output from the brain, Edinburgh, UK: ACM Press.
recent efforts are also focusing on input channels. Mason, S. G., & Birch, G. E. (In press). A general framework for brain-
Much work also remains to be done on appropriate computer interface design. IEEE Transactions on Neural Systems
and Rehabilitation Technology.
mappings to control signals. Moore, M., Mankoff, J., Mynatt, E., & Kennedy, P. (2002). Nudge
As work in the field continues, mainstream ap- and shove: Frequency thresholding for navigation in direct brain-
plications for BCIs may emerge, perhaps for peo- computer interfaces. In Proceedings of SIGCHI 2001Conference on
ple in situations of imposed disability, such as jet Human Factors in Computing Systems (pp. 361362). New York:
ACM Press.
pilots experiencing high G-forces during maneuvers, Perelmouter, J., & Birbaumer, N. (2000). A binary spelling interface
or for people in situations that require hands-free, with random errors. IEEE Transactions on Rehabilitation Engineer-
heads-up interfaces. Researchers in the BCI field are ing, 8(2), 227232.
just beginning to explore the possibilities of real- Pfurtscheller, G., Neuper, C., Guger, C., Harkam, W., Ramoser, H.,
Schlgl, A., et al. (2000). Current trends in Graz brain-computer in-
world applications for brain signal control. terface (BCI) research. IEEE Transactions on Rehabilitation Engi-
neering, 8(2), 216218.
Melody M. Moore, Adriane B. Davis, Tomori, O., & Moore, M. (2003). The neurally controllable Internet
browser. In Proceedings of SIGCHI 03 (pp. 796798).
and Brendan Allison Wolpaw, J. R., Birbaumer, N., McFarland, D., Pfurtscheller, G., &
Vaughan, T. (2002). Brain-computer interfaces for communica-
See also Physiology tion and control. Clinical Neurophysiology, 113, 767791.
Wolpaw, J. R., McFarland, D. J., & Vaughan, T. M. (2000). Brain-com- this ethnographic style of contextually immersed in-
puter interface research at the Wadsworth Center. IEEE Transactions vestigation is that of Michael Byrne and his colleagues
on Rehabilitation Engineering, 8(2), 222226.
(1999), who used their observations to create a
taxonomy of Web-browsing tasks. Their method in-
volved videotaping eight people whenever they used
a browser in their work. The participants were en-
BROWSERS couraged to continually articulate their objectives
and tasks, essentially thinking aloud. A total of five
For millions of computer users worldwide, a browser hours of Web use was captured on video and tran-
is the main interface with the World Wide Web, the scribed, and a six-part taxonomy of stereotypical
worlds foremost Internet information exchange serv- tasks emerged:
ice. Banking, shopping, keeping in contact with
1. Use information: activities relating to the use
friends and family through e-mail, accessing news,
of information gathered on the Web;
looking words up in the dictionary, finding facts and
2. Locate on page: searching for particular in-
solving puzzlesall of these activities and many
formation on a page;
more can be carried out on the Web.
3. Go to: the act of trying to get the browser to dis-
After the 1993 release of the first graphical user
play a particular URL (Web address);
interface Web browser (NCSA Mosaic), the Web rap-
4. Provide information: sending information to a
idly evolved from a small user base of scientists ac-
website through the browser (for example, pro-
cessing a small set of interlinked text documents to
viding a billing address or supplying search
approximately 600 million users accessing billions
terms to a search engine);
of webpages that make use of many different media,
5. Configure browser: changing the configuration
including text, graphics, video, audio, and anima-
of the browser itself; and
tion. Economies of scale clearly apply to the effec-
6. React to environment: supplying information
tiveness of Web browsers.
required for the browser to continue its opera-
Although there has been substantial work on the
tion (for example, responding to a dialog box
webification of sources of information (for exam-
that asks where a downloaded file should be
ple, educational course materials), there has been
saved).
surprisingly little research into understanding and
characterizing Web users tasks, developing better Although these results were derived from only a
browsers to support those tasks, and evaluating the few hours of Web use by a few people, they provide
browsers success. But ethnographic and field stud- initial insights into the tasks and actions accom-
ies can give us a contextual understanding of Web plished using a browser.
use, and longitudinal records of users actions make Another approach to studying how people use
possible long-term quantitative analyses, which in the Web is to automatically collect logs of users
turn are leading to low-level work on evaluating and actions. The logs can then be analyzed to provide a
improving browsers. wide variety of quantitative characterizations of Web
use. Although this approach cannot provide insights
into the context of the users actions, it has the ad-
What Do Web Users Do? vantage of being implementable on a large scale.
The best way to understand fully what users do with Months or years of logged data from dozens of users
their browsers, why they do it, and the problems they can be included in an analysis.
encounter is to observe and question users directly Two approaches have been used to log Web-use
as they go about their everyday work. Unfortunately data. Server-side logs collect data showing which pages
this approach puts inordinate demands on re- were served to which IP address, allowing Web de-
searchers time, so it is normally used only with small signers to see, for instance, which parts of their sites
sets of participants. The study that best demonstrates are particularly popular or unpopular. Unfortunately,
BROWSERS 81
server-side logs only poorly characterize Web usability of participants). Cockburn and McKenzies log analy-
issues. sis suggested that bookmark use had evolved, with
The second approach uses client-side logs, which users either maintaining large bookmark collections
are established by equipping the Web browser (or a or almost none: The total number of bookmarks in
client-side browser proxy) so that it records the exact participants collections ranged from 0 to 587, with
history of the users actions with the browser. The first a mean of 184 and a high standard deviation of 166.
two client-side log analyses of Web use were both con- A final empirical characterization of Web use
ducted in 1995 using the then-popular XMosaic from Cockburn and McKenzies log analysis is that
browser. The participants in both studies were pri- Web browsing is surprisingly rapid, with many or
marily staff, faculty, and students in university com- most webpages being visited for only a very brief pe-
puting departments. Lara Catledge and James Pitkow riod (less than a couple of seconds). There are two
logged 3 weeks of use by 107 users in 1995, while Linda main types of browsing behavior that can explain
Tauscher and Saul Greenberg analyzed 5 to 6 weeks the very short page visits. First, many webpages are
of use by 23 users in 1995. The studies made several simply used as routes to other pages, with users
important contributions to our understanding of what following known trails through the series of links
users do with the Web. In particular, they revealed that that are displayed at known locations on the pages.
link selection (clicking on links in the Web browser) Second, users can almost simultaneously display a
accounts for approximately 52 percent of all webpage series of candidate interesting pages in inde-
displays, that webpage revisitation (returning to pendent top-level windows by shift-clicking on the
previously visited webpages) is a dominant naviga- link or by using the links context menu. For exam-
tion behavior, that the Back button is very heavily used, ple, the user may rapidly pop up several new win-
and that other navigation actions, such as typing URLs, dows for each of the top result links shown as a result
clicking on the Forward button, or selecting book- of a Google search.
marked pages, were only lightly used. Tauscher and
Greenberg also analyzed the recurrence rate of page
visitsthe probability that any URL visited is a re- Improving the Web Browser
peat of a previous visit, expressed as a percentage
(Tauscher and Greenberg 1997, 112). They found a User Interface
recurrence rate of approximately 60 percent, mean- The studies reported above inform designers
ing that on average users had previously seen approxi- about what users do with the current versions of their
mately three out of five pages visited. In a 2001 study, browsers. Naturally, there is a chicken-and-egg prob-
Andy Cockburn and Bruce McKenzie showed that the lem in that stereotypical browser use is strongly af-
average recurrence rate had increased to approximately fected by the support provided by browsers. Browser
80 percentfour out of five pages a user sees are ones interfaces can be improved both by designing to bet-
he or she has seen previously. Given these high recur- ter support the stereotypes and by innovative design
rence rates, it is clearly important for browsers to pro- that enables previously difficult or impossible tasks.
vide effective tools for revisitation. The problems of hypertext navigation were
The 1995 log analyses suggested that people rarely well known long before the Web. As users navigate
used bookmarks, with less than 2 percent of user through the richly interconnected information nodes
actions involving bookmark use. However, a survey of the Web (or any hypertextual information space)
conducted the following year (Pitkow, 1996) indicates their short-term memory becomes overloaded with
that users at least had the intention of using book- the branches made, and they become lost in hy-
marks, with 84 percent of respondents having more perspace. In the late 1980s many researchers were
than eleven bookmarks. Pitkow reported from a experimenting with graphical depictions of hyper-
survey of 6,619 users that organizing retrieved in- text spaces in order to help users orient themselves:
formation is one of the top three problems people re- For example, the popular Apple language Hypercard
port relating to using the Web (reported by 34 percent provided a thumbnail graphical representation of the
recent cards displayed, and gIBIS provided a network Another problem users have with current Web
diagram of design argumentation. Soon after the Web browsers is that they misunderstand the behavior
emerged in 1991, similar graphical techniques were of the Back button. An experiment showed that eight
being constructed to aid Web navigation. Example of eleven computer scientists incorrectly predicted
systems included MosaicG, which provided the behavior of Back in simple Web navigation tasks.
thumbnail images of the visited pages arranged in a The problem stems from users believing that Back
tree hierarchy, WebNet, which drew a hub-and-spoke provides access to a complete history of previously
representation of the pages users visited and the links visited pages, rather than the stack-based subset that
available from them, and the Navigational View can actually be accessed. Cockburn and his colleagues
Builder, which could generate a wide variety of describe the behavior and make an evaluation of a
two-dimensional and three-dimensional represen- true history-based Back system, but results indicate
tations of the Web. that the pros and cons of the technique are closely
Despite the abundance of tools that provide balanced, such that the advantages do not outweigh
graphical representations of the users history, none the difficulties inherent in making a switch from cur-
have been widely adopted. Similarly, log analyses of rent behavior.
Web use show that users seldom use the history tools The World Wide Web revolution has been a great
provided by all of the main Web browsers. Given that success in bringing computer technology to the
Web revisitation is such a common activity, why masses. The widespread adoption and deployment
are these history tools so lightly used? The best ex- of the Web and the browsers used to access it hap-
planation seems to be that these tools are not needed pened largely without input from researchers in hu-
most of the time, so they are unlikely to be on per- man-computer interaction. Those researchers are
manent display, where they would compete with other now improving their understanding of the usability
applications for screen real estate. Once iconified, the issues associated with Web browsers and browsing.
tools are not ready to hand, and it is overhead for As the technology and understanding matures, we
users to think of using them, take the actions to can expect browser interfaces to improve, enhanc-
display them, orient themselves within the informa- ing the efficiency of Web navigation and reducing
tion they display, and make appropriate selections. the sensation of becoming lost in the Web.
While the projects above focus on extending
browser functionality, several other research projects Andy Cockburn
have investigated rationalizing and improving browsers
current capabilities. The interface mechanisms for re- See also Mosaic; Website Design
turning to previously visited pages have been a par-
ticular focus. Current browsers support a wide range
of disparate facilities for revisitation, including the FURTHER READING
Back and Forward buttons and menus, menus that al-
low users to type or paste the URLs of websites the user Abrams, D., Baecker R., & Chignell, M. (1998). Information archiv-
wants to visit, the history list, bookmarks or lists of fa- ing with bookmarks: Personal Web space construction and or-
ganization. In Proceedings of CHI'98 Conference on Human Factors
vorites, and the links toolbar. Of these utilities, log in Computing Systems (pp. 4148). New York: ACM Press.
analyses suggest that only the Back button is heavily Ayers, E., & Stasko, J. (1995). Using graphic history in browsing the
used. The WebView system and Glabster both demon- World Wide Web. In Proceedings of the Fourth International World
strate how history facilities and bookmarks can be en- Wide Web Conference (pp. 451459). Retrieved January 19, 2004,
from http://www.w3j.com/1/ayers.270/paper/270.html
hanced and integrated within the Back menu, providing Bainbridge, L. (1991). Verbal protocol analysis. In J. Wilson & E. Corlett
a powerful and unified interface for all revisitation (Eds.), Evaluation of human work: A practical ergonomics method-
tasks. Both WebView and Glabster automatically cap- ology (pp. 161179). London: Taylor and Francis.
Byrne, M., John, B., Wehrle, N., & Crow, D. (1999). The tangled
ture thumbnail images of webpages, making it easier Web we wove: A taskonomy of WWW Use. In Proceedings of CHI'99
for the user to identify previously visited pages from Conference on Human Factors in Computing Systems (pp. 544551).
the set displayed within the back menu. New York: ACM Press.
BROWSERS 83
Catledge, L., & Pitkow, J. (1995). Characterizing browsing strategies in the Conklin, J., & Begeman, M. (1988). gIBIS: A hypertext tool for ex-
World Wide Web. In Computer systems and ISDN systems: Proceedings ploratory discussion. ACM Transactions on Office Information
of the Third International World Wide Web Conference, 27, 10651073). Systems, 6(4), 303313.
Chi, E., Pirolli, P., & Pitkow, J. (2000). The scent of a site: A system for Coulouris, G., & Thimbleby, H. (1992). HyperProgramming. Woking-
analyzing and predicting information scent, usage, and usability ham, UK: Addison-Wesley Longman.
of a Web site. In Proceedings of CHI'2000 Conference on Human Fischer, G. (1998). Making learning a part of life: Beyond the 'gift-
Factors in Computing Systems (pp.161168). New York: ACM Press. wrapping' approach of technology. In P. Alheit & E. Kammler (Eds.),
Cockburn, A., Greenberg, S., McKenzie, B., Jason Smith, M., & Kaasten, Lifelong learning and its impact on social and regional development
S. (1999). WebView: A graphical aid for revisiting Web pages. In (pp. 435462). Bremen, Germany: Donat Verlag.
Proceedings of the 1999 Computer Human Interaction Specialist Kaasten, S., & Greenberg, S. (2001). Integrating Back, History and
Interest Group of the Ergonomics Society of Australia (OzCHI'91) bookmarks in Web browsers. In Proceedings of CHI'01 (pp. 379380).
(pp. 1522). Retrieved January 19, 2004, from http://www.cpsc.ucal- New York: ACM Press.
gary.ca/Research/grouplab/papers/1999/99-WebView.Ozchi/ Mukherjea, S., & Foley, J. (1995). Visualizing the World Wide Web with
Html/webview.html the navigational view builder. Computer Systems and ISDN Systems,
Cockburn, A., & Jones, S. (1996). Which way now? Analysing and eas- 27(6), 10751087.
ing inadequacies in WWW navigation. International Journal of Nielsen, J. (1990). The art of navigating through HyperText: Lost in
Human-Computer Studies, 45(1), 105129. hyperspace. Communications of the ACM, 33(3), 296310.
Cockburn, A., & McKenzie, B. (2001). What do Web users do? An em- Pirolli, P., Pitkow, J., & Rao, R. (1996). Silk from a sow's ear: Extracting
pirical analysis of Web use. International Journal of Human- usable structures from the Web. In R. Bilger, S. Guest, & M. J. Tauber
Computer Studies, 54(6), 903922. (Eds.), Proceedings of CHI'96 Conference on Human Factors in
Cockburn, A., McKenzie, B., & Jason Smith, M. (2002). Pushing Back: Computing Systems (pp. 118125). New York: ACM Press.
Evaluating a new behaviour for the Back and Forward buttons Pitkow, J. (n.d.). GVU's WWW User Surveys. Retrieved January 19,
in Web browsers. International Journal of Human-Computer Studies, 2004, from http://www.gvu.gatech.edu/user_surveys/
57(5), 397414. Tauscher, L., & Greenberg, S. (1997). How people revisit Web pages:
Conklin, J. (1988). Hypertext: An introduction and survey. In I. Greif Empirical findings and implications for the design of history
(Ed.), Computer supported cooperative work: A book of readings systems. International Journal of Human Computer Studies, 47(1),
(pp. 423475). San Mateo, CA: Morgan-Kauffman. 97138.
CATHODE RAY TUBES
CAVE
CHATROOMS
CHILDREN AND THE WEB
CLASSROOMS
C
CLIENT-SERVER ARCHITECTURE
COGNITIVE WALKTHROUGH
COLLABORATORIES
COMPILERS
COMPUTER-SUPPORTED COOPERATIVE WORK
CONSTRAINT SATISFACTION
CONVERGING TECHNOLOGIES
CYBERCOMMUNITIES
CYBERSEX
CYBORGS
sion and radar over the next fifty years provided the
CATHODE RAY TUBES impetus for continual improvements. With the emer-
gence of desktop computing in the 1980s, the CRT
The cathode ray tube (CRT) has been the dominant market expanded, and its performance continued to
display technology for decades. Products that utilize evolve. As portability has come to be more and more
CRTs include television and computer screens in the important in the consumer electronics industry, the
consumer and entertainment market, and electronic CRT has been losing ground. The development of
displays for medical and military applications. CRTs flat panel technologies such as liquid crystal displays
are of considerable antiquity, originating in the and plasma displays for portable products, computer
late nineteenth century when William Crookes screens, and television makes the CRT very vulner-
(18321919) studied the effects of generating an elec- able. Because of the CRTs maturity and compara-
trical discharge in tubes filled with various gases. tively low cost, however, its application will be assured
(The tubes were known as discharge tubes.) It was for many years to come.
over thirty years later in 1929 that the CRT was
utilized to construct actual imagery for television
applications by Vladimir Zworykin (18891982) of How Cathode Ray Tubes Work
Westinghouse Electric Corporation. The further de- A CRT produces images when an electron beam
velopment and optimization of the CRT for televi- is scanned over a display screen in a pattern that is
85
determined by a deflection mechanism. The display must be and on when the beam should be moved
screen is coated with a thin layer of phosphor that across different portions of the screen.
luminesces under the bombardment of electrons.
By this means the display screen provides a two-
dimensional visual display, corresponding to in- Displaying Color
formation contained in the electron beam. There are One of the most important tasks of the modern dis-
four major components of a CRT display: the play is rendering full-color images. Shadow-masking
vacuum tube, the electron source (known as the elec- configurations are by far the most successful way
tron gun), the deflection mechanism, and the phos- to create full color images in CRT displays. The
phor screen. shadow mask CRT typically uses three electron beams
The tube (sometimes referred to as a bulb) is deflected by one coil (the simplest configuration).
maintained at a very high vacuum level to facili- The electron beams traverse a perforated metal mask
tate the flow of electrons in the electron beam. The (shadow mask) before impinging on selected
front surface of the tube defines the visual area phosphor materials (there are three sorts of phos-
of the display, and it is this front surface that is cov- phor that can emit red, green, and blue light). The
ered with phosphor, which is in turn covered by shadow mask apertures are typically configured as
the anode (the electron-collecting electrode). The stripes, circles, or slots. The arrangement of the elec-
tube has three main sections: the front surface, the tron optics and the deflection system is such that
funnel, and the neck. The entire tube is typically three electron beams converge onto the screen af-
made of glass so that very high vacuums can be ter passing through the shadow mask, each beam
sustained, but in some cases the funnel and neck impinging on one phosphor, which, when bom-
can be fabricated from metal or ceramic. For de- barded with electrons, emits red, green, or blue
manding applications that require additional ro- visible light. The red, green, and blue phosphors are
bustness, an implosion-proof faceplate may be spatially arranged on the viewing screen.
secured to the front tube surface for durability. This The Tr init ron desig n, invented by Sony
typically comes at the expense of optical through- Corporation, uses vertical stripe arrays rather than
put, but antireflection coatings are often used to circular or slotted apertures. These arrays alternate
improve contrast and to compensate for the trans- red, green, and blue when viewed from the faceplate
mission losses. side of the tube. There is a single electron source,
The electron source, a hot cathode at the far end rather than three, which eliminates the problem of
from the front surface, generates a high-density beam convergence. The Trinitron also has superior
electron beam whose current can be modulated. resolution in the vertical direction since its apertures
The electron beam can be focused or reflected are not limited in that direction. The only negative
deflectedby electrostatic or magnetic methods, attribute of the Trinitron is that the mask is not self-
and this deflection steers the electron beam to des- supporting, which ultimately limits the size of the
ignated positions of the front surface to create visual vacuum tube. The advantages of CRT displays in-
imagery. clude their maturity, their well-understood manu-
The phosphor screen on the inside front surface facturing process, their ability to provide full-color
of the tube converts the electron beam into visible and high-resolution imaging, and the comparatively
light output. On top of the phosphor particles is the low cost for high information content. CRTs are vul-
thin layer of conducting material (usually aluminum) nerable to competition from liquid crystal displays
that serves as the anode, drawing the electrons toward and plasma displays (both of which make possible
the screen. The directions on how to manipulate the flat-panel displays), however, because CRTs are bulky,
electron stream are contained in an electronic sig- heavy, and big power consumers. In addition to
nal called a composite video signal. This signal con- the utility of flat-panel display for portable applica-
tains information on how intense the electron beam tions for which CRTs could never be considered, flat-
CAVE 87
panel displays have made significant inroads into left eye image should be seen by the left eye and are
desktop monitors and large-area televisions. As the opaque otherwise. Similarly, the right eye gets the right
price of flat-panel displays continues to plummet, image. Images need to be generated at 100 to 120 hertz
they are certain to capture even more of the CRT so each eye can get a flicker-free 50- to 60-hertz dis-
market in the future. play. All screens need to be synchronized so that each
eye sees the same phase stereo image on every screen,
Gregory Philip Crawford a requirement that until 2003 meant that only the
most expensive SGI (Silicon Graphics, Inc.) computer
See also Liquid Crystal Displays graphics systems could be used. Synchronizing PC
graphics cards now reduce the cost of CAVE com-
puting and image generation by 90 percent.
FURTHER READING The CAVEs projection onto the screens does not
need to keep up with the viewers head motion nearly
Castelliano, J. (1992). Handbook of display technology. San Diego, CA: as much as is required in a head-mounted VR dis-
Academic Press.
Keller, P. A. (1997). Electronic display measurement. New York:
play (HMD), which needs to have small screens at-
Wiley SID. tached in front of the eyes. Of course, any movement
MacDonald, L. W., & Lowe, A. C. (1997). Display systems: Design of the viewers body within the space requires up-
and applications. New York: Wiley SID. dating the scene perspective, but in normal inves-
tigative use, the CAVE needs to keep up only with
body motion, not head rotation; the important re-
sult is that the delay of trackers is dramatically less
of a problem with CAVEs than with HMDs. In ad-
CAVE dition, although only one viewer is tracked, other
people can share the CAVE visuals at the same time;
The CAVE is a virtual reality (VR) room, typically 3 their view is also in stereo and does not swing with
by 3 by 3 meters in size, whose walls, floor, and some- the tracked users head rotation, although their per-
times ceiling are made entirely of computer-projected spective is still somewhat skewed. Often the person
screens. Viewers wear a six-degree-of-freedom loca-
tion sensor called a tracker so that when they move
within the CAVE, correct viewer-centered perspec-
tives and surround-stereo projections are produced
fast enough to give a strong sense of 3D visual im-
mersion. Viewers can examine details of a com-
plex 3D object simply by walking up to and into it.
The CAVE was invented in 1991 for two reasons:
to help scientists and engineers achieve scientific in-
sight without compromising the color and distortion-
free resolution available then on workstations and
to create a medium worthy of use by fine artists.
CAVE viewers see not only projected computer-
generated stereo scenes but also their own arms and
bodies, and they can interact easily with other people.
The CAVE uses active stereo, which produces differ-
ent perspective views for the left and right eyes of the The CAVE is a multi-person, room-sized, high-
viewer in synchrony with special electronic shutter resolution, 3D video and audio environment. Photo cour-
glasses that go clear in front of the left eye when the tesy of National Center for Supercomputing Applications.
CAVE Variants
Variants of the CAVE include the ImmersaDesk,
a drafting-table-size rear-projected display with a
screen set at an angle so that the viewer can look
down as well as forward into the screen; looking
down gives a strong sense of being in the scene.
PARIS uses a similarly angled half-silvered screen
that is projected from the top; the viewers hands
work under the screen and are superimposed on the
3D graphics (rather than blocking them, as with nor-
mal projections).
The CAVE originally used three-tube stereo pro-
The Personal Augmented Reality Immersive System jectors with special phosphors to allow a 100- to 120-
(PARIS) has a half-silvered mirror at an angle in front hertz display without ghosting from slow green phos-
of the user. The screen, above the desk facing down, phor decay. Tube projectors are now rather dim by
superimposes a stereo image on the users hands work- modern standards, so the CAVE was rebuilt to use
ing beyond the mirror. Photo courtesy of the Electronic bright digital mirror-based projectors, like those used
Visualization Laboratory. in digital cinema theaters. Projectors require sig-
nificant alignment and maintenance; wall-sized flat-
in the role of guide or instructor handles the con- panel screens will be welcomed since they need no
trols (a 3D mouse called Wanda) and the student alignment and have low maintenance and no pro-
wears the tracker to get the best view, a mode of jection distance.
usage that is quite functional for both learning and The GeoWall, a passive stereo device, works
demonstrations. differently, polarizing the output of two projectors
The CAVE uses a rear-screen projection for the onto a single screen. Viewers wear the throw-away
walls so the viewer does not block the light and cast polarized glasses used in 3D movies to see stereo.
shadows. The floor is typically projected down from In addition to visual immersion, the CAVE has
the top, which creates a small shadow around the synchronized synthetic and sampled surround sound.
viewers feet. A CAVE with three walls and a floor min- The PARIS system features a PHANTOM tactile de-
imally requires a 13- by 10-meter space with a ceiling vice, which is excellent for manipulating objects the
4.5 meters high. Six-sided CAVEs have rear projec- size of a bread box or smaller.
tions from every direction, which require much higher
ceilings, more elaborate support structures, and floor
screens that can withstand the weight of several people. CAVEs for Tele-Immersion
Someday, 3-square-meter flat-panel displays sus- The CAVE was originally envisioned as a tele-
pended as a ceiling, positioned vertically as walls, and immersive device to enable distance collaboration
tough enough to walk on would allow CAVEs in nor- between viewers immersed in their computer-
mal rooms. However, current technology panel- generated scenes, a kind of 3D phone booth. Much
displays refresh too slowly to use shutter glasses, so work has gone into building and optimizing ultra-
they must be otherwise modified for stereo display. high-speed computer networks suitable for shar-
The Varrier method involves placing a barrier screen ing gigabits of information across a city, region,
so that the computed views to each eye are seen nation, or indeed, the world. In fact, scientists, en-
through perfectly placed thin black bars, that is, the gineers, and artists in universities, museums, and
correctly segmented image is placed in dynamic commercial manufacturing routinely use CAVEs and
perspective behind the barrier in real time. Varrier variants in this manner.
viewers wear no special glasses since the image sepa-
ration is performed spatially by the barrier screen. Tom DeFanti and Dan Sandin
CHATROOMS 89
See also Virtual Reality; Telepresence; Three-

Dimensional Graphics CHATROOMS
Defined most broadly, chatrooms are virtual spaces
FURTHER READING where conversations occur between two or more users
in a synchronous or nearly synchronous fashion. Many
Cruz-Neira, C., Sandin, D., & DeFanti, T. A. (1993). Virtual reality: different types of chat spaces exist on the Internet.
The design and implementation of the CAVE. Proceedings of the
SIGGRAPH 93 Computer Graphics Conference, USA, 135142.
One type is Internet Relay Chat (IRC), a multiuser
Czernuszenko, M., Pape, D., Sandin, D., DeFanti, T., Dawe, G. L., & synchronous chat line often described as the citi-
Brown, M. D. (1997). The ImmersaDesk and Infinity Wall zens band radio of the Internet. Another type of vir-
projection-based virtual reality displays [Electronic version]. tual space where computer-mediated communication
Computer Graphics, 31(2), 4649.
DeFanti, T. A., Brown M. D., & Stevens, R. (Eds.). (1996). Virtual re- (CMC) takes place is Multi-User Domains (MUDs,
ality over high-speed networks. IEEE Computer Graphics & sometimes called Multi-User Dungeons, because of
Applications, 16(4), 1417, 4284. their origin as virtual locations for a Dungeons and
DeFanti, T., Sandin, D., Brown, M., Pape, D., Anstey, J., Bogucki, M.,
et al. (2001). Technologies for virtual reality/tele-immersion ap-
Dragons role-playing type of networked gaming).
plications: Issues of research in image display and global net- MUDs were initially distinguished from IRC by their
working. In R. Earnshaw, et al. (Eds.), Frontiers of Human-Centered persistence, or continued existence over time, and their
Computing, Online Communities and Virtual Environments malleability, where users may take part in the build-
(pp. 137159). London: Springer-Verlag.
Johnson, A., Leigh, J., & Costigan, J. (1998). Multiway tele-immersion
ing of a community or even a virtual world, de-
at Supercomputing 97. IEEE Computer Graphics and Applications, pending on the tools and constraints built into the
18(4), 69. architecture of their particular MUD. Web-based cha-
Johnson, A., Sandin, D., Dawe, G., Qiu, Z., Thongrong, S., & Plepys, trooms are a third type of chat space where users may
D. (2000). Developing the PARIS: Using the CAVE to prototype
a new VR display [Electronic version]. Proceedings of IPT 2000, converse synchronously in a persistent location hosted
CD-ROM. Korab H., & Brown, M. D. (Eds.). (1995). Virtual by Internet Service Providers (ISPs) or websites, which
Environments and Distributed Computing at SC95: GII Testbed and may be either large Web portals like Yahoo.com or
HPC Challenge Applications on the I-WAY. Retrieved November 5,
2003, from http://www.ncsa.uiuc.edu/General/Training/SC95/GII
small individual sites.
.HPCC.html
Lehner, V. D., & DeFanti, T. A. (1997). Distributed virtual reality:
Supporting remote collaboration in vehicle design. IEEE Computer UNIX was not designed to stop people from doing stu-
Graphics & Applications (pp. 1317).
Leigh, J., DeFanti, T. A., Johnson, A. E., Brown, M. D., & Sandin,
pid things, because that would also stop them from do-
D. J. (1997). Global tele-immersion: Better than being there. ICAT ing clever things.
97, 7th Annual International Conference on Artificial Reality and Doug Gwyn
Tele-Existence, pp. 1017. University of Tokyo, Virtual Reality Society
of Japan.
Leigh, J., Johnson, A., Brown, M., Sandin, D., & DeFanti, T. (1999).
Tele-immersion: Collaborative visualization in immersive envi- Another type of chat function on the Internet is
ronments. IEEE Computer Graphics & Applications (pp. 6673). instant messaging (IM), which allows users to chat
Sandin, D. J., Margolis, T., Dawe, G., Leigh, J., and DeFanti, T. A. (2001).
The Varrier auto-stereographic display. Proceedings of Photonics
with individuals (or invited groups) in real time,
West 2001: Electronics Imaging, SPIE. Retrieved on November 5, provided that they know a persons user name. Instant
2003, from http://spie.org/web/meetings/programs/pw01/ messaging is distinguished from other chat functions
home.html in that it is often used to hold multiple, simultane-
Stevens, R., & DeFanti, T. A. (1999). Tele-immersion and collabora-
tive virtual environments. In I. Foster & C. Kesselman (Eds.),
ous, private one-on-one chats with others. IM is also
The grid: Blueprint for a new computing infrastructure (pp. 131158). unusual in that the user can also monitor a list of
San Francisco: Morgan Kaufmann. online friends to see when they are logged in to the
instant messaging service. IM chats also differ
from other types of chatrooms in that they are not
persistentthat is, a user cannot log in to the same
chat after the last chatterer has logged off. Instant
message chats are more likely to occur among a group nated by a hatch mark (#) and a number. Because
of people with some personal or professional con- that proved difficult to use as IRC expanded, each
nection than among a group of strangers with only channel was also given a text label, like #hottub
one shared interest who happen to be in the same or #gayboston. IRC channels were originally not
virtual space at the same time. persistentanyone could create a channel on any
conceivable topic, and when the last person logged
out of that channel it ceased to exist. Only with the
History of Internet Chat introduction in 1996 of Undernet and later DalNet
The first function that allowed for synchronous or did it become possible to create persistent channels.
nearly synchronous communication over a network IRC runs through client softwarethe client soft-
was Talk, available on UNIX machines and the ware is what allows the user to see the text in the chat
networks that connected them. Developed in the channel that theyre using and to see who else is cur-
early 1980s, Talk allowed for the nearly synchronous rently in that channel. The most popular client is
exchange of text between two parties; however, un- mIRC, a windows-compatible client; others include
like its descendents, it displayed text messages as they Xircon and Pirch.
were written, character by character, rather than as IRC does not have a central organizing system;
completed messages posted to the discussion all at organizations like universities and research groups
once. Talk and its sister program Phone fell into dis- simply run the software on their servers and make it
use after the introduction of the World Wide Web in available to their users. In the late 1990s, IRC de-
1991 and the introduction of graphical and multi- centralized architecture contributed to a system
user interfaces. breakdown. In mid-1996, when one IRC server op-
erator, based in North America, started abusing
the IRC system, other North American IRC server
Home computers are being called upon to perform many operators expelled the abuser; however, when he
new functions, including the consumption of homework disconnected his server they discovered that
formerly eaten by the dog. he was also the main link between North American
Doug Larson and European IRC networks. After weeks of negoti-
ations between North American and European
IRC server operators, who disagreed over the han-
Internet Relay Chat dling of the expulsion, the impasse was not resolved.
Jarkko Oikarinen, a Finnish researcher, developed While interconnectivity between continents has been
IRC in 1988 based on the older Bulletin Board restored, the two IRC networks remain separate (IRC
System (BBS). BBSs were central locations where net and Efnet [Eris Free Net]); they have their own
users could dial in to a central server using a mo- separate channels and have developed separately.
dem and leave messages and hold discussions on Other networks, including DALnet and Undernet,
this central server, usually dedicated to a certain have developed since the separation.
topic or interest group. Oikarinen wrote the IRC
program to allow users to have real-time discus- MUDs
sions not available on the BBS. First implemented Pavel Curtis, a researcher at Xerox PARC who spe-
on a server at the University of Oulu where Oikarinen cializes in virtual worlds, gives this definition of a
worked, IRC quickly spread to other Finnish uni- Multi-User Domain: A MUD is a software program
versities, and then to universities and ISPs through- that accepts connections from multiple users across
out Scandinavia and then the World. some kind of network and provides to each user ac-
Each channel on IRC (the name was taken from cess to a shared database of rooms, exits and other
the Citizens Band radio community) represents a objects. Each user browses and manipulates this data-
specific topic. Initially each channel was desig- base from inside one of the rooms. A MUD is a kind
CHATROOMS 91
A Personal StoryLife Online
In the mid 1990s, I went to visit my first online chatroom as part of a larger project on computer-mediated communi-
cation. I had no idea what to expectwhether the people would be who they said they were, whether Id have anything
in common with other visitors, or what it would be like to interact in a text-based medium. I found myself enjoying the
experience of talking to people from all over the world and came to spend much time in this virtual community.
I soon learned that the community was much larger than the chatroom I had visited, connected by telephone, e-mail,
letters, and occasional face-to-face visits. Over the past five years, Ive spoken or emailed with many new acquaintances,
and have had the pleasure of meeting my online friends in person when my travels take me to their part of the country.
Participation in a virtual community has provided me opportunities to talk in depth with people from around the
world, including Australia, New Zealand, South America, Mexico, Europe, and even Thailand. The virtual community
also brings together people from a wide range of socioeconomic backgrounds that might ordinarily never have mixed.
Its been fascinating to get to know such a diverse group of individuals.
My personal experiences in an online community have helped shape my research into the societal dimensions of
computing and computer-mediated communication. One of my current projects investigates the effects of participation
in online support communities on peoples mental and physical well-being. In addition, the success with which Ive been
able to meet and become acquainted with others using a text-only medium has had a strong impact on my theories about
how technologies can successfully support remote communication and collaboration.
Susan R. Fussell
of virtual reality, an electronically represented place Web-Based Chat

that users can visit (Warschauer 1998, 212). Web-based chatting became possible after the World
MUDs provide more virtual places to visit, hang Wide Web was first released onto the Internet in
out, socialize, play games, teach, and learn than IRC December 1990, but it didnt became popular until
or Web-based chatrooms do. Some MUDs have been after the release of the Java programming language
used to hold meetings or conferences because they a few years later. Java allowed developers to create
allow participants to convene without travel hassles user-friendly graphical interfaces to chat spaces on
virtual conferences may have different rooms for dif- websites or ISP portals that could function across
ferent topics and a schedule of events similar to that different computing and Internet browsing plat-
of a geographically located conference. forms. Web-based chatting, like IRC, tends to be
Two British graduate students, Richard Bartle and based around common themes, issues, or specific
Roy Trubshaw, developed the first MUD in 1979, as a discussion topicsit has given rise to rooms like
multiuser text-based networked computer game. Love@AOL or sports- or hobby-themed rooms
Other MUDs followed, and new subtypes grew, in- like The Runners Room on Yahoo Chats. Other chat
cluding MOOs (Multiuser domains Object Oriented), spaces may be on an individual website devoted to a
used primarily for educational purposes, and MUSHs common theme (like the chat on the Atlantic County
(Multi-user Shared Hallucinations). MOOs allow for Rowing Association site, hosted by ezteams).
greater control because the users of the virtual space
can build objects and spaces as well as contribute text.
Because MUDs are complex virtual environments that Chatroom Users
users visit to master commands and understand pro- All the different iterations of chatrooms discussed
tocols, rules, and mores, their use and appeal has been here have some common elements to help users nav-
limited to a tech-savvy group of users. igate and quickly understand how to use the software.
After entering a chatroom, channel, or domain, a Web-based chat, however, there is the expectation
user is confronted with a screen that is split into two that users are presenting themselves honestly.
or more parts: One side, usually the left, shows the Nevertheless, all chat spaces give users the op-
discussion in progress. In another box on the screen portunity to explore portions of their identity,
is a list of who is logged in to the room. Generally whether it is by choosing to have the opposite gen-
below these is the box where the user enters text or der, a different race, or a different set of personal
commands to begin the conversation or move about experiences, or in the case of some games, by ex-
the space (in the case of MUDs). In some chat spaces, ploring what it is like to be something other than
users can create their own private chat with a sin- human. Anonymity or pseudonymity on line gives
gle individual from the chatroom. In some Web- many users a feeling of freedom and safety that al-
based tools, the chatroom is designed to use an lows them to explore identities that they dare not
instant messaging program to conduct one-on- assume in the offline world. Users are separated by
one chats. In others the private chat tool is built-in geographic distances so it is unlikely that actions
in MUDs, a user uses the whisper command to taken or phrases uttered will come back to haunt
direct a comment or conversation to a particular in- them later. And finally, in chat environments with-
dividual, and in some Web-based chats a private chat out audio or video, communication is mediated by
may be opened in another smaller window in the the technology so there are none of the cues that can
same chatting interface. make a conversation emotional. All of this leads to
In a survey in the summer of 2002, the Pew lower levels of inhibitions, which can either create
Internet & American Life Project found that only one- greater feelings of friendship and intimacy among
quarter of Internet users had ever visited a chatroom chat participants or lead to a greater feeling of ten-
or participated in an online discussion, and only 4 per- sion and lend an argumentative, even combative qual-
cent had visited a chatroom on a typical day. Men ity to a chat space.
are more likely to use chatrooms than women, as are
those who are less well off; those earning less than
$50,000 a year are much more likely to chat than those The Future of Chat
earning more. Younger people are also more likely to In 1991 researchers at Cornell University created
chat, particularly those between eighteen and twenty- CUSeeMe, the first video chat program to be dis-
nine, although among teens, particularly adolescent tributed freely online. Video and audio chat did not
girls, chatting is frequently perceived as unsafe. truly enter mainstream use until the late 1990s, and
Nevertheless, in spite of (or because of) chats rep- with the advent of Apples iChat and Microsofts im-
utation, 55 percent of young people between twelve proved chatting programs and web cams, video chat
and seventeen have visited a chatroom. utilizing speakers and web cams looks to be the fu-
Chatrooms have become the favorite play- ture direction of chatting. Today Yahoo.com and
grounds of many Internet users because they enable other portal-provided Web-based chatrooms allow
them to assume a character or a role different audio and video chat in their rooms, though the
from the one they play in their offline life. As social number of users taking advantage of the technology
psychologist Erving Goffman noted in his 1959 book is still relatively small. A users bandwidth and hard-
Presentation of Self in Everyday Life, we present ware capabilities are still limiting factors in the use
different images of ourselves to different people, and of the bandwidth-intensive video chat, but as broad-
some theorists have described chatrooms as band Internet connectivity percolates through the
spaces of performance where an identity is per- population, the quality of video Web-based chatting
formed for the audience of other chatters. In cer- available to most users will improve, and its adop-
tain chatrooms, like MUDs, where gaming or tion will undoubtedly become more widespread.
role-playing is often the reason users go there, it is MUDs and MOOs are also moving into HTML-
expected that visitors do not bear any resem- based environments, which will make it much eas-
blance to their selves at the keyboard. In IRC and ier for the average Internet user to adopt them,
CHILDREN AND THE WEB 93
and will perhaps move Multi-User Domains from Taylor, T.L. (1999). Life in virtual worlds: Plural existence, multi-
the subculture of academics and devotees into modalities and other online research challenges. American Behavioral
Scientist, 4(3).
everyday use. Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
Amanda Lenhart Warshauer, S. C. (1998). Multi-user environment studies: Defining
a field of study and four approaches to the design of multi-user
environments. Literary and Linguistic Computing, 13(4).
See also E-mail, MUDs Young, J. R. (1994). Textuality and cyberspace: MUDs and written
experience. Retrieved July 31, 2003, from http://ftp.game.org/pub/
mud/text/research/textuality.txt
FURTHER READING
Bartle, R. (1990). Early MUD History. Retrieved July 31, 2003, from
http://www.ludd.luth.se/mud/aber/mud-history.html CHILDREN
Bevan, P. (2002). The circadian geography of chat. Paper presented at the
conference of the Association of Internet Researchers, Maastricht,
Netherlands.
AND THE WEB
Campbell, J. E. (2004). Getting it on online: Cyberspace, gay male sex-
uality and embodied identity. Binghamton, NY: The Haworth Press. Children are among the millions of people who have
Dibbell, J. (1998). A rape in cyberspace. In My tiny life: Crime and pas- been introduced to new ways of accessing infor-
sion in a virtual world. Owl Books, chapter 1. Retrieved July 31, mation on the World Wide Web, which was launched
2003, from http://www.juliandibbell.com/texts/bungle.html
IRC.Net. IRC net: Our history. Retrieved July 30, 2003, from http:// in 1991 and began to become popular with the adop-
www.irc.net/ tion of a graphical user interface in 1993. The fact
Hudson, J. M., & Bruckman, A. S. (2002). IRC Francais: The cre- that the Web utilizes hypertext (content with active
ation of an Internet-based SLA community. Computer Assisted
Language Learning, 1(2), 109134.
links to other content) and a graphical user interface
Kendall, L. (2002). Hanging out in the virtual pub: Masculinities and have made it more congenial and much easier to use
relationships online. Berkeley: University of California Press. than earlier menu-driven, text-based interfaces (i.e.,
Lenhart, A., et al. (2001). Teenage life online: The rise of the instant Gopher, Jughead, Veronica) with the Internet.
message generation and the Internets impact on friendships and
family relationships. Pew Internet & American Life Project, retrieved
August 21, 2003, from http://www.pewinternet.org/
Murphy, K. L., & Collins, M. P. (1997). Communication conven- Childrens Web Use
tions in instructional electronic chats. First Monday, 11(2). Children use the Web inside and outside the class-
Pew Internet & American Life Project. (2003). Internet activities (Chart).
Retrieved July 31, 2003, from http://www.pewinternet.org/reports/ room, and they navigate it to find information for
index.asp both simple and complex projects. They recognize
Pew Internet & American Life Project. (2003). Unpublished data from the Web as a rich source of up-to-date information,
June-July 2002 on chatrooms. Author.
Reid, E. M. (1994). Cultural formation in text-based virtual realities.
hard-to-find information, and compelling images.
Unpublished doctoral dissertation, University of Melbourne, Research by Dania Bilal (2000) and Jinx Watson
Australia. Retrieved July 31, 2003, from http://www.aluluei.com/ (1998) has revealed that children who use the Web
cult-form.htm have a sense of independence, authority, and con-
Rheingold, H. (1993). The virtual community: Homesteading on the
electronic frontier. Cambridge, MA: MIT Press.
trol. They are motivated, challenged, and self-
Rheingold, H. (1998). Building fun online learning communities. confident. They prefer the Web to print sources due
Retrieved July 30, 2003, from http://www.rheingold.com/texts/ to the vast amount of information available and their
education/moose.html ability to search by keyword and browse subject hi-
Schaap, F. (n.d.). Cyberculture, identity and gender resources (online
hyperlinked bibliography). Retrieved July 31, 2003, from http:// erarchies quickly. Research conducted for the Pew
fragment.nl/resources/ Internet & American Life Project revealed that both
Surkan, K. (n.d.). The new technology of electronic text: Hypertext and parents and children believe that the Internet helps
CMC in virtual environments. Retrieved July 31, 2003, from http://
english.cla.umn.edu/GraduateProfiles/Ksurkan/etext/etable.html
with learning. While these positive perceptions of
Talk mode (n.d.). The jargon file. Retrieved November 1, 2002, from the Internet are encouraging, childrens success in
http://www.tuxedo.org/~esr/jargon/html/entry/talk-mode.html finding information on the Web is questioned. Given
Two of the many books available that

educate children on the perils and
promise of the Web. Safety on the
Internet is geared to ages 69, while
Cyber Space is meant for ages 912.
the Webs increasing complexity and the abundance cult vocabulary for elementary schoolchildren to un-
of information available there, it is worth asking how derstand. Children in that age range preferred sites
well children handle the challenges of using the Web. with high visual content, animation, and short, sim-
Researchers from library and information sci- ple textual content. In 1998 the researchers John
ence, educational psychology, sociology, cognitive Schacter, Gregory Chung, and Aimee Dorr studied
science, and human-computer interaction have stud- the effect of types of tasks on the success of fifth and
ied childrens interaction with the Web. In the field sixth graders in finding information. They found
of information science, researchers have investigated that children browsed more than they searched by
childrens search strategies, their relative preferences keyword and performed better on open-ended (com-
for browsing and searching, their successes and fail- plex) than factual (simple) tasks. By contrast, in 2000
ures, the nature of tasks and success, Web design, and Terry Sullivan and colleagues found that middle and
childrens navigational skills, relevance judgment, high school students were more successful on simple
and affective states (feelings, perception, motivation). tasks than complex ones. Results from Dania Bilals
Findings and conclusions from these studies have research in 20002002 echoed Sullivans results
begun to provide a rich framework for improving and revealed that middle school students were most
system design and developing more effective Web successful on tasks that they chose themselves than
training programs. they were on tasks that were assigned.
The first study in library and information sci- In 1999 Andrew Large, Jamshid Beheshti, and
ence appeared in 1997 when Jasmine Kafai and Haidar Moukdad examined the Web activities of
Marcia Bates examined elementary schoolchildrens Canadian sixth graders. These researchers found that
Web literacy skills. They found that children were children browsed more than they searched by key-
enthusiastic about using the Web and were able to word, had difficulty finding relevant information,
scroll webpages and use hyperlinks effectively. and, although they had been given basic Web
However, the researchers perceived that many web- training, lacked adequate navigational skills. The
sites had too much text to read and too much diffi- childrens use of the Netscape Back command to
CHILDREN AND THE WEB 95
return to the previous page, for example, accounted tion, despite the fact that most of the concepts they
for 90 percent of their total Web moves; they acti- employed were appropriate. The failure to find re-
vated online search help only once. In fact, frequent sults can be attributed largely to the poor indexing
use of the Back command is common among chil- of the Yahooligans! database. Overall, the children
dren and young adults. Various studies in the late took initiative and attempted to counteract their
1990s and early 2000s found similar results. In a information retrieval problems by browsing subject
follow-up to a 1999 study, Andrew Large and Jamshid categories. Indeed, they were more successful when
Beheshti (2000) concluded that children valued they browsed than when they searched by keyword.
the Web for finding information on hard topics, speed Childrens low success rates on the assigned tasks
of access, and the availability of color images, but were attributed to their lack of awareness of the
perceived it as more difficult to use than print sources. difference between simple and complex tasks, es-
Children expressed frustration with information pecially in regard to the approach to take to fulfill
overload and with judging relevance of the retrieved the assignments requirements. On the complex
results. Information overload and problems deter- assigned task, for example, children tended to seek
mining relevance seem to be widespread among chil- specific answers rather than to develop an under-
dren and young adults using the Web; a study of standing of the information found. On the positive
elementary, middle, and high school students in side, children were motivated and persistent in us-
England corroborated Large and Beheshtis finding. ing the Web. When asked about reasons for their mo-
Most children assume that the Web is an efficient tivation and persistence, children cited convenience,
and effective source for all types of information. challenge, fun, and ease of use. Ease of use was de-
Consequently, they rarely question the accuracy and scribed as the ability to search by keyword. On the
authority of what they find. If they retrieve results negative side, children expressed frustration at both
that are close enough to the topic, they may cease to information overload and the zero retrieval that
pursue their initial inquiry and take what they get at resulted from keyword searching. Indeed, this fea-
face value. ture was central to most of the search breakdowns
Most studies focused on using the Web as a whole children experienced. Although Yahooligans! is de-
and on search engines that are developed for adult signed for children aged seven through twelve, neither
users rather than children. Bilal has investigated the its interface design nor its indexing optimized chil-
information-seeking behavior of children who used drens experience. Childrens inadequate knowledge
Yahooligans!, a search engine and directory specif- of how to use Yahooligans! and their insufficient
ically designed for children aged seven through knowledge of the research process hindered their
twelve. She found that 50 percent of the middle success in finding information.
school children were successful on an assigned, fact-
based task, 69 percent were partially successful on
an assigned, research-based task, and 73 percent were Optimizing the Web for Children
successful on tasks they selected themselves. The flex- Childrens experiences with the Web can be greatly
ibility children had in topic selection and modifica- improved by designing Web interfaces that build
tion combined with their satisfaction with the results on their cognitive developmental level, information-
may have influenced their success rate on the self- seeking behaviors, and information needs. Since 2002,
selected task. Children were more motivated, stimu- Bilal (working in the United States) and Large,
lated, and engaged in completing their tasks when Beheshti, and Tarjin Rahman (working together in
they selected topics of personal interest. Canada), have begun projects that involve children
The children used concrete concepts (selected in the design of such interfaces. Both groups have
from the search questions) in their searches and, when concluded that children are articulate about their in-
these concepts failed to generate relevant informa- formation needs and can be effective design partners.
tion, they utilized abstract ones (synonyms or related Based on the ten interfaces that children designed for
terms). The children had trouble finding informa- search engines, Bilal was able to identify the types
of information architecture and visual design chil- in its use. Children, too, should be taught how to use
dren needed and the information architecture, func- the Web effectively and efficiently.With critical-thinking
tionality, and visual design they sought. skills and an understanding of how to manipulate the
In sum, both Bilal and the Canadian-based team Web, children can move from being active explorers
concluded that children are creative searchers who of the Web to becoming discerning masters of it.
are more successful when they browse than when In discussing how usable Web interfaces are for
they search by keyword. Typically, children prefer children, Jacob Neilsen notes that existing Web
keyword searching but resort to browsing when they [interfaces] are based at best by insights gleaned from
experience continued information-retrieval prob- when designers observe their own children, who are
lems. Children do not take advantage of the search hardly representative of average kids, typical Internet
features provided in search engines and rarely acti- skills, or common knowledge about the Web
vate the help file for guidance. (Neilsen 2002, 1). Thus, it is not surprising to find
The research also revealed that children have both that children experience difficulty in using the Web.
positive and negative feelings when it comes to the System developers need to design interfaces that ad-
Web. They associate the Web with motivation, chal- dress childrens cognitive developmental level,
lenge, convenience, fun, authority, independence, information needs, and information-seeking be-
and self-control, but also with frustration, dissatis- haviors. Developing effective Web interfaces for chil-
faction, and disappointment caused by information dren requires a team effort involving information
overload, lack of success in searches, and inability to scientists, software engineers, graphic designers, and
make decisions about document relevance. educational psychologists, as well as the active par-
As to information literacy, children possess ticipation of representative children.
inadequate information-seeking skills, nave Web We have a growing understanding of the
navigational skills, and an insufficient conceptual un- strengths and weaknesses of the Web as a tool for
derstanding of the research process. These problems teaching and learning. We also know much about
cry out to teachers and information specialists to pro- childrens perceptions of and experiences with the
vide more effective Web training and to design in- Web, as well as their information-seeking behavior
structional strategies that successfully integrate the on the Web. The rapid progress made in these ar-
Web into effective learning. eas of study is commended.
With regard to system design, it appears that web- However, research gaps remain to be filled. We do
sites, Web directories, and search engines are not easy not have sufficient research on working with children
for children to use. Too much text, difficult vocabu- as partners in designing Web interfaces. We have in-
lary, long screen display, deep subject hierarchies, investigated childrens information-seeking behavior in
effective help files, poor indexing, and misleading formal settings, such as schools, to meet instructional
hyperlink titles all hinder childrens successful use. needs, but with the exception of Debra J. Slones 2002
study, we have little information on childrens Web be-
havior in informal situations, when they are using it to
Education, Design, and meet social or entertainment needs. We also lack a
sophisticated model that typifies childrens information-
Future Research seeking behavior.We need to develop a model that more
Use of the Web in school and its increased use at home fully represents this behavior so that we can predict suc-
does not ensure that children possess effective skills in cessful and unsuccessful outcomes, diagnose problems,
using it. Information professionals (e.g., school and and develop more effective solutions.
public librarians) who serve children need to collab-
orate with teachers to identify how the Web can ef- Dania Bilal
fectively support meaningful learning. Teachers cannot
make the Web an effective learning and research tool See also Classrooms; Graphical User Interface; Search
unless they first receive effective, structured training Engines
CLASSROOMS 97
Washington, DC: Pew Internet and American Life Project. Retrieved

FURTHER READING January 4, 2004, from http://www.pewinternet.org/reports/
pdfs/PIP_Schools_Report.pdf
Neilsen, J. (2002). Kids corner: Website usability for children. Retrieved
Bilal, D. (1998). Childrens search processes in using World Wide Web
Januar y 4, 2004, from http://www.useit.com/aler tbox/
search engines: An exploratory study. Proceedings of the 61st
20020414.html
ASIS Annual Meeting, 35, 4553.
Schacter, J., Chung, G. K. W. K., & Dorr, A. (1998). Childrens Internet
Bilal, D. (1999). Web search engines for children: A comparative study
searching on complex problems: Performance and process
and performance evaluation of Yahooligans!, Ask Jeeves for
analyses. Journal of the American Society for Information Science,
Kids, and Super Snooper. Proceedings of the 62nd ASIS Annual
49(9), 840849.
Meeting, 36, 7082.
Shenton, A. K., & Dixon, P. (2003). A comparison of youngsters use
Bilal, D. (2000). Childrens use of the Yahooligans! Web search engine,
of CD-ROM and the Internet as information resources. Journal of
I: Cognitive, physical, and affective behaviors on fact-based
the American Society for Information Science and Technology, 54(11),
tasks. Journal of the American Society for Information Science, 51(7),
10492003.
646665.
Slone, D. J. (2002). The influence of mental models and goals on search
patterns during Web interaction. Journal of the American Society
II: Cognitive and physical behaviors on research tasks. Journal of
for Information Science and Technology, 53(13), 11521169.
the American Society for Information Science and Technology, 52(2),
Wallace, R. M., Kupperman, J., and Krajcik, J. (2002). Science on the
118137.
Web: Students on-line in a sixth-grade classroom. The Journal of
the Learning Sciences, 9(1), 75104.
III: Cognitive and physical behaviors on fully self-generated tasks.
Watson, J. S. (1998). If you dont have it, you cant find it: A close look
Journal of the American Society for Information Science and
at students perceptions of using technology. Journal of the American
Technology, 53(13), 11701183.
Society for Information Science, 49(11), 10241036.
Bilal, D. (2003). Draw and tell: Children as designers of Web inter-
faces. Proceedings of the 66th ASIST Annual Meeting, 40,
135141.
CLASSROOMS
Bilal, D. (In press). Research on childrens use of the Web. In C. Cole
& M. Chelton (Eds.), Youth Information Seeking: Theories, Models,
and Approaches. Lanham, MD: Scarecrow Press.
Druin, A., Bederson, B., Hourcade, J. P., Sherman, L., Revelle, G., Platner,
M., et al. (2001). Designing a digital library for young children. In People have regarded electronic technology through-
Proceedings of the first ACM/IEEE-CS Joint Conference on Digital out its evolution as an instrument for improving
Libraries (pp. 398405). New York: ACM Press. learning in classrooms. Television and video were
Fidel, R., Davies, R. K., Douglass, M. H., Holder, J. K., Hopkins, C. J.,
Kushner, E. J., et al. (1999). A visit to the information mall: Web
examples of early electronic technology used in class-
searching behavior of high school students. Journal of the American rooms, and now personal computers have shown
Society for Information Science, 50(1), 2437. how electronic technology can enhance teaching
Hirsh, S. G. (1999). Childrens relevance criteria and information seek- and learning. Some people have lauded the new
ing on electronic resources. Journal of the American Society for
Information Science, 50(14), 12651283. kinds of learning activities afforded by electronic
Kafai, Y. B., & Bates, M. J. (1997). Internet Web-searching instruc- technology; but other people maintain that such
tion in the elementary classroom: Building a foundation for technology can be detrimental in classrooms. Despite
information literacy. School Library Media Quarterly, 25(2),
103111.
such criticisms, researchers in different fields
Large, A., & Beheshti, J. (2000). The Web as a classroom resource: education, computer science, human-computer
Reactions from the users. Journal of the American Society for interactioncontinue to explore how such tech-
Information Science And Technology, 51(12), 10691080. nology, paired with innovative curricula and teacher
Large, A., Beheshti J., & Moukdad, H. (1999). Information seeking on
the Web: Navigational skills of grade-six primary school students.
training, can improve classrooms.
Proceedings of the 62nd ASIS Annual Meeting, 36, 8497.
Large, A., Beheshti, J., & Rahman, R. (2002). Design criteria for chil-
drens Web portals: The users speak out. Journal of the American
Society for Information Science and Technology, 53(2), 7994. Early Visions of
Lenhart, A., Rainie, L., Oliver, L. (2003). Teenage life online: The rise of
the instant-message generation and the Internets impact on Learning Technologies
friendships and family relationships. Washington, DC: Pew Internet Early visions of how technology could be applied to
and American Life Project. Retrieved January 4, 2004, from http://
www.pewinternet.org/reports/pdfs/PIP_Teens_Report.pdf
learning included so-called behaviorist teaching ma-
Lenhart, A., Simon, M., & Graziano, M. (2001). The Internet and ed- chines inspired by the U.S. psychologist B. F. Skinner
ucation: Findings of the Pew Internet & American Life Project. in the 1960s. Skinner believed that classrooms
History Comes Alive in Cyberspace
OLD DEERFIELD, Mass. (ANS)On a blustery spring computer labs that are being established throughout
morning, 18 students from Frontier Regional High district schools. But as the trip to Old Deerfield demon-
School made their way down Main Street here in this colo- strated, students will also add to the pool of knowledge
nial village, jotting down notes on the Federal and Italianate and contribute original data to the web site as well.
architecture and even getting a look at an early 18th-century This is not just an electronic test book, said Tim
kitchen. Neumann, executive director of Pocumtuck Valley
But this was no ordinary field trip. The students Memorial Association and one of the projects designers.
were gathering information for an Internet-based project Students are not just surfing the web but actively en-
that is integrating state-of-the-art computer technology gaging with the text and images on the screen. Stu-
with the social studies curriculum throughout this rural dents also address questions posed by teachers and then
western Massachusetts school district. conduct research via the Internet, as well as other field
The project, titled Turns of the Centuries, focuses on studies, he said.
life at the turns of the last three centuries, beginning in Building the web sites, from teachers notes and class-
1700 and continuing through 1900. Its an unusual room lesson plans, are students and technicians at the
partnership between three distinct entitiesa secondary University of Massachusetts Center for Computer-Based
school, a university and a museum. Instructional Technology.
In the project, the primary sources of the Pocumtuck The students in Friday mornings expedition were re-
Valley Memorial Association, a nationally recognized mu- sponding to an assignment to choose a colonial family and
seum of frontier life in this region, will be available to stu- track them over time, using the resources at the museum.
dents through a web site that teachers, students and Those results will eventually be incorporated into the Turns
researchers are putting together. of the Centuries web site, where other students through-
Central to the project are the over 30,000 museum ar- out the K-12 district will be able to access them.
tifactsdiaries, letters and other primary sourcesmade In addition to helping acquaint students with emerg-
available to students through the developing web site. The ing technologies, the Turns of the Centuries project in-
marriage of technology with the museum archives has structs teachers how to teach social studies with a web-based
made possible new opportunities for inquiry-based ed- curriculum, and how to access these resources in their class-
ucation, which focuses on developing the student as active rooms, as well as exploring the potential partnerships
learner. among school and communities linked by the information
In essence, the educational project here is a cyberspace highway.
version of the museum, enabling students to access archives Robin Antepara
through the Internet, either from their homes or through Source: Students learn about history with classroom computers of to-
morrow. American News Service, June 17, 1999.
suffered from a lack of individual attention and that Incorrect responses would prevent advancement
individualized instruction would improve learning. to the next level of questions, giving students the op-
The idea was that individual students could use a portunity to consider how they could correct their
computer that would teach and test them on dif- responses. Software adopting this approach is fre-
ferent topics. Students would receive positive rein- quently called drill and practice software, but few
forcement from the computer through some reward examples of such software exist outside of educa-
mechanism (e.g., praise and advancement to the next tional games and other kinds of flash card pro-
level of instruction) if they gave correct responses. grams that teach skills such as spelling and arithmetic.
CLASSROOMS 99
A different vision is found in the work of concrete design information to guide software de-
Seymour Papert, an MIT professor who has explored velopers in developing and assessing effective soft-
technology in education since the 1960s, advocating ware for learning. Many software projects had little
learning theories proposed by the Swiss psycholo- real grounding in learning theories and the nature
gist Jean Piaget. Paperts vision uses computers of children. Thus, for every successful software
as tools that children use for exploratory and con- project, many others had educational content that
structive activities. Through these activities children was lacking and classroom use that was less than
create and shape their own understanding of con- successful. For instance, many software projects in-
cepts. Papert incorporated these ideas in the Logo volved the development of educational games
programming language, which was intended to (sometimes called edutainment software) whose
let children write programs to create computer graph- educational content was dubious and whose ini-
ics by exploring deeper concepts, such as the math- tial appeal to children soon wore off. Other ap-
ematical concepts needed to draw geographical proaches involved tools such as HyperCard, which
concepts. allowed laypeople to create software with the hope
A related vision came from Alan Kay, a renowned that educators could create software for their
computer science researcher, who proposed the students. However, although the idea of teacher-
Dynabook concept during the early 1970s. The created software was well intentioned and although
Dynabook was envisioned as a device similar to to- teachers have educational knowledge, they lack
days laptop computer that children could use in software design knowledge, again resulting in
information-rich and constructive activities. The few major successes.
Dynabook would have basic software core func- Other issues were contextual. Many early at-
tionality (using the Smalltalk computer language). tempts at educational computing were techno-
However, children could extend their Dynabooks centric and lacked a full understanding of the
functionality through Smalltalk programming. This support needed in classrooms. Inadequate training
would allow children to create new tools for creative for busy teachers to use electronic technology can
expression, information gathering, simulation, and become a large enough burden that teachers simply
so forth by learning not only programming but also bypass it. Furthermore, technology has been in-
the fundamentals of the underlying content (e.g., to troduced into the classroom without a good un-
create a music tool, students would need to learn derstanding by teachers (and sometimes by
musical concepts). researchers developing the technology) of how the
technology interacts with the classroom curriculum
and learning goals. Again, technology has little
Initial Attempts at impact if it is not a good fit with the activities that
teachers desire. Finally, schools have lacked adequate
Technology-Enhanced Classrooms technological resources to make full use of tech-
Each technological vision has brought promises of nology, so disparities in the number of computers
how technology can improve classrooms and learn- in classrooms and in network connectivity have hin-
ing. With the advent of personal computers, edu- dered effective use of technology.
cators rushed to place computers in classrooms with
the hope of implementing different visions of learn-
ing technologies. However, many initial attempts Designing
of technology-enhanced classrooms fell short of their
promise because of technological and contextual Learner-Centered Technology
issues in classrooms. Simply developing stand-alone technology for class-
One issue was that although people had some rooms is not enough. If effective technology-
ideas about what kinds of learning activities and enhanced classrooms are to become a reality, then
goals computers might support, people had little designers must design an overall learning system
that integrates three factors: technology, curricu- enough challenge that learners still work in the mind-
lum, and teacher support and development. During ful manner needed for real learning to occur.
the last ten to fifteen years designers have developed Teacher support and development are also key
many learning systems in the classroom by con- for technology-enhanced classrooms. Teacher sched-
sidering these three factors. Research in many ules and the classroom environment are busy, and
content areas, such as science education, is shedding introducing technology into classrooms can make
light on effective technology-enhanced classrooms. matters more complex for teachers. Teachers need
In such educational approaches technology acts as support and development to show them how tech-
a cognitive tool to support learners as they engage nology works, how they can integrate technology
in curricular activities. For example, many educa- into their classroom activities, and how they can use
tional approaches in science education use an technology effectively in the classroom.
inquiry-based technique in which students engage
in the same kinds of scientific activityfinding sci-
entific articles, gathering and visualizing scientific New Visions of
data, building scientific models, and so forthin
which experts engage. For example, learners can use Technology-Enhanced Classrooms
software to search digital libraries for informa- Current classroom technology includes primarily
tion, use handheld computers with probes to gather desktop-based software. Some software implements
data in different locations, use software to build scaffolding features that support learners by ad-
graphs and scientific models, and so forth. Such tech- dressing the difficulties they encounter in their learn-
nology should be designed to support learners in ing activities. For example, one particular software
mindfully engaging in curricular activities so that feature implementing a scaffolding approach would
learners can meet the learning goals that their teach- be a visual process map that displays the space of ac-
ers have outlined. tivities that learners should perform (e.g., the activ-
Given this motivation, the approach for design- ities in a science investigation) in a way that helps
ing learne-rcentered technologies shifts from sim- them understand the structure of the activities. Other
ply designing technologies whose hallmark is ease classroom software includes intelligent tutoring sys-
of use to designing technologies that learners can tems that oversee students as they engage in new ac-
use in new, information-rich activities. Developing tivity. Intelligent tutoring systems can sense when
learner-centered technologies requires designers to students have encountered problems or are working
understand the kinds of work that learners should incorrectly and can provide just-in-time advice to
engage in (i.e., curricular activities) and the learn- help them see their errors and understand their tasks.
ing goals of those learners. Then designers need to Aside from traditional desktop-based software,
understand the areas where learners may face dif- researchers are exploring new technology. For ex-
ficulties in performing those kinds of work (e.g., ample, handheld computers (e.g., Palm or PocketPC
learners may not know what kinds of activities com- computers) are becoming more pervasive among
prise a science investigation or how to do those ac- students. The mobility of handheld computers lets
tivities) so that designers can create support features students take them to any learning context, not
that address those difficulties. Furthermore, such just the classroom. Thus, researchers are exploring
support features differ from usability-oriented tra- how to develop learning tools for handheld com-
ditional software design. Although ease of use is still puters. An example of such tools is probes that can
important, learner-centered technologies should not be attached to handhelds for scientific data gather-
necessarily make tasks as easy as possible. Rather, just ing (e.g., probes to measure oxygen levels in a
as a good teacher guides students toward an an- stream). Handhelds with wireless networking capa-
swer without giving the answer outright, learner- bility can be used to gather information (e.g., access
centered technologies must provide enough support digital libraries) from a range of locations outside of
to make tasks accessible to novice learners but leave a classroom. Additionally, handhelds can be part of
CLASSROOMS 101
A Personal StoryLearning through Multimedia
When I talk about why I became interested in exploring computers in education, I like to tell a story from my early grad-
uate school days in the late 1990s. My research group was working in a local Michigan high school using the MediaText
software they had developed earlier. MediaText was a simple text editor that made it possible to incorporate different
media objects, such as images, sounds, or video, into the text one was writing.
In one class, students had been given an assignment to explain a series of physics terms. One particular student some-
times had difficulty writing, but with MediaText, she could use other media types for her explanations. For example,
using video clips from the movie Who Framed Roger Rabbit? she explained potential energy with a clip of a cartoon baby
sitting on top of a stack of cups and saucers, swaying precariously without falling over. Then she explained kinetic ener-
gy with a clip of the same baby sliding across the floor of a room.
What struck me was that it was clear from her choice of video clips that she understood those physics concepts. If she
had been confined to textual explanations, she might not have been able to convey as much understanding. But because
she had a choice of media types, she was able to successfully convey that she knew those concepts.
This episode helped me realize how computers could impact learners by giving them a range of different media types
for self-expression. Now sometimes this story gets me in trouble with people who say that if you give students all these
alternatives to writing, theyll never learn to write correctly. Ill buy thatto a certain extent. But people are diverse
they learn differently and they express themselves differently. My response to the naysayers is that effectively incorporat-
ing different media in software tools isnt for flash, but to give people different languages to learn from and use. By
offering these alternatives, we open new educational doors, especially for todays diverse, tech-savvy kids. After all, if one
student can explain physics terms using a movie about a cartoon rabbit, then multimedia in the classroom is working.
Chris Quintana
new kinds of learning activities called participatory Meeting the Challenge

simulations in which groups of students can use the Technology-enhanced classrooms have had fail-
beaming feature of wireless handhelds to be part ures as researchers have struggled to understand not
of a simulation in which they exchange information. only the kinds of effective learning technologies, but
For example, students can explore epidemiological also the role of technology in classrooms and the
crises in a simulation in which they meet other support needed for effective technology-enhanced
people by exchanging information with their hand- classrooms. Critics of technology in classrooms still
helds. During the simulation a students handheld exist. Education professor Larry Cuban has written
might announce that it is sick, at which point extensively on the problems and failures of technol-
students would engage in a discussion to understand ogy in classrooms. Scientist and author Clifford Stoll
how disease might spread through a community. has also written about the possible adverse effects of
Researchers are also exploring the other end of technology and the caution that must be taken for
the spectrum, looking at how large displays and vir- children.
tual reality can be used as learning tools. Such However, successes and new visions of how
tools can help students explore virtual worlds and technology-enhanced classrooms can support learn-
engage in activities such as virtual expeditions. ers also exist. Designers of learning technologies need
Students can explore environments that might be to understand that designing software for ease of use
difficult or impossible to explore in person (e.g., dif- is not enough. Designers must understand learning
ferent ecosystems), thus allowing them to engage in theories, the nature of learners, and the classroom
inquiry-based activities throughout a range of loca- context to design cognitive learning technologies that
tions and gather otherwise inaccessible information. students use to mindfully engage in substantive
learning activities. People implementing technology- puter (PC) explosion, client-server architecture pro-
enhanced classrooms need to consider other issues, vides a distributed synthesis of the highly interactive
such as classroom curriculum augmenting technol- personal computer (the client) with a remotely
ogy and teachers having the support and develop- located computer providing data storage and com-
ment that they need in order to understand and make putation (the server). The goal of client-server archi-
full use of technology. tecture is to create structures and communication
As new technology arises, people will always protocols between the client computer and the server
attempt to see how that technology can be used to computer in order to optimize the access to a set of
enhance learning. By understanding the classroom computational resources.
context and the local design issues involved in de-
veloping learner-centered technology, the human-
computer interaction community can make Motivating Example
significant contributions to realizing the promise of To understand client-server architecture, one can
technology-enhanced classrooms. consider a manufacturing company using computer
technology to support day-to-day business opera-
Chris Quintana tions and long-range strategic planning. Product
orders come from the sales department, inventory is
See also Children and the Internet; Psychology maintained by the manufacturing department,
and HCI and the raw materials orders are generated by the
planning department. Furthermore, the accounting
department tracks the money, and the chief execu-
FURTHER READING tive officer (CEO) wants a perspective on all aspects
of the company.
Bransford, J. D., Brown, A. L., & Cocking, R. R. (Eds.). (2000). How To be judged successful, the software solution
people learn: Brain, mind, experience, and school (Exp. ed.).
Washington, DC: National Academy Press.
implemented should provide data storage and up-
Cuban, L. (1986). Teachers and machines: The classroom use of tech- date capability for all aspects of the company oper-
nology since 1920. New York: Teachers College Press. ation. Further, the appropriate database segments
Kay, A., & Goldberg, A. (1977). Personal dynamic media. IEEE should be accessible by all of the employees based
Computer, 10(3), 3141.
Papert, S. (1980). Mindstorms. New York: Basic Books. on their particular job responsibility, regardless of
Quintana, C., Soloway, E., & Krajcik, J. (2003). Issues and approaches where they are physically located. Finally, the ap-
for developing learner-centered technology. In M. Zelkowitz (Ed.), plication views of the database should be highly
Advances in computers: Volume 57. Information Repositories
(pp. 272323). New York: Academic Press.
usable, interactive, and easy to build and update to
Reiser, B. J. (2002). Why scaffolding should sometimes make tasks reflect ongoing business growth and development.
more difficult for learners. Proceedings of CSCL 2002, 255264.
Soloway, E., Guzdial, M., & Hay, K. E. (1994). Learner-centered de-
sign: The challenge for HCI in the 21st century. Interactions,
1(2), 3648.
Conicting Goals
One key feature of any software application is the
database, the dynamic state of the application. For
example, the status of inventory and orders for a fac-
tory would be maintained in a database management
CLIENT-SERVER system (DBMS). Modern database management tech-
nology is quite well developed, supporting database
ARCHITECTURE lookup and update in a secure, high performance
fashion. DBMS computers, therefore, are typically
Client-server architecture is one of the many ways high-performance, focused on the task, and have
to structure networked computer software. large permanent storage capacity (disk) and large
Developed during the 1980s out of the personal com- working memory. The cost of this hardware, the crit-
CLIENT-SERVER ARCHITECTURE 103
ical need for consistency, and the complexity of sys- Common Object Request Broker Architecture
tem management dictate that the DBMS be centrally (CORBA) or the Component Object Model (COM).
located and administered. This goal was realized in Returning to our motivating example, the soft-
the mainframe architecture of the 1960s and the ware solution would include a separate interactive
time-sharing architecture of the 1970s. PC application designed for each business function:
On the other hand, personal computer appli- sales manufacturing, accounting, planning, and
cations such as the spreadsheet program VisiCalc, the CEO. Each of these individual PC applications
introduced in 1979, demonstrate the power of highly would use an RPC call for each query or update
interactive human-computer interfaces. Responding operation to the company database server. This par-
instantly to a users every keystroke and displaying titioning of function is effective both in terms of
results using graphics as well as text, the PC has hardware cost performance (relatively inexpensive
widened the scope and number of users whose pro- client computers for each user versus a relatively ex-
ductivity would be enhanced by access to comput- pensive database server computer shared between
ing. These inexpensive computers bring processing all users) and end-user application design.
directly to the users but do not provide the same scal- As the number of simultaneous users grows,
able, high-performance data-storage capability of the portion of a servers computation time spent man-
the DBMS. Furthermore, the goal of information aging client-server sessions grows as well. To mitigate
management and security is counter to the personal this processing overhead, it is useful to introduce
computer architecture, in which each user operates an intermediary server to help handle the client-server
on a local copy of the database. requests. Called a message queuing server, this soft-
The networkthe tie that binds together the ware system accepts operations to be performed on
DBMS and the human-computer interfacehas the database and manages the request queues asyn-
evolved from proprietary system networks, such as chronously. Priority information allows intelligent
IBM System Network Architecture (SNA), intro- management and scheduling of the operations. Result
duced in 1974, to local area networks, such as queues, returning answers back to the requesting
Ethernet, developed at Xeroxs Palo Alto Research client, provide for asynchronous delivery in the other
Center (PARC) and introduced in 1976, to the direction as well. Through a message server the queu-
Internet, which began as the U.S. Department of ing operations are offloaded from the database server,
Defenses Advanced Research Projects Agency net- providing enhanced throughput (output). The mes-
work (Arpanet) in 1972 and continues to evolve. sage server also leads to increased flexibility be-
A networking infrastructure allows client software, cause the message queuing provides a layer of
operating on a PC, to make requests of the server for translation and independence between the client soft-
operations on the users behalf. In other words, the ware and the DBMS server.
network provides for the best of both worlds:
high-performance, high-reliability components pro-
viding centralized data computation and user in- Business Processes
terface components located on the personal computer Although PC-client access to a server-based DBMS was
providing high interactivity and thereby enhanced an early client-server scenario and continues to be im-
usability. portant, other client-server architectures include other
Furthermore, by defining standardized message- types of network services. For example, an application
passing protocols for expressing the requests from server hosts computation rather than data storage, as
client to server, a level of interoperability is achieved. with a DBMS server. The business processes for an en-
Clients and servers coming from different vendors terprise may be implemented using an application
or implementing different applications may com- server. Like the message queuing server, the application
municate effectively using protocols such as Remote server sits between the client software and the DBMS,
Procedure Call (RPC) or Standard Query Language encapsulating functions that may be common across
(SQL), together with binding services such as the many clients, such as policies and procedures.
The Future of based system. It is crucial to design systems for

ease of learning, because people generally learn to
Client-Server Computing use new computer-based systems by exploration.
Client-server computing will continue to be impor- People resort to reading manuals, using help systems,
tant long into the future. PCs continue to drop in price, or taking formal training only when they have
and new networked devices such as personal data as- been unsuccessful in learning to do their tasks by ex-
sistants (PDAs) and the World Wide Web are driving ploration. CW has been applied to a wide variety
network accessibility to a broader audience. The client- of systems, including automatic teller machines
server architecture, which lives on the network through (ATMs), telephone message and call forwarding sys-
standardized messaging protocols, will continue to tems, websites, computerized patient-record systems
have wide applicability, especially in business. for physicians, programming languages, multimedia
authoring tools, and computer-supported coopera-
Mark R. Laff tive work systems. HCI researcher Andrew J. Ko and
his associates innovatively applied CW (in lieu of pi-
See also Peer-to-Peer Architecture lot experiments) to predict problems that experi-
mental participants might have with the instructions,
procedures, materials, and interfaces used in ex-
FURTHER READING periments for testing the usability of a system (the
system was a visual programming language).
Berson, A. (1992). Client/server architecture. New York: McGraw-Hill.
Berson, A. (1995). Sybase and client/server computing. New York:
McGraw-Hill.
Comer, D. (1994). Internetworking with TCP/IP: Vol. 3. Client-server Cognitive Walkthrough
programming and applications. Englewood Cliffs, NJ: Prentice Hall.
C o r b i n , J. R . ( 1 9 9 1 ) . T h e a r t o f d i s t r i b u t e d a p p l i c a t i o n s : Methodology
Programming techniques for remote procedure calls. New York: The CW approach was invented in 1990 and has
Springer-Verlag.
Edelstein, H. (1994). Unraveling client/server architecture. Redwood
evolved into a cluster of similar methods with the
City, CA: M & T Publishing. following four defining features:
Hall, C. (1994). Technical foundations of client/server systems. New
York: Wiley. 1. The evaluation centers on particular users and
IBM Corporation. (2002). Websphere MQ application message inter- their key tasks. Evaluators start a CW by care-
face. (SC34-6065-00). Armonk, NY: International Business fully analyzing the distinctive characteristics of
Machines Corporation.
Krantz, S. R. (1995). Real world client server: Learn how to success-
a particular user group, especially the relevant
fully migrate to client/server computing from someone whos actu- kinds of background knowledge these users can
ally done it. Gulf Breeze, FL: Maximum Press. call upon when learning to perform tasks on
Metcalfe, R. M., & Boggs, D. R. (1976). Ethernet: Distributed packet the system. Next, CW evaluators select a set of
switching for local computer networks. Communications of the
ACM, 19(5), 395404.
key tasks that members of the user group will
Sims, O. (1994). Business objects: Delivering cooperative objects for client- do on the system. Key tasks are tasks users do
server. New York: McGraw-Hill. frequently, tasks that are critical even if done
infrequently, and tasks that exhibit the core ca-
pabilities of the system.
2. The steps designers prescribe for doing tasks are
COGNITIVE evaluated. For each key task, CW evaluators
record the full sequence of actions necessary to
WALKTHROUGH do the task on the current version of the sys-
tem. Then CW evaluators walk through the
The cognitive walkthrough (CW) is a usability eval- steps, simulating users action selections and
uation approach that predicts how easy it will be for mental processes while doing the task. The sim-
people to learn to do particular tasks on a computer- plest CW version asks two questions at each
COGNITIVE WALKTHROUGH 105
step: (1) Is it likely that these particular users designers can readily learn CW, but they have
will take the right actionmeaning the ac- a shallower grasp of the underlying theory than
tion designers expect them to takeat this step? usability experts trained in cognitive psychol-
and (2) If these particular users do the right ogy and consequently find less than half as many
action and get the feedback the system pro- usability problems. A group CW, including at
vides (if any), will they know they made a good least one usability expert trained in cognitive
choice and realize that their action brought psychology, can find a higher percentage of us-
them closer to accomplishing their goal? To an- ability problems than an individual evaluator
swer each question evaluators tell a believable up to 50 percent of the problems that appear in
success story or failure story. They record fail- usability tests of the system.
ure stories and have the option of adding sug-
CW was one of the several evaluation methods
gestions for how to repair the problems and
pioneered in the early 1990s to meet a practical need,
turn failures into successes. Anchoring the eval-
the need to identify and repair usability problems
uation to the steps specified by designers com-
early and repeatedly during the product develop-
municates feedback to designers in their own
ment cycle. The cost of repairing usability problems
terms, facilitating design modifications that re-
rises steeply as software engineers invest more time
pair the usability problems.
in building the actual system, so it is important to
3. Evaluators use theory-based, empirically verified
catch and fix problems as early as possible. For a
predictions. The foundation for CW is a theory
product nearing completion the best evaluation
of learning by exploration that is supported by
method is usability testing with end users (the people
extensive research done from the 1960s to the
who will actually use the system), but CW is ap-
1980s on how people attempt to solve novel
propriate whenever it is not possible to do usabil-
problems when they lack expert knowledge
ity testing. Early versions of CW were tedious to
or specific training. According to this theory,
perform, but the 1992 cognitive jogthrough and
learning to do tasks on a computer-based sys-
streamlined CW of 2000, which still preserve all
tem requires people to solve novel problems by
the essential CW features, are much quicker to
using general problem-solving methods, gen-
perform.
eral reading knowledge, and accumulated ex-
perience with computers. The key idea is
that correct actions are chosen based on their Transforming CW to Faster and
perceived similarity to the users current goal
(Wharton et al. 1994, 126). For software appli- More Accurately Predict User Actions
cations, the theory predicts that a user scans The cognitive walkthrough for the Web (CWW) has
available menu item labels on the computer transformed the CW approach by relying on Latent
screen and picks the menu item label that is Semantic Analysis (LSA)instead of on the subjec-
most similar in meaning to the users current tive judgments of usability experts and software en-
goal. CW evaluators answer the first question gineersto predict whether users are likely to select
with a success story if the right action desig- the right action. LSA is a computer software sys-
nated by the designer is highly similar in mean- tem that objectively measures semantic similarity
ing to the users goal and if the menu item labels similarity in meaningbetween any two passages of
on the screen use words familiar to the user. text. LSA also assesses how familiar words and phrases
4. Software engineers can easily learn how to are for particular user groups.
make CW evaluations. It is crucial to involve While analyzing the distinctive characteristics of
software engineers and designers in CW, be- the particular user group, CWW evaluators choose
cause they are the individuals responsible for the LSA semantic space that best represents the back-
revising the design to repair the problems. There ground knowledge of the particular user group
is strong evidence that software engineers and the space built from documents that these users
are likely to have read. For example, CWW currently about the experimental design and statistics of these
offers a college-level space for French and five spaces studies.
that accurately represent general reading knowledge Relying on LSA opens the door to fully auto-
for English at college level and at third-, sixth-, mating CWW and increasing its cost-effectiveness.
ninth-, and twelfth-grade levels. If other CW methods start to rely on LSA they,
CWW uses LSA to measure the semantic simi- too, could be automated. The streamlined CW is
larity between a users information search goal more efficient than earlier CW methods, but it still
(described in 100 to 200 words) and the text labels consumes the time of multiple analysts and relies on
for each and every subregion of the web page and subjective judgments of uncertain accuracy.
for each and every link appearing on a web page.
CWW then ranks all the subregions and link labels
in order of decreasing similarity to the users goal. Objectively Predicting Actions for
CWW predicts success if the right action is the
highest-ranking link, if that link is nested within the Diverse Users
highest-ranking subregion, and if the right action Relying on LSA makes it possible for CWW to do
link label and subregion avoid using words that are something that even usability experts trained in
liable to be unfamiliar to members of the user group. cognitive psychology can almost never do: objec-
Relying on LSA produces the same objective an- tively predict action selections for user groups whose
swer every time, and laboratory experiments con- background knowledge is very different from
firm that actual users almost always encounter serious the background knowledge of the human evaluators.
problems whenever CWW predicts that users will For example, selecting the sixth-grade semantic space
have problems doing a particular task. Furthermore, enables LSA to think like a sixth grader, because
using CWW to repair the problems produces two- the sixth-grade LSA semantic space contains only
to-one gains in user performance. So far, CWW re- documents likely to have been read by people who
searchers have tested predictions and repairs only have a sixth-grade education. In contrast, a college-
for users with college-level reading knowledge of educated analyst cannot forget the words, skills, and
English, but they expect to prove that CWW gives technical terms learned since sixth grade and can-
comparably accurate predictions for other user not, therefore, think like a sixth grader.
groups and semantic spaces. Since CW was invented in 1990, the number and
diversity of people using computers and the Internet
have multiplied rapidly. Relying on LSA will enable
APPLICATION A software program that performs a ma- the CW approach to keep pace with these changes.
jor computing function (such as word processing or Web In cases where none of the existing LSA semantic
browsing). spaces offers a close match with the background
knowledge of the target user group, new semantic
spaces can be constructed for CWW (and potentially
Research by cognitive psychologist Rodolfo Soto for CW) analysesin any language at any level of
suggests that CW evaluations of software applica- ability in that language. Specialized semantic spaces
tions would be improved by relying on LSA, but to can also be created for bilingual and ethnic minor-
date CW has consistently relied on subjective judg- ity user groups and user groups with advanced back-
ments of human evaluators. Consequently the agree- ground knowledge in a specific domain, such as
ment between any two CW evaluators is typically the domain of medicine for evaluating systems used
low, raising concerns about the accuracy of CW pre- by health professionals.
dictions. Many studies have tried to assess the accu-
racy and cost-effectiveness of CW compared to Marilyn Hughes Blackmon
usability testing and other evaluation methods. The
results are inconclusive, because there is controversy See also Errors in Interactive Behavior; User Modeling
COLLABORATORIES 107
Spencer, R. (2000). The streamlined cognitive walkthrough method,

FURTHER READING working around social constraints encountered in a software de-
velopment company. In CHI 2000: Proceedings of the Conference
on Human Factors in Computing Systems, 353359.
Blackmon, M. H., Kitajima, M., & Polson, P. G. (2003). Repairing
Wharton, C., Rieman, J., Lewis, C., & Polson, P. (1994). The cognitive
usability problems identified by the cognitive walkthrough for the
walkthrough method: A practitioners guide. In J. Nielsen &
web. In CHI 2003: Proceedings of the Conference on Human Factors
R. L. Mack (Eds.), Usability inspection methods (pp. 105140). New
in Computing Systems, 497504.
York: Wiley.
Blackmon, M. H., Polson, P. G., Kitajima, M., & Lewis, C. (2002).
Cognitive walkthrough for the Web. In CHI 2002: Proceedings of
the Conference on Human Factors in Computing Systems, 463470.
Desurvire, H. W. (1994). Faster, cheaper!! Are usability inspection
COLLABORATIVE
methods as effective as empirical testing? In J. Nielsen & R. L. Mack
(Eds.), Usability inspection methods (pp. 173202). New York: Wiley.
Gray, W. D., & Salzman, M. D. (1998). Damaged merchandise? A re-
view of experiments that compare usability evaluation methods.
Human-Computer Interaction, 13(3), 203261. INTERFACE
Hertzum, M., & Jacobsen, N. E. (2003). The evaluator effect: A chill-
ing fact about usability evaluation methods. International Journal See Multiuser Interfaces
of Human Computer Interaction, 15(1), 183204.
John B. E., & Marks, S. J. (1997). Tracking the effectiveness of us-
ability evaluation methods. Behaviour & Information Technology,
16(4/5), 188202.
John, B. E., & Mashyna, M. M. (1997). Evaluating a multimedia au-
thoring tool. Journal of the American Society for Information Science,
COLLABORATORIES
48(11), 10041022.
Ko, A. J., Burnett, M. M., Green, T. R. G., Rothermel, K. J., & Cook, A collaboratory is a geographically dispersed or-
C. R. (2002). Improving the design of visual programming lan- ganization that brings together scientists, instru-
guage experiments using cognitive walkthroughs. Journal of Visual
Languages and Computing, 13, 517544.
mentation, and data to facilitate scientific research.
Kushniruk, A. W., Kaufman, D. R., Patel, V. L., Lvesque, Y., & In particular, it supports rich and recurring hu-
Lottin, P. (1996). Assessment of a computerized patient record sys- man interaction oriented to a common research area
tem: A cognitive approach to evaluating medical technology. M D and provides access to the data sources, artifacts, and
Computing, 13(5), 406415.
Lewis, C., Polson, P., Wharton, C., & Rieman, J. (1990). Testing a walk-
tools required to accomplish research tasks. Collab-
through methodology for theory-based design of walk-up-and- oratories have been made possible by new com-
use interfaces. In CHI 90: Proceedings of the Conference on Human munication and computational tools that enable
Factors in Computing Systems, 235242. more flexible and ambitious collaborations. Such
Lewis, C., & Wharton, C. (1997). Cognitive walkthroughs. In
M. Helander, T. K. Landauer, & P. Prabhu (Eds.), Handbook of collaborations are increasingly necessary. As science
human-computer interaction (2nd ed., revised, pp. 717732). progresses, the unsolved problems become more
Amsterdam: Elsevier. complex, the need for expensive instrumentation in-
Pinelle, D., & Gutwin, C. (2002). Groupware walkthrough: Adding
context to groupware usability evaluation. In CHI 2002: Proceedings
creases, larger data sets are required, and a wider
of the Conference on Human Factors in Computing Systems, 455462. range of expertise is needed. For instance, in high-
Polson, P., Lewis, C., Rieman, J., & Wharton, C. (1992). Cognitive walk- energy physics, the next generation of accelerators
throughs: A method for theory-based evaluation of user interfaces. will require vast international collaborations and will
International Journal of Man-Machine Studies, 36, 741773.
Rowley, D. E., & Rhoades, D. G. (1992). The cognitive jogthrough: A
have a collaboratory model for remote access. At least
fast-paced user interface evaluation procedure. In CHI 92: 150 collaboratories representing almost all areas of
Proceedings of the Conference on Human Factors in Computing science have appeared since the mid-1980s.
Systems, 389395. Collaboratories offer their participants a num-
Sears, A., & Hess, D. J. (1999). Cognitive walkthroughs: Understanding
the effect of task description detail on evaluator performance. ber of different capabilities that fall into five broad
International Journal of Human-Computer Interaction, 11(3), categories: communication (including tools such as
185200. audio or video conferencing, chat, or instant mes-
Soto, R. (1999). Learning and performing by exploration: Label qual-
ity measured by Latent Semantic Analysis. In CHI 99: Proceedings
saging), coordination (including tools relating to
of the Conference on Human Factors and Computing Systems, access rights, group calendaring, and project man-
418425. agement), information access (including tools for
accessing online databases, digital libraries, and doc- As the frontiers of science are pushed back, the in-
ument repositories), computational access (includ- strumentation required for advances becomes more
ing access to supercomputers), and facility access and more esoteric, and therefore usually more and
(including tools for remotely accessing specialized more expensive. Alternatively, certain scientific in-
facilities or instruments, such as a particle acceler- vestigations require instrumentation in specific geo-
ator or a high-powered microscope). graphic settings, such as an isolated or inhospitable
Research on collaboratories has focused mostly area. A typical example is the Keck Observatory,
on solving technical problems. However, substantial which provides access to an astronomical observa-
gains in the practice of science are likely to be the tory on the summit of Mauna Kea in Hawaii to a
combined effect of social and technical transforma- consortium of California universities.
tions. The gap between the raw performance capa-
bility of collaboratory tools (based on bandwidth, Community Data System
storage capacity, processor speed, and so forth) An especially common collaboratory type is one in
and the realized performance (usage for scientific which a geographically dispersed community agrees
purposes, which is limited by factors such as usabil- to share their data through a federated or central-
ity and fit to the work and culture) can limit the po- ized repository. The goal is to create a more power-
tential of collaboratories. This point will be discussed ful data set on which more sophisticated or powerful
in greater detail later. analyses can be done than would be possible if the
parts of the data set were kept separately. A typical
example of a community data system is the Zebrafish
Types of Collaboratories: Information Network (ZFIN), an online aggregation
of genetic, anatomical, and methodological infor-
Research-Focused Collaboratories mation for zebra fish researchers.
There are a number of different kinds of collabora-
tories. A collaboratory that satisfies all elements of the Open-Community Contribution System
definition given above is a prototypical collaboratory Open-community contribution systems are an emerg-
a distributed research center. Other kinds of collabo- ing organizational type known as a voluntary asso-
ratories are missing one or more of the elements of ciation. Interested members of a community (usually
that definition. The following four types of collabo- defined quite broadly) are able to make small con-
ratories focus on enabling geographically distributed tributions (the business scholar Lee Sproull calls them
research. microcontributions) to some larger enterprise. These
contributions are judged by a central approval or-
Distributed Research Center ganization and placed into a growing repository. The
This type of collaboratory functions like a full-fledged classic example is open-source software development,
research center or laboratory, but its users are geo- which involves hundreds or even thousands of con-
graphically dispersedthat is, they are not located tributors offering bug fixes or feature extensions to a
at the research center. It has a specific area of inter- software system. In science, such schemes are used to
est and a general mission, with a number of specific gather data from a large number of contributors. Two
projects. A good example of a distributed research examples will help illustrate this. The NASA Ames
center is the Alliance for Cellular Signaling, a large, Clickworkers project invited members of the public
complex distributed organization of universities to help with the identification of craters on images
whose goal is to understand how cells communicate from a Viking mission to Mars. They received 1.9 mil-
with one another to make an organism work. lion crater markings from over 85,000 contribu-
tors, and the averaged results of these community
Shared Instrument contributions were equivalent in quality to those of
A shared-instrument collaboratory provides access expert geologists. A second example is MITs Open
to specialized or geographically remote facilities. Mind Common Sense Initiative, which is collecting
COLLABORATORIES 109
examples of commonsense knowledge from mem- collaboratory is successful. What follow are some of
bers of the public to help make computers smarter the most important factors.
(Singh n.d.).
Readiness for Collaboration
Participants must be ready and willing to collaborate.
Types of Collaboratories: Science is by its very nature a delicate balance of
cooperation and competition. Successful collabora-
Practice-Focused Collaboratories tions require cooperation, but collaboration is very
The next two collaboratory types support the pro- difficult and requires extra effort and motivation.
fessional practice of science more broadly, as op- Technologies that support collaboration will not be
posed to supporting the conduct of research itself. used if the participants are not ready or willing to col-
laborate. Various fields or user communities have
Virtual Community of Practice quite different traditions of sharing. For instance,
This is a network of individuals who share a research upper-atmospheric physicists have had a long tra-
area of interest and seek to share news of profes- dition of collaboration; the Upper Atmospheric
sional interest, advice, job opportunities, practical Research Collaboratory (UARC) began with a col-
tips on methods, and the like. A good example of laborative set of users. On the other hand, several ef-
this kind of collaboratory is Ocean US, which forts to build collaboratories for biomedical research
supports a broad community of researchers inter- communities (for instance, for researchers studying
ested in ocean observations. A listserv is another HIV/AIDS or depression) have had difficulty in part
mechanism that is used to support a virtual com- because of the competitive atmosphere. Readiness
munity of practice, but much more common for collaboration can be an especially important fac-
these days are websites and wikis. tor when the collaboratory initiative comes from
an external source, such as a funding agency.
Virtual Learning Community
This type of collaboratory focuses on learning that Technical Readiness
is relevant to research, but not research itself. A good The participants, the supporting infrastructure, and
example is the Ecological Circuitry Collaboratory, the design of the tools must be at a threshold tech-
whose goal is to train doctoral students in ecology nical level. Some communities are sufficiently col-
in quantitative-modeling methods. laborative to be good candidates for a successful
collaboratory, but their experience with collaborative
technologies or the supporting infrastructure is not
Evolution and sufficient. Technical readiness can be of three kinds.
Success of Collaboratories INDIVIDUAL TECHNICAL READINESS People in various

Collaboratories that last more than a year or two tend organizations or fields have different levels of ex-
to evolve. For example, a collaboratory may start perience with collaboration tools. A specific new
as a shared-instrument collaboratory. Those who technology such as application sharing may be a leap
share the instrument may add a shared database com- for some and an easy step for others. It is impor-
ponent to it, moving the collaboratory toward a tant to take account of users specific experience when
community data system. Then users may add com- introducing new tools.
munication and collaboration tools so they can plan
experiments or data analyses, making the collabo- INFRASTRUCTURE READINESS Collaborative technolo-
ratory more like a distributed research center. gies require good infrastructure, both technical and
Some collaboratories are quite successful, while social. Poor networks, incompatible workstations, or
others do not seem to work very well. There are a a lack of control over different versions of software
number of factors that influence whether or not a can cause major problems. It is also very important
to have good technical support personnel, especially and also created for the community a spirit of gen-
in the early phases of a collaboratory. The Worm erosity and collaboration. Although goodwill among
Community System (WCS) was a very early collab- the community of researchers has been a sufficient
oratory project, intended to support a community of incentive for participation, ZFIN is now expanding
researchers who studied the organism c. elegans (a its participation beyond its founders, and it will be in-
type of nematode). Sophisticated software was de- teresting to see how successful the goodwill incentive
veloped for the WCS on a UNIX platform that was is in the context of the expanded community.
not commonly used in the laboratories of the sci-
entists. Since the tools were thus not integrated GOODWILL PLUS KARMA POINTS Slashdot is a very large
with everyday practice, they were seldom used. and active community of open-source software
Furthermore, the necessary technical support was not developers who share and discuss news. Slashdot re-
generally present in the lab, so when there were prob- wards those who make the most informative con-
lems, they were showstoppers. tributions by bringing them more into the center of
attention and allocating them karma points. Karma
SOCIAL ERGONOMICS OF TOOLS The social interac- points are allocated in accordance with how highly
tions that take place in teams are affected both by the a contributors postings are rated by others. These
characteristics of team members and by the tools karma points give contributors some additional priv-
that are used. The study of the impact of technology ileges on the site, but their main value is as a tangi-
characteristics on this process may be called social ble measure of community participation and status.
ergonomics (ergonomics is the application of knowl- Karma points are a formalization of goodwill, valu-
edge about humans to the design of things). For ex- able primarily because the members of the com-
ample, video conferencing systems often ignore such munity value them as an indicator of the quality of
details as screen size, display arrangement in relation the sharing done by specific individuals.
to participants, camera angle, and sound volume.
But it turns out that these details can have social REQUIRING CONTRIBUTION AS A PREREQUISITE FOR OTHER
effects. For example, a study conducted by the re- ACTIVITY In order to get the details of gene sequences
searchers Wei Huang, Judith Olson, and Gary Olson out of published articles in journals, a consortium
found that the apparent height of videoconference of high-prestige journals in biology requires that
participants, as conveyed via camera angle, influ- those who submit articles to the consortiums
enced a negotiation task. The apparently taller per- journals have a GenBank accession number indi-
son was more influential in shaping the final outcome cating that they have stored their gene sequences
than the apparently shorter person. in the shared database.
Aligned Incentives NEW FORMS OF PUBLICATION The Alliance for Cellular

Aligning individual and organizational incentives is Signaling has taken a novel approach to providing
an important element of successful collaborations. researchers with an incentive to contribute molecule
Consider the incentives to participate in a commu- pages to the Alliances database. Because the mole-
nity data system: What motivates a researcher to con- cule pages represent a lot of work, the Alliance has
tribute data to a shared database? By contributing, worked out an agreement with Nature, one of the
the researcher gives up exclusive access to the data high-prestige journals in the field, to count a mol-
he or she has collected. There are a variety of incentive ecule page as a publication in Nature. Nature
schemes for encouraging researchers to collaborate. coordinates the peer reviews, and although mole-
cule-page reviews do not appear in print, the mol-
GOODWILL ZFIN has relied on the goodwill of its mem- ecule pages are published online and carry the
bers. Most of the members of this community had a prestige of the Nature Publishing Group. The
connection to one specific senior researcher who both Alliances editorial director has written letters in sup-
pioneered the use of zebra fish as a model organism port of promotion and tenure cases indicating that
COLLABORATORIES 111
molecule page contributions are of journal- projects there are tensions between users, who want
publication quality. This agreement is a creative at- reliable tools that do what they need done, and com-
tempt to ensure that quality contributions will be puter scientists, who are interested in technical inno-
made to the database; it also represents an interest- vations and creative software ideas. There is little
ing evolution of the scholarly journal to include new incentive for the computer scientists to go beyond
forms of scholarly publication. the initial demonstration versions of tools to the re-
liable and supported long-term operational infra-
Data Issues structure desired by the users. In some fields, such
Data are a central component of all collaborations. as high-energy physics, this tension has been at least
There are numerous issues concerning how data are partially resolved. The field has used advanced soft-
represented and managed; how these issues are re- ware for so long that it is understood that the extra
solved affects collaboratory success. For example, costs associated with having production versions
good metadatadata about dataare critical as of tools must be included in a project. Other fields
databases increase in size and complexity. Library are only just discovering this. The organization of
catalogs and indexes to file systems are examples of the George E. Brown, Jr., Network for Earthquake
metadata. Metadata are key to navigation and search Engineering Simulation (NEES) project represents
through databases. an innovation in this regard. The National Science
Information about the provenance or origins of Foundation, which funds the project, established it
the data is also important. Data have often been in two phases, an initial four-year system-integration
highly processed, and researchers will want to know phase in which the tools are developed and tested,
what was done to the original raw data to arrive at and a ten-year operational phase overseen by a NEES
the processed data currently in the database. Two consortium of user organizations.
related collaboratories in high-energy physics, Any large organization faces difficult manage-
GriPhyN and iVDGL, are developing schemes for ment issues, and practicing scientists may not always
showing investigators the paths of the transfor- have the time or the skills to properly manage a com-
mations that led to the data in the database. This plex enterprise. Management issues get even more
will help researchers understand the data and will complicated when the organization is geographically
also help in identifying and correcting any errors in distributed. Many large collaboratories have faced
the transformations. difficult management issues. For instance, the two
For some kinds of collaboratories, the complex physics collaboratories mentioned earlier, GriPhyN
jurisdictional issues that arise when data are com- and iVDGL, found that it was necessary to hire a full-
bined into a large database pose an interesting new time project manager for each collaboratory in or-
issue. The BIRN project is facing just such an issue der to help the science project directors manage
as it works to build up a database of brain images. the day-by-day activities of the projects. The Alliance
The original brain images were collected at different for Cellular Signaling has benefited from a charis-
universities or hospitals under different institutional matic leader with excellent management skills who
review boards, entities that must approve any hu- has set up a rich management structure to oversee
man data collection and preservation, and so the the project. The BIRN collaboratory has an explicit
stipulations under which the original images were governance manual that contains guidelines for a
collected may not be the same in every case. host of tricky management issues; it also has a steer-
ing committee that is responsible for implementing
these management guidelines.
Other Issues
Many collaboratory projects involve cooperation
between domain scientists, who are the users of Collaboratories in the Future
the collaboratory, and computer scientists, who are Geographically distributed research projects are be-
responsible for the development of the tools. In many coming commonplace in all the sciences. This
proliferation is largely driven by what is required to Sproull, L., Conley, C., & Moon, J. Y. (in press). Pro-social behavior on
work at the frontiers of science. In the future, widely the net. In Y. Amichai-Hamburger (Ed.), The social net: The social
psychology of the Internet. New York: Oxford University Press.
shared knowledge about how to put together suc- Sproull, L. & Kiesler, S. (in press). Public volunteer work on the Internet.
cessful collaboratories will be essential. Of course, In B. Kahin & W. Dutton (Eds.), Transforming enterprise. Cam-
scientists are not alone in attempting geographically bridge, MA: MIT Press.
distributed collaborations. Similar issues are faced Star, S. L., & Ruhleder, K. (1994). Steps towards an ecology of infra-
structure: Complex problems in design and access for large-scale
in industry, education, government, and the non- collaborative systems. In Proceedings of CSCW 94 (pp. 253264).
profit sector. Good tools for collaboration and the New York: ACM Press.
social and organizational knowledge to make effec- Teasley, S., & Wolinsky, S. (2001). Scientific collaborations at a dis-
tance. Science, 292, 22542255.
tive use of them will be critical in all domains. Torvalds, L., & Diamond, D. (2001). Just for fun: The story of an acci-
dental revolutionary. New York: Harper Business.
Gary M. Olson Wulf, W.A. (1993). The collaboratory opportunity. Science, 261,
854855.
See also Computer-Supported Cooperative Work;
Groupware
COMPILERS
FURTHER READING
Compilers are computer programs that translate
Aldhous, P. (1993). Managing the genome data deluge. Science, 262, one programming language into another. The orig-
5023. inal program is usually written in a high-level lan-
Birnholtz, J., & Bietz, M. (2003). Data at work: Supporting sharing in guage by a programmer and then translated into a
science and engineering. In Proceedings of Group 2003. New
York: ACM Press.
machine language by a compiler. Compilers help
Cinkosky, M. J., Fickett, J. W., Gilna, P., & Burks, C. (1991). Electronic programmers develop user-friendly systems by al-
data publishing and GenBank. Science, 252, 12731277. lowing them to program in high-level languages,
Finholt, T. A. (2002). Collaboratories. In B. Cronin (Ed.), Annual which are more similar to human language than ma-
Review of Information Science and Technology, 36, 74107.
Washington, DC: American Society for Information Science and
chine languages are.
Technology.
Finholt, T. A., & Olson, G. M. (1997). From laboratories to collabo-
ratories: A new organizational form for scientific collaboration.
Psychological Science, 8(1), 2836.
Background
Huang, W., Olson, J. S., & Olson, G. M. (2002). Camera angle affects Of course, the first compilers had to be written in
dominance in video-mediated communication. In Proceedings of machine languages because the compilers needed to
CHI 2002, short papers (pp. 716717). New York: ACM Press. operate the computers to enable the translation
National Science Foundation. (2003) Revolutionizing science and en-
gineering through cyberinfrastructure: Report of the National Science
process. However, most compilers for new computers
Foundation blue-ribbon panel on cyberinfrastructure. Retrieved are now developed in high-level languages, which
December 24, 2003, from http://www.communitytechnology.org/ are written to conform to highly constrained syntax
nsf_ci_report/ to ensure that there is no ambiguity.
Olson, G. M., Finholt, T. A., & Teasley, S. D. (2000). Behavioral aspects
of collaboratories. In S. H. Koslow & M. F. Huerta (Eds.), Electronic
Compilers are responsible for many aspects of
collaboration in science (pp. 114). Mahwah, NJ: Lawrence Erlbaum information system performance, especially for
Associates. the run-time performance. They are responsible
Olson, G. M., & Olson, J. S. (2000). Distance matters. Human-Computer for making it possible for programmers to use the
Interaction, 15(23), 139179.
Raymond, E. S. (1999). The cathedral and the bazaar: Musing on Linux full power of programming language. Although com-
and open source by an accidental revolutionary. Sebastopol, CA: OReilly. pilers hide the complexity of the hardware from
Schatz, B. (1991). Building an electronic community system. Journal ordinary programmers, compiler development re-
of Management Information Systems, 8(3), 87107.
Singh, Push (n.d.). Open mind common sense. Retrieved December
quires programmers to solve many practical algo-
22, 2003, from http://commonsense.media.mit.edu/cgi-bin/ rithmic and engineering problems. Computer
search.cgi hardware architects constantly create new challenges
COMPILERS 113
for compiler developers by building more complex takes the source code as input and then examines the
machines. source program to check its conformity to the syn-
Compilers translate programming languages and tactic and semantic constraints of the language in
the following are the tasks performed by each spe- which the program was written. During the synthetic
cific compiler type: process, the object code in the target language is gen-

erated. Each major process is further divided. The
Assemblers translate low-level language in-
analytic process, for example, consists of a character
structions into machine code and map low-level
handler, a lexical analyzer, a syntax analyzer, and a
language statements to one or more machine-
constraint analyzer. The character handler identifies
level instructions.
Compilers translate high-level language in-
characters in the source text, and the lexical analyzer
groups the recognized characters into tokens such
structions into machine code. High-level lan-
as operators, keywords, strings, and numeric con-
guage statements are translated into more than
stants. The syntax analyzer combines the tokens into
one machine-level instruction.
Preprocessors usually perform text substitutions
syntactic structures, and the constraint analyzer
checks to be sure that the identified syntactic
before the actual translation occurs.
High-level translators convert programs written
structures meet scope and type rules.
The synthetic process consists of an intermedi-
in one high-level language into another high-
ate code generator, a code optimizer, and a code gen-
level language. The purpose of this translation is
erator. An intermediate code generator produces code
to avoid having to develop machine-language-
that is less specific than the machine code, which will
based compilers for every high-level language.
Decompilers and disassembers translate the ob-
be further processed by another language translator.
A code optimizer improves the intermediate code
ject code in a low-level language into the source
with respect to the speed of execution and the com-
code in a high-level language. The goal of this
puter memory requirement. A code generator
translation is to regenerate the source code.
takes the output from the code optimizer and then
In the 1950s compilers were often synonymous generates the machine code that will actually be
with assemblers, which translated low-level language executed on the target computer hardware.
instructions into directly executable machine code.
The evolution from an assembly language to a high-
level language was a gradual one, and the FORTRAN Interpreters and
compiler developers who produced the first successful
high-level language did not invent the notion of pro- Interpretive Compilers
gramming in a high-level language and then com- In general, compilers produce the executable ob-
piling the source code to the object code. The first ject code at the full speed, and compilers are usually
FORTRAN compiler was designed and written be- designed to compile the entire source code before
tween 1954 and 1957 by an IBM team led by John executing the resulting object code. However, it is
W. Backus, but it had taken about eighteen person- common for programmers to expect to execute one
years of effort to develop. The main goal of the team or more parts of a program before completing the
led by Backus was to produce object code that could program. In addition, many programmers want to
execute as efficiently as human machine coders could. write programs using a trial-and-error or what-if
strategy. These cases call for the use of an interpreter
in lieu of a traditional compiler because an inter-
Translation Steps preter, which executes one instruction at a time, can
Programming language translators, including com- take the source program as input and then execute
pilers, go through several steps to accomplish their the instructions without generating any object code.
task, and use two major processesan analytic pro- Interpretive compilers generate simple inter-
cess and a synthetic process. The analytic process mediate code, which satisfies the constraints of the
practical interpreters. The intermediate code is then control flow is manipulated by instances of regular
sent as input to an interpreter, which executes the al- expressions in the input stream. Regular expressions
gorithm embedded in the source code by utilizing a consist of normal characters, which include upper-
virtual machine. Within the virtual machine setting, and lower-case letters and digits, and metacharac-
the intermediate code plays the role of executable ters, which have special meanings. For example, a
machine code. dot is a metacharacter, which matches any one char-
acter other than the new-line character. There is also
a table of regular expressions and their associated
Famous Compiler: program pieces, called Lex source, and the resulting
program is a translation of the table. The program
GNU Compiler Collection (GCC) reads the input stream and generates the output
Many high-level language compilers have been im- stream by partitioning the input into strings that
plemented using the C programming language match the given regular expression.
and generating C code as output. Because almost all Yacc is a general tool for describing the source
computers come with a C compiler, source code writ- code to a program. After the Yacc user specifies the
ten in C is very close to being truly hardware- structures to be recognized and the corresponding
independent and portable. The GNU Compiler codes to be invoked, Yacc finds the hierarchical struc-
Collection (GCC) provides code generation for many tures and transforms their specifications into sub-
programming languages such as C, C++, and Java, routines that process the input.
and supports more than two hundred different soft-
ware and hardware platforms. The source code of
GCC is free and open, based on GNU General Public The Future of Compilers
License, which allows people to distribute the Proebstrings Law states that compiler advances
compilers source code as long as the original double computing power every 18 years (Proebsting,
copyright is not violated and the changes are pub- n.d., 1). This implies that compiler-optimization
lished under the same license. This license enables work makes a very minor contribution because it
users to port GCC to their platform of choice. means that while the processing power of computer
Presently almost all operating systems for per- hardware increases by about 60 percent per year, the
sonal computers are supported by GCC and ship the compiler optimization increases by only 4 percent.
compiler as an integrated part of the platform. For Furthermore, some people claim that compilers
example, Apples Mac OS X is compiled using will become obsolete with the increased use of script-
GCC 3.1. Other companies such as Sun and The ing languages, which rely on interpreters or inter-
Santa Cruz Operation also offer GCC as their stan- pretive compilers. Scripting languages, such as
dard system compiler. These examples show the flex- Python, are popular among new programmers and
ibility and portability of GCC. people who do not care about minute efficiency dif-
ferences. However, there are arguments for the con-
tinued existence of compilers. One of the arguments
Compiler Constructor: Lex and Yacc is that there has to be a machine code on which the
Roughly speaking, compilers work in two stages. The interpreters rely in order for a programmers intended
first stage is reading the source code to discover its algorithm to be executed. In addition, there will al-
structure. The second stage is generating the exe- ways be new and better hardware, which will then
cutable object code based on the identified structure. rely on new compilers. It will also be impossible to
Lex, a lexical-analyzer generator, and Yacc, a com- extinguish the continuing desire to achieve even
piler-compiler, are programs used to discover the minute performance improvements and compile-
structure of the source code. Lex splits the source time error-detection capability. One of the proposed
code into tokens and then writes a program whose future directions for compilers is to aid in increas-
COMPUTER-SUPPORTED COOPERATIVE WORK 115
ing the productivity of programmers by optimizing Pizka, M. (1997). Design and implementation of the GNU INSEL
the high-level code. Another possible direction is Compiler gic. Technical Report TUMI 9713. Munich, Germany:
Munich University of Technology.
to make compilers smarter by making them self- Proebstring, T. (n.d.). Todd Proebstings home page. Retrieved January
steering and self-tuning, which would allow them to 20, 2004, from http://research.microsoft.com/~toddpro/
adapt to input by incorporating artificial-intelligence Rice compiler group. (n.d.). Retrieved January 20, 2004, from http://
techniques. www.cs.rice.edu/CS/compilers/index.html
Terry, P. D. (1997). Compilers and compiler generators an introduc-
tion with C++. London: International Thomson Computer Press.
Woojin Paik The comp.compilers newsgroup. (2002). Retrieved January 20, 2004,
from http://compilers.iecc.com/index.html
The Lex and Yacc page. (n.d.). Retrieved January 20, 2004, from
See also Programming Languages http://dinosaur.compilertools.net/
Why compilers are doomed. (April 14, 2002). Retrieved January 20,
2004, from http://www.equi4.com/jcw/wiki.cgi/56.html
FURTHER READING
Aho, A. V., Sethi, R., & Ulman, J. D. (1986). Compilers: principles, tech-
niques and tools. Reading, MA: Addison-Wesley.
COMPUTER-SUPPORTED
Aho, A. V., & Ulman, J. D. (1977). Principles of compiler design. Reading,
MA: Addison-Wesley.
Bauer, A. (2003). Compilation of functional programming languages
using GCCTail Calls. Retrieved January 20, 2004, from
http://home.in.tum.de/~baueran/thesis/baueran_thesis.pdf COOPERATIVE WORK
A Brief History of FORTRAN /fortran. (1998). Retrieved January 20,
2004, from http://www.ibiblio.org/pub/languages/FORTRAN/ Computer-supported cooperative work (CSCW) is
ch1-1.html
Catalog of free compilers and interpreters. (1998). Retrieved January
the subarea of human-computer interaction con-
20, 2004, from http://www.idiom.com/free-compilers/ cerned with the communication, collaboration,
Clodius W. (1997). Re: History and evolution of compilers. Retrieved and work practices of groups, organizations, and com-
January 20, 2004, from http://compilers.iecc.com/comparch/article/ munities, and with information technology for
97-10-008
Compiler Connection. (2003). Retrieved January 20, 2004, from
groups, organizations, and communities. As the
http://www.compilerconnection.com/index.html Internet and associated networked computing activ-
Compiler Internet Resource List. (n.d.). Retrieved January 20, 2004, ities have become pervasive, research in CSCW has
from http://www.eg3.com/softd/compiler.htm expanded rapidly, and its central concepts and vo-
Cooper, K., & Torczon, L. (2003). Engineering a Compiler. Burlington,
MA: Morgan Kaufmann. cabulary are still evolving. For the purposes of this
Cooper, K., Kennedy, K., and Torczon, L. (2003). COMP 412 Overview discussion, we understand cooperative work as any
of the course. Retrieved January 20, 2004, from http://www.owl- activity that includes or is intended to include the co-
net.rice.edu/~comp412/Lectures/L01Intro.pdf
Cranshaw, J. (1997). Lets build a compiler. Retrieved January 20, 2004,
ordinated participation of at least two individuals;
from http://compilers.iecc.com/crenshaw/ we take computer support of such work to be any in-
GCC Homepage. (January 26, 2004). Retrieved January 26, 2004, from formation technology used to coordinate or carry out
http://gcc.gnu.org/ the shared activity (including archiving of the records
Free Software Foundation. (1991). GNU General Public License.
Retrieved January 20, 2004, from http://www.fsf.org/licenses/
of an activity to allow subsequent reuse by another).
gpl.html Several themes dominate research and practice
Joch, A. (January 22, 2001). Compilers, interpreters and bytecode. in CSCW: studies of work, in which activities and
Retrieved January 20, 2004, from http://www.computerworld.com/ especially tool usage patterns are observed, analyzed,
softwaretopics/software/story/0,10801,56615,00.html
Lamm, E. (December 8, 2001). Lambda the Great. Retrieved January and interpreted through rich qualitative descriptions;
20, 2004, from http://lambda.weblogs.com/2001/12/08 design and use of computer-mediated communica-
Mansour, S. (June 5, 1999). A Tao of Regular Expressions. Retrieved tion (CMC) systems and of groupware, designed
January 20, 2004, from http://sitescooper.org/tao_regexps.html
Manzoor, K. (2001). Compilers, interpreters and virtual machines.
to aid with collaborative planning, acting, and sense
Retrieved January 20, 2004, from http://homepages.com.pk/ making; and analyses of the adoption and adapta-
kashman/jvm.htm tion of CSCW systems.
A Personal StorySocial Context in Computer-Supported

Cooperative Work (CSCW)
In the early 1980s, our research group at the IBM Watson Research Center focused on the early stages of learning word pro-
cessing systems, like the IBM Displaywriter. We carried out an extensive set of studies over several years. In these investi-
gations, we noticed that people tried to minimize the amount of rote learning they engaged in, preferring to adopt
action-oriented approaches in their own learning.
Eventually, we developed a description of the early stages of learning to use computer applications that helped to de-
fine new design approaches and learning support. But this work also made us wonder what more advanced learning might
be like.
To investigate this, my colleague John Gould and I visited an IBM customer site, to observe experienced users of
Displaywriters as they worked in their everyday environments. These individuals were competent and confident in their
use of the software. However we observed a pattern of distributed expertise: Each member of the staff had mastered one
advanced function. Whenever someone needed to use an advanced function, she contacted the corresponding expert for
personal, one-on-one coaching. This was a win-win situation: the requestors received customized help, and the specialized
experts earned an increase in status. These field observations taught us the importance of peoples social context in the use
and evaluation of information technology, something we now take for granted in CSCW.
Mary Beth Rosson
Studies of Work and learning support and flexibility in the roles and
A fundamental objective of CSCW is to understand responsibilities available to human workers.
how computers can be used to support everyday work Studies of work often employ ethnographic
practices. Early research in the 1980s focused on methods adapted from anthropology. In ethno-
workflow systems. This approach codifies existing graphic research, the activities of a group are ob-
business procedures (for example, relating to the hir- served over an extended period of time. This allows
ing of a new employee) in a computer model and collaborative activity to be seen in context. Thus,
embeds the model in a tracking system that moni- tasks are not characterized merely in terms of the
tors execution of the procedures, providing reminders, steps comprising procedures, but also in terms of
coordination across participants, and assurance that who interacts with whom to carry out and im-
appropriate steps are followed. Computerized work- provise procedures, what tools and other arti-
flow systems are highly rational technological tools facts are used, what information is exchanged and
whose goal is to support the effective execution of created, and the longer-term collateral outcomes
normative procedures. Ironically, a major lesson that of activity, such as personal and collective learn-
emerged from building and studying the use of these ing and the development of group norms and mu-
systems is that exceptions to normative business tual trust. This work has demonstrated how, for
procedures are pervasive in real activity, and that example, the minute interdependencies and
handling such exceptions characteristically involves personal histories of doctors, nurses, patients,
social interactions that need to be fluid and nuanced administrators, and other caregivers in the func-
in order to succeed. Indeed, the failure of direct and tioning of a hospital must be analyzed to properly
rational workflow support was to a considerable ex- understand actions as seemingly simple as a doc-
tent the starting point for modern CSCW, which now tor conveying a treatment protocol to a nurse on
emphasizes balance between structured performance the next shift.
Sometimes the observer tries to be invisible in ambiguous, entailing clarifications and confirma-
ethnographic research, but sometimes the investi- tions. And of course informal interactions are also
gator joins the group as a participant-observer. often unproductive. Balancing direct support for
Typically video recordings of work activities are work activities with broader support for building and
made, and various artifacts produced in the course maintaining social networks is the current state of
of the work are copied or preserved to enable later the classic workflow systems challenge.
analysis and interpretation.
Ethnographic methods produce elaborate and
often voluminous qualitative descriptions of com- Computer-Mediated Communication
plex work settings. These descriptions have be- The central role of communication in the behavior
come central to CSCW research and have greatly of groups has led to intense interest in how tech-
broadened the notion of context with respect to un- nology can be used to enable or even enhance com-
derstanding human activity. Theoretical frameworks munication among individuals and groups. Much
such as activity theory, distributed cognition, and attention has been directed at communication among
situated action, which articulate the context of ac- group members who are not colocated, but even
tivity, have become the major paradigms for science for people who share an office, CMC channels such
and theory in CSCW. as e-mail and text chat have become pervasive. Indeed
Much of what people do in their work is guided e-mail is often characterized as the single most suc-
by tacit knowledge. A team of engineers may not re- cessful CSCW application, because it has been inte-
alize how much they know about one anothers grated so pervasively into everyday work activities.
unique experience, skills, and aptitudes, or how well The medium used for CMC has significant con-
they recruit this knowledge in deciding who to call sequences for the communicators. Media richness
when problems arise or how to phrase a question or theory suggests that media supporting video or voice
comment for best effect. But if an analyst observes are most appropriate for tasks that have a subjective
them at work, queries them for their rationale dur- or evaluative component because the nonverbal cues
ing problem-solving efforts, and asks for reflections provided by a communicators visual appearance or
on why things happen, the tacit knowledge that is voice tone provide information that helps partici-
uncovered may point to important trade-offs in pants better understand and evaluate the full impact
building computerized support for their work of one anothers messages. In contrast, text-based
processes. For instance, directing a question to an media like e-mail or chat are better for gathering and
expert colleague provides access to the right infor- sharing objective information. Of course, even text-
mation at the right time, but also establishes and re- based channels can be used to express emotional con-
inforces a social network. Replacing this social tent or subjective reactions to some extent; a large
behavior with an automated expert database may and growing vocabulary of character-based icons
answer the query more efficiently, but may cause em- and acronyms are used to convey sadness, happiness,
ployees to feel more disconnected from their surprise, and so on.
organization. Use of CMC has also been analyzed from the per-
A persistent tension in CSCW studies of work spective of the psychologist Herbert Clarks theory
springs from the scoping of activities to be sup- of common ground in languagethe notion that
ported. Many studies have shown how informal language production, interpretation, and feedback
communicationdropping by a coworkers office, relies extensively on communicators prior knowl-
encountering someone in the hall, sharing a coffee edge about one another, the natural language they
can give rise to new insights and ideas and is essen- are using, the setting they are in, and their group and
tial in creating group cohesion and collegiality, social cultural affiliations. In CMC settings some of this
capital to help the organization face future challenges. information may be missing. Furthermore, many of
But communication is also time consuming and often the acknowledgement and feedback mechanisms that
humans take for granted in face-to-face conversa- and to build trusting relationships. Indeed, there have
tion (for example, head nods and interjected uh- been a number of well-publicized episodes of cruel
huhs and so on) become awkward or impossible to behavior in CMC environments such as chatrooms
give and receive in CMC. The theory of common and MUDs (multiuser dungeons or domains).
ground argues that these simple acknowledgement During the 1990s, cell phones, pagers, personal
mechanisms are crucial for fluid conversation be- digital assistants, and other mobile devices rendered
cause they allow conversation partners to monitor people and their work activities more mobile. As a
and track successful communication: A head nod or consequence, the context of CMC became quite var-
an uh-huh tells the speaker that the listener under- ied and unpredictable. A research area that has de-
stands what the speaker meant, is acknowledging veloped in response to users changing environments
that understanding, and is encouraging the speaker is context-aware computing, wherein the technology
to continue. is used not only to support work activities, but also
Despite the general acknowledgement that text- to gather information about the users situation. For
based CMC media such as e-mail and chat are rel- example, it is relatively straightforward to set up dis-
atively poor at conveying emotion and subjective tinct settings for how a cell phone will operate
content, these channels have advantages that make (e.g., ring tone or volume) at work, home, outdoors,
them excellent choices for some tasks. E-mail, for ex- and so on, but it takes time and attention to re-
ample, is usually composed and edited in advance of member to activate and deactivate them as needed.
sending the message; it can be read and reviewed Thus the goal is to build devices able to detect changes
multiple times; and it is very easily distributed to in peoples environment and to activate the appro-
large groups. E-mail is also easy to archive, and its priate communication options or tasks. Whether such
text content can be processed in a variety of ways mode changes take place automatically or are man-
to create reusable information resources. Because aged by the individual, the resulting context infor-
e-mail is relatively emotion-free, it may be appro- mation can be important to collaborators, signaling
priate for delicate or uncomfortable communication if and when they can initiate or return to a shared
tasks. With so many CMC options, people are now activity.
able to make deliberate (or tacit) choices among
CMC channels, using a relatively informal and un-
obtrusive medium like text chat for low-cost inter- Groupware
action, more formally composed e-mail for business CSCW software is often categorized by the timing
memos, and video or audio conferencing for im- of the collaboration it supports: Synchronous group-
portant decision-making tasks. ware supports interaction at the same point in time,
The relative anonymity of CMC (particularly while asynchronous groupware supports collabo-
with text-based channels) has provoked considerable ration across time. Another distinction is the col-
research into the pros and cons of anonymous com- laborators relative location, with some groupware
munication. Communicators may use their real designed for colocated interaction and some for dis-
names or screen names that only loosely convey their tributed activities. For example, group decision sup-
identity; in some situations virtual identities may be port systems are typically used for synchronous and
adopted explicitly to convey certain aspects of an in- colocated interaction: As part of a face-to-face meet-
vented personality or online persona. Anonymity ing, group members might use a shared online en-
makes it easier to express sensitive ideas and so can vironment to propose, organize, and prioritize ideas.
be very effective when brainstorming or discussion In contrast, an online forum might be used for asyn-
is called for but social structures would otherwise in- chronous discussions among distributed group
hibit a high degree of sharing. However the same fac- members.
tors that make anonymity an aid to brainstorming A longstanding goal for many groupware devel-
also can lead to rude or inappropriate exchanges and opers has been building support for virtual meetings
may make it difficult to establish common ground synchronous group interactions that take place
A Personal StoryInternet Singing Lessons
Having immigrated to the United States from India at an early age, I have always had a problem mastering the fine melodic
nuances required to sing traditional Hindi songs. This problem has limited my singing repertoire.
Last winter, during a religious meeting of Indian immigrants living in the Ann Arbor area, I was struck with how
well a young Indian man sang a haunting Hindu chant. Later that evening I asked him to help me improve how I sang
Hindu chants, which he did willingly. However, he soon informed me that he was returning to India the following week
as he was in the U.S. on a temporary work visa.
Because I was disappointed in losing such a willing teacher, my friend suggested a technological solution. He suggested
that I set up an account with Yahoo! Messenger, and to buy a stereo headset through which we could continue our music
interaction. Yahoo! Messenger is an instant messaging system that enables logged-in users to exchange text messages, and
to talk free of charge on the Internet.
When my friend returned to India, we had to deal with two problems. First, we had to deal with the time-difference.
India is 10 hours ahead of the U. S. Second, we had to deal with the problem that my friend only had access to an Internet
connection at the office where he worked. This is because computers and Internet connections are still quite expensive
for the average Indian. We therefore decided that the best time for undisturbed instant voice messaging would be at 7:30 a.m.
Indian Standard Time when other employees had not yet arrived in my friends office. This time also work out well for
me because it would be 9:00 p.m. (EST), the time when I liked to pluck on my guitar and sing.
The above plan worked wellon February 8th, 2004, I had my first transcontinental singing lesson. Despite a slight
delay in sound transmission due to the Internet bandwidth problem, my friend was able to correct the fine melodic nu-
ances that I missed when I sang my favorite Hindu chant. I can now sing a Hindu chant with nuances approved by a singing
teacher sitting in front of a computer many oceans away.
Suresh Bhavnani
entirely online as a substitute for traditional face-to- viewing, and revising of content that is a natural con-
face meetings. As businesses have become increas- sequence of working together online.
ingly international and distributed, support for When collaborators meet online, participant au-
virtual meetings has become more important. A vir- thentication is an important issue. Many work sit-
tual meeting may use technology as simple as a tele- uations have policies and procedures that must be
phone conference call or as complex as a collaborative respected; for example, meetings may have a spec-
virtual environment (CVE) that embodies attendees ified attendee list or restricted documents, or de-
and their work resources as interactive objects in a cisions may require the approval of a manager.
three-dimensional virtual world. Because virtual Enforcing such restrictions creates work for both
meetings must rely on CMC, attendees have fewer the organizer of the activity (who must activate the
communication cues and become less effective at appropriate controls) and the participants (who
turn taking, negotiation, and other socially rich in- must identify themselves if and when required).
teraction. It is also often difficult to access and in- Depending on a groups culture and setting, the
teract with meeting documents in a CVE, particularly meeting organizers may choose to make no restric-
when the meeting agenda is open and information tions at all (for example, they may meet in an on-
needs to evolve during the meeting. Some researchers line chatroom and rely on group members to
have argued that online meetings will never equal self-enforce relevant policies and group behavior),
face-to-face interaction, and that researchers should or they may rely on a set of roles (such as leader, at-
focus instead on the special qualities offered by a tendee, or scribe) built into the groupware system
virtual mediumfor example, the archiving, re- to manage information access and interaction.
A significant technical challenge for synchronous know which group members are around, available
groupware is ensuring data consistency. When col- for interaction, and so on. Social awareness can be
laborators are able to communicate or edit shared provided through mechanisms such as buddy lists,
data in parallel, there is the possibility that simul- avatars (online representations of group members),
taneous requests will conflict: One participant might or even regularly updated snapshots of a person in
correct the spelling of a word at the same time that their work setting. For a shared project that takes
another member deletes a phrase containing the place over weeks or months, collaborators need ac-
word, for example. The simplest technique for avoid- tivity awareness: They must be aware of what proj-
ing consistency problems is to implement a floor ect features have changed, who has done what, what
control mechanism that permits only one partici- goals or plans are currently active, and how to con-
pant at a time to have the virtual pen, with others tribute. However, promoting activity awareness re-
waiting until it is passed to them. Because such mech- mains an open research topic; considerable work is
anisms can be awkward and slow, many groupware needed to determine how best to integrate across
systems have explored alternatives, including implicit synchronous and asynchronous interactions, what
locking of paragraphs or individual words, and fully information is useful in conveying status and
optimistic serialization, which processes all input in progress, and how this information can be gathered
the order in which it is received, with the assump- and represented in a manner that supports rather
tion that well-learned social protocols of turn tak- than interrupts collaborative activities.
ing and coordination will reduce conflict and ensure
smooth operation.
Many other technical challenges plague the Adoption and Adaptation of
smooth operation of groupware. For instance, it is
quite common for collaborators to be interacting CSCW Systems
with rather different hardware and software plat- Even when great care is taken in the design and
forms. Although work groups may settle on a stan- implementation of a CSCW system, there is no guar-
dard set of software, not all group members may antee that it will be successfully adopted and inte-
follow all aspects of the standard, and beyond the grated into work practicesor that when it is
work group, there may be no standards. Thus inter- adopted it will work as originally intended. Many
operability of data formats, search tools, editing or case studies point to a sociotechnical evolution cy-
viewing software, and analysis tools is a constant con- cle: Initially, delivered CSCW systems do not fit onto
cern. As work settings have become more mobile and existing social and organizational structures and
dynamic, the variety of technical challenges has in- processes. During a process of assimilation and ac-
creased: Some members at a virtual meeting may commodation, the organization changes (for exam-
join by cell phone, while others may use a dedicated ple, a new role may be defined for setting up and
broadband network connection. It is increasingly facilitating virtual meetings) in concert with the tech-
common for groupware systems to at least provide nology (for example, a set of organization-specific
an indicator of such variation, so that collabora- templates may be defined to simplify agenda setting
tors can compensate as necessary (for example, by and meeting management).
recognizing that a cell phone participant may not be Several implications follow from this view of
able to see the slides presented at a meeting). CSCW adoption. One is that participatory design of
The general goal of promoting awareness dur- the software is essentialwithout the knowledge of
ing CSCW interactions has many facets. During syn- praxis provided by the intended users, the software
chronous work, groupware often provides some form will not be able to evolve to meet their specific needs;
of workspace awareness, with telepointers or minia- furthermore if users are included in the design
turized overviews showing what objects are selected process, introduction of the CSCW system into the
or in view by collaborators. In more extended col- workplace will already have begun by the time the
laborations, partners depend on social awareness to system is deployed. Another implication is that
CSCW software should have as open an architecture the new technology, much of the sociotechnical evo-
as possible, so that when the inevitable need for lution has taken place, context-specific procedures
changes is recognized months or years after deploy- have been developed and refined in situ, and there
ment, it will be possible to add, delete, or other- are local experts to assist new users.
wise refine existing services. A third implication is As more and more of an organizations activities
that organizations seeking CSCW solutions should take place onlinewhether through e-mail or video-
be ready to change their business structures and conferencing or shared file systemsvia CSCW tech-
processesand in fact should undertake business nology, the amount of online information about the
process reengineering as they contribute to the de- organization and its goals increases exponentially.
sign of a CSCW system. A frequent contributing fac- The increased presence of organizational informa-
tor in groupware failure is uneven distribution of tion online has generated great interest in the
costs and benefits across organizational roles and re- prospects for organizational memory or knowledge
sponsibilities. There are genuine costs to collabo- management. The hope is that one side effect of carry-
ration: When an individual carries out a task, its ing out activities online will be a variety of records
subtasks may be accomplished in an informal and about how and why tasks are decomposed and ac-
ad hoc fashion, but distributing the same task among complished, and that these records can provide guid-
individuals in a group is likely to require more ad- ance to other groups pursuing similar goals. Of
vance planning and negotiation, recordkeeping, and course once again, there are important cost-benefit
explicit tracking of milestones and partial results. issues to consider: Recording enough information
Collaboration implies coordination. Of course the to be helpful to future groups takes time, especially
benefits are genuine as well: One can assign tasks if it is to be stored in any useful fashion, and the ben-
to the most qualified personnel, one gains multiple efit in most cases will be enjoyed by other people.
perspectives on difficult problems, and social recog- One solution is to give computers the job of record-
nition and rewards accrue when individuals coming, organizing, and retrieving. For example, even
bine efforts to reach a common goal. Unfortunately, a coarse-grained identification of speakers making
the costs of collaboration are often borne by work- comments in a meeting can simplify subsequent
ers, who have new requirements for online planning browsing of the meeting audiotape.
and reporting, while its benefits are enjoyed by man-
agers, who are able to deliver on-time results of higher
quality. Therefore, when designing for sociotechni- Research Directions
cal evolution, it is important to analyze the expected Much of the active research in CSCW is oriented
costs and benefits and their distribution within the toward new technologies that will enhance aware-
organization. Equally important are mechanisms for ness, integrate multiple devices, populations, and
building social capital and trust, such that individ- activities, and make it possible to visualize and share
uals are willing to contribute to the common good, rich data sets and multimedia documents. The need
trusting that others in the group will reward or to interconnect people who are using diverse devices
care for them when the time comes. in diverse settings entails many research challenges,
Critical mass is another determinant of success- some related to the general issues of multiplatform
ful adoptionthe greater the proportion of indi- computing and others tied to understanding and
viduals within an organization who use a technology, planning for the social and motivational differences
the more sense it makes to begin using it oneself. associated with varied work settings. The rapidly
A staged adoption process is often effective, with a expanding archives in organizations offer many re-
high-profile individual becoming an early user and search opportunities related to data processing and
advocate who introduces the system to his or her analysis as well as information visualization and re-
group. This group chronicles its adoption experience trieval. At the same time, these digital storehouses
and passes the technology on to other groups, and raise important questions about individual privacy
so on. By the time the late adopters begin to use and identitythe more information an organization
collects about an individual, the more opportunity Baecker, R. M. (1993). Readings in groupware and computer-supported
there is for inappropriate access to and use of this cooperative work: Assisting human-human collaboration. San
Francisco: Morgan-Kaufmann.
information. Beaudouin-Lafon, M. (Ed). (1999). Computer supported co-operative
A methodological challenge for CSCW is the de- work. Chichester, UK: John Wiley & Sons.
velopment of effective evaluation methods. Field Bikson, T. K., & Eveland, J. D. (1996). Groupware implementation:
Reinvention in the sociotechnical frame. In Proceedings of the
studies and ethnographic analyses yield very rich
Conference on Computer Supported Cooperative Work: CSCW 96
data that can be useful in understanding system (pp. 428437). New York: ACM Press.
requirements and organizational dynamics. But Carroll, J. M., Chin, G., Rosson, M .B., & Neale, D. C. (2000). The
analyzing such detailed records to answer precise development of cooperation: Five years of participatory design in
the virtual school. In Designing interactive systems: DIS 2000
questions is time consuming and sometimes im- (pp. 239251). New York: ACM Press.
possible due to the complexity of real-world settings. Carroll, J. M., & Rosson, M.B. (2001). Better home shopping or new
Unfortunately, the methods developed for study- democracy? Evaluating community network outcomes. In
ing individual computer use do not scale well to the Proceedings of Human Factors in Computing Systems: CHI 2001
(pp. 372379). New York: ACM Press.
evaluation of multiple users in different locations. Dourish, P., & Bellotti, V. (1992). Awareness and coordination in shared
Because social and organizational context are a key workspaces. In Proceedings of the Conference on Computer Supported
component of CSCW activities, it is difficult to sim- Cooperative Work: CSCW 92 (pp. 107114). New York: ACM Press.
Grudin, J. (1994). Groupware and social dynamics: Eight challenges
ulate shared activities in a controlled lab setting. for developers. Communications of the ACM, 37(1), 92105.
Groupware has been evolving at a rapid rate, so there Gutwin, C., & Greenberg, S. (1999). The effects of workspace aware-
are few if any benchmark tasks or results to use for ness support on the usability of real-time distributed group-
comparison studies. One promising research di- ware. ACM Transactions on Computer-Human Interaction, 6(3),
243281.
rection involves fieldwork that identifies interesting Harrison, S., & Dourish, P. (1996). Re-placing space: The roles of place
collaboration scenarios; these are then scripted and space in collaborative systems. In Proceedings of the Conference
and simulated in a laboratory setting for more sys- on Computer Supported Cooperative Work: CSCW 96 (pp. 6776).
New York: ACM Press.
tematic analysis. Hughes, J., King, V., Rodden, T., & Andersen, H. (1994). Moving out
In the 1980s human-computer interaction from the control room: Ethnography in system design. In
focused on solitary users finding and creating in- Proceedings of the Conference on Computer Supported Cooperative
formation using a personal computer. Today, the Work: CSCW 94 (pp. 429439). New York: ACM Press.
Hutchins, E. (1995). Distributed cognition. Cambridge, MA: MIT Press.
focus is on several to many people working together Malone, T. W., & Crowston, K. (1994). The interdisciplinary study
at a variety of times and in disparate places, relying of coordination. ACM Computing Surveys, 26(1), 87119.
heavily on the Internet, and communicating and col- Markus, M. L. (1994). Finding a happy medium: Explaining the
negative effects of electronic communication on social life at work.
laborating more or less continually. This is far more ACM Transactions on Information Systems, 12(2), 119149.
than a transformation of human-computer inter- Nardi, B. A. (1993). A small matter of programming. Cambridge,
action; it is a transformation of human work and MA: MIT Press.
activity. It is still under way, and CSCW will continue Nardi, B. A. (Ed). (1996). Context and consciousness: Activity theory
and human-computer interaction. Cambridge, MA: MIT Press.
to play a large role. Olson, G. M., & Olson, J. S. (2000). Distance matters. Human Computer
Interaction, 15(23), 139179.
Mary Beth Rosson and John M. Carroll Orlikowski, W. J. (1992). Learning from notes: Organizational issues
in groupware implementation. In Proceedings of the Conference on
Computer Supported Cooperative Work: CSCW 92 (pp. 362369).
See also Collaboratories; Ethnography; MUDs; Social New York: ACM Press.
Psychology and HCI Roseman, M., & Greenberg, S. (1996). Building real time groupware
with Groupkit, a groupware toolkit. ACM Transactions on Computer
Human Interaction, 3(1), 66106.
Sproull, L., & Kiesler, S. (1991). Connections: New ways of working in
FURTHER READING the networked organization. Cambridge, MA: MIT Press.
Streitz, N. A., Geiler, J., Haake, J., & Hol, J. (1994). DOLPHIN:
Ackerman, M. S. (2002). The intellectual challenge of CSCW: The gap Integrated meeting support across local and remote desktop en-
between social requirements and technical feasibility. In J. M. Carroll vironments and liveboards. In Proceedings of the Conference on
(Ed.), Human-computer interaction in the new millennium Computer Supported Cooperative Work: CSCW 94 (pp. 345358).
(pp. 303324). New York: ACM Press. New York: ACM Press.
CONSTRAINT SATISFACTION 123
Suchman, L. (1987). Plans and situated actions: The problem of human- straints specify that exactly four of the decision vari-
machine communication. Cambridge, UK: Cambridge University ables have value 1 (queen in this square) and
Press.
Sun, C., & Chen, D. (2002). Consistency maintenance in real-time col-
that there cannot be two queens in the same row, col-
laborative graphics editing systems. ACM Transactions on Computer umn, or diagonal. Because there are sixteen variables
Human Interaction, 9(1), 141. (one for each square) and each can take on two pos-
Tang, J., Yankelovich, N., Begole, J., Van Kleek, M., Li, F., & Bhalodia, sible values, there are a total of 2 (65,536) possible
J. (2001). Connexus to awarenex: Extending awareness to mobile
users. In Proceedings of Human Factors in Computing Systems: CHI assignments of values to the decision variables. There
2001 (pp. 221228). New York: ACM Press. are other ways of modeling the 4-queen problem
Winograd, T. (1987/1988). A language/action perspective on the de- within the CSP framework. One alternative is to treat
sign of cooperative work. Human-Computer Interaction, 3(1), 330.
each row on the board as a decision variable. The
values that can be taken by each variable are the four
column positions in the row. This formulation yields
4 (256) possibilities. This example illustrates how
CONSTRAINT the initial formulation or model affects the number
of possibilities to be examined, and ultimately the
SATISFACTION performance of problem solving.
Constraint satisfaction refers to a set of representa-

tion and processing techniques useful for model- CSP Representations
ing and solving combinatorial decision problems; A CSP is often represented as an undirected graph
this paradigm emerged from the artificial intelligence (or network), which is a set of nodes connected by
community in the early 1970s. A constraint satis- a set of edges. This representation opens up the
faction problem (CSP) is defined by three elements: opportunity to exploit the properties and algorithms
(1) a set of decisions to be made, (2) a set of choices developed in graph theory for processing and solv-
or alternatives for each decision, and (3) a set of con- ing CSPs. In a constraint graph, the nodes represent
straints that restrict the acceptable combinations the variables and are labeled with the domains of the
of choices for any two or more decisions. In general, variables. The edges represent the constraints and
the task of a CSP is to find a consistent solution link the nodes corresponding to the variables to
that is, a choice for every decision such that all the which the constraints apply. The arity of a constraint
constraints are satisfied. More formally, each deci- designates the number of variables to which the con-
sion is called a variable, the set of alternative choices straint applies, and the set of these variables consti-
for a given variable is the set of values or domain of tutes the scope of the constraint. Constraints that apply
the variable, and the constraints are defined as the set to two variables are called binary constraints and are
of allowable combinations of assignments of values represented as edges in the graph. Constraints that
to variables. These combinations can be given in ex- apply to more than two variables are called non-
tension as the list of consistent tuples, or defined in binary constraints. While, early on, most research
intention as a predicate over the variables. has focused on solving binary CSPs, techniques for
solving nonbinary CSPs are now being investigated.
The 4-Queen Problem

A familiar example of a CSP is the 4-queen problem. The Role of CSPs in Science
In this problem, the task is to place four queens on Beyond puzzles, CSPs have been used to model
a 44 chessboard in such a way that no two queens and solve many tasks (for example, temporal rea-
attack each other. One way to model the 4-queen soning, graphical user interfaces, and diagnosis) and
problem as a CSP is to define a decision variable for have been applied in many real-world settings (for
each square on the board. The square can be either example, scheduling, resource allocation, and prod-
empty (value 0) or have a queen (value 1). The con- uct configuration and design). They have been used
in various areas of engineering, computer science, domain of B is the interval [7, 11]), and B oc-
and management to handle decision problems. curred one hour after A (B-A 1). It is easy to in-
A natural extension of the CSP is the constrained fer that the domains of A and B must be restricted
optimization problem (COP), where the task is to to [8, 10] and [9, 11] respectively, because B can-
find an optimal solution to the problem given a set not possibly occur before 9, or A after 10, without
of preferences and optimization criteria. The prob- violating the constraint between A and B. This fil-
lems and issues studied in the constraint processing tering operation considers every combination of two
(CP) community most obviously overlap with those variables in a binary CSP. It is called 2-consistency.
investigated in operations research, satisfiability and A number of formal properties have been proposed
theoretical computer science, databases, and pro- to characterize the extent to which the alternative
gramming languages. The 1990s have witnessed a combinations embedded in a problem description
sharp increase in the interactions and cross-fertilization are likely to yield consistent solutions, as a mea-
among these areas. sure of how close is the problem to being solved.
A special emphasis is made in CP to maintain These properties characterize the level of consistency
the expressiveness of the representation. Ideally, a of the problem (for example, k-consistency, mini-
human user should be able to naturally express the mality, and decomposability).
various relations governing the interactions among Algorithms for achieving these properties, also
the entities of a given problem without having to re- known as constraint propagation algorithms, remain
cast them in terms of complex mathematical mod- the subject of intensive research. Although the cost
els and tools, as would be necessary in mathematical of commonly used constraint propagation algorithms
programming. The area of constraint reformulation is a polynomial function of the number of variables
is concerned with the task of transforming the prob- of the CSP and the size of their domains, solving the
lem representation in order to improve the perfor- CSP remains, in general, an exponential-cost process.
mance of problem solving or allow the use of available An important research effort in CP is devoted to find-
solution techniques. Sometimes such transforma- ing formal relations between the level of consistency
tions are truthful (that is, they preserve the essence in a problem and the cost of the search process used
of the problem), but often they introduce some for solving it. These relations often exploit the topol-
sufficient or necessary approximations, which may ogy of the constraint graph or the semantic proper-
or may not be acceptable in a particular context. ties of the constraint. For example, a tree-structured
constraint graph can be solved backtrack-free after
ensuring 2-consistency, and a network of constraints
Solution Methods of bounded differences (typically used in temporal
The techniques used to solve a CSP can be divided reasoning) is solved by ensuring 3-consistency.
into two categories: constraint propagation (or infer-
ence) and search. Further, search can be carried out Systematic Search
as a systematic, constructive process (which is ex- In systematic search, the set of consistent combi-
haustive) or as an iterative repair process (which of- nations is explored in a tree-like structure starting
ten has a stochastic component). from a root node, where no variable has a value, and
considering the variables of the CSP in sequence.
Constraint Propagation The tree is typically traversed in a depth-first man-
Constraint propagation consists in eliminating, from ner. At a given depth of the tree, the variable under
the CSP, combinations of values for variables that consideration (current variable) is assigned a value
cannot appear in any solution to the CSP. Consider from its domain. This operation is called variable in-
for example two CSP variables A and B representing stantiation. It is important that the value chosen for
two events. Assume that A occurred between the current variable be consistent with the instan-
8 a.m. and 12 p.m. (the domain of A is the interval tiations of the past variables. The process of check-
[8, 12]), B occurred between 7 a.m. and 11 a.m. (the ing the consistency of a value for the current variable
CONSTRAINT SATISFACTION 125
with the assignments of past variables is called back- state to another and attempting to find a state where
checking. It ensures that only instantiations that are all constraints are satisfied. This move operator and
consistent (partial solutions) are explored. If a the state evaluation function are two important com-
consistent value is found for the current variable, ponents of an iterative-repair search. The move is
then this variable is added to the list of past variables usually accomplished by changing the value of one
and a new current variable is chosen from among variable (thus the name local search). However, a
the un-instantiated variables (future variables). technique operating as a multiagent search allows
Otherwise (that is, no consistent value exists in the any number of variables to change their values. The
domain of the current variable), backtracking is ap- evaluation function measures the cost or quality of
plied. Backtracking undoes the assignment of the a given state, usually in terms of the number of bro-
previously instantiated variable, which becomes the ken constraints. Heuristics, such as the min-conflict
current variable, and the search process attempts heuristic, are used to choose among the states reach-
to find another value in the domain of this variable. able from the current state (neighboring states).
The process is repeated until all variables have The performance of iterative-repair techniques
been instantiated (thus yielding a solution) or back- depends heavily on their ability to explore the so-
track has reached the root of the tree (thus proving lution space. The performance is undermined by the
that the problem is not solvable). Various techniques existence in this space of local optima, plateaux, and
for improving the search process itself have been pro- other singularities caused by the nonconvexity of the
posed. For systematic search, these techniques in- constraints. Heuristics are used to avoid falling
clude intelligent backtracking mechanisms such as into these traps or to recover from them. One heuris-
backjumping and conflict-directed backjumping. tic, a breakout strategy, consists of increasing the
These mechanisms attempt to remember the reasons weight of the broken constraints until a state is
for failure and exploit them during search in order reached that satisfies these constraints. Tabu search
to avoid exploring barren portions of the search maintains a list of states to which search cannot move
space, commonly called thrashing. The choices of the back. Other heuristics use stochastic noise such as
variable to be instantiated during search and that of random walk and simulated annealing.
the value assigned to the variable are handled, re-
spectively, by variable and value ordering heuristics, Blending Solution Techniques
which attempt to reduce the search effort. Such Constraint propagation has been successfully com-
heuristics can be applied statically (that is, before the bined with backtrack search to yield effective look-
search starts) or dynamically (that is, during the ahead strategies such as forward checking. Combining
search process). The general principles that guide constraint propagation with iterative-repair strate-
these selections are the most constrained variable gies is less common. On the other hand, randomiza-
first and the most promising value first. Examples tion, which has been for a long time utilized in
of the former include the least domain heuristic local search, is now being successfully applied in back-
(where the variable with the smallest domain is cho- track search.
sen for instantiation) and the minimal-width heuris-
tic (where the variables are considered in the ordering
of minimal width of the constraint graph). Research Directions
The use of constraint processing techniques is wide-
Iterative-Repair Search spread due to the success of the constraint pro-
In iterative repair (or iterative improvement) search, gramming paradigm and the increase of commercial
all the variables are instantiated (usually randomly) tools and industrial achievements. While research on
regardless of whether or not the constraints are the above topics remains active, investigations are also
satisfied. This set of complete instantiations, which invested in the following directions: user interaction;
is not necessarily a solution, constitutes a state. discovery and exploitation of symmetry relations;
Iterative-repair search operates by moving from one propagation algorithms for high-arity constraints
and for continuous domains; preference modeling Montanari, U. (1974). Networks of constraints: Fundamental proper-
and processing; distributed search techniques; em- ties and application to picture processing. Information Sciences, 7,
95132.
pirical assessment of problem difficulty; and statis- Prosser, P. (1993). Hybrid algorithms for the constraint satisfaction
tical evaluation and comparison of algorithms. problem. Computational Intelligence, 9(3), 268299.
Rgin, J.-C. (1994). A filtering algorithm for constraints of difference
Berthe Y. Choueiry in constraint satisfaction problems. In Proceedings from the National
Conference on Artificial Intelligence (AAAI 1994) (pp. 362437).
Seattle, WA.
See also Artificial Intelligence; N-grams Revesz, P. (2002). Introduction to constraint databases. New York:
Springer.
Stuckey, K. M. (1998). Programming with constraints: An introduction.
Cambridge, MA: MIT Press.
FURTHER READING Tsang, E. (1993). Foundations of constraint satisfaction. London, UK:
Academic Press.
Yokoo, M. (1998). Distributed constraint satisfaction. New York:
Bistarelli, S., Montanari, U., & Rossi, F. (1997). Semiring-based con-
Springer.
straint satisfaction and optimization. Journal of the ACM, 44(2),
201236.
Borning, A., & Duisberg, R. (1986). Constraint-based tools for build-
ing user interfaces. ACM Transactions on Graphics, 5(4), 345374.
CONVERGING
Cohen, P. R. (1995). Empirical methods for artificial intelligence.
Dechter, R. (2003). Constraint processing. San Francisco: Morgan
Kaufmann.
Ellman, T. (1993). Abstraction via approximate symmetry. In TECHNOLOGIES
Proceedings of the 13th IJCAI (pp. 916921). Chambry, France.
Freuder, E. C. (1982). A sufficient condition for backtrack-free search. Human-computer interaction (HCI) is a multi-
JACM, 29(1), 2432.
Freuder, E. C. (1985). A sufficient condition for backtrack-bounded
disciplinary field arising chiefly in the convergence
search. JACM, 32(4), 755761. of computer science, electrical engineering, infor-
Freuder, E. C. (1991). Eliminating interchangeable values in constraint mation technology, and cognitive science or psy-
satisfaction problems. In Proceedings of AAAI-91 (pp. 227233). chology. In the future it is likely to be influenced
Anaheim, CA.
Gashnig, J. (1979). Performance measurement and analysis of certain
by broader convergences currently in progress, reach-
search algorithms. Pittsburgh, PA: Carnegie-Mellon University. ing out as far as biotechnology and nanotechnology.
Glaisher, J. W. L. (1874). On the problem of the eight queens. Together, these combined fields can take HCI to new
Philosophical Magazine, 4(48), 457467. levels where it will unobtrusively but profoundly en-
Glover, F. (1989). Tabu SearchPart I. ORSA Journal on Computing,
1(3), 190206. hance human capabilities to perceive, to think, and
Gomes, C. P. (2004). Randomized backtrack search. In M. Milano to act with maximum effectiveness.
(Ed.), Constraint and Integer Programming: Toward a Unified
Methodology (pp. 233291). Kluwer Academic Publishers.
Haralick, R. M., & Elliott, G. L. (1980). Increasing tree search efficiency
for constraint satisfaction problems. Artificial Intelligence, 14, 263313. The Basis for Convergence
Hogg, T., Huberman, B. A., & Williams, C. P. (Eds.). (1996). Special During the twentieth century a number of interdis-
volume on frontiers in problem solving: Phase transitions and ciplinary fields emerged, bridging the gaps be-
complexity. Artificial Intelligence, 81(12). Burlington, MA: Elsevier
Science.
tween separate traditionally defined sciences. Notable
Hooker, J. (2000). Logic-based methods for optimization: Combining examples are astrophysics (astronomy plus physics),
optimization and constraint satisfaction. New York: Wiley. biochemistry (biology plus chemistry), and cogni-
Hoos, H. H., & Sttzle, T. (2004). Stochastic local search. San Francisco: tive science (psychology plus neurology plus com-
Morgan Kaufmann.
Kirkpatrick, S., Gelatt, J. C. D., & Vecchi, M. P. (1983). Optimization puter science). Many scientists and engineers believe
by simulated annealing. Science, 220(4598), 671680. that the twenty-first century will be marked by a
Liu, J., Jing, H., & Tang, Y. Y. (2002). Multi-agent oriented constraint broader unification of all of the sciences, permitting
satisfaction. Artificial Intelligence, 136(1), 101144.
Minton, S., et al. (1992). Minimizing conflicts: A heuristic repair
a vast array of practical breakthroughsnotably
method for constraint satisfaction and scheduling problems. in the convergence of nanotechnology, biotechnology,
Artificial Intelligence, 58, 161205. information technology, and cognitive technology
CONVERGING TECHNOLOGIES 127
based on the unification of nanoscience, biology, seem remote from HCI because the human senses
information science, and cognitive science. HCI itself operate at a much larger scale. However, we can al-
stands at the junction between the last two of these ready identify a number of both direct and in-
four, and it has the potential to play a major role in direct connections, and as work at the nanoscale
the emergence of converging technologies. promotes convergence between other fields it will
A number of scientific workshops and conferences, create new opportunities and challenges for HCI.
organized by scientists and engineers associated with The largest single atoms, such as those of ura-
the U.S. National Science Foundation and building nium, are just smaller than 1 nanometer. The struc-
upon the United States National Nanotechnology tures of complex matter that are fundamental to
Initiative, have concluded that nanoscience and nano- all sciences originate at the nanoscale. That is the
technology will be especially important in conver- scale at which complex inorganic materials take on
gence. Nanoscience and nanotechnology concern the characteristic mechanical, electrical, and
scientific research and engineering (respectively) chemical properties they exhibit at larger scales. The
at the nanoscale, the size range of physical structures nanoscale is where the fundamental structures of life
between about 1 nanometer and 100 nanometers in arise inside biological cells, including the human
shortest dimension. A nanometer is 1 billionth of a DNA (deoxyribonucleic acid) molecule itself. The
meter, or 1 millionth of a millimeter, and a millimeter double helix of DNA has the proportions of a twisted
is about the thickness of a dime (the thinnest U.S. piece of string, about 2.5 nanometers thick but as
coin). Superficially, nanoscience and nanotechnology much as 4 centimeters (40 million nanometers) long
The BRLESC-II, a solid-state digital computer introduced in 1967. It was designed to be

200 times faster than the ORDVAC computer it replaced. Photo courtesy of the U.S. Army.
if uncoiled. The synaptic gaps between neurons in Wide Web, both hardware and software. Many of the
the human brain, and the structures that contain the early applications have been new ways of accom-
neurotransmitter chemicals essential to their func- plishing old tasks, for example, word processors, dig-
tioning, are on the order of 20 to 50 nanometers. ital music and television, and more recently digital
Nanotechnology and nanoscience are chiefly a libraries. The integration of mobile computing with
partnership of physics, chemistry, and materials sci- the Internet is expected to unleash a wave of radi-
ence (an interdisciplinary field at the intersection of cally different innovations, many of which cannot
physics, chemistry, and engineering that deals with even be imagined today, connected to ubiquitous
the properties of materials, including composite ma- availability of information and of knowledge tools.
terials with complex structures). In the near term Cognitive science is the study of intelligence,
nanotechnology offers engineering a host of new whether human, nonhuman animal, or machine,
materials, including powders with nanoscale gran- including perception, memory, decision, and un-
ules, thin coatings that transform the properties of derstanding. It is itself a convergence of fields, draw-
surfaces, and composite materials having nanoscale ing upon psychology, social psychology, cultural
structure that gives them greater strength, durabil- anthropology, linguistics, economics, sociology,
ity, and other characteristics that can be precisely de- neuroscience, artificial intelligence, and machine
signed for many specific uses. In the midterm to long learning. The fundamental aim is a profound un-
term, nanotechnology is expected also to achieve derstanding of the nature of the human mind. By
practical accomplishments with complex nano- the beginning of the twenty-first century a new uni-
structures, including new kinds of electronic com- verse of cognitive technologies clearly was opening
ponents and nanoscale machines. up, especially in partnerships between humans and
Biotechnology applies discoveries in biology to computers. The result could be technologies that
the invention and production of products that are overcome breakdowns in human awareness, analy-
valuable for human health, nutrition, and economic sis, planning, decision making, and communication.
well-being. The traditional application areas for Each of these four fields is a fertile field of sci-
biotechnology are medicine and agriculture, in- entific research and technological development, but
cluding the production of chemicals and construc- in combination they can achieve progress much more
tion materials having organic origins. Biotechnology rapidly and broadly than they can alone. Following
has a long history, extending back thousands of years are examples of the science and engineering oppor-
to ancient industries such as fermentation of alco- tunities in each of the six possible pairs.
hol, tanning of hides, dyeing of clothing, and bak-
ing of bread. The pace of innovation accelerated NanotechnologyBiotechnology
throughout the nineteenth and twentieth centuries, Research at the nanoscale can reveal the detailed, dy-
leading to the latest developments in genomics (a namic geometry of the tiny structures that carry out
branch of biotechnology concerned with applying metabolism, movement, and reproduction inside the
the techniques of genetics and molecular biology to living cell, thereby greatly expanding biological sci-
the genetic mapping and DNA sequencing of sets of ence. Biology provides conceptual models and prac-
genes or the complete genomes of selected organ- tical tools for building inorganic nanotechnology
isms) and a growing understanding of the structures structures and machines of much greater complex-
and processes inside the living cell. ity than currently possible.
Information technology is a creation of the
second half of the twentieth century, revolutioniz- NanotechnologyInformation Technology
ing traditional communication technologies through Nanoelectronic integrated circuits will provide the
the introduction of electronic computation. It fast, efficient, highly capable hardware to support
comprises computers, information systems, and com- new systems for collecting, managing, and distrib-
munication networks such as Internet and the World uting information wherever and whenever it is
needed. Advances in information technology will be ization tools to help people understand biology and
essential for the scientific analysis of nanoscale struc- biotechnology correctly.
tures and processes and for the design and manu-
facture of nanotechnology products. Information TechnologyCognitive Technology
Experiments on human and nonhuman animal
NanotechnologyCognitive Technology behavior depend upon computerized devices for data
New research methods based on nanoscale sensor collection and on information systems for data analy-
arrays will enable neuroscientists to study the fine sis, and progress can be accelerated by sharing in-
details of neural networks in the brain, including the formation widely among scientists. Discoveries by
dynamic patterns of interaction that are the basis of cognitive scientists about the ways the human mind
human thought. Cognitive science will help nano- carries out a variety of judgments provide models
scientists and educators develop the most readily for how machines could do the same work, for ex-
intelligible models of nanoscale structures and the ample, to sift needed information from a vast as-
innovative curriculum needed for students to un- sembly of undigested data.
derstand the world as a complex hierarchy of sys-
tems built up from the nanoscale.
HCI Contributions to Convergence
BiotechnologyInformation Technology Attempting to combine two scientific disciplines
Principles from evolutionary biology can be applied would be futile unless they have actually moved into
to the study of human culture, and biologically in- adjacent intellectual territories and proper means
spired computational methods such as genetic al- can be developed to bridge between them. Disciplines
gorithms (procedures for solving a mathematical typically develop their own distinctive assumptions,
problem in a finite number of steps that frequently terminologies, and methodologies. Even under the
involve repetition of an operation) can find mean- most favorable conditions, transforming tools are
ingful patterns in vast collections of information. needed, such as new concepts that can connect the
Bioinformatics, which consists of biologically ori- disparate assumptions of different disciplines,
ented databases with lexicons for translating from ontologiescategory schemes and lexicons of
one to another, is essential for managing the huge concepts in a particular domainthat translate lan-
trove of data from genome (the genetic material of guage across the cultural barriers between disciplines,
an organism) sequencing, ecological surveys, and research instrumentation or mathematical analy-
large-scale medical and agricultural experiments, sis techniques that can be applied equally well in
and systematic comparisons of evolutionary con- either discipline. Because many of these trans-
nections among thousands of species. forming tools are likely to be computerized, human-
computer interaction research will be essential for
BiotechnologyCognitive Technology scientific and technological convergence.
Research techniques and instruments developed in One of the key ways of developing fresh scientific
biotechnology are indispensable tools for research conceptualizations, including models and metaphors
on the nature and dynamics of the nervous system, that communicate successfully across disciplinary
in both humans and nonhuman animals, under- barriers, is computer visualizations. For example,
stood as the products of millions of years of bio- three-dimensional graphic simulations can help stu-
logical evolution. Human beings seem to have great dents and researchers alike understand the structures
difficulty thinking of themselves as parts of com- of complex molecules at the nanoscale, thus bridg-
plex ecological systems and as the products of ing between nanoscience and molecular biology,
evolution by natural selection from random evolu- including genomics and the study of the structures
tion, so advances will be needed to design fresh inside the living cell. In trying to understand the
approaches to scientific education and new visual- behavior of protein molecules, virtual reality (VR)
may incorporate sonification (the use of sounds to Researchers in many diverse sciences already have
represent data and information) in which a established a shared data infrastructure, such as in-
buzzing sound represents ionization (the dissocia- ternational protein structure and genomics databases
tion of electrons from atoms and molecules, thus giv- and the online archives that store thousands of so-
ing them an electric charge), and haptics (relating to cial and behavioral science questionnaire datasets.
the sense of touch) may be used to represent the at- The development of digital libraries has expanded
traction between atoms by providing a counteract- the range of media and the kinds of content that can
ing force when a VR user tries to pull them apart. For be provided to scholars, scientists, and engineers over
data that do not have a natural sensory representa- the Internet. Grid computing, which initially served
tion, a combination of psychology and user-centered the supercomputing community by connecting geo-
design, focusing on the needs and habitual thought graphically distributed heavy iron machines, is ma-
patterns of scientists, will identify the most success- turing into a vast, interconnected environment of
ful forms of data visualization, such as information shared scientific resources, including data collection
spaces that map across the conceptual territories of instrumentation, information storage facilities, and
adjacent sciences. major storehouses of analytic tools. As more and
HCI is relevant not only for analyzing statistical more research traditions join the grid world, they
or other data that have already been collected and will come to understand each other better and find
computerized, but also for operating scientific in- progressively more areas of mutual interest. This con-
struments in real time. Practically every kind of sci- vergence will be greatly facilitated by advances in
entific research uses computerized instruments today. human-computer interaction research.
Even amateur astronomical telescopes costing under
$500 have guidance computers built into them. In
the future expensive computerized instruments used Implications for Computing
in nanoscience, such as atomic force microscopes Because HCI already involves unification of infor-
(tools for imaging individual atoms on a surface, al- mation and cognitive technologies, distinctive effects
lowing one to see the actual atoms), may provide of convergence will primarily occur in unification
haptic feedback and three-dimensional graphics to with the two other realms: nanotechnology and bio-
let a user virtually feel and see individual atoms when technology. Nanotechnology is likely to be especially
manipulating them, as if they have been magnified crucial because it offers the promise of continued
10 million times. improvement in the performance of computer com-
In any branch of science and engineering, HCI- ponents. Already a nanoscale phenomenon called
optimized augmented cognition and augmented re- the giant magnetoresistance (GMR) effect has been
ality may play a useful role, and after scientists and used to increase the data density on mass produc-
engineers in different fields become accustomed to tion computer hard disks, giving them much greater
the same computer methods for enhancing their abil- capacity at only slight cost. The two key components
ities, they may find it easier to communicate and thus of a computer hard disk are a rotatable magnetic disk
collaborate with each other. For example, primate and a read-and-write head that can move along
cognitive scientists, studying the behavior of baboons, the radius of the disk to sense the weak magnetism
may collaborate with artificial-intelligence re- of specific tiny areas on the disk, each of which
searchers, and both can employ augmented reality represents one bit (a unit of computer information
to compare the behavior of a troop of real animals equivalent to the result of a choice between two al-
with a multiagent system designed to simulate them. ternatives) of data. Making the active tip of the read-
Internet-based scientific collaboratories can not only and-write head of precisely engineered materials
provide a research team at one location with a va- constructed in thin (nanoscale) layers significantly
riety of transforming tools, but also let researchers increases its sensitivity. This sensitivity, in turn, al-
from all around the world become members of the lows the disk to be formatted into a larger number
team through telepresence. of smaller areas, thereby increasing its capacity.
Since the beginning of the human-computer in- that perform logical operations) and carbon nano-
teraction field, progress in HCI has depended not tube transistors (transistors made of nanoscale tubes
only on the achievements of its researchers, but also composed of carbon).
on the general progress in computer hardware. For If successful, these radically new approaches
example, early in the 1970s the Altair computer pio- require development of an entire complex of fresh
neered the kind of graphic user interface employed technologies and supporting industries; thus, the
by essentially all personal computers at the end of cost of shifting over to them may be huge. Only a
the twentieth century, but its memory chips were too host of new applications could justify the massive
expensive, and its central processing unit was investments, by both government and industry, that
too slow. A decade later the chips had evolved to the will be required. Already people in the computer in-
point where Apple could just barely market the dustry talk of performance overhang, the possi-
Macintosh, the first commercially successful com- bility that technical capabilities have already
puter using such an interface. Today many areas outstripped the needs of desirable applications. Thus,
of HCI are only marginally successful, and along with a potential great benefit for HCI becomes also a great
HCI research and development, increased power and challenge. If HCI workers can demonstrate that a
speed of computers are essential to perfect such ap- range of valuable applications is just beyond the reach
proaches as virtual reality, real-time speech recogni- of the best computers that the old technology can
tion, augmented cognition, and mobile computing. produce, then perhaps people will have sufficient
Since the mid-1960s the density of transistors on motivation to build the entire new industries that
computer chips has been doubling roughly every eight- will be required. Otherwise, all of computer sci-
een months, and the cost of a transistor has been drop- ence and engineering may stall.
ping by half. So long as this trend continues, HCI can During the twentieth century several major tech-
count on increasingly capable hardware. At some nologies essentially reached maturity or ran into
point, possibly before 2010, manufacturers will no social, political, or economic barriers to progress.
longer be able to achieve progress by cramming more Aircraft and automobiles have changed little in recent
and more components onto a chip of the tradi- years, and they were certainly no faster in 2000 than
tional kind. HCI progress will not stop the next day, in 1960. The introduction of high-definition tele-
of course, because a relatively long pipeline of research vision has been painfully slow, and applications of
and development exists and cannot be fully exploited haptics and multimodal augmented reality outside
before several more years pass. Progress in other areas, the laboratory move at a snails pace. Space flight
such as parallel processing and wireless networking, technology has apparently stalled at about the tech-
will still be possible. However, HCI would benefit nical level of the 1970s. Nuclear technology has either
greatly if electronic components continued to become been halted by technical barriers or blocked by po-
smaller and smaller because this miniaturization litical opposition, depending on how one prefers
means they will continue to get faster, use progres- to analyze the situation. In medicine the rate of in-
sively less power, and possibly also be cheaper. troduction of new drugs has slowed, and the great
Here is where nanotechnology comes in. Actually, potential of genetic engineering is threatened by in-
the transistors on computer chips have already shrunk creasing popular hostility. In short, technological
into the nanoscale, and some of them are less than civilization faces the danger of stasis or decline un-
50 nanometers across. However, small size is only less something can rejuvenate progress.
one of the important benefits of nanotechnology. Technological convergence, coupled with ag-
Equally important are the entirely new phenomena, gressive research at the intersections of technical
such as GMR, that do not even exist at larger scales. fields, may be the answer. Because HCI is a conver-
Nanotechnologists have begun exploring alterna- gent field itself and because it can both benefit from
tives to the conventional microelectronics that we and promote convergence, HCI can play a central
have been using for decades, notably molecular logic role. In addition to sustaining progress as tradition-
gates (components made of individual molecules ally defined, convergence enables entirely new
applications. For example, nanotechnology provides explanationsstatements about how and why re-
the prospect of developing sensors that can instantly wards may be obtained and costs are incurred. In
identify a range of chemicals or microorganisms in the language of computer science, such explana-
the environment, and nano-enabled microscale sen- tions are algorithms.
sor nets can be spread across the human body, a rain Some algorithms are very specific and apply only
forest, and the wing of an experimental aircraft to under certain narrowly defined circumstances. If one
monitor their complex systems of behavior. wants meat, one takes a big stick from the forest, goes
into the meadow, and clobbers one of the sheep graz-
ing there. If one wants water, one goes to the brook
Paradigm Transformation at the bottom of the valley. These are rather spe-
Convergence is not just a matter of hiring a multi- cific explanations, assuming that only one meadow,
disciplinary team of scientists and engineers and one kind of animal, one brook, and one valley exist.
telling them to work together. To do so they need ef- As the human mind evolved, it became capable of
fective tools, including intellectual tools such as com- working out much more general algorithms that ap-
prehensive theories, mathematical techniques for plied to a range of situations. If one wants meat, one
analyzing dynamic systems, methods for visualizing takes a club, goes to any meadow, and sees what one
complex phenomena, and well-defined technical can clobber there. If one wants water, the bottoms
words with which to talk about them. of deep valleys are good places to look. In the terms
Decades ago historian Thomas Kuhn described of artificial intelligence, the challenge for human in-
the history of science as a battle between old ways of telligence was how to generalize, from a vast com-
thought and new paradigms (frameworks) that may plexity of experience, by reasoning from particular
be objectively better but inevitably undergo oppo- cases to develop rules for solving particular broad
sition from the old-guard defenders of the prevail- kinds of problems.
ing paradigm. His chief example was the so-called Stark and Bainbridge noted how difficult it is for
Copernican Revolution in astronomy, when the no- human beings to invent, test, and perfect very gen-
tion that the Earth is the center of the universe was eral explanations about the nature of the universe
displaced by a new notion that the sun is the cen- and thereby to find empirically good algorithms for
ter of the solar system and of a vast, centerless uni- solving the problems faced by our species. In other
verse far beyond. The problem today is that many words, science and technology are difficult enter-
paradigms exist across all branches of science and prises that could emerge only after ten thousand years
engineering. Some may be equivalent to each other, of civilization and that cannot be completed for many
after their terms are properly translated. Others may decades to come. In the absence of a desired reward,
be parts of a larger intellectual system that needs people often will accept algorithms that posit at-
to be assembled from them. However, in many areas tainment of the reward in the distant future or in
inferior paradigms that dominate a particular dis- some other non-verifiable context. Thus, first sim-
cipline will need to be abandoned in favor of one ple magic and then complex religious doctrines
that originated in another discipline, and this process emerged early in human history, long before humans
is likely to be a hard-fought and painful one taking had accurate explanations for disease and other dis-
many years. asters, let alone effective ways of dealing with
The human intellectual adventure extends back them. If the full convergence of all the sciences and
tens of thousands of years. In their research and technologies actually occurs, as it may during the
theoretical work on the origins of religion, Rodney twenty-first century, one can wonder what will be-
Stark and William Sims Bainbridge observed that come not only of religion but of all other forms of
human beings seek rewards and try to avoid costs unscientific human creativity, what are generally
a commonplace assumption in economics and other called the humanities.
branches of social science. To solve the problems The U.S. entomologist and sociobiologist Edward
they faced every day, ancient humans sought O. Wilson has written about the convergence that
CYBERCOMMUNITIES 133
is far advanced among the natural sciences, calling human minds of all ages. Possibly no such com-
it consilience, and has wondered whether the hu- prehensive explanation of reality (an algorithm for
manities and religion will eventually join in to be- controlling nature) is possible. Or perhaps the in-
come part of a unified global culture. Here again tellectuals and investors who must build this fu-
human-computer interaction may have a crucial role ture world may not be equal to the task. Thus,
to play because HCI thrives exactly at the boundary whether it succeeds or fails, the technological con-
between humans and technology. vergence movement presents a huge challenge for
During the first sixty years of their existence, the field of human-computer interaction, testing
computers evolved from a handful of massive ma- how well we can learn to design machines and in-
chines devoted to quantitative problems of engi- formation systems that help humans achieve their
neering and a few physical sciences to hundreds of maximum potential.
millions of personal tools, found in every school
or library, most prosperous peoples homes, and William Sims Bainbridge
many peoples pockets. Many people listen to music
or watch movies on their computers, and thousands See also Augmented Cognition; Collaboratories
of works of literature are available over the Internet.
A remarkable number of digital libraries are devoted
to the humanities, and the U.S. National Endowment FURTHER READING
for the Humanities was one of the partner agencies
in the Digital Library Initiative led by the U.S. Atkins, D. E., Drogemeier, K. K., Feldman, S. I., Garcia-Molina, H.,
Klein, M. L., Messerschmitt, D. G., Messina, P., Ostriker, J. P., &
National Science Foundation. Wright, M. H. (2003). Revolutionizing science and engineering
The same HCI methods that are used to help sci- through cyberinfrastructure. Arlington, VA: National Science
entists visualize complex patterns in nature can Foundation.
become new ways of comprehending schools of art, Kuhn, T. (1962). The structure of scientific revolutions. Chicago:
University of Chicago Press.
tools for finding a desired scholarly reference, or even Roco, M. C., & Bainbridge, W. S. (2001). Societal implications of
new ways of creating the twenty-second-century nanoscience and nanotechnology. Dordrecht, Netherlands: Kluwer.
equivalents of paintings, sculptures, or symphonies. Roco, M. C., & Bainbridge, W. S. (2003). Converging technologies for
improving human performance. Dordrecht, Netherlands: Kluwer.
The same virtual reality systems that will help sci- Roco, M. C., & Montemagno, C. D. (Eds.). (2004). The coevolution of
entists collaborate across great distances can become human potential and converging technologies. Annals of the New
a new electronic medium, replacing television, in York Academy of Sciences, 1013. New York: New York Academy of
which participants act out roles in a drama while Sciences.
Stark, R., & Bainbridge, W. S. (1987). A theory of religion. New York:
simultaneously experiencing it as theater. Cyber- Toronto/Lang.
infrastructure resources such as geographic infor- Wilson, E. O. (1998). Consilience: The unity of knowledge. New York:
mation systems, automatic language translation Knopf.
machines, and online recommender systems can
be used in the humanities as easily as in the sciences.
The conferences and growing body of publica-
tions devoted to converging technologies offer a pic- CYBERCOMMUNITIES
ture of the world a decade or two in the future when
information resources of all kinds are available at all For many people, the primary reason for interacting
times and places, organized in a unified but malleable with computers is the ability, through computers, to
ontology, and presented through interfaces tai- communicate with other people. People form cyber-
lored to the needs and abilities of individual users. communities by interacting with one another
Ideally, education from kindergarten through grad- through computers. These cybercommunities are
uate school will be organized around a coherent conceived of as existing in cyberspace, a concep-
set of concepts capable of structuring reality in ways tual realm created through the networking and in-
that are simultaneously accurate and congenial to terconnection that computers make possible.
Cybercommunity Definition and History able on the Internet first made available in 1991 and
The prefix cyber first appeared in the word cyber- given a graphical interface in 1993). People also
netics, popularized by Norbert Weiner (18941964) began using the Internet in the early 1980s to run
in the 1940s to refer to the science of control and bulletin board services such as Usenet, which, unlike
communication in the animal and the machine (the the earlier local BBSs, could now be distributed to
subtitle of Weiners 1948 book Cybernetics). Since a much larger group of people and accessed by people
that time, cyber has prefixed many other words to in widely dispersed geographical locations. Usenet
create new terms for various interconnections be- expanded to include many different cybercommu-
tween computers and humans. One of the terms, cy- nities, most based around a common interest such
berspace, has become a popular metaphor for the as Linux programming or soap operas.
perceived location of online interactions. Coined by
William Gibson in his 1984 novel Neuromancer, Existing Cybercommunities
cyberspace originally referred to a graphical repre- Cybercommunities have risen in number with the
sentation of computerized data to which people con- increasing availability and popularity of the Internet
nected through direct electrical links to the brain. and the World Wide Web. Even within the short over-
Since then, the term has come to mean any virtual all history of cybercommunities, some cyber-
forum in which people communicate through com- communities have been short-lived. However, there
puters, whether the form of communication involves are several, begun in the early days of computer net-
text, graphics, audio, or combinations of those. working, that still exist online and therefore present
Cybercommunities predate widespread use of a useful view of factors involved in the formation
the Internet, with the first forming in localized sys- and maintenance of online communities.
tems called bulletin board services (BBSs). BBSs One of the oldest still-extant cybercommunities
usually ran on a single computer, and participants is The WELL, which began in 1985 as a local BBS
connected through modems and a local phone line. in the San Francisco Bay Area in California. Laurence
This meant that most participants lived within a lim- Brilliant, a physician with an interest in computer
ited geographical area. Thus many BBSs were able to conferencing, and Stewart Brand, editor of the Whole
hold occasional face-to-face get-togethers, enhanc- Earth Review and related publications, founded The
ing community relationships. Communication on WELL with the explicit goal of forming a virtual
BBSs was usually asynchronous; that is, people logged community. One savvy method the founders used
on at different times and posted messages in various to attract participants was to give free accounts to
topical forums for others to read and respond to later. local journalists, many of whom later wrote about
(E-mail and similar bulletin boards now available their participation, generating further interest and
on the World Wide Web are also asynchronous forms publicity. In the early years, when most participants
of online communication, while the various types of lived in the same geographical area, The WELL held
online chat and instant messaging are considered monthly face-to-face meetings. Currently owned by
to be synchronous forms of communication, since Salon.com, The WELL is now accessible through the
participants are present on a forum simultane- World Wide Web.
ously and can spontaneously respond to each others Another venerable cybercommunity, Lambda-
communications.) MOO, also began as an experiment in online com-
From the earliest days of the Internet and its munity. In contrast to The WELL, LambdaMOO
military-funded precursor, the Arpanet (established provided a forum for synchronous communica-
in 1969), online participants began forming cyber- tion and allowed people to create a virtual envi-
communities. E-mail immediately emerged as the ronment within which to interact. LambdaMOO
largest single use of the Internet, and remained so is an example of a type of program called a MUD,
until 2002, when it was matched by use of the World (for multiuser dimension or multiuser dungeon).
Wide Web (an information-exchange service avail- MUDs are similar to online chatrooms, but also allow
participants to create their own virtual spaces and

objects as additions to the program, with which they Welcome to LamdaMOO
and others can then interact. These objects enhance
B
elow is an introduction to the cybercommunity
the feel of being in a virtual reality. Created by the
LamdaMOO, as presented on www.lamdamoo.info:
computer scientist Pavel Curtis as a research project
for Xerox, LambdaMOO opened in 1990. A 1994 ar- LambdaMOO is sort of like a chat room. Its a
ticle about it in Wired magazine led to a significant text-only based virtual community of thousands
increase in interest in it and to dramatic population of people from all over the world. Its comprised
growth. Pavel Curtis has moved on to other projects, of literally thousands of rooms that have
and LambdaMOO is no longer associated with Xerox. been created by the users of LambdaMOO, and
you endlessly navigate (walk around) north,
But although it has undergone considerable social
south, etc. from room to room, investigating, and
changes over the years, it still attracts hundreds of meeting people that you can interact with to your
participants. hearts content.
MUDs began as interactive text-based role- You get there not thru an HTML browser like
playing games inspired by similar face-to-face Netscape or IE but through another program
roleplaying games such as Dungeons and Dragons called TELNET (search). Your computer most
likely has Telnet but enhanced versions can be
(hence dungeon in one expansion of the acronym).
found. (Telnet address: telnet://lambda.moo
More recently, similar online games have become .mud.org:8888/). You can try the Lambda button
available with the enhancement of a graphical in- at the top of this page to see if all goes well. If so,
terface. People have used MMORPGs (massively a window will open and youll be able to log in.
multiplayer online role-playing games) such as Ever- When you get the hang of it, you can create
quest as forums for socializing as well as gaming, and a character who has a name and a physical de-
scription, and who can be seen by all who meet
cybercommunities are forming amongst online
you. As you walk around from room to room you
gamers. are given a description of the room and a list of
Web logs, or blogs, are a relatively new and in- contents (including other people). You can look
creasingly popular platform for cybercommunities. at each person to get a more detailed description
Blogs are online journals in which one can post ones and when you do, they see a message stating that
thoughts, commentary, or reflections, sometimes you just checked them out. You can talk to
them and they see your words in quotes, like read-
also allowing others to post comments or reactions
ing spoken words in a book. You can also
to these entries. Many blogs provide a forum for emote (communicate with body language) using
amateur (or, in some cases, professional) journal- gestures such as a smile or a nod of the head. In
ism, however others more closely resemble online time youll learn to create your own rooms or
personal diaries. There are many different programs other objects, which are limited only by your
available for blogging, and some give specific at- imagination.
There are many people to meet up with
tention to community formation. LiveJournal, for
and build cyber-friendships with. When you
instance, enables each participant to easily gather first get there youll be asked to log in. First timers
other journals onto a single page, making it easy to can sign in as a guest. After that you can apply for
keep up with friends journals. Links between par- a permanent character name and password. Give
ticipants are also displayable, enabling people to see it a try and if you see me around say hi.
who their friends friends are and to easily form and Felis~Rex
expand social networks. Status: Programmer/108/33%Fogy/PC
Parent: Psychotic Class of Players
Community Networks Seniority: 1320/4093, (33%)
MOO-age: 108 months. (1995 January 28,
Some cybercommunities grow out of existing offline
Saturday)
communities. In particular, some local municipali-
ties have sought to increase citizen participation in
the local government and community by forming As more and more online sites use multimedia
community computer networks that allow people and bandwidth-intensive enhancements (that is,
access to government officials and provide forums enhancements that can only be successfully trans-
for community discussion. The first of these, the mitted across a wide rangeor bandof electro-
Public Electronic Network (PEN), started in Santa magnetic frequencies), speed of access has also become
Monica, California, in 1989. It was particularly a crucial issue. People with older equipmentslower
successful in providing access to low-income citizens modems and computer processorsare disadvan-
who might not otherwise have had access to com- taged in their ability to access online materials, es-
puters or computer networks. pecially at multimedia sites. Some governments,
In more recent years, some offline communities notably in South Korea and Japan, have sought to
have gone beyond providing an online forum specif- address that problem by subsidizing the develop-
ically related to the community and have also sought ment of broadband networks, enabling widespread
to promote computer use and connectivity in gen- relatively inexpensive access in those countries to
eral. For instance, the town of Blacksburg, Virginia, high-speed Internet connections.
with the help of Virginia Polytechnic Institute and In addition to access to equipment and networks,
State University (known as Virginia Tech and located people need the skills that enable them to use that
in town) and local businesses, is attempting to access. Research has also shown that people are
provide the infrastructure necessary to bring Internet unlikely to take advantage of the availability of
connectivity to every household in town and in the computers and the Internet if they do not consider
surrounding rural area. This project, called the computer-related activities useful and do not have
Blacksburg Electronic Village (BEV) has had several social support for such activities from people they
goals, including expanding the local economy know, especially their peers. This is particularly
through the promotion of high-tech industry, in- apparent in wealthier nations such as the United
creasing citizen access to online resources, and States, where the usefulness of and accessibility to
promoting a stronger sense of community. Recent online resources is taken for granted by more af-
evaluations by project leaders indicate that BEV fluent members of society but where such online re-
has been more successful in the first two areas than sources are less likely to be perceived as desirable
in the third. by members of less affluent communities. To address
that problem, several nonprofit groups in the United
Access Issues States have set up community computing centers in
As the BEV project recognized, in order to partici- poorer neighborhoods, where they provide both train-
pate in cybercommunities, people need access to ing in necessary computer skills and a community-
computers and to computer networks, especially based context for valuing such skills.
to the Internet and the World Wide Web. Although Another approach to broadening community ac-
such access has been expanding rapidly, people in cess to the Internet has been to integrate Internet
poorer nations and disadvantaged populations in connections into the construction of new build-
more affluent countries still have limited access to ings or entire neighborhoods. However, these
the Internet, if they have it at all. types of developments also benefit only those who
This issue has been particularly salient for com- can afford to buy into them.
munity networks, which are often created with the
specific goal of making it possible for disadvantaged Interfaces
groups to access and influence their local govern- The direct-brain interfaces envisioned by Gibson,
mental structures. Thus many community net- if possible at all, are likely many years in the future
works, in addition to setting up websites, have (although there have been some promising early ex-
provided publicly accessible terminals for the use periments in recent years, including one in which a
of those who do not have access to computers at blind person was given partial sight through a video
home or at work. feed wired to the optical nerve). Most people cur-
rently access and participate in cybercommunity dimensional virtual spaces, has slowed the develop-
through personal computers. Usually, these comment of cybercommunities using three-dimensional
puters are connected to the Internet by a modem or spaces and avatars. One such community, Active
other wired connection to an Internet service pro- Worlds (introduced in 1995), provides a three-
vider. However, wireless services are increasing, and dimensional view of the environment similar to those
in some countries, most notably Japan, cell phones first used in first-person shooter computer games
are commonly used to access the Internet and to (games in which you see on the screen what your
communicate textually with others. In other coun- character sees, rather than watching your character
tries, including the United States, people are also be- move about) such as Doom and Quake. Technical
ginning to use cell phones and personal digital considerations, including the simple problem of the
assistants (PDAs) for these purposes. amount of real estate available on a computer
Most communication in cybercommunities oc- screen, meant that participants in the early years of
curs through text, although some forums use graph- Active Worlds could see only the twelve closest
ics or voice communication, often supplemented by avatars. This contrasts with text-only interactive
text. Some of the oldest existing cybercommunities forums such as MUDs and chat, in which thirty to
are still text-only and therefore require a high level fifty participants can be simultaneously involved
of literacy as well as comfort with computers. Early in overlapping textual conversations. Graphical in-
text-based forums were not always particularly easy terfaces provide both limitations and enhancements
to use, either. The WELLs original interface was no- to online communications.
toriously difficult to work with. This meant that only
those with an understanding of computers and a Identity in Cybercommunities
strong interest in the possibilities of cybercommu- Aside from the more technical aspects of interface
nity had the motivation and ability to participate. design, cybercommunities have also had to grapple
Currently, The WELL has a much more accessible with the question of self-representation. How do par-
Web interface and a concomitantly more diverse pop- ticipants appear to one another? What can they know
ulation of users. about one another at the outset, and what can they
As available Internet bandwidth and computer find out? How accountable are cybercommunity
processing speeds have increased, cybercommuni- members for their words and behavior within the
ties are able to use graphical representations of people virtual space?
and objects within the cyberspace. One of the ear- In purely text-based systems such as chat forums
liest examples, from 1985, was Habitat, a role-playing or MUDs, participants are generally expected to pro-
game and socializing space that emulated an of- vide some sort of description or personal informa-
fline community. Habitat featured a local economy tion, although on some systems it is understood that
(based on points rather than real money) and such this information may be fanciful. On LamdaMOO,
social structures as a church and sheriff s office. for instance, many participants describe themselves
Habitat used two-dimensional cartoon-like draw- as wizards, animals, or creatures of light. However,
ings to represent people and objects within the each participant is required to choose a gender for
forum. Many current graphical worlds also use flat their character, partly in order to provide pronoun
cartoon-type representations. Habitat originated the choice for text generated by the MUD program, but
use of the term avatar to refer to the representa- also indicating the assumed importance of this
tion of people in such graphical worlds, and that term aspect of identity. In a divergence from real life,
has persisted in most such systems. LambdaMOO provides ten choices for gender iden-
The technical difficulties inherent in rendering tification. Despite this, most participants choose
three-dimensional spaces through which characters either male or female. LambdaMOO participants
can move and in which people can manipulate vir- choose what other personal information they wish
tual objects, along with the high level of computer to reveal. On other MUDs, especially those intended
processing power required to make possible three- as professional spaces or as forums for discussions
relating to real life (offline life), participants may face that can recover well from erroneous input.
be required to provide e-mail addresses or real names. Names and accountability are another crucial issue
In graphical forums, participants are represented for community forums. PEN found that anonymity
both by the textual information they provide tended to facilitate and perhaps even encourage flam-
about themselves and by their avatar. Design choices ing (caustic criticism or verbal abuse) and other an-
involved in avatar creation in different virtual spaces tisocial behavior, disrupting and in some cases
often reveal important underlying social assump- destroying the usefulness of the forums for others.
tions, as well as technical limitations. In the early
years on Active Worlds, for instance, participants Conflict Management and Issues of Trust
were required to choose from a limited number of Cybercommunities, like other types of communi-
existing predesigned avatars. In part this stemmed ties, must find ways to resolve interpersonal conflicts
from the difficulties of rendering even nominally and handle group governance. In the early years of
human-seeming avatars in the three-dimensional the Internet, users of the Internet were primarily
space. However, the particular avatars available also white, male, young, and highly educated; most were
revealed biases and assumptions of the designers. In connected to academic, government, or military in-
contrast to MUDs such as LambdaMOO, all avatars stitutions, or to computing-related businesses.
were human. At one point, participants exploited a However, in the mid-1990s the Internet experienced
programming loophole to use other objects, such as a great increase in participation, especially from
trees and walls, as personal representations, but this groups who had previously been on private systems
loophole was quickly repaired by the designers, who not connected to the Internet, notably America
felt strongly that human representations promoted Online (AOL). This sudden change in population
better social interaction. Active Worlds avatars and increase in diversity of participants created ten-
also displayed a very limited range of human vari- sions in some existing cybercommunities.
ation. Most were white, and the few non-white avatars In one now-famous Usenet episode in 1993, par-
available tended to display stereotypical aspects. ticipants in a Usenet newsgroup called alt.tasteless,
For instance, the single Asian avatar, a male, used a forum for tasteless humor frequented primarily by
kung-fu moves, the female avatars were all identifi- young men, decided to stage an invasion of another
able by their short skirts, and the single black male newsgroup, rec.pets.cats, whose participants, atypi-
avatar sported dreadlocks. Since then, programming cally for Usenet newsgroups at the time, were largely
improvements and feedback from users has enabled women, older than Usenet participants in general,
Active Worlds to improve their graphics (the avatars and in many cases relatively new to the Internet. The
now have distinct facial features) and expand their alt.tasteless participants flooded rec.pets.cats with
representational offerings. gross stories of cat mutilation and abuse, disrupting
In two-dimensional graphical environments such the usual discussions of cat care and useful infor-
as Worlds Away (introduced in 1995), variation mation about cats. Some of the more computer-savvy
tended to be greater from the beginning, and par- participants on rec.pets.cats attempted to deal
ticipants were given the ability to construct avatars with the disruption through technical fixes such as
from components. They could even change avatar kill files (which enable a participant to automatically
appearance at will by (for example) buying new heads eliminate from their reading queue messages posted
from the head shop. In some systems, participants by particular people), but this was difficult for par-
can also import their own graphics to further cus- ticipants with less understanding of the somewhat
tomize their online self-representation. arcane Usenet system commands. The invaders,
Cybercommunities with greater ties to offline meanwhile, found ways around those fixes. The con-
communities also have to deal with interface and rep- flict eventually spread to peoples offline lives, with
resentations issues. In order to provide community some rec.pets.cats participants receiving physical
access to as wide a range of townspeople as possi- threats, and at least one alt.tasteless participant hav-
ble, networks such as PEN need an easy-to-use inter- ing their Internet access terminated for abusive be-
havior. Eventually, the invaders tired of their sport presentsomething not really possible offline. As a
and rec.pets.cats returned to normal. Some news- positive contribution, this command can allow com-
groups have sought to avoid similar problems by es- munity members to discuss approaches to dealing
tablishing a moderator, a single person who must with a disruptive participant. However, the com-
approve all contributions before they are posted to mand can also have negative consequences.
the group. In high traffic groups, however, the task The use of avatars in graphical forums presents
of moderation can be prohibitively time-consuming. another set of potential conflicts. In the two-
LambdaMOO experienced a dramatic popula- dimensional space of Worlds Away, participants
tion surge in the late 1990s, causing not only social found that they could cause another participant to
tensions, but also technical problems as the completely disappear from view by placing their own
LambdaMOO computer program attempted to avatar directly on top of the others. With no avail-
process the increasing numbers of commands. able technical fix for this problem, users had to
LambdaMOO community members had to come up counter with difficult-to-enforce social sanctions
with social agreements for slowing growth and for against offenders.
limiting commands that were particularly taxing on
the server. For instance, they instituted a limit on the Trust
numbers of new participants that could be added The potential for conflicts in cybercommunities is
each day, started deleting (reaping) the characters probably no greater than that in offline communi-
and other information of participants who had been ties. On the one hand, physical violence is not pos-
inactive for several months, and set limits on the sible online (although in theory escalating online
number of new virtual objects and spaces that par- conflicts can lead to offline violence). On the other
ticipants could build. This created some tension as hand, the difficulty in completely barring offend-
community members attempted to find fair ways to ers from a site (since people can easily reappear us-
determine who would be allowed to build and how ing a different e-mail address) and the inability to
much. The solution, achieved through vote by par- otherwise physically enforce community standards
ticipants, was to create a review board elected by the has increased cybercommunities vulnerability to
community that would reject or approve proposed disruption. In some cases, the greater potential for
projects. Designers of cybercommunity forums have anonymity or at least pseudonymity online has
also had to consider what types of capabilities to give also facilitated antisocial behavior. Many cyber-
participants and what the social effects of those ca- communities have therefore tried to find ways to en-
pabilities might be. For instance, Active Worlds orig- hance trust between community members.
inally did not allow participants to have private Some have sought to increase accountability by
conversations that were not visible to all other par- making participants e-mail addresses or real life
ticipants in the same virtual space. The designers felt names available to other participants. Others have
that such conversations were antisocial and might set rules for behavior with the ultimate sanction be-
lead to conflicts. However, participants continued to ing the barring of an individual from the forum
request a command that enabled such whispered (sometimes technologically tricky to implement).
conversations, and also implemented other programs, LambdaMOO, for instance, posts a set of rules for
such as instant messaging, in order to work around polite behavior. Because it is also one of the most fa-
the forums limitations. The designers eventually ac- mous (and most documented) cybercommunities,
quiesced and added a whisper command. LambdaMOOs opening screen also displays rules of
Similarly, some MUDs have a command known conduct for journalists and academic researchers vis-
as mutter. Rather than letting you talk only to one iting the site.
other person, as is the case with whisper, mutter lets LiveJournal requires potential participants to ac-
you talk to everyone else in the virtual room except quire a code from an existing user in order to become
a designated person; in other words, it enables you a member, which it is hoped ensures that at least one
to talk behind a persons back while that person is person currently a member of the community
vouches for the new member. LiveJournal is con- vious prejudices might be left behind, enabling a
sidering abandoning this practice in favor of a utopian meeting of minds and ideas.
complex system of interpersonal recommendations So far, it appears that cybercommunities tend to
that give each participant a trust rating, theoretically augment rather than supplant peoples other social
an indication of their trustworthiness and status connections. They appear to contain many of the
within the community. same positive and negative social aspects present
Although perhaps not as complex, similar sys- in offline communities. Further, many cybercom-
tems are in use at other online forums. Slashdot, a munities emerge from existing offline groups, also
bulletin board service focusing primarily on com- include an offline component (including face-to-
puter-related topics, allows participants to rank post- face contact between at least some participants), or
ings and then to filter what they read by aggregated utilize other technologies such as the telephone to
rank. A participant can, for instance, decide to enhance connections. Whatever form cybercommu-
read only messages that achieve the highest aver- nities take in the future, their presence and popu-
age rating, as averaged from the responses of other larity from the earliest days of computer networks
participants. makes it clear that such interconnections will con-
The online auction site eBay has a feedback tinue to be a significant part of human-computer
system through which buyers and sellers rate one an- interaction.
others performance after each transaction, result-
ing in a numerical score for each registered member. Lori Kendall
Each instance of positive feedback bestows a point,
and each instance of negative feedback deletes one. See also Avatars; Digital Divide; MUDs
A recent change in the way auctions are displayed
now lists a percentage of positive feedback for each
seller. Users can also read the brief feedback mes- FURTHER READING
sages left for other users. These features are intended
to allow users to evaluate a persons trustworthiness Baym, N. K. (2000). Tune in, log on: Soaps, fandom, and online com-
munity. Thousand Oaks, CA: Sage.
prior to engaging in transactions with that person. Belson, K., & Richtel, M. (2003, May 5). Americas broadband
The degree to which these types of trust- dream is alive in Korea. The New York Times, p. C1.
promotion systems work to foster and enhance Benedikt, M. (Ed.). (1992). Cyberspace: First steps. Cambridge, MA:
community is unclear. Participants in various cyber- MIT Press.
Blackburg Electronic Village. (n.d.) About BEV. Retrieved August
communities continue to consider issues of trust and 12, 2003, from http://www.bev.net/about/index.php
to work on technological enhancements to the virtual Cherny, L. (1999). Conversation and community: Chat in a virtual
environment that will help suppress antisocial be- world. Stanford, CA: CSLI Publications.
Damer, B. (1998). Avatars! Berkeley, CA: Peachpit Press.
havior and promote greater community solidarity. Dibbell, J. (1998). My tiny life: Crime and passion in a virtual world.
New York: Henry Holt and Company.
Future Directions Gibson, W. (1984). Neuromancer. New York: Ace Books.
As cybercommunities first developed, mainstream Hafner, K. (2001). The Well: A Story of love, death & real life in the sem-
inal online community. Berkeley, CA: Carroll & Graf.
media commentary discussed a variety of hyperbolic Hampton, K. (2001). Living the wired life in the wired suburb: Netville,
fears and hopes. People feared that cybercommu- glocalization and civil society. Unpublished doctoral dissertation,
nities would replace and supplant other forms of University of Toronto, Ontario, Canada.
community and that cybercommunities were less Herring, S. C., with D. Johnson & T. DiBenedetto. (1995). This dis-
cussion is going too far! Male resistance to female participation
civilized, with greater potential for rude and anti- on the Internet. In M. Bucholtz & K. Hall (Eds.), Gender articu-
social behavior. On the other hand, people also hoped lated: Language and the socially constructed self (pp. 6796). New
that cybercommunities might provide forms of York: Routledge.
Jones, S. (Ed.). (1995). Cybersociety: Computer-mediated communica-
interconnectedness that had otherwise been lost in tion and community. Thousand Oaks, CA: Sage.
modern life. Some people also suggested that cyber- Jones, S. (Ed.). (1997). Virtual culture: Identity and communication in
communities could provide a forum in which pre- cybersociety. London: Sage.
CYBERSEX 141
Kavanaugh, A., & Cohill, A. (1999). BEV research studies, 1995 Internet, visiting sex-related websites, masturbating
1998. Retrieved August 12, 2003, from http://www.bev.net/about/ to sexual media from the Internet, engaging in sex-
research/digital_library/docs/BEVrsrch.pdf
Kendall, L. (2002). Hanging out in the virtual pub. Berkeley, CA:
ualized videoconferencing activities, creating sexual
University of California Press. materials for use/distribution on the Internet, and
Kiesler, S. (1997). Culture of the Internet. Mahwah, NJ: Lawrence using the Internet to obtain/enhance offline sexual
Erlbaum Associates. behaviors.
McDonough, J. (1999). Designer selves: Construction of technologically-
mediated identity within graphical, multi-user virtual environ- A broader term used to describe Internet sex-
ments. Journal of the American Society for Information Science, 50(10), ual behavior is online sexual activity (OSA), which
855869. includes using the Internet for any sexual purpose,
McDonough, J. (2000). Under construction. Unpublished doctoral dis-
sertation, University of California at Berkeley.
including recreation, entertainment, exploration, or
Morningstar, C., & Farmer, F. R. (1991). The lessons of Lucasfilms education. Examples of OSA are using online ser-
Habitat. In M. Benedikt (Ed.), Cyberspace: First steps (pp. 273302). vices to meet individuals for sexual /romantic pur-
Cambridge, MA: The MIT Press. poses, seeking sexual information on the Internet
Porter, D. (1997). Internet culture. New York: Routledge.
Renninger, K. A., & Shumar, W. (Eds.). (2002). Building virtual com-
(for instance, about contraception and STDs), and
munities. Cambridge, UK: Cambridge University Press. purchasing sexual toys/paraphernalia online. What
Rheingold, H. (1993). The virtual community: Homesteading on the distinguishes cybersex from OSA is that cybersex
electronic frontier. Reading, MA: Addison-Wesley. involves online behaviors that result in sexual arousal
Smith, M., & Kollock, P. (Eds.). (1999). Communities and cyber-
space. New York: Routledge. or gratification, while other online sexual activities
Taylor, T. L. (2002). Living digitally: Embodiment in virtual worlds. may lead to offline sexual arousal and gratifica-
In R. Schroeder (Ed.), The social life of avatars: Presence and inter- tion. Sexual arousal from cybersex is more immedi-
action in shared virtual environments. London: Springer Verlag.
Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
ate and is due solely to the online behavior.
Wellman, B. (2001). The persistence and transformation of community:
From neighbourhood groups to social networks. Report to the Law Venues
Commission of Canada. Retrieved August 12, 2003, from http:// Many people assume that the World Wide Web is the
www.chass.utoronto.ca/~wellman/publications/lawcomm/ main venue for cybersex. In fact, the Web repre-
lawcomm7.htm
Wellman, B., & Haythornthwaite, C. (Eds.). (2002). The Internet in
sents only a small portion of the places where cyber-
everyday life. Oxford, UK: Blackwell. sex activities can occur. Other areas of the Internet
Wellman, B., Boase, J., & Chen, W. (2002). The networked nature of where cybersex may take place include the following:
community online and offline. IT & Society, 1(1), 151165.
Weiner, N. (1948). Cybernetics, or control and communication in the Newsgroups This area serves as a bulletin
animal and the machine. Cambridge, MA: MIT Press. board where individuals can post text or multi-
WELL, The. (2002). About the WELL. Retrieved August, 2003, from
http://www.well.com/aboutwell.html
media messages, such as sexual text, pictures,
sounds, and videos;
E-mailE-mail can be used for direct com-
munication with other individuals or groups
of individuals. In the case of cybersex, the mes-
CYBERSEX sage may be a sexual conversation, story, picture,
sound, or video;
The term cybersex is a catch-all word used to describe ChatroomsBoth sexualized conversation
various sexual behaviors and activities performed and multimedia can be exchanged in chatrooms.
while on the Internet. The term does not indicate Casual users are familiar with Web-based chat-
that a particular behavior is good or bad, only that ting such as Yahoo Chat or America Online
the sexual behavior occurred in the context of the (AOL) Chat. Most Web-based chat areas have
Internet. Examples of behaviors or activities that may sections dedicated to sexual chats. However,
be considered cybersex include sexual conversations the largest chat-based system is the Internet Relay
in Internet chatrooms, retrieving sexual media (for Chat (IRC), an area largely unfamiliar to most
example, photographs, stories, or videos) via the casual users. In addition to text-based chatting,
IRC contains a number of chatrooms specifically that from 11 to 17 percent of individuals who en-
dedicated to the exchange of pornography gaged in cybersex report some consequences in their
through file servers; life and score moderately high on measures of gen-
Videoconferencing/Voice ChattingThe use of eral sexual compulsivity. In addition, approximately
these areas is rapidly increasing. As technology 6 percent report feeling out of control with their
improves and connection speeds increase, the Internet sexual behavior and scored high on mea-
use of the Internet for live cybersex sessions sures of sexual compulsivity.
will become commonplace. Videoconferencing
combined with voice chat constitutes a high-tech
version of a peep show mixed with an obscene Healthy Versus
phone call; and
Peer-to-Peer File SharingSoftware packages Problematic Cybersex
such as Napster and Kazaa have made file shar- One of the difficulties in defining cybersex as
ing a popular hobby. Casual users of this soft- either healthy or problematic is the fact that there
ware know its use for exchanging music files, but are few agreed-upon definitions about what con-
any file can be shared on the network, including stitutes sexually healthy behavior. Society has clearly
sexual images, sounds, and videos. delineated some behaviors as unhealthy, for exam-
ple, sex with children or other non-consenting part-
ners. However, people disagree about whether
Statistics masturbation, multiple affairs, bondage, and fetishes
Although the term cybersex often has negative con- are healthy or unhealthy. In the world of cybersex,
notations, research in this area suggests that these same gray areas exist between healthy and un-
nearly 80 percent of individuals who engage in healthy and are often even more difficult to define
Internet sex report no significant problems in their since the behavior does not include actual sexual
lives associated with their online sexual activities. contact.
Although this may be an underestimate since the re- It is also important not to assume that frequency
search relied on the self-reports of respondents, it is the key factor in determining whether an indi-
is safe to assume that the majority of individuals vidual is engaged in unhealthy cybersex. Some indi-
who engage in cybersex behavior report this activ- viduals engage in cybersex at a high frequency and
ity to be enjoyable and pleasurable, with few nega- have few problems, while others who engage in it
tive consequences. only a few hours a week have significant negative
However, there are individuals who engage in consequences.
cybersex who do report significant negative conse- Physician and researcher Jennifer Schneider pro-
quences as a result of their online sexual behavior. posed three criteria to help determine if some-
These individuals often report that their occupa- ones behavior has become compulsivethat is,
tional, social, or educational life areas have been neg- whether the person has crossed the line from a
atively impacted or are in jeopardy as a result of their recreational to a problematic user of cyber-
sexual use of the Internet. Often these individuals sex. The three criteria are (1) loss of freedom to
report a sense of being out of control or compulsive choose whether to stop the behavior; (2) negative
in their sexual use of the Internet and often compare consequences as a result of the behavior; and (3) ob-
it to addictions like gambling, eating, shopping, or sessive thinking about engaging in the behavior. The
working. Internet Sex Screening Test (ISS) described by coun-
Several large-scale studies estimate the per- seling professor David Delmonico and professor of
centage of individuals who are negatively impacted school psychology Jeffrey Miller can be used to con-
by cybersex behaviors. While exact numbers are im- duct initial screening of whether an individual has
possible given the size of the Internet, estimates are a problem with cybersex.
CYBERSEX 143
The Appeal of the Internet

With an estimated 94 million users accessing it Cybersex Addiction
regularly, it is difficult to dispute the Internets wide-
T
he Center for Online and Internet Addiction (www
spread appeal. In 2001 Delmonico, Moriarity, and .netaddiction.com) offers the following test to help
marriage and family therapist Elizabeth Griffin, pro- diagnose cybersex addiction:
posed a model called the Cyberhex for under-
1. Do you routinely spend significant amounts of time
standing why the Internet is so attractive to its users.
in chat rooms and private messaging with the sole
Their model lists the following six characteristics:
purpose of finding cybersex?
Integral: The Internet is nearly impossible to avoid.
2. Do you feel preoccupied with using the Internet to
Even if a cybersex user decided to never use the Internet
find on-line sexual partners?
again, the integral nature of the Internet would make
3. Do you frequently use anonymous communica-
that boundary nearly impossible, since many need the
tion to engage in sexual fantasies not typically car-
Internet for work, or to access bank information, and
ried out in real-life?
so on. In addition, public availability, the use of e-mail,
4. Do you anticipate your next on-line session with the
and other activities like shopping and research make
expectation that you will find sexual arousal or
the Internet a way of life that is integrated into our daily
gratification?
routines.
5. Do you find that you frequently move from cyber-
Imposing: The Internet provides an endless sup-
sex to phone sex (or even real-life meetings)?
ply of sexual material 7 days a week, 365 days a year.
6. Do you hide your on-line interactions from your sig-
The amount of information and the imposing nature
nificant other?
of marketing sexual information on the Internet
7. Do you feel guilt or shame from your on-line use?
contributes to the seductiveness of the world of cybersex.
8. Did you accidentally become aroused by cybersex at
Inexpensive: For a relatively small fee, twenty to
first, and now find that you actively seek it out when
forty dollars per month, a user can access an intoxi-
you log on-line?
cating amount of sexual material on the Internet. In
9. Do you masturbate while on-line while engaged in
the offline world such excursions can be cost-prohibitive
erotic chat?
to many.
10. Do you provide less investment with your real-life
Isolating: Cybersex is an isolating activity. Even
sexual partner only to prefer cybersex as a primary
though interpersonal contact may be made during the
form of sexual gratification?
course of cybersex, these relationships do not require
Source: Are you addicted to cybersex. Center for Online and Internet
the same level of social skills or interactions that offline Addiction. Retrieved March 23, 2004, from http://www.netaddiction.com/
behaviors require. The Internet becomes a world in it- resources/cybersexual_addiction_test.htm
self, where it is easy to lose track of time, consequences,

and real-life relationships. The isolation of cybersex of-
ten provides an escape from the real world, and while
everyone takes short escapes, cybersex often becomes Intoxicating: This is what happens when the pre-
the drug of choice to anesthetize any negative feelings ceding five elements are added together. This com-
associated with real-life relationships. bination makes for an incredibly intoxicating
Interactive: While isolating in nature, the Internet experience that is difficult for many to resist. The
also hooks individuals into pseudorelationships. These intoxication of the Internet is multiplied when cyber-
pseudorelationships often approximate reality with- sex is involved since behaviors are reinforced with
out running the risks of real relationshipslike emo- one of the most powerful rewards, sex.
tional and physical vulnerability and intimacy. This Any single aspect of the Internet can be pow-
close approximation to reality can be fuel for the fan- erful enough to entice a cybersex user. However,
tasy life of those who experience problems with their it is typically a combination of these six factors that
cybersex behaviors. draws problematic cybersex users into their rituals
and leads to their loss of control over their cyber- be useful in educating children and adolescents about
sex use. sexuality, it can also be a dangerous venue for the de-
velopment of compulsive behavior and victimiza-
tion by online predators. Although the effect of
Special Populations hardcore, explicit pornography on the sexual devel-
opment of children and adolescents has yet to be re-
Engaged in Cybersex searched, early exposure to such pornography may
The following subgroups of cybersex users have been impact their moral and sexual development.
studied in some detail: Physically or Developmentally Challenged
Males and Females: In the early to mid-1990s People: Only recently have questions been raised
there were three times as many males online as fe- about the appropriate use of the Internet for sex-
males. Recent research shows that the gap has closed ual and relational purposes among physically chal-
and that the split between male and female Internet lenged individuals. This area warrants more research
users is nearly fifty-fifty. As a result, research on cyber- and exploration, but initial writings in this area sug-
sex behavior has also included a significant number gest that the Internet can confer a tremendous ben-
of females who engage in cybersex. Most of this re- efit for sexual and relationship exploration for
search suggests that men tend to engage in more vi- persons with disabilities.
sual sex (for example, sexual media exchange), while While sex on the Internet can be a positive ex-
women tend to engage in more relational sex (for perience for these subpopulations, it can also intro-
example, chatrooms and e-mail). Females may duce the people in these groups to the same problems
find the Internet an avenue to sexual exploration and associated with cybersex that other groups report.
freedom without fear of judgment or reprisal from
society. In this way, the Internet can have genuine
benefits. Implications
Gays and Lesbians: Researchers have reported Cybersex is changing sexuality in our culture. The
that homosexuals tend to engage in cybersex at higher positive side is that sexual behavior is becoming more
levels than heterosexuals, which may be because they open and varied, and better understood. The nega-
dont have to fear negative cultural responses or even tive implications are that sexuality may become ca-
physical harm when they explore sexual behaviors sual, trivial, and less relational.
and relationships on the Internet. Some homosexu- The pornography industry continues to take ad-
als report that cybersex is a way to engage in sexual vantage of the new technologies with the primary
behavior without fear of HIV or other sexually trans- goal of profit, and these new technologies will allow
mitted diseases. By offering homosexuals a safe for faster communication to support better video
way to explore and experience their sexuality, the and voice exchanges. The eventual development of
Internet gives them freedom from the stigma often virtual reality technologies online will further en-
placed on them by society. hance the online sexual experience, and perhaps make
Children and Adolescents: Studies conducted by the sexual fantasy experience more pleasurable than
AOL and Roper Starch revealed that children use the real life. These technological advances will continue
Internet not only to explore their own sexuality and to alter the way we interact and form relationships
relationships, but also to gather accurate sexual health with others.
information. Since many young adults have grown Researchers are just starting to realize the impli-
up with the Internet, they often see it through a cations of sex on the Internet. Theories like Cyberhex
different lens than adults. Children, adolescents, and are helpful in understanding why people engage in
young adults use the Internet to seek answers to a cybersex, but the best methods for helping those strug-
multitude of developmental questions, including gling with cybersex have yet to be discovered. However,
sexuality, which they may be afraid to address di- society will continue to be impacted by the Internet
rectly with other adults. Although the Internet can and cybersex. Parents, teachers, and others who
CYBORGS 145
have not grown up with the Internet will fail future

generations if they discount the significant impact it CYBORGS
can have on social and sexual development. Continued
research and education will be necessary to help in- A cyborg is a technologically enhanced human be-
dividuals navigate the Internet and the world of cyber- ing. The word means cybernetic organism. Because
sex more safely. many people use the term cybernetics for computer
science and engineering, a cyborg could be the fu-
David L. Delmonico and Elizabeth J. Griffin sion of a person and a computer. Strictly speaking,
however, cybernetics is the science of control pro-
See also Chatrooms; Cybercommunities cesses, whether they are electronic, mechanical, or
biological in nature. Thus, a cyborg is a person, some
of whose biological functions have come under tech-
FURTHER READING nological control, by whatever means. The standard
term for a computer-simulated person is an avatar,
Carnes, P. J. (1983). Out of the shadows. Minneapolis, MN: CompCare. but when an avatar is a realistic copy of a specific real
Carnes, P. J., Delmonico, D. L., Griffin, E., & Moriarity, J. (2001). In
the shadows of the Net: Breaking free of compulsive online behavior.
person, the term cyclone is sometimes used, a cyber-
Center City, MN: Hazelden Educational Materials. netic clone or virtual cyborg.
Cooper, A. (Ed.). (2000). Sexual addiction & compulsivity: The jour-
nal of treatment and prevention. New York: Brunner-Routledge.
Cooper, A. (Ed.). (2002). Sex and the Internet: A guidebook for clini-
cians. New York: Brunner-Routledge. Imaginary Cyborgs
Cooper, A., Delmonico, D., & Burg, R. (2000). Cybersex users, abusers, The earliest widely known cyborg in literature,
and compulsives: New findings and implications. Sexual Addiction dating from the year 1900, is the Tin Woodman in
and Compulsivity: Journal of Treatment and Prevention, 7, 529.
Cooper, A., Scherer, C., Boies, S. C., & Gordon, B. (1999). Sexuality on
The Wonderful Wizard of Oz by L. Frank Baum.
the Internet: From sexual exploration to pathological expres- Originally he was a man who earned his living chop-
sion. Professional Psychology: Research and Practice, 30(2), 154164. ping wood in the forest. He wanted to marry a beau-
Delmonico, D. L. (1997). Internet sex screening test. Retrieved tiful Munchkin girl, but the old woman with whom
August 25, 2003, from http://www.sexhelp.com/
Delmonico, D. L., Griffin, E. J., & Moriarity, J. (2001). Cybersex un-
the girl lived did not want to lose her labor and pre-
hooked: A workbook for breaking free of compulsive online behavior. vailed upon the Wicked Witch of the East to enchant
Wickenburg, AZ: Gentle Path Press. his axe. The next time he went to chop wood, the axe
Delmonico, D. L., & Miller, J. A. (2003). The Internet sex screening chopped off his left leg instead. Finding it incon-
test: A comparison of sexual compulsives versus non-sexual com-
pulsives. Sexual and Relationship Therapy, 18(3), 261276. venient to get around without one of his legs, he went
Robert Starch Worldwide, Inc. (1999). The America Online/Roper to a tinsmith who made a new one for him. The
Starch Youth Cyberstudy. Author. Retrieved on December 24, 2003, axe then chopped off his right leg, which was then
from http://www.corp.aol.com/press/roper.html/
Schneider, J. P. (1994). Sex addiction: Controversy within mainstream
also replaced by one made of tin. This morbid process
addiction medicine, diagnosis based on the DSM-III-R and physi- continued until there was nothing left of the origi-
cian case histories. Sexual Addiction & Compulsivity: The Journal nal man but his heart, and when that finally was
of Treatment and Prevention, 1(1), 1944. chopped out, he lost his love for the Munchkin girl.
Schneider, J. P., & Weiss, R. (2001). Cybersex exposed: Recognizing the
obsession. Center City, MN: Hazelden Educational Materials.
Still, he missed the human emotions that witchcraft
Tepper, M. S., & Owens, A. (2002). Access to pleasure: Onramp to and technology had stolen from him, and was ready
specific information on disability, illness, and other expected to join Dorothy on her journey to the Emerald City,
changes throughout the lifespan. In A. Cooper (Ed.), Sex and on a quest for a new heart.
the Internet: A guidebook for clinicians. New York: Brunner-
Routledge. This story introduces one of the primary themes
Young, K. S. (1998). Caught in the Net. New York: Wiley. associated with cyborgs: the idea that a person ac-
Young, K. S. (2001). Tangled in the web: Understanding cybersex from cepts the technology to overcome a disability. That
fantasy to addiction. Bloomington, IN: 1st Books Library.
is, the person is already less than complete, and the
technology is a substitute for full humanity, albeit an
inferior one. This is quite different from assimilating
new technology in order to become more than hu- become its enemy. This reflects a second persistent
man, a motive severely criticized by the Presidents literary theme associated with cyborgs: They re-
Council on Bioethics in 2003. A very different flect the evils of an oppressive society in which tech-
viewpoint on what it means to be disabled has been nology has become a tool by which the masters
expressed by Gregor Wolbring, a professor at the enslave the majority.
University of Calgary. Who decides the meanings By far the most extensive treatment of the idea
of disability and normality is largely a political issue, that cyborg technology is wicked can be found in the
and Wolbring argues that people should generally Dalek menace from the long-running BBC televi-
have the power to decide for themselves. He notes the sion series, Dr. Who. Sometimes mistaken for robots,
example of children who are born without legs be- Daleks are metal-clad beings that resemble huge salt
cause their mothers took thalidomide during preg- shakers, wheeled trash cans, or British post boxes.
nancy, then forced to use poorly designed artificial They became extremely popular villains since their
legs because that makes them look more normal, first appearance in 1963. Two low-budget feature
when some other technology would have given them films that retold the first TV serials added to their
far better mobility. fame, Dr. Who and the Daleks (1965) and Dr. Who:
Daleks Invasion Earth 2150 A.D. (1966). Inside a
Daleks metal shell lurks a helpless, sluggish creature
C makes it easy to shoot yourself in the foot. C++ makes with vestigial claws, yet the combination of biol-
it harder, but when you do, it blows away your whole leg. ogy and technology gave it the possibility of con-
Bjarne Stroustrup quering the universe. Their motto describes how they
treat all other living creatures: Exterminate. The
secret of their origins is revealed in the 1975 serial,
A common variation on the disability theme is Genesis of the Daleks. The protagonist of Dr. Who,
the hero who suffers a terrible accident, is rebuilt, The Doctor, lands his time machine on the battle-
and becomes a cyborg superhero. A well-known ex- scarred planet Skaro, just as the nuclear war between
ample is The Six Million Dollar Man, a television se- the Thals and the Kaleds reaches its climax. Davros,
ries that aired 19731978 and was based on the 1972 the evil (and disabled) Kaled scientist, recognizes
novel Cyborg by Martin Caidin. Test pilot Steve Austin that chemical weapons are causing his people to mu-
is severely injured in a plane crash, then rebuilt with tate horribly, and rather than resist this trend, he
bionic (biological plus electronic) technology. A spin- accelerates it, transforming humans into the vile
off series, The Bionic Woman (19761978) focuses Dalek cyborgs.
on tennis player Jaime Sommers who is similarly dis-
abled in a parachute accident. Both become super-
hero special agents, perhaps to justify the heavy Real Cyborg Research
investment required to insert and maintain their Since human beings began wearing clothing, the
bionics. An especially striking example is the motion boundary between ourselves and our technology has
picture Robocop (1987). Policeman Alex Murphy lives blurred. Arguably, everybody who wears a wristwatch
in a depressing future Detroit, dominated by a single, or carries a cell phone is already a cyborg. But the
exploitative corporation. To control the increasingly usual definition implies that a human body has been
violent population, the corporation develops robot modified, typically by insertion of some nonbiological
police possessing overwhelming firepower but lack- technology. In the early years of the twentieth cen-
ing the judgment to interact successfully with human tury, when surgeons first gained technological
beings. Thus, when Murphy is blown to pieces by control over pain and infection, many brave or irre-
criminals, the corporation transforms him into a cy- sponsible doctors began experimenting with im-
borg that combines human judgment with machine provements to their patients. Sir William Arbuthnot
power. The corporation denies Murphy the right Lane, the British royal physician, theorized that many
to be considered human, thereby forcing him to illnesses were caused by a sluggish movement of food
CYBORGS 147
through the bowels that supposedly flooded the sys- safely, and how to interface active or sensory com-
tem with poisonous toxins. Diagnosing this chronic ponents to the human nervous system. Several re-
intestinal stasis in many cases, Lane performed searchers, such as Miguel Nicolelis of Duke
surgery to remove bands and adhesions, and free the University, have been experimenting with brain im-
intestines to do their job. Some of his colleagues op- plants in monkeys that allow them to operate artifi-
erated on neurotic patients, believing that moving cial arms, with the hope that this approach could be
the abdominal organs into their proper places could applied therapeutically to human beings in the near
alleviate mental disorders. Later generations of doc- future.
tors abandoned these dangerous and useless pro-
cedures, but one of Lanes innovations has persisted.
He was the first to plate a bonethat is to screw Visions of the Future
a supportive metal plate onto a broken bone. Today Kevin Warwick, professor of Cybernetics at Reading
many thousands of people benefit from artificial hip University in Britain, is so convinced of the near-term
and knee joints. prospects for cyborg technology, that he has experi-
In World War I, even before the introduction mented on his own body. In 1998, he had surgeons
of antibiotics, rigorous scientific techniques were implant a transponder in his left arm so a computer
sufficiently effective to prevent death from infection could monitor his movements. His first implant
in most wounded cases, thereby vastly increasing the merely consisted of a coil that picked up power from
number of people who survived with horrendous a transmitter and reflected it back, letting the com-
war-caused disabilities. The Carrel-Dakin technique puter know where he was so it could turn on lights
was especially impressive, employing an antiseptic when he entered a room. In 2002 he had neuro-
solution of sodium hypochlorite in amazingly rig- surgeons connect his nervous system temporarily
orous procedures. Suppose a soldiers leg had been to a computer for some very modest experiments, but
badly torn by an artillery shell. The large and irreg- in the future he imagines that implants interfacing
ular wound would be entirely opened up and cleaned. between computers and the human nervous system
Then tubes would be placed carefully in all parts of will allow people to store, playback, and even share
the wound to drip the solution very slowly, for experiences. He plans someday to experiment with
days and even for weeks. Daily, a technician takes the stored perceptions associated with drinking wine,
samples from every part of the wound, examining to see if playing them back really makes him feel
them under the microscope, until no more microbes drunk. His wife, Irena, has agreed that someday
are seen and the wound can be sewed up. Restorative they both will receive implants to share feelings
plastic surgery and prosthetics could often help the such as happiness, sexual arousal, and even pain. In
survivors live decent lives. the long run, Warwick believes, people will join
In the second half of the twentieth century, much with their computers to become superhuman cyborgs.
progress was achieved with transplants of living In so doing, they will adopt a radically new concep-
tissuesuch as kidneys from donors and coronary tion of themselves, including previously unknown
artery bypass grafts using material from the patient. understandings, perceptions, and desires.
Inorganic components were also successfully in- Natasha Vita-More, an artist and futurist, has
troduced, from heart values to tooth implants. sketched designs for the cyborg posthuman she
Pacemakers to steady the rhythm of the heart and calls Primo, based on aesthetics and general techno-
cochlear transplants to overcome deafness are among logical trends. Although she is not building proto-
the relatively routine electronic components inserted types or experimenting with components at the
into human bodies, and experiments are being present time, she believes that her general vision could
carried out with retina chips to allow the blind to be achieved within this century. Primo would be age-
see. There are many difficult technical challenges, less rather than mortal, capable of upgrades when-
notably how to power artificial limbs, how to connect ever an organ wore out or was made obsolete by
large components to the structure of the human body technical progress, and able to change gender
whenever (s)he desires. Nanotechnology would give technology takes me above and beyond myself. A per-
Primo 1,000 times the brainpower of a current hu- son who is thoroughly plugged in experiences radi-
man, and thus capable of running multiple viewpoints cally transformed consciousness: I construct, and I
in parallel rather than being locked into one narrow am constructed, in a mutually recursive process
frame of awareness. Primos senses would cover a that continually engages my fluid, permeable bound-
vastly wider bandwidth, with sonar mapping onto the aries and my endlessly ramifying networks. I am a
visual field at will, an internal grid for navigating and spatially extended cyborg (Mitchell 2003, 39).
moving anywhere like an acrobatic dancer with
perfect sense of direction, and a nervous system Williams Sims Bainbridge
that can transmit information from any area of the
body to any other instantly. Primos nose could iden-
tify any chemical or biological substance in the en- FURTHER READING
vironment, and smart skin will not only protect the
Bainbridge, W. S. (1919). Report on medical and surgical developments
body, but provide vastly enhanced sensations. Instead of the war. Washington, DC: Government Printing Office.
of the depression and envy that oppress modern Barnes, B. A. (1977). Discarded operations: Surgical innovation by
humans, (s)he would be filled with ecstatic yet real- trial and error. In J. P. Bunker, B. A. Barnes, & F. Mosteller (Eds.),
istic optimism. The old fashioned bodys need to elim- Costs, risks, and benefits of surgery (pp. 109123). New York: Oxford
University Press.
inate messy wastes will be transcended by Primos Baum, L. F. (1900). The wonderful wizard of Oz. Chicago: G. M. Hill.
ability to recycle and purify. Bentham, J. (1986). Doctor Who: The early years. London: W. H. Allen.
William J. Mitchell, the director of Media Arts and Caidin, M. (1972). Cyborg. New York: Arbor House.
Haining, P. (Ed.). (1983). Doctor Who: A celebration. London:
Sciences at the Massachusetts Institute of Technology, W. H. Allen.
argues that we have already evolved beyond tradi- Mitchell, W. J. (2003). ME++: The cyborg self and the networked city.
tional homo sapiens by become embedded in a Cambridge, MA: MIT Press.
ubiquitous communication network. The title of Nicolelis, M. A. L., & Srinivasan, M. A. (2003). Human-machine
interaction: Potential impact of nanotechnology in the design of
his book, ME++: The Cyborg Self and the Networked neuroprosthetic devices aimed at restoring or augmenting human
City (2003), offers a nice metaphor derived from performance. In M. C. Roco & W. S. Bainbridge (Eds.), Converging
the C language for programming computers. C was technologies for improving human performance (pp. 251255).
originally developed by a telephone company (Bell Dordrecht, Netherlands: Kluwer.
Presidents Council on Bioethics. (2003). Beyond therapy: Biotechnology
Labs) and has become possibly the most influential and the pursuit of happiness. Washington, DC: Presidents Council
language among professional programmers, especially on Bioethics.
in the modular version called C++. In C (and in the Warwick, K. (2000). Cyborg 1.0. Wired, 8(2), 144151.
Wolbring, G. (2003). Science and technology and the triple D (Disease,
Java language as well), ++ means to increment a Disability, Defect). In M. C. Roco & W. S. Bainbridge (Eds.),
number by adding 1 to it. Thus, C++ is one level more Converging technologies for improving human performance
than C, and ME++ is one level more than me, in which (pp. 232243). Dordrecht, Netherlands: Kluwer.
DATA MINING
DATA VISUALIZATION
DEEP BLUE
DENIAL-OF-SERVICE ATTACK
DESKTOP METAPHOR
D
DIALOG SYSTEMS
DIGITAL CASH
DIGITAL DIVIDE
DIGITAL GOVERNMENT
DIGITAL LIBRARIES
DRAWING AND DESIGN
future data samples. Only then can the rules and mod-
DATA MINING els obtained be considered meaningful. The discov-
ered patterns should also be novel, that is, not already
Data mining is the process of automatic discovery known to experts; otherwise, they would yield very
of valid, novel, useful, and understandable patterns, little new understanding. Finally, the discoveries
associations, changes, anomalies, and statistically sig- should be useful as well as understandable.
nificant structures from large amounts of data. It is Typically data mining has two high-level goals:
an interdisciplinary field merging ideas from sta- prediction and description. The former answers the
tistics, machine learning, database systems and data- question of what and the latter the question of
warehousing, and high-performance computing, as why. For prediction, the key criterion is the accuracy
well as from visualization and human-computer in- of the model in making future predictions; how
teraction. It was engendered by the economic and the prediction decision is arrived at may not be
scientific need to extract useful information from important. For description, the key criterion is the
the data that has grown phenomenally in all spheres clarity and simplicity of the model describing the
of human endeavor. data in understandable terms. There is sometimes
It is crucial that the patterns, rules, and models a dichotomy between these two aspects of data min-
that are discovered be valid and generalizable not only ing in that the most accurate prediction model for
in the data samples already examined, but also in a problem may not be easily understandable, and the
149
most easily understandable model may not be highly mining algorithmsassociation rule discovery, se-
accurate in its predictions. quence mining, classification tree induction, clus-
tering, and so onto analyze the data.
Interpret, evaluate, and visualize patterns: After
Steps in Data Mining the algorithms have produced their output, it is still
Data mining refers to the overall process of discov- necessary to examine the output in order to inter-
ering new patterns or building models from a pret and evaluate the extracted patterns, rules, and
given dataset. There are many steps involved in the models. It is only by this interpretation and evalu-
mining enterprise. These include data selection, data ation process that new insights on the problem be-
cleaning and preprocessing, data transformation and ing analyzed can be derived.
reduction, data mining task and algorithm selection,
and finally, postprocessing and the interpretation
of discovered knowledge. Here are the most im- Data Mining Tasks
portant steps: In verification-driven data analysis the user postu-
Understand the application domain: A proper lates a hypothesis, and the system tries to validate it.
understanding of the application domain is neces- Common verification-driven operations include
sary to appreciate the data mining outcomes desired querying and reporting, multidimensional analysis,
by the user. It is also important to assimilate and take and statistical analysis. Data mining, on the other
advantage of available prior knowledge to maximize hand, is discovery driventhat is, it automatically
the chance of success. extracts new hypotheses from data. The typical data
Collect and create the target dataset: Data min- mining tasks include the following:
ing relies on the availability of suitable data that Association rules: Given a database of trans-
reflects the underlying diversity, order, and structure actions, where each transaction consists of a set of
of the problem being analyzed. Therefore, it is cru- items, association discovery finds all the item sets
cial to collect a dataset that captures all the possi- that frequently occur together, and also the rules
ble situations relevant to the problem being analyzed. among them. For example, 90 percent of people
Clean and transform the target dataset: Raw data who buy cookies also buy milk (60 percent of gro-
contain many errors and inconsistencies, such as cery shoppers buy both).
noise, outliers, and missing values. An important el- Sequence mining: The sequence-mining task is
ement of this process is the unduplication of data to discover sequences of events that commonly oc-
records to produce a nonredundant dataset. Another cur together. For example, 70 percent of the people
important element of this process is the normal- who buy Jane Austens Pride and Prejudice also buy
ization of data records to deal with the kind of Emma within a month.
pollution caused by the lack of domain consistency. Similarity search: An example is the problem
Select features and reduce dimensions: Even where a person is given a database of objects and a
after the data have been cleaned up in terms of elim- query object, and is then required to find those
inating duplicates, inconsistencies, missing values, objects in the database that are similar to the query
and so on, there may still be noise that is irrelevant object. Another example is the problem where a per-
to the problem being analyzed. These noise attrib- son is given a database of objects, and is then required
utes may confuse subsequent data mining steps, pro- to find all pairs of objects in the databases that are
duce irrelevant rules and associations, and increase within some distance of each other.
computational cost. It is therefore wise to perform Deviation detection: Given a database of objects,
a dimension-reduction or feature-selection step to find those objects that are the most different from
separate those attributes that are pertinent from those the other objects in the databasethat is, the out-
that are irrelevant. liers. These objects may be thrown away as noise, or
Apply data mining algorithms: After perform- they may be the interesting ones, depending on the
ing the preprocessing steps, apply appropriate data specific application scenario.
DATA MINING 151
Classification and regression: This is also called stream. Extracted knowledge thus needs to be con-
supervised learning. In the case of classification, some- stantly updated.
one is given a database of objects that are labeled with Database integration: The various steps of the
predefined categories or classes. They are required mining process, along with the core data mining
to develop from these objects a model that separates methods, need to be integrated with a database
them into the predefined categories or classes. system to provide common representation, storage,
Then, given a new object, the learned model is applied and retrieval. Moreover, enormous gains are possi-
to assign this new object to one of the classes. In the ble when these are combined with parallel data-
more general situation of regression, instead of pre- base servers.
dicting classes, real-valued fields have to be predicted. Privacy and security issues in mining: Privacy-
Clustering: This is also called unsupervised learn- preserving data mining techniques are invaluable in
ing. Here, given a database of objects that are usu- cases where one may not look at the detailed data,
ally without any predefined categories or classes, the but one is allowed to infer high-level information.
individual is required to partition the objects into This also has relevance for the use of mining for na-
subsets or groups such that elements of a group share tional security applications.
a common set of properties. Moreover, the partition Human interaction: While a data mining algo-
should be such that the similarity between members rithm and its output may be readily handled by a
of the same group is high and the similarity between computer scientist, it is important to realize that the
members of different groups is low. ultimate user is often not the developer. In order for
a data mining tool to be directly usable by the ulti-
mate user, issues of automationespecially in the
Challenges in Data Mining sense of ease of usemust be addressed. Even for
Many existing data mining techniques are usually ad computer scientists, the use and incorporation of
hoc; however, as the field matures, solutions are prior knowledge into a data mining algorithm is often
being proposed for crucial problems like the incor- a challenge; they too would appreciate data mining
poration of prior knowledge, handling missing data, algorithms that can be modularized in a way that fa-
adding visualization, improving understandability, cilitates the exploitation of prior knowledge.
and other research challenges. These challenges in- Data mining is ultimately motivated by the
clude the following: need to analyze data from a variety of practical
Scalability: How does a data mining algorithm applications from business domains such as
perform if the dataset has increased in volume and finance, marketing, telecommunications, and man-
in dimensions? This may call for some innovations ufacturing, or from scientific fields such as biology,
based on efficient and sufficient sampling, or on a geology, astronomy, and medicine. Identifying new
trade-off between in-memory and disk-based pro- application domains that can benefit from data min-
cessing, or on an approach based on high-perfor- ing will lead to the refinement of existing techniques,
mance distributed or parallel computing. and also to the development of new methods where
New data formats: To date, most data mining recurrent tools are inadequate.
search has focused on structured data, because it is
the simplest and most amenable to mining. However, Mohammed J. Zaki
support for other data types is crucial. Examples in-
clude unstructured or semistructured (hyper)text,
temporal, spatial, and multimedia databases. Mining FURTHER READING
these is fraught with challenges, but it is necessary
because multimedia content and digital libraries pro- Association for Computing Machinerys special interest group on
knowledge discovery and data mining. Retrieved August 21, 2003,
liferate at astounding rates. from http://www.acm.org/sigkdd.
Handling data streams: In many domains the Dunham, M. H. (2002). Data mining: Introductory and advanced top-
data changes over time and/or arrives in a constant ics. Upper Saddle River, NJ: Prentice Hall.
Han, J., & Kamber, M. (2000). Data mining: Concepts and techniques. ping, rendering, and viewing. The data-generation
San Francisco: Morgan Kaufman. step can be a numerical simulation, a laboratory
Hand, D. J., Mannila, H., & Smyth, P. (2001). Principles of data mining.
experiment, a collection of sensors, an image scan-
Kantardzic, M. (2002). Data mining: Concepts, models, methods, and ner, or a recording of Web-based business trans-
algorithms. Somerset, NJ: Wiley-IEEE Press. actions. Filtering removes noise, extracts and
Witten, I. H., & Frank, E. (1999). Data mining: Practical machine learn- enhances features, or rescales data. Mapping derives
ing tools and techniques with Java implementations. San Francisco:
Morgan Kaufmann. appropriate representations of data for the render-
ing step. The representations can be composed of
geometric primitives such as point, lines, poly-
gons, and surfaces, supplemented with properties
DATA VISUALIZATION such as colors, transparency, textures. Whereas the
visualization of a computerized tomography (CT)
Data visualization is a new discipline that uses com- scan of a fractured bone should result in an image
puters to make pictures that elucidate a concept, of a bone, plenty of room for creativity exists when
phenomenon, relationship, or trend hidden in a making a visual depiction of the trend of a stock mar-
large quantity of data. By using interactive three- ket or the chemical reaction in a furnace.
dimensional (3D) graphics, data visualization goes Rendering generates two-dimensional or
beyond making static illustrations or graphs and em- three-dimensional images based on the mapping re-
phasizes interactive exploration. sults and other rendering parameters, such as the
The pervasive use of computers in all fields of lighting model, viewing position, and so forth. Finally,
science, engineering, medicine, and commerce has the resulting images are displayed for viewing. Both
resulted in an explosive growth of data, presenting photorealistic and nonphotorealistic rendering tech-
people with unprecedented challenges in under- niques exist for different purposes of visual com-
standing data. Data visualization transforms raw data munication. Nonphotorealistic rendering, which
into pictures that exploit the superior visual pro- mimics how artists use brushes, strokes, texture, color,
cessing capability of the human brain to detect layout, and so forth, is usually used to increase the
patterns and draw inferences, revealing insights hid- clarity of the spatial relationship between objects,
den in the data. For example, data visualization al- improve the perception of an objects shape and size,
lows us to capture trends, structures, and anomalies or give a particular type of media presentation.
in the behavior of a physical process being modeled Note that the filtering and mapping steps are
or in vast amounts of Internet data. Furthermore, it largely application dependent and often require do-
provides us with a visual and remote means to com- main knowledge to perform. For example, the fil-
municate our findings to others. tering and mapping steps for the visualization of
Since publication of a report on visualization in website structure or browsing patterns would be
scientific computing by the U.S. National Science quite different from those of brain tumors or
Foundation in 1987, both government and industry bone fractures.
have invested tremendous research and development A data-visualization process is inherently itera-
in data-visualization technology, resulting in advances tive. That is, after visualization is made, the user
in visualization and interactive techniques that have should be able to go back to any previous steps,
helped lead to many scientific discoveries, better en- including the data-generation step, which consists
gineering designs, and more timely and accurate o f a nu m e r i c a l o r p hy s i c a l m o d e l , to m a ke
medical diagnoses. changes such that more information can be obtained
from the revised visualization. The changes may be
made in a systematic way or by trial and error.
Visualization Process The goal is to improve the model and understand-
A typical data-visualization process involves mul- ing of the corresponding problem via this visual
tiple steps, including data generation, filtering, map- feedback process.
DATA VISUALIZATION 153
Computational Steering visualization can offer better 3D spatial acuity than

Data visualization should not be performed in iso- humans have, and computer-assisted surgery is more
lation. It is an integral part of data analysis and the reliable and reproducible. However, computer-
scientific discovery process. Appropriate visualiza- assisted surgery has several challenges. First, the en-
tion tools integrated into a modeling process can tire visualization processconsisting of 3D
greatly enhance scientists productivity, improve the reconstruction, segmentation, rendering, and image
efficiency of hardware utilization, and lead to scien- transport and displaymust be an integrated part
tific breakthroughs. The use of visualization to drive of the end-to-end surgical planning and procedure.
scientific discovery processes has become a trend. Second, the visualization must be in real time, that
However, we still lack adequate methods to achieve is, flicker free. Delayed visual response could lead to
computational steeringthe process of inter- dangerous outcomes for a surgery patient. Most im-
acting with as well as changing states, parameters, or portant, the visualization must attain the required
resolution of a numerical simulationand to be able accuracy and incorporate quantitative measuring
to see the effect immediately, without stopping or mechanisms.
restarting the simulation. Consequently, the key to Telesurgery, which allows surgeons at remote sites
successful data visualization is interactivity, the abil- to participate in surgery, will be one of the major ap-
ity to effect change while watching the changes plications of virtual reality and augmented reality
take effect in real time on the screen. (where the virtual world and real world are allowed
If all the steps in the modeling and visualization to coexist and a superimposed view is presented to
processes can be performed in a highly interactive the user). Due to the distance between a patient and
fashion, steering can be achieved. The ability to steer surgeons, telesurgery has much higher data visual-
a numerical model makes the visualization process ization, hardware, and network requirements.
a closed loop, becoming a scientific discovery process Nevertheless, fast-improving technology and de-
that is self-contained. Students can benefit from such creasing costs will make such surgery increasingly
a process because they can more easily move from appealing. Whether and how much stereoscopic
concepts to solutions. Researchers can become much viewing can benefit surgeons remains to be investi-
more productive because they can make changes ac- gated. The most needed advance, however, is in inter-
cording to the interactive graphical interpretation of face software and hardware.
the simulation states without restarting the simu-
lation every time.
Computational steering has been attempted in User Interfaces for Data Visualization
only a few fields. An example is the SCIRun system Most data visualization systems supply the user with
used in computational medicine. To adopt compu- a suit of visualization tools that requires the user
tational steering, researchers likely must redesign the to be familiar with both the corresponding user in-
computational model that is required to incorporate terfaces and a large visualization parameter space (a
feedback and changes needed in a steering process. multidimensional space which consists of those
More research is thus needed to make computational input variables used by the visualization program).
steering feasible in general. Intuitive and intelligent user interfaces can greatly
assist the user in the process of data exploration. First,
the visual representation of the process of data ex-
Computer-Assisted Surgery ploration and results can be incorporated into the
During the past ten years significant advances have user interface of a visualization system. Such an inter-
been made in rendering software and hardware tech- face can help the user to keep track of the visualiza-
nologies, resulting in higher fidelity and real-time tion experience, use it to generate new visualizations,
visualization. Computer-assisted surgery is an and share it with others. Consequently, the interface
application of such advanced visualization tech- needs to display not only the visualizations but
nologies with a direct societal impact. Computer also the visualization process to the user. Second, the
task of exploring large and complex data and visu- by visual perception study, and user studies for mea-
alization parameter space during the mapping step suring the usability of visualization tools and the suc-
can be delegated to an intelligent system such as a cess of visualizations.
neural network. One example is to turn the 3D
segmentation problem into a simple 2D painting Kwan-Liu Ma
process for the user, leaving the neural network to
classify the multidimensional data. As a result, the See also Information Spaces; Sonification
user can focus on the visualizations rather than on
the user interface widgets (e.g., a color editor, plot-
ting area, or layout selector) for browsing through FURTHER READING
the multidimensional parameter space. Such next-
generation user interfaces can enhance data un- Johnson, C., & Parker, S. (1995). Applications in computational med-
icine using SCIRun: A computational steering programming
derstanding while reducing the cost of visualization environment. The 10th International Supercomputer Conference
by eliminating the iterative trial-and-error process (pp. 219).
of parameter selection. For routine analysis of large- Ma, K. L. (2000). Visualizing visualizations: Visualization viewpoints.
scale data sets, the saving can be tremendous. IEEE Computer Graphics and Applications, 20(5), 1619.
Ma, K.-L. (2004). VisualizationA quickly emerging field. Computer
Graphics, 38(1), 47.
McCormick, B., DeFanti, T., & Brown, M. (1987). Visualization in sci-
Research Directions entific computing. Computer Graphics, 21(6)
The pervasiveness of the World Wide Web in aver-
age peoples lives has led to a data explosion. Some
data are relevant to some peoples needs, but most
are not. Nevertheless, many people do their every- DEEP BLUE
day jobs by searching huge databases of information
distributed in locations all over the world. A large In 1997, the chess machine Deep Blue fulfilled a
number of computer services repeatedly operate on long-standing challenge in computer science by de-
these databases. Information visualization, a branch feating the human world chess champion, Garry
of visualization, uses visual-based analysis of data Kasparov, in a six-game match. The idea that a com-
with no spatial references, such as a large amount of puter could defeat the best humanity had to offer in
text and document. A data mining step (the proce- an intellectual game such as chess brought many
dure to reduce the size, dimensionality, and/or com- important questions to the forefront: Are com-
plexity of a data set), which may be considered as the puters intelligent? Do computers need to be intelli-
filtering step, usually precedes the picture-making gent in order to solve difficult or interesting
step of visualization. The mapping step often con- problems? How can the unique strengths of humans
verts reduced relations into graphs or charts. Most and computers best be exploited?
information visualizations are thus about display-
ing and navigating 2D or 3D graphs. People need
new reduction, mapping, and navigation methods Early History
so that they can manage, comprehend, and use the Even before the existence of electronic computers,
fast-growing information on the World Wide Web. there was a fascination with the idea of machines
Other important research directions in data that could play games. The Turk was a chess-playing
visualization include improving the clarity of visu- machine that toured the world in the eighteenth and
alizations, multidimensional and multivariate data nineteenth centuries, to much fanfare. Of course the
(a data set with a large number of dependent vari- technology in the Turk was mainly concerned with
ables) visualization, interaction mechanisms for large concealing the diminutive human chess master hid-
and shared display space, visualization designs guided den inside the machine.
DEEP BLUE 155
In 1949, the influential mathematician Claude a complex evaluation function to assess the good-
Shannon (19162001) proposed chess as an ideal ness of a chess position, and
domain for exploring the potential of the then-new a strong emphasis on intelligent exploration
electronic computer. This idea was firmly grasped (selective search) of the possible move sequences.
by those studying artificial intelligence (AI), who
The first two factors allowed the full Deep Blue
viewed games as providing an excellent test bed for
system to examine 100200 million chess positions
exploring many types of AI research. In fact, chess
per second while selecting a move, and the complex
has often been said to play the same role in the field
evaluation function allowed Deep Blue to make more
of artificial intelligence that the fruit fly plays in ge-
informed decisions. However, a naive brute-force ap-
netic research. Although breeding fruit flies has
plication of Deep Blues computational power would
no great practical value, they are excellent subjects
have been insufficient to defeat the top human chess
for genetic research: They breed quickly, have suffi-
players. It was essential to combine the computers
cient variation, and it is cheap to maintain a large
power with a method to focus the search on move
population. Similarly, chess avoids some aspects
sequences that were important. Deep Blues selective
of complex real-world domains that have proven
search allowed it to search much more deeply on the
difficult, such as natural-language understanding,
critical move sequences.
vision, and robotics, while having sufficient com-
Deep Blue first played against world champion
plexity to allow an automated problem solver to
Garry Kasparov in 1996, with Kasparov winning the
focus on core AI issues such as search and knowl-
six-game match by a score of 4-2. A revamped Deep
edge representation.
Blue, with improved evaluation and more compu-
Chess programs made steady progress in the
tational power, won the 1997 rematch by a score of
following decades, particularly after researchers aban-
3.52.5.
doned the attempt to emulate human thought
processes and instead focused on doing a more thor-
ough and exhaustive exploration of possible move
sequences. It was soon observed that the playing Human and Computer
strength of such brute-force chess programs cor-
related strongly with the speed of the underlying Approaches to Chess
computer, and chess programs gained in strength It is clear that systems like Deep Blue choose moves
both from more sophisticated programs and faster using methods radically different from those em-
computers. ployed by human experts. These differences result
in certain characteristic strengths of the two types
of players. Computers tend to excel at the short-
Deep Blue range tactical aspects of a game, mainly due to an
The Deep Blue computer chess system was devel- extremely thorough investigation of possible move
oped in 19891997 by a team (Murray Campbell, A. sequences. Human players can only explore perhaps
Joseph Hoane, Jr., Feng-hsiung Hsu) from IBMs a few dozen positions while selecting a move, but
T. J. Watson Research Center. Deep Blue was a leap can assess the long-term strategic implications of
ahead of the chess-playing computers that had gone these moves in a way that has proven difficult for
before it. This leap resulted from a number of fac- a computer.
tors, including: The combination of human and computer play-
a computer chip designed specifically for high- ers has proven to be very powerful. High-level chess
speed chess calculations, players routinely use computers as part of their prepa-
a large-scale parallel processing system, with more ration. One typical form of interaction would have the
than five hundred processors cooperating to se- human player suggest strategically promising moves
lect a move, that are validated tactically by the computer player.
Research Areas Newborn, M. (2003). Deep Blue: An artificial intelligence milestone.

While computer programs such as Deep Blue achieve New York: Springer-Verlag.
Schaeffer, J. (2001). A gamut of games. AI Magazine, 22(3), 2946.
a very high level of play, most of the knowledge in Schaeffer, J., & van den Herik, J. (Eds.). (2002). Chips challenging cham-
such systems is carefully crafted by human experts. pions: Games, computers, and artificial intelligence. New York:
Much research is needed to understand how to have Elsevier.
future systems learn the knowledge necessary to play Shannon, C. (1950). Programming a computer for playing chess.
Philosophical Magazine, 41, 256275.
the game without extensive human intervention. Standage, T. (2002). The Turk: The life and times of the famous
TD-gammon, a neural-network system that plays eighteenth-century chess-playing machine. New York: Walker &
world-class backgammon, was an early leader in Company.
this area.
Determining the best mode for humans and
computers to interact in a more cooperative man-
ner (for example, with one or the other acting as as- DENIAL-OF-SERVICE
sistant, trainer, or coach) is another area worthy of
further research. Game-playing programs that are ATTACK
based on large-scale searches have problems in trans-
lating the search results into forms that humans can A denial-of-service (DoS) attack causes the con-
deal with easily. sumption of a computing systems resources
Some types of games, such as the Chinese typically with malicious intenton such a scale as
game Go, have too many possible moves to allow a to compromise the ability of other users to interact
straightforward application of the methods used for with that system. Virgil Gligor coined the term denial-
chess, and pose significant challenges. Games with of-service attack in reference to attacks on operating
hidden information and randomness, such as poker systems (OS) and network protocols. Recently the
or bridge, also require new and interesting ap- term has been used specifically in reference to attacks
proaches. Interactive games, which employ com- executed over the Internet.
puter-generated characters in simulated worlds, can As governments and businesses increasingly rely
be more realistic and entertaining if the characters on the Internet, the damage that a DoS attack can
can behave in intelligent ways. Providing such cause by the disruption of computer systems has
intelligent characters is a key goal for future AI provided incentive for attackers to launch such at-
researchers. tacks and for system operators to defend against such
attacks.
Murray Campbell
See also Artificial Intelligence Evolution of

Denial-of-Service Attacks
FURTHER READING DoS vulnerabilities occur when a poor resources-allo-
cation policy allows a malicious user to allocate so
Campbell, M., Hoane, A. J., & Hsu, F. (2002). Deep Blue. Artificial many resources that insufficient resources are left for
Intelligence, 134(12), 5783. legitimate users. Early DoS attacks on multiuser
Frey, P. W. (Ed.). (1983). Chess skill in man and machine. New York: operating systems involved one user spawning a large
Springer-Verlag.
Hsu, F. (2002). Behind Deep Blue: Building the computer that number of processes or allocating a large amount of
defeated the world chess champion. Princeton, NJ: Princeton memory, which would exhaust the memory avail-
University Press. able and result in operating system overload.
Laird, J. E., & van Lent, M. (2000). Human-level AIs killer applica-
tion: Interactive computer games. AI Magazine, 22(2), 1526.
Early network DoS attacks took advantage of the
Marsland, T. A., & Schaeffer, J. (Eds.). (1990). Computers, chess, and fact that the early Internet was designed with implicit
cognition. New York: Springer-Verlag. trust in the computers connected to it. The unin-
DENIAL-OF-SERVICE ATTACK 157
tended result of this trust was that users paid little and the DDoS variant was that attackers were be-
attention to handling packets (the fundamental unit ginning to use multiple computers in each attack,
of data transferred between computers on the thus amplifying the attack. Internet software imple-
Internet) that did not conform to standard Internet mentation and protocol attacks did not require mul-
protocols. When a computer received a malformed tiple attackers to be successful, and effective defenses
packet that its software was not equipped to handle, were designed (in some cases) against them. A DDoS
it might crash, thus denying service to other users. attack, however, did not require a software imple-
These early DoS attacks were relatively simple and mentation or protocol flaw to be present. Rather, a
could be defended against by upgrading the OS soft- DDoS attack would consist of an attacker using mul-
ware that would identify and reject malformed pack- tiple computers (hundreds to tens of thousands)
ets. However, network DoS attacks rapidly increased to send traffic at the maximum rate to a victims com-
in complexity over time. puter. The resulting flood of packets was sometimes
A more serious threat emerged from the implicit enough to either overload the victims computer
trust in the Internets design: The protocols in the (causing it to slow to a crawl or crash) or overload
Internet themselves could be exploited to execute a the communication line from the Internet to that
DoS attack. The difference between exploiting an computer. The DDoS attacker would subvert con-
Internet software implementation (as did the pre- trol of other peoples computers for use in the attack,
vious class of DoS attacks) and exploiting an Internet often using flaws in the computers control code sim-
protocol itself was that the former were easy to iden- ilar to Internet software implementation DoS attacks
tify (malformed packets typically did not occur out- or simply attaching the attack control codes in an
side of an attack) and once identified could be e-mail virus or Internet worm. The presence of at-
defended against, whereas protocol-based attacks tacking computers on many portions of the Internet
could simply look like normal traffic and were dif- gave this class of attacks its name.
ficult to defend against without affecting legiti-
mate users as well. An example of a protocol attack
was TCP SYN flooding. This attack exploited the fact Defense against DoS Attacks
that much of the communication between comput- Defending against DoS attacks is often challenging
ers over the Internet was initiated by a TCP hand- because the very design of the Internet allows
shake where the communicating computers them to occur. The Internets size requires that even
exchanged specialized packets known as SYN pack- the smallest change to one of its fundamental pro-
ets. By completing only half of the handshake, an tocols be compatible with legacy systems that do not
attacker could leave the victim computer waiting for implement the change. However, users can deploy
the handshake to complete. Because computers could effective defenses without redesigning the entire
accept only a limited number of connections at Internet. For example, the defense against Internet
one time, by repeating the half-handshake many software implementation DoS attacks is as simple as
times an attacker could fill the victim computers updating the software on a potential victims com-
connection capacity, causing it to reject new con- puter; because the packets in this type of attack are
nections by legitimate users or, even worse, causing usually malformed, the compatibility restriction is
the OS to crash. A key component of this attack was easy to meet.
that an attacker was able to hide the origin of his Defending against a protocol-level attack is more
or her packets and pose as different computer difficult because of the similarity of the attack it-
users so the victim had difficulty knowing which self to legitimate traffic. Experts have proposed several
handshakes were initiated by the attacker and which mechanisms, which mostly center on the concept of
were initiated by legitimate users. forcing all computers initiating a handshake to show
Then another class of DoS attack began to ap- that they have performed some amount of work
pear: distributed denial-of-service (DDoS) attacks. during the handshake. The expectation is that an
The difference between traditional DoS attacks attacker will not have the computing power to
impersonate multiple computers making handshake We must wait to see whether DoS attacks will further
requests. Unfortunately, this class of defenses requires threaten the Internet, provoking the acceptance of
a change in the Internet protocol that must be im- radical defense proposals, or will simply fade into
plemented by all computers wanting to contact the the background and become accepted as a regular
potential victims computer. Moreover, more aspect of the Internet.
protocol-compliant solutions involve placing be-
tween the victims computer and the Internet spe- Adrian Perrig and Abraham Yaar
cialized devices that are designed to perform many
handshakes at once and to pass only completed hand- See also Security; Spamming
shakes to the victim.
The DDoS variant of DoS attacks is the most dif-
ficult to defend against because the attack simply FURTHER READING
overwhelms the victims computer with too many
packets or, worse, saturates the victims connection Aura, T., Nikander, P., & Leiwo, J. (2000). DoS-resistant authentica-
tion with client puzzles. Security Protocols8th International
to the Internet so that many packets are dropped be- Workshop.
fore ever reaching the victims computer or network. Gligor, V. D. (1983). A note on the denial of service problem. Proceedings
Some businesses rely on overprovisioning, which of 1983 Symposium on Security and Privacy (pp. 139149).
is the practice of buying computer resources far in Gligor, V. D. (1986). On denial of service in computer networks.
Proceedings of International Conference on Data Engineering
excess of expected use, to mitigate DDoS attacks; this (pp. 608617).
practice is expensive but raises the severity of an Gligor, V. D. (2003). Guaranteeing access in spite of service-flooding
attack that is necessary to disable a victim. Proposed attacks. Proceedings of the Security Protocols Workshop.
Savage, S., Wetherall, D., Karlin, A., & Anderson, T. (2000). Practical
defenses against this type of attackmore so than network support for IP traceback. Proceedings of ACM SIGCOMM
proposed defenses against other types of attacks 2000 (pp. 295306).
have focused on changing Internet protocols. Many Wang, X., & Reiter, M. K. (2003). Defending against denial-of-service
proposals favor some type of traceback mechanism, attacks with puzzle auctions. Proceedings of the 2003 IEEE
Symposium on Security and Privacy (pp. 7892).
which allows the victim of an attack to determine Yaar, A., Perrig, A., & Song, D. (2003). Pi: A path identification mech-
the identity and location of the attacking comput- anism to defend against DDoS attacks. IEEE Symposium on Security
ers, in the hope that filters can be installed in the and Privacy (pp. 93107).
Internet to minimize the flood of traffic while
leaving legitimate traffic unaffected. At the time of
this writing, no DDoS defense proposal has been ac-
cepted by the Internet community. DESKTOP METAPHOR
The desktop metaphor is being used when the inter-
The Future face of an interactive software system is designed such
DoS attacks are likely to trouble the Internet for that its objects and actions resemble objects and
the foreseeable future. These attacks, much like ur- actions in a traditional office environment. For ex-
ban graffiti, are perpetrated by anonymous attack- ample, an operating system designed using the desk-
ers and require a substantial investment to defend top metaphor represents directories as labeled folders
against, possibly requiring a fundamental change in and text documents as files. In graphical user inter-
the Internets protocols. Although several DoS at- faces (GUIs), the bitmap display and pointing devices
tacks have succeeded in bringing down websites of such as a mouse, a trackball, or a light pen are used
well-known businesses, most attacks are not as wildly to create the metaphor: The bitmap display pres-
successful, nor have all businesses that have been ents a virtual desk, where documents can be created,
victimized reported attacks for fear of publicizing stored, retrieved, reviewed, edited, and discarded.
exactly how weak their computing infrastructure is. Files, folders, the trash can (or recycle bin) and so
DESKTOP METAPHOR 159
forth are represented on the virtual desktop by graph- Historical Overview

ical symbols called icons. Users manipulate these icons In the 1960s and 1970s, several innovative con-
using the pointing devices. With pointing devices, the cepts in the area of HCI were originated and im-
user can select, open, move, or delete the files or fold- plemented using interactive time-shared computers,
ers represented by icons on the desktop. graphics screens, and pointing devices.
Users can retrieve information and read it on the
desktop just as they would read actual paper docu- Sketchpad
ments at a physical desk. The electronic document Sketchpad was a pioneering achievement that opened
files can be stored and organized in electronic the field of interactive computer graphics. In 1963,
folders just as physical documents are saved and man- Ivan Sutherland used a light pen to create engineer-
aged in folders in physical file cabinets. Some of ing drawings directly on the computer screen for his
the accessories one finds in an office are also present Ph.D. thesis, Sketchpad: A Man-Machine Graphical
on the virtual desktop; these include the trash can, Communications System. His thesis initiated a totally
a clipboard, a calendar, a calculator, a clock, a new way to use computers. Sketchpad was exe-
notepad, telecommunication tools, and so on. cuted on the Lincoln TX-2 computer at MIT. A light
The metaphor of the window is used for the pen and a bank of switches were the user interface
graphical boxes that let users look into informa- for this first interactive computer graphics system.
tion in the computer. Multiple windows can be open Sketchpad also pioneered new concepts of mem-
on the desktop at once, allowing workers to alternate ory structures for storing graphical objects, rubber-
quickly between multiple computer applications (for banding of lines (stretching lines as long as a user
example, a worker may have a word processing ap- wants) on the screen, the ability to zoom in and
plication, a spreadsheet application, and an Internet out on the screen, and the ability to make perfect
browser open simultaneously, each in its own win- lines, corners, and joints.
dow). Computer users can execute, hold, and resume
their tasks through multiple windows. NLS
The scientist and inventor Doug Engelbart and his
colleagues at Stanford University introduced the oN
BITMAP An array of pixels, in a data file or structure, Line System (NLS) to the public in 1968. They also
which correspond bit for bit with an image. invented the computer mouse. The NLS was
equipped with a mouse and a pointer cursor for the
first time; it also was the first system to make use
Beginning in the late 1970s, as personal computers of hypertext. Among other features, the system pro-
and workstations became popular among knowledge vided multiple windows, an online context-sensitive
workers (people whose work involves developing and help system, outline editors for idea development,
using knowledgeengineers, researchers, and teach- two-way video conferencing with shared workspace,
ers, for example), the usability of the computers and word processing, and e-mail.
the productivity of those using them became im-
portant issues. The desktop metaphor was invented Smalltalk
in order to make computers more usable, with the In the early 1970s, Alan Kay and his colleagues at
understanding that more-usable computers would Xeroxs Palo Alto Research Center (PARC) invented
increase users productivity. The desktop metaphor an object-oriented programming language called
enabled users to work with computers in a more fa- Smalltalk. It was the first integrated programming
miliar, more comfortable manner and to spend less environment, and its user interface was designed us-
time learning how to use them. The invention of the ing the desktop metaphor. It was designed not only
desktop metaphor greatly enhanced the quality of for expert programmers of complex software, but
human-computer interaction. also for the novice users, including children: Its
designers intended it to be an environment in which Personalized interaction: A user could set up

users learned by doing. attributes of the system in order to customize the
In 1981 Xerox PARC integrated the innovations interaction with the system. The personalized inter-
in the fields of human-computer symbiosis, personal action did not interfere with the standard interaction
computing, objected-oriented programming lan- methods.
guages, and local-area networks and arrived at the Multiple tasks: Because workers in office envi-
Xerox 8010 Star information system. It was the ronments perform many tasks simultaneously and
first commercial computer system that implemented are often interrupted in their work, Lisa was designed
the desktop metaphor, using a mouse, a bitmap dis- to be able to hold the current work while users at-
play, and a GUI. In interacting with the system, users tended to those interruptions and to other business.
made use of windows, icons, menus, and pointing The idea was that the user should be able to switch
devices (WIMP). Most of the workstations and per- from one task to another freely and instantly.
sonal computers that were developed subse- Apples Lisa provided knowledge workers with a
quently, including Apple Computers Lisa (1983) and virtual desktop environment complete with manip-
Macintosh (1984) and Microsofts Windows (1985), ulable documents, file folders, calculators, electronic
were inspired by Star; like the Star system, they too paper clips, wastebasket, and other handy tools. The
adopted the desktop metaphor. documents and other office-based objects were rep-
Apples Lisa was designed to be a high-quality, resented by naturalistic icons. The actions defined
easy-to-use computer for knowledge workers such for the icons, such as selecting, activating, moving,
as secretaries, managers, and professionals in gen- and copying, were implemented by means of mouse
eral office environments. Its design goals were: operations such as clicking, double-clicking, and
User friendliness: The developers of the Lisa dragging. Lisa users did not have to memorize com-
wanted users to use the computer not only because mands such as delete (del), remove (rm),
doing so was part of their job, but also because it was or erase in order to interact with the system.
fun to use. The users were expected to feel comfort- The first version of Microsoft Windows was in-
able because the user interface resembled their work- troduced in 1985. It provided an interactive software
ing environment. environment that used a bitmap display and a mouse.
Standard method of interaction: A user was pro- The product included a set of desktop applica-
vided with consistent look and feel in the system and tions, including a calendar, a card file, a notepad, a
all applications, which meant that learning time could calculator, a clock, and telecommunications pro-
be dramatically decreased and training costs lowered. grams. In 1990 Windows 3.0 was introduced, the first
Gradual and intuitive learning: A user should be real GUI-based system running on IBM-compatible
able to complete important tasks easily with mini- PCs. It became widely popular. The Windows oper-
mal training. The user should not be concerned with ating system evolved through many more incarna-
more sophisticated features until they are necessary. tions, and in the early 2000s was the most popular
Interaction with the computer should be intuitive; operating system in the world.
that is, the user should be able to figure out what
he or she needs to do.
Error protection: A user should be protected from Research Directions
obvious errors. For example, Lisa allowed users to The desktop metaphor, implemented through a
choose from a collection of possible operations that graphical user interface, has been the dominate
were proper for the occasion and the object. By lim- metaphor for human-computer interfaces since the
iting the choices, fatal errors and obvious errors could 1980s. What will happen to the human-computer
be avoided. Any error from a user should be pro- interaction paradigm in the future? Will the desktop
cessed in a helpful manner by generating a warn- metaphor continue to dominate? It is extremely dif-
ing message or providing a way of recovering from ficult to predict the future in the computer world.
the error. However, there are several pioneering researchers
DESKTOP METAPHOR 161
exploring new interaction paradigms that could mensional views of the MIT geographical model.
replace the desktop-based GUI. Users could physically control the phicons by
grasping and placing them so that a two dimensional
map of the MIT campus appears on the desk surface
OPERATING SYSTEM Software (e.g., Windows 98, UNIX, beneath the phicons. The locations of the Dome and
or DOS) that enables a computer to accept input and the Media Lab buildings on the map should match
produce output to peripheral devices such as disk drives with the physical locations of the phicons on the desk.
and printers.
Ubiquitous Computing
In 1988 Mark Weiser at Xerox PARC introduced a
A Tangible User Interface computing paradigm called ubiquitous computing.
At present, interactions between human and com- The main idea was to enable users to access com-
puters are confined to a display, a keyboard, and a puting services wherever they might go and when-
pointing device. The tangible user interface (TUI) ever they might need them. Another requirement
proposed by Hiroshi Ishii and Brygg Ullmer at MITs was that the computers be invisible to the users, so
Media Lab in 1997 bridges the space between human the users would not be conscious of them. The users
and computers in the opposite direction. The user do what they normally do and the computers in the
interface of a TUI-based system can be embodied in background recognize the intention of the users and
a real desk and other real objects in an office envi- provide the best services for them. It means that the
ronment. Real office objects such as actual papers users do not have to learn how to operate comput-
and pens could become meaningful objects for the ers, how to type a keyboard, how to access the
user interface of the system. Real actions on real ob- Internet, etc. Therefore, the paradigm requires
jects can be recognized and interpreted as operations that new types of computing services and computer
applied to the objects in the computer world, so that, systems be created. New technologies such as con-
for example, putting a piece of paper in a waste- text awareness, sensors, and intelligent distributed
basket could signal the computer to delete a docu- processing. are required. Their interaction methods
ment. This project attempts to bridge the gap between must be based on diverse technologies such as face
the computer world and the physical office envi- recognition, character recognition, gesture recogni-
ronment by making digital information tangible. tion, and voice recognition.
While the desktop metaphor provides the users with
a virtual office environment, in a TUI the physical
office environment, including the real desktop, be- OPEN-SOURCE SOFTWARE Open-source software per-
comes the user interface. mits sharing of a programs original source code with
Ishii and Ullmer designed and implemented users, so that the software can be modified and redis-
a prototype TUI called metaDESK for Tangible tributed to other users.
Geospace, a physical model of landmarks such as the
MIT campus. The metaDESK was embodied in real-
world objects and regarded as a counterpart of the As new computing services and technologies are
virtual desktop. The windows, icons, and other graph- introduced, new types of computing environments and
ical objects in the virtual desktop corresponded to new interaction paradigms will emerge. The desktop
physical objects such as activeLENS (a physically em- metaphor will also evolve to keep pace with techno-
bodied window), phicon (a physically embodied logical advances. However, the design goals of the user
iconin this case, models of MIT buildings such as interfaces will not change much. They should be de-
the Great Dome and the Media Lab building), and so signed to make users more comfortable, more effec-
forth. In the prototype system, the activeLENS was tive, and more productive in using their computers.
equivalent to a window of the virtual desktop and
was used in navigating and examining the three di- Jee-In Kim
See also Alto; Augmented Reality; Graphical User are more resource intensive; however, despite the extra
Interface costs, the need to reach a broader community of users
and the desire to create systems that can perform tasks
that require collaboration with users has led to a shift
FURTHER READING toward more dialog-based systems.
Speech is a good modality for remote database
Goldberg, A. (1984). Smalltalk-80: The interactive programming envi- access systems, such as telephone information ser-
ronment. Reading, MA: Addison-Wesley.
Ishii, H., & Ullmer, B. (1997). Tangible bits: Towards seamless inter-
vices, which would otherwise require a human op-
faces between people, bits and atoms. In Proceedings of CHI 97 erator or a tedious sequence of telephone keystrokes.
(pp. 234241), New York: ACM Press. Spoken interaction is also useful when a users hands
Kay, A. (1993) The early history of Smalltalk. ACM SIGPLAN Notices, are busy with other tasks, such as operating me-
28(3), 6995.
Kay, A., & Goldberg, A (1977). Personal dynamic media. IEEE chanical controls, or for tasks for which the sound
Computer, 10(3), 3142. of the users speech is important, such as tutoring
Myers, B., Ioannidis, Y., Hollan, J., Cruz, I., Bryson, S., Bulterman, speakers in oral reading. Speech interaction can also
D., et al. (1996). Strategic directions in human computer inter-
action. ACM Computing Survey, 28(4), 794809.
make computers more accessible to people with
Perkins, R., Keller, D., & Ludolph, F. (1997). Inventing the Lisa user vision impairments.
interface. Interactions, 4(1), 4053. Speech dialog systems have been used for tutor-
Shneiderman, B. (1998). Designing the user interface: Strategies for ing in oral reading; for providing information about
effective human-computer interaction (3rd ed.). Reading, MA:
Addison-Wesley.
public transportation, train schedules, hotels, and
Weiser, M. (1991). The computer for the 21st century. Scientific sight seeing; for making restaurant and real estate
American, 256(3), 94104. recommendations; for helping people diagnose fail-
ures in electronic circuits; and for making travel reser-
vations. Spoken dialog is most successful when the
scope of the task is well-defined and narrow, such as
DIALOG SYSTEMS providing airline reservations or train schedules, be-
cause the task creates expectations of what people
Speech dialog systems are dialog systems that use will sayand the more limited the scope, the more
speech recognition and speech generation to allow limited the expectations. These expectations are
a human being to converse with a computer, usually needed for the system to interpret what has been said;
to perform some well-defined task such as making in most cases the same group of speech sounds
travel reservations over the telephone. will have several different possible interpretations,
A dialog is a two-way interaction between two but the task for which the dialog system is used makes
agents that communicate. Dialogs are incremental one of the interpretations by far the most likely.
and can be adapted dynamically to improve the ef-
fectiveness of the communication.
While people communicate efficiently and effec- The Architecture of
tively using dialog, computers do not typically engage
in dialogs with people. More common are presen- Speech Dialog Systems
tation systems, which are concerned with the effec- Speech dialog systems include the following
tive presentation of a fixed content, subject to a limited components or processes: speech recognition,
number of constraints. Unlike dialogs, presentations natural-language parsing, dialog management,
are planned and displayed in their entirety (with- natural-language generation, and speech synthesis.
out intermediate feedback from the user) and thus There is also an application or database that provides
do not allow the system to monitor the effective- the core functionality of the system (such as book-
ness of the presentation or allow the user to interrupt ing a travel reservation) and a user interface to trans-
and request clarification. Dialog systems have been mit inputs from the microphone or telephone to
less common than presentation systems because they the speech-recognition component.
DIALOG SYSTEMS 163
Speech Recognition its style of dialog interaction. In either case, a vo-

Understanding speech involves taking the sound in- cabulary and a language model can be obtained that
put and mapping it onto a command, request, or may only require a few thousand (and possibly only
statement of fact to which the application can re- a few hundred) words.
spond. Speech recognition is the first step, which in-
volves mapping the audio signal into words in the Natural-Language Parsing
target language. Early approaches, such as HEARSAY II Natural-language parsing maps the sequence of
and HARPY, which were developed in the 1970s, were words produced by the speech recognizer onto com-
based on rule-based artificial intelligence. They were mands, queries, or propositions that will be meaning-
not very successful. Current approaches are based ful to the application. There are a variety of approaches
on statistical models of language that select the most to parsing; some try to identify general-purpose lin-
probable interpretation of each sound unit given im- guistic patterns, following a so-called syntactic gram-
mediately preceding or following ones and the con- mar, while others look for patterns that are specific
text of the task (which determines the vocabulary). to the domain, such as a semantic grammar. Some
For spoken-dialog systems, the level of speech systems use simpler approaches, such as word
recognition quality that is desired is known as tele- spotting, pattern matching, or phrase spotting.
phone quality, spontaneous speech (TQSS). This level The output of the parser will typically be a slot-and-
of recognition is necessary if spoken-dialog applica- filler-based structure called a case frame, in which
tions such as reservation services are to be success- phrases in the input are mapped to slots correspon-
ful over a telephone line. TQSS is more difficult to ding to functions or parameters of the application.
understand than face-to-face speech because over A key requirement of parsing for spoken dialog
the telephone the audio signal normally includes is that the parser be able to handle utterances that
background and channel noise, acoustical echo and do not form a complete sentence or that contain the
channel variations, and degradation due to band- occasional grammatical mistake. Such parsers are
width constraints. Moreover, spontaneous speech termed robust. Syntactic and semantic parsers work
includes pauses, disfluencies (such as repetitions and best when the input is well formed structurally.
incomplete or ill-formed sentences), pronunciation Simpler methods such as pattern matching and
variations due to dialects, as well as context-dependent phrase spotting can be more flexible about structural
formulations and interruptions or overlapping speech ill-formedness, but may miss important syntactic
(known as barge-in). variations such as negations, passives, and topical-
One way that speech recognition is made more izations. Also, the simpler approaches have little
accurate is by limiting the vocabulary that the sys- information about how to choose between two
tem allows. To determine a sublanguage that will be different close matches. To be useful in practice, an-
sufficiently expressive to allow people to use the ap- other requirement is that the parser be fast enough
plication effectively and comfortably, two techniques to work in real time, which is usually only possible
are generally used: One approach is to observe or if the analysis is expectation driven. By the late 1990s,
stage examples of two people engaged in the domain most spoken-dialog systems still focused on get-
task; the other approach is to construct a simulated ting the key components to work together and had
man-machine dialog (known as a Wizard of Oz not achieved real-time behavior (interpretation and
[WOZ] simulation) in which users try to solve the generation of speech)with only a few systems
domain task. In a WOZ simulation, users are led to using real-time behavior.
believe they are communicating with a function-
ing system, while in reality the output is generated Dialog Management
by a person who simulates the intended functional- Dialog management involves interpreting the rep-
ity of the system. This approach allows the design- resentations created by the natural-language
ers to see what language people will use in response parser and deciding what action to take. This process
to the limited vocabulary of the proposed system and is often the central one and drives the rest of the
system. Dialog management may involve following Historically, natural-language generation com-
a fixed pattern of action defined by a grammar (for ponents have not run in real time, with the realiza-
example, answering a question, or it may involve rea- tion component being an important bottleneck.
soning about the users or the systems current knowl- These systems can be slow if they follow an approach
edge and goals to determine the most appropriate that is essentially the inverse of parsingtaking a
next step. In this second instance, then, dialog man- structural description of a sentence, searching for
agement may also keep track of the possible goals of grammar rules that match the description, and then
the users and their strategies (plans) for achieving applying each of the rules to produce a sequence of
them. It may also try to identify and resolve break- words. As a result, many spoken-dialog systems have
downs in communication caused by lack of under- relied on preformulated answers (canned text). More
standing, misunderstanding, or disagreement. recently, real-time approaches to text generation have
One factor that distinguishes dialog managers is been developed that make use of fixed patterns or
the distribution of control between the system and templates that an application can select and thereby
the user. This has been referred to as initiative, or the bypass the need to perform a search within the
mode of communication, with the mode being con- generation grammar.
sidered from the perspective of the computer sys-
tem. When the computer has complete control, it Speech Synthesis
is responsible for issuing queries to the user, col- Speech synthesis allows the computer to respond to
lecting answers, and formulating a response. This the user in spoken language. This may involve se-
has been called directive mode. At the opposite ex- lecting and concatenating pieces of prerecorded
treme, some systems allow the user to have complete speech or generating speech two sounds at a time, a
control, telling the system what the user wants to do method known as diphone-based synthesis. (Diphone
and asking the system to provide answers to specific refers to pairs of sounds.) Databases of utterances to
queries. This is known as passive mode. In the be prerecorded for a domain can be determined by
middle are systems that share initiative with the user. analyzing the utterances produced by a human per-
The system may begin by issuing a query to the user forming the same task as the information system and
(or receiving a query from the user) but control may then selecting the most frequent utterances. Diphone-
shift if either party wishes to request clarification or based synthesis also requires a database of prerecorded
to obtain information needed for a response. Control sound; however instead of complete utterances the
may also shift if one party identifies a possible break- database will contain a set of nonsense words (that
down in communication or if one party disagrees have examples of all pairs of sounds), containing all
with information provided by the other. Dialogs have phone-phone transitions for the target output lan-
been shown to more efficient if control can shift to guage. Then when the synthesizer wants to gener-
the party with the most information about the ate a pair of sounds, it selects a word that contains
current state of the task. the sound-pair (diphone) and uses the correspon-
ding portion of the recording.
Natural-Language Generation Although these basic components of speech
Natural-language generation is used to generate an- dialog systems can be combined in a number of
swers to the users queries or to formulate queries ways, there are three general approaches: pipelined
for the user in order to obtain the information needed architectures, agent-based architectures, and hub-
to perform a given task. Natural-language genera- and-spoke-based architectures. In a pipelined archi-
tion involves three core tasks: content selection (de- tecture, each component in the sequence processes its
ciding what to say), sentence planning (deciding how input and initiates the next component in the
to organize what to say into units), and realization sequence. Thus, the audio interface would call the
(mapping the planned response onto a grammati- speech recognizer, which would call the natural-
cally correct sequence of words). language parser, and so on, until the speech synthesis
DIALOG SYSTEMS 165
component is executed. In an agent-based approach, infrastructure and is distributed with a working

a centralized component (typically the dialog man- implementation for a travel-planning domain. It can
ager) initiates individual components and deter- be downloaded freely from the Internet. The group
mines what parameters to provide them. This may also maintains a telephone number connected to
involve some reasoning about the results provided their telephone-based travel-planning system that
by the components. In a hub-and-spoke architecture anyone can try.
there is a simple centralized component (the hub) The Center for Spoken Language Research (CSLR)
which brokers communication among the other at the University of Colorado in Boulder distributes
components, but performs no reasoning. Since 1994, the Conversational Agent Toolkit. This toolkit in-
a hub-and-spoke architecture called Galaxy cludes modules that provide most of the function-
Communicator has been under development. It has ality needed to build a spoken-dialog system,
been proposed as a standard reference architec- although code must be written for the application
ture that will allow software developers to combine itself. As a model, CSLR distributes their toolkit
plug-and-play-style components from a variety with a sample (open-source) application for the travel
of research groups or commercial vendors. The domain that can be used as a template; it is based on
Galaxy Communicator effort also includes an open- the Galaxy Communicator hub architecture.
source software infrastructure. TRINDIKIT is a toolkit for building and exper-
imenting with dialog move engines (mechanisms for
updating what a dialog system knows, based on di-
Dialog System Toolkits alog moves (single communicative actions such as
Creating speech dialog systems is a major under- giving positive feedback) and information states
taking because of the number and complexity of the (information stored by the dialog system). It has been
components involved. This difficulty is mitigated by developed in the TRINDI and SIRIDUS projects, two
the availability of a number of software toolkits that European research projects that investigate human-
include many, if not all, of the components needed machine communication using natural language.
to create a new spoken-dialog system. Currently such TRINDIKIT specifies formats for defining informa-
toolkits are available both from commercial vendors tion states, rules for updating the information state,
(such as IBM, which markets ViaVoice toolkits for types of dialog moves, and associated algorithms.
speech recognition and speech synthesis) and aca-
demic institutions. Academic institutions generally
distribute their software free for noncommercial use, The Evaluation of
but sell licenses for commercial applications. Below,
we consider a few of the major (academically avail- Speech Dialog Systems
able) speech dialog toolkits. In addition to these Prior to 1990, methods of evaluating speech dialog
toolkits, a number of institutions distribute indi- systems concentrated on the number of words that
vidual components useful in building speech dialog the speech recognizer identified correctly. In the early
systems. For example, the Festival Speech Synthesis 1990s, there was a shift to looking at the quality of
System developed by the University of Edinburgh the responses provided by spoken-dialog systems.
has been used in a number of applications. For example, in 1991, the U.S. Defense Advanced
The Communicator Spoken Dialog Toolkit, Research Projects Agency community introduced a
developed by researchers at Carnegie Mellon Uni- metric that evaluates systems based on the number
versity, is an open-source toolkit that provides a com- of correct and incorrect answers given by the system.
plete set of software components for building and Systems are rewarded for correct answers and
deploying spoken-language dialog systems for penalized for bad answers, normalized by the total
both desktop and telephone applications. It is built number of answers given. (The effect is that it is bet-
on top of the Galaxy Communicator software ter to give a nonanswer such as I do not understand
or please rephrase your request than to give an in- and government. Among its goals are promoting,
correct answer.) This approach relies on the exis- developing, and distributing reusable discourse-
tence of a test database with a number of sample processing components; encouraging empirical meth-
sentences from the domain along with the correct ods in research; sharing resources and data among
answer, as well as a set of answers from the system the international community; exploring techniques
to be evaluated. for evaluating dialog systems; promoting stan-
Starting in the late 1990s, approaches to evalu- dards for discourse transcription, segmentation, and
ating dialog success have looked at other measures, annotation; facilitating collaboration between de-
such as task-completion rates and user satisfaction velopers of various system components; and en-
(as determined by subjective questionnaires). couraging student participation in the discourse and
Subjective factors include perceived system-response dialog community.
accuracy, likeability, cognitive demand (how much
effort is needed to understand the system), habit- Susan W. McRoy
ability (how comfortable or natural the system is
to use), and speed. There has also been success in See also Natural-Language Processing, Open Source
predicting user satisfaction or task completion on Software, Speech Recognition, Speech Synthesis
the basis of objectively observable features of the di-
alog, such as task duration, the number of system
words per turn, the number of user words per FURTHER READING
turn, the number of overlapping turns, sentence er-
ror rates, and perceived task completion. Statistical Allen, J. F., Schubert, L. K., Ferguson, G., Heeman, P., Hwang, C. H.,
methods such as multiple regression models and clas- Kato, T., et al. (1995). The TRAINS project: A case study in
building a conversational planning agent. Journal of Experimental
sification trees are then used to predict user satis- and Theoretical Artificial Intelligence, 7, 748.
faction and task-completion scores. Bernsen, N. O., Dybkjaer, H., & Dybkjaer, L. (1998). Designing inter-
active speech systems: From first ideas to user testing. New York:
Springer Verlag.
Fraser, N. (1997). Assessment of interactive systems. In D. Gibbon,
The Research Community for R. Moore, and R. Winski (Eds.), Handbook of standards and re-
sources for spoken language systems (pp. 564614). New York:
Speech Dialog Systems Mouton de Gruyter.
Research on speech dialog systems is interdiscipli- Grosz, B. J., & Sidner, C. (1986). Attention, intention, and the struc-
ture of discourse. Computational Linguistics, 12(3), 175204.
nary, bringing together work in computer science, Haller, S., Kobsa, A., & McRoy, S. (Eds.). (1999). Computational mod-
engineering, linguistics, and psychology. There are els for mixed-initiative interaction. Dordrect, Netherlands: Kluwer
a number of journals, conferences, and workshops Academic Press.
Huang X. D., Alleva, F., Hon, H. W., Hwang, M. Y., Lee, K. F., and
through which researchers and developers of spoken- Rosenfeld, R. (1993). The Sphinx II Speech Recognition System:
dialog systems disseminate their work. Important An overview. Computer Speech and Language, 7(9), 137148.
journals include Computer Speech and Language and Jurafsky, D., & Martin, J. (2000). Speech and language processing: An
Natural Language Engineering. Conferences most fo- introduction to natural language processing, computational linguis-
tics, and speech recognition. Upper Saddle River, NJ: Prentice-Hall.
cused on such systems include Eurospeech and Larsson, S., & Traum, D. (2000). Information state and dialogue man-
Interspeech (the International Conference on Spoken agement in the TRINDI Dialogue Move Engine Toolkit [Special
Language Processing). In addition, the Special Interest issue on best practice in spoken dialogue systems]. Natural Language
Group on Discourse and Dialog (SIGdial) organizes Engineering, 6(34), 323340.
Luperfoy, S. (Ed.). (1998). Automated spoken dialog systems. Cambridge,
an annual workshop. SIGdial is a Special Interest MA: MIT Press.
Group (SIG) of both the Association for Com- McRoy, S. W. (Ed.). (1998). Detecting, repairing, and preventing
putational Linguistics and the International Speech human-machine miscommunication [Special issue]. International
Journal of Human-Computer Studies, 48(5).
Communication Association (ISCA). SIGdial is an McRoy, S. W., Channarukul, S., & Ali, S. S. (2001). Creating natural
international, nonprofit cooperative organization language output for real-time applications intelligence. Intelligence:
that includes researchers from academia, industry, New Visions of AI in Practice, 12(2), 2134.
DIGITAL CASH 167
McRoy, S. W., Channarukul, S., & Ali, S. S. (2003). An augmented 1. Token-based systems store funds as tokens that
template-based approach to text realization. Natural Language can be exchanged between parties. Traditional
Engineering, 9(2), 140.
Minker, W., Bhler, D., & Dybkjr, L. (2004). Spoken multimodal
currency falls in this category, as do many types
human-computer dialog in mobile environments. Dordrect, of stored-value payment systems, such as sub-
Netherlands: Kluwer. way fare cards, bridge and highway toll systems
Mostow, J., Roth, S. F., Hauptmann, A., & Kane, M. (1994). A proto- in large metropolitan areas (e.g., FastPass,
type reading coach that listens. In Proceedings of the Twelfth National
Conference on Artificial Intelligence (AAAI-94) (pp. 785792). Seattle, EasyPass), and electronic postage meters. These
WA: AAAI Press. systems store value in the form of tokens, either
Pellom, B., Ward, W., Hansen, J., Hacioglu, K., Zhang, J., Yu, X., & a physical token, such as a dollar bill, or an elec-
Pradhan, S. (2001, March). University of Colorado dialog systems
for travel and navigation. Paper presented at the Human Language
tronic register value, such as is stored by a sub-
Technology Conference (HLT-2001), San Diego, CA. way fare card. During an exchange, if the full
Roe, D. B., & Wilpon, J. G. (Eds.). (1995). Voice communication be- value of a token is not used, then the remain-
tween humans and machines. Washington, D.C.: National Academy der is returned (analogous to change in a cur-
Press.
Seneff, S., Hurley, E., Lau, R. Pau, C., Schmid, P., & Zue, V. (1998).
rency transaction)either as a set of smaller
Galaxy II: A reference architecture for conversational system de- tokens or as a decremented register value.
velopment. Proceedings of the 5th International Conference on Spoken Generally, if tokens are lost (e.g., if ones wallet
Language Processing, 931934. is stolen or one loses a subway card), the tokens
Smith, R., & Hipp, D. R. (1995). Spoken natural language dialog sys-
tems: A practical approach. New York: Oxford University Press. cannot be recovered.
Smith, R., & van Kuppevelt, J. (Eds.). (2003). Current and new direc- 2. Account-based systems charge transactions to
tions in discourse and dialogue. Dordrect, Netherlands: Kluwer. an account. Either the account number or a ref-
van Kuppevelt, J, Heid, U., & Kamp, H. (Eds.). (2000). Best practice
in spoken dialog systems [Special issue]. Natural Language
erence to the account is used to make payment.
Engineering, 6(34). Examples include checking accounts, credit card
Walker, M., Litman, D., Kamm, C., & Abella, A. (1998). Evaluating accounts, and telephone calling cards. In some
spoken dialogue agents with PARADISE: Two case studies. instances, the account is initially funded and
Computer Speech and Language, 12(3), 317347.
Walker, M. A., Kamm, C. A., & Litman, D. J. (2000). Towards devel-
then spent down (e.g., checking accounts); in
oping general models of usability with PARADISE [Special issue other instances, debt is increased and periodi-
on best practice in spoken dialogue systems]. Natural Language cally must be paid (e.g., credit cards). In most
Engineering, 6(34). account-based systems, funds (or debt) are
Wilks, Y. (Ed.). (1999). Machine conversations. Dordrect, Netherlands:
Kluwer. recorded by a trusted third party, such as a bank.
The account can be turned off or renumbered
if the account number is lost.
The more complex an electronic payment sys-
tem is, the less likely consumers are to use it. (As
an example, a rule of thumb is that merchants of-
DIGITAL CASH fering one-click ordering for online purchases en-
joy twice the order rate of merchants requiring
that payment data be repeatedly entered with each
The use of digital cash has increased in parallel with
purchase.)
the use of electronic commerce; as we purchase
items online, we need to have ways to pay for them
electronically. Many systems of electronic payment Electronic Payment
exist.
Using Credit Cards
The most common form of electronic payment on
Types of Money the Internet today is credit card payment. Credit cards
Most systems of handling money fall into one of two are account based. They are issued by financial
categories: institutions to consumers and in some cases to
organizations. A consumer presents the credit card auctions. As a result, a market has opened for third-
number to a merchant to pay for a transaction. On party payment processors. Today, the largest third-
the World Wide Web credit card account numbers party payment processor is PayPal, owned by the
are typically encrypted using the Secure Socket Layer eBay auction service. Third-party payment proces-
(SSL) protocol built into most Web browsers. The sor systems are account based.
merchant often attempts to verify the card holder by Consumers can pay for third-party purchases in
performing address verification (checking numbers three ways: by paying from an account maintained
appearing in an address) or by using a special veri- with the third party, by paying from a credit card
fication code (typically printed on the reverse side account, and by paying from a checking account.
of the credit card). In the United States credit card Merchants rates for accepting funds from a credit card
users typically enjoy strong rights and can reverse account are slightly higher than their rates for ac-
fraudulent transactions. cepting funds from a conventional credit card account.
Although the SSL protocol (in typical configu- Third-party payment accounts are convenient
rations) provides strong encryption preventing third because they are simple to use and provide consumers
parties from observing the transaction, risks still ex- with protection against being overcharged. However,
ist for the credit card holder. Many merchants apply they tend not to provide the same degree of pro-
inadequate security to their database of purchases, tection that a credit card-funded purchase provides.
and attackers have gained access to large numbers of Because third-party payment accounts are widely
credit cards stored online. Moreover, some merchants used with auction systems, where fraud rates are un-
charge incorrect amounts (or charge multiple times) usually high, the degree of protection is a serious
for credit card transactions. Although fraudulent consideration.
transactions are generally reversible for U.S. resi-
dents, time and effort are required to check and
amend such transactions. In some instances, crim- Smartcards and
inals engage in identity theft to apply for addi-
tional credit by using the identity of the victim. Other Stored-Value Systems
To reduce these risks, some experts have proposed Stored-value systems store value on a card that is
a system that uses third parties (such as the bank that used as needed. Smartcards are a token-based pay-
issued the card) to perform credit card transactions. ment system. Many smartcards use an integrated cir-
A notable example of this type of system is Verified cuit to pay for purchases. They are widely used in
by Visa. However, the additional work required to con- Europe for phone cards and in the GSM cellular tele-
figure the system has deterred some consumers, and phone system. Mondex is a consumer-based system
as a result Verified by Visa and similar systems remain for point-of-sale purchases using smartcards. Use of
largely unused. The most elaborate of these systems smartcards is limited in Asia and largely unused in
was the Secure Electronic Transactions (SET) pro- North America. (In North America only one major
tocol proposed by MasterCard International and Visa vendor, American Express, has issued smartcards to
International; however, the complexity of SET led to large numbers of users, and in those cards the smart-
its being abandoned. In these systems credit card pur- card feature is currently turned off.)
chases are usually funded with a fee that is charged to Experts have raised a number of questions about
the merchant. Although rates vary, typical fees are fifty the security of smartcards. Successful attacks con-
cents plus 2 percent of the purchase amount. ducted by security testers have been demonstrated
against most smartcard systems. Experts have raised
even deeper questions about the privacy protec-
Third-Party Payment Accounts tion provided by these systems. For example, in
A merchant must be able to process credit card pay- Taiwan, where the government has been moving to
ments. This processing is often inconvenient for small switch from paper records to a smartcard system for
merchants, such as people who sell items in online processing National Health Insurance payments,
DIGITAL CASH 169
considerable public concern has been raised about Micropayments

potential privacy invasions associated with the use of One of the most interesting types of electronic pay-
health and insurance records on a smartcard system. ment is micropayments. In many instances consumers
A number of devices function like a smartcard wish to purchase relatively small-value items. For ex-
but have different packaging. For example, some ur- ample, consider a website that vends recipes. Each
ban areas have adopted the FastPass system, which recipe might be sold for only a few cents, but sold in
allows drivers to pay bridge and highway tolls us- volume, their value could be considerable. (Similarly,
ing radio link technology. As a car passes over a sen- consider a website that offers online digital recordings
sor at a toll booth, value stored in the FastPass device of songs for ninety-nine cents each.) Currently, mak-
on the car is decremented to pay the toll. The state ing small payments online using traditional payment
of California recently disclosed that it uses the same methods is not feasible. For example, as mentioned,
technology to monitor traffic flow even when no toll credit card companies typically charge merchants a
is charged. The state maintains that it does not gather processing fee of fifty cents plus 2 percent of the pur-
personal information from FastPass-enabled cars, chase amount for credit card transactionsclearly
but experts say that it is theoretically possible. making credit card purchases for items that cost less
than fifty cents impractical. Most merchants refuse to
deal with small single-purchase amounts and require
Anonymous Digital Cash that consumers either buy a subscription or purchase
A number of researchers have proposed anonymous the right to buy large numbers of items. For example,
digital cash payment systems. These would be token- newspaper websites that offer archived articles typi-
based systems in which tokens would be issued by cally require that consumers purchase either a sub-
a financial institution. A consumer could blind scription to access the articles or purchase a minimum
such tokens so that they could not be traced to the number of archived articlesthey refuse to sell
consumer. Using a cryptographic protocol, a con- archived articles individually.
sumer could make payments to merchants without To enable small single purchases, a number of
merchants being able to collect information about researchers have proposed micropayment systems
the consumer. However, if a consumer attempted to that are either token based or account based. An ex-
copy a cryptographic token and use it multiple times, ample of an account-based micropayment system is
the cryptographic protocol would probably allow the NetBill system designed at Carnegie Mellon
the consumers identity to be revealed, allowing University. This system provides strong protection
the consumer to be prosecuted for fraud. for both consumers and merchants and acts as an
Anonymous digital cash payment systems have aggregator of purchase information. When purchases
remained primarily of theoretical interest, although across a number of merchants exceed a certain thresh-
some tr ials have been made (notably of the old amount, that amount is charged in a single credit
Digicash system pioneered by David Chaum). card purchase.
Anonymous payment for large purchases is illegal An example of a token-based micropayment sys-
in the United States, where large purchases must be tem is the PepperCoin system proposed by Ron Rivest
recorded and reported to the government. Moreover, and Silvio Micali and currently being commercial-
consumers generally want to record their purchases ized. Peppercoin uses a unique system of lottery
(especially large ones) to have maximum consumer tickets for purchases. For example, if a consumer
protection. Some researchers have demonstrated that wishes to make a ten-cent purchase, he might use a
anonymous digital cash payment systems are not com- lottery ticket that is worth ten dollars with a prob-
patible with atomic purchases (that is, guaranteed ex- ability of 1 percent. The expected value paid by the
change of goods for payment). The principal demand consumer would be the same as the items he pur-
for anonymous payment appears to be for transactions chased; but any single charge would be large enough
designed to evade taxes, transactions of contraband, to justify being charged using a traditional payment
and transactions of socially undesirable material. mechanism (such as a credit card).
Despite the promise of micropayment systems,

they remain largely unused. Most merchants prefer DIGITAL DIVIDE
to support small-value items by using Web-based
advertising or subscriptions. Nonetheless, advocates There is both optimism and pessimism about the ul-
of micropayment systems maintain that such sys- timate impact of the digital revolution on individ-
tems enable new classes of electronic commerce. ual, societal, and global well-being. On the optimistic
side are hopes that access to information and com-
munication technologies, particularly the Internet,
Challenges for Digital Cash will facilitate a more equitable distribution of social,
Although digital cash is being increasingly used, a economic, and political goods and services. On the
number of challenges remain. The principal chal- pessimistic side are beliefs that lack of access to these
lenge is associating payment with delivery of goods technologies will exacerbate existing inequalities,
(this challenge is often known as the atomic swap both globally and among groups within societies.
or fair exchange problem.) Merchants also need to The phrase digital divide was coined to refer to this
be protected from using stolen payment information, gap between the technology haves and have-nots
and consumers need to be protected from merchants between those who have access to information and
who inadequately protect payment information (or, communications technologies, most notably the
even worse, engage in fraud.) Finally, effective pay- Internet, and those who do not. The overriding
ment methods need to be developed and accepted to concern is that a world divided by geographic, re-
support both large and small purchases. A balance ligious, political, and other barriers will become
must be reached between consumers who want further divided by differing degrees of access to dig-
anonymous purchases and government authorities ital technologies.
who want to tax or record purchases. These challenges
make digital cash a rapidly developing research area.
Evidence of a Digital Divide
J. D. Tygar Evidence supporting the existence of a global digi-
tal divide is overwhelming. Of the estimated 430 mil-
See also E-business lion people online in 2001, 41 percent resided in the
United States and Canada. The remaining Internet
users were distributed as follows: 25 percent in
FURTHER READING Europe, 20 percent in the Asian Pacific (33 percent
of this group in Asia, 8 percent in Australia and New
Chaum, D., Fiat, A., & Naor, M. (1990). Untraceable electronic cash. Zealand), 4 percent in South America, and 2 percent
In G. Blakley & D. Chaum (Eds.), Advances in cryptolog y
(pp. 319327). Heidelberg, Germany: Springer-Verlag.
in the Middle East and Africa. Even among highly
Electronic Privacy Information Center. (2003). Privacy and human developed nations there are vast differences in
rights 2003. Washington, DC: Author. Internet access. For example, in Sweden 61 percent
Evans, D., & Schmalensee, R. (2000). Paying with plastic: The digi- of homes have Internet access compared to 20 per-
tal revolution in buying and borrowing. Cambridge, MA: MIT
Press.
cent of homes in Spain.
Kocher, P., Jaffe, J., & Jun, B. (1999). Differential power analysis. In M. In light of the global digital divide evidence
Weiner (Ed.), Advances in cryptology (pp. 388397). Heidelberg, that Internet use is rapidly increasing takes on ad-
Germany: Springer-Verlag. ditional significance. According to data compiled
Mann, R., & Winn, J. (2002). Electronic commerce. Gaithersburg, MD:
Aspen Publishers. by a variety of sources, the rise in Internet use
OMahony, D., Peirce, M., & Tewari, H. (2001). Electronic payment sys- extends throughout both the developed and de-
tems for e-commerce (2nd ed.). Norwood, MA: Artech House. veloping world. The rapidly increasing global reach
Tygar, J. D. (1998). Atomicity in electronic commerce. Networker, 2(2),
2343.
of the Internet intensifies concerns about its po-
Wayner, P. (1997). Digital cash: Commerce on the Net (2nd ed.). San tential to exacerbate existing global economic and
Francisco: Morgan-Kaufmann. social disparities.
DIGITAL DIVIDE 171
HomeNetToo Tries to Bridge Digital Divide
B e g u n i n t h e f a l l o f 2 0 0 0 , Hom e Ne t To o w a s a n If Im stressed out or depressed or the day

eighteen-month field study of home Internet use in low- is not going right, I just get on the computer
and just start messing around and I come
income families. Funded by an Information Technology
up with all sorts of things like okay, wow.
Research grant from the National Science Foundation, the
project recruited ninety families who received in-home You get a lot of respect because you have
a computer in your house. I think people
instruction on using the Internet, and agreed to have their
view you a little differently.
Internet use recorded and to complete surveys on their ex-
periences. In exchange, each family received a new home A lot of times Im real busy, and it was hard
for me to get a turn on the computer too.
computer, Internet access, and in-home technical support.
My best chance of getting time on the com-
The comments of the HomeNetToo participants about
puter is I get up at 6 AM and the rest of
their computer use provide a broad range of views about the family gets up at seven. So if I finish my
the pleasures and problems of computer interactions: bath and get ready quickly I can get on
before anyone else is up. And I can have an
When somebodys on the computer what-
hour space to do whatever I want while
ever it is theyre doing on that computer at
theyre sleeping and getting up and dressed
that time, thats the world theyre inits
themselves.
another world.
I feel like I dont have time ...who has time
With the computer I can do thingswell,
to watch or play with these machines.
I tell the computer to do things nobody else
Theres so much more in life to do.
will ever know about, you know what I am
saying? I have a little journal that I keep that Instead of clicking, I would like to talk to
actually nobody else will know about unit and then say Can I go back please?
less I pull it up.
They talk in computer technical terms. If
I escape on the computer all the time...I they could talk more in laymans terms, you
like feeling connected to the world and I know, we could understand more and solve
can dream. our own problems.
Source: Jackson, L. A., Barbatsis, G., von Eye, A., Biocca, F. A., Zhao, Y.,
& Fitzgerald, H. E. (2003c). Implications for the digital divide of
Internet use in low-income families. IT & Society., 1(5), 219244.
Evidence for a digital divide within the United to the following factors have been observed in
States is a bit more controversial, and has shifted all surveys to date:
from irrefutable in 1995 to disputable in 2002.
In its first Internet repor t in 1995, the U.S. Income: Income is the best predictor of Internet
Department of Commerce noted large disparities access. For example, only 25 percent of house-
in Internet access attributable to income, educa- holds with incomes of less than $15,000 had
tion, age, race or ethnicity, geographic location, Internet access in 2001, compared to 80 per-
and gender. In its fifth Internet report in 2002, all cent of households with incomes of more than
disparities had shrunk substantially. However, only $75,000.
a few disappeared entirely. Although 143 million Education: Higher educational attainment is as-
U.S citizens now have access to the Internet sociated with higher rates of Internet use. For ex-
(54 percent of the population), gaps attributable ample, among those with bachelors degrees or
better, over 80 percent use the Internet, compared African-Americans are less likely than whites to
to 40 percent of those with only a high school say the Internet helps them to stay connected
diploma. to family and friends.
Age: Internet use rates are highest between the Women and parents are driving the growth of
ages of twelve and fifty; they drop precipitously the African-American Internet population.
after age fifty-five. Mirroring the pattern of gender differences in the
Race or ethnicity: Asian/Pacific Islanders and general population, African-American women
whites are more likely to use the Internet (71 per- are much more likely than African-American men
cent and 70 percent, respectively) than are to search for health, job, and religious informa-
Hispanics and African-Americans (32 percent tion online. African American men are much
and 40 percent, respectively). However, growth more likely than African-American women to
in Internet use has been greater among the lat- search for sports and financial information and
ter than the former groups. to purchase products online.
Compared with older African-Americans, those
The gender gap so evident in the 1995 U.S.
under age thirty are more likely to participate in
Department of Commerce survey disappeared by the
chat rooms, play games, and use multimedia
2002 survey. However, gender-related differences re-
sources. Older African-Americans are more likely
main. Among those over sixty years old, men had
to search for religious information than are
higher Internet use rates than did women. Among
younger African-Americans.
the twenty-to-fifty-year-old group, women had higher The gap in Internet access between African-
Internet use rates than did men. Also diminishing if
Americans and whites is closing, but African-
not disappearing entirely are gaps related to geo-
Americans still have a long way to go. Moreover,
graphic location. Internet use rates in rural areas
those with access to the Internet do not go on-
climbed to 53 percent in 2002, almost as high as the
line as often on a typical day as do whites, and
national average, but use rates for central-city resi-
online African-Americans do not participate on
dents was only 49 percent, compared to 57 percent
a daily basis in most Web activities at the same
for urban residents outside the central city.
level as do online whites.
In addition to the five Internet reports by the U.S.
Department of Commerce, a number of other or-
A number of researchers have also been inter-
ganizations have been tracking Internet use and is-
ested in race differences in U.S. Internet access.
sues related to the digital divide. The Pew Internet
Thomas Hoffman and Donna Novak, professors of
and American Life Project devoted one of its several
management at Vanderbilt University, examined the
reports to African-Americans and the Internet,
reasons for race differences in Internet access and
focusing on how African-Americans Internet use
concluded that income and education cannot fully
differ from whites use. These differences are im-
explain them. Even at comparable levels of income
portant to understanding the racial digital divide in
and education, African-Americans were less likely to
the United States and are potentially important to
have home PCs and Internet access than were whites.
understanding global digital-divide issues that
The psychologist Linda Jackson and her colleagues
may emerge as access to the Internet becomes less
have found race differences in Internet use among
problematic. The Pew Internet and American Life
college students who had similar access to the
Project reported the following findings:
Internet.
African-Americans are more likely than whites The United States is not the only country to re-
to use the Internet to search for jobs, places to port a domestic digital divide. In Great Britain the
live, entertainment (for example, music and digital divide separates town and country, according
videos), religious or spiritual information and to a 2002 joint study by IBM and Local Futures, a re-
health care information, and as a means to search and strategy consultancy. According to the
pursue hobbies and learn new things. studys findings, Britains digital divide may soon
DIGITAL DIVIDE 173
grow so wide that it will not be bridgeable. People in The Divide between
Great Britains rural areas currently do not have
the same degree of access to new technologies, Digital Use and Nonuse
such as cell phones, as do people in cities and the Why do individuals choose to use or not use the
areas surrounding them. Internet, assuming they have access to it? A number
of studies have examined peoples motivations for
using or not using the Internet. According to the uses
Why Is There a and gratifications model of media use, individuals
should use the Internet for the same reasons they use
Digital Divide? other media, namely, for information, communi-
The global digital divide appears to have an obvious cation, entertainment, escape, and transactions.
cause. In the absence of evidence to the contrary, it Research generally supports this view, although
is reasonable to assume that the divide is attributa- the relative importance of these different motiva-
ble to differing degrees of access to digital tech- tions varies with demographic characteristics of the
nologies, especially the Internet. Of course there are user and changes in the Internet itself. For example,
a host of reasons why access may be lacking, including older users are more likely to use the Internet for in-
the absence of necessary infrastructure, government formation, whereas younger users are more likely to
policy, and abject poverty. Regardless of the specific use it for entertainment and escape. Entertainment
factor or factors involved, the access explanation as- and escape motives are more important today than
sumes that if access were available, then the global they were when the World Wide Web was first
divide would disappear. In other words, Internet ac- launched in 1991.
cess would translate readily into Internet use. A report issued in 2000 by the Pew Internet
Explaining the U.S. digital divide in terms of ac- and American Life Project focused specifically on
cess to digital technologies is a bit more problem- why some Americans choose not to use the Internet.
atic. Indeed, some have argued that there is no digital The authors noted that 32 percent of those currently
divide in the U.S. and that the so-called information without Internet access said they would definitely
have-nots are really information want-nots. Those not be getting accessabout 31 million people.
advocating this perspective view the U.S. Department Another 25 percent of non-Internet users said they
of Commerce 2002 report as evidence that indi- probably would not get access. Reasons for not go-
viduals without access have exercised their free choice ing online centered on beliefs that the Internet is a
to say no to the Internet in favor of higher priorities. dangerous place (54 percent), that the online world
Moreover, those who argue that the divide is dis- has nothing to offer (51 percent), that Internet ac-
appearing say that because the growth rate in Internet cess is too expensive (39 percent), and that the on-
use is much higher for low-income groups than it is line world is confusing and difficult to navigate
for high-income groups (25 percent as opposed to (36 percent). The strongest demographic predictor
15 percent), the gap between rich and poor will even- of the decision not to go online was age. Older
tually be negligible without any intervention from Americans apparently perceived few personal ben-
government or the private sector. efits to participating in the online world; 87 percent
Those who argue that a digital divide persists of those sixty-five and older did not have Internet
in the United States despite increasing low-income access, and 74 percent of those over fifty who were
access suggest that the divide be reconceptualized to not online said they had no plans to go online. In
focus on use rather than access. This reconceptual- contrast, 65 percent of those under fifty said they
ization highlights the importance of understanding planned to get Internet access in the near future.
peoples motivations for Internet use and nonuse, Ipsos-Reid, a research firm, used an international
an understanding that will be even more impor- sample to examine peoples reasons for not going
tant if the global digital divide proves to be more online. Their findings, published in 2000, were sim-
than a matter of access to digital technologies. ilar to the Pew report findings: Thirty-three percent
of respondents said they had no intention of going olution to organize a world summit on the infor-
online. Their reasons included lack of need for the mation society, the first in Geneva in 2003, and the
online world (40 percent), lack of a computer (33 per- second in Tunisia in 2005. The summits are expected
cent), lack of interest in going online (25 percent), to promote universal access to the information,
lack of necessary technical skills, and general cost knowledge, and communications technologies
concerns (16 percent). needed for social and economic development.
The Childrens Partnership, which also published In April 2002, Erkki Liikanen, the European com-
a report in 2000 on why people do not go online, of- missioner for the Enterprise Directorate General and
fered four reasons why low-income and underserved the Information Society Directorate General, argued
Americans may choose to stay away from the that developing countries must be included in the
Internet. First, the Internet may lack the local infor- shift to a networked, knowledge-based global econ-
mation of interest to low-income and underserved omy. He stressed the importance of strong political
Americans; second, there may be literacy barriers; leadership, top-level involvement and contributions
third, there may be language barriers; and fourth, from both the public and private sectors. In 2000,
the lack of cultural diversity on the Internet may keep the European Commission launched an action plan,
them from participating. Lack of local information the goal of which was to bring all of Europe online
disproportionately affects users living on limited by 2002. As a result of this action plan, decision mak-
incomes. Literacy barriers come into play because ing on telecommunications and e-commerce regu-
online content is often directed at more educated lation accelerated and Internet access has moved to
Internet users, particularly users who have discre- the top of the political agenda in all European Union
tionary money to spend online. Reading and un- member countries. In the coming years the focus will
derstanding Web content may be especially difficult move to the user and usage of the Internet. The goal
for the less educated and those for whom English is to encourage more profound and inclusive use
is a second language (32 million Americas). An es- of the Internet. In the United States a number of non-
timated 87 percent of the documents on the Internet profit organizations have looked to the federal
are in English. The lack of cultural diversity on the government to address the digital divide. For exam-
Internet may be rendering the Internet less interest- ple, upon release of the U.S. Department of Com-
ing to millions of Americans. merces fifth digital divide report in 2002, the Benton
Others have argued that access alone may not be Foundation issued a policy brief stating that
enough to produce equity in Internet use in the Targeted [government] funding for community
United States. Gaps will persist due to differences in technology is essential to maintain national digital
education, interest in Web topics, and interpersonal divide leadership (Arrison 2002). The government,
contact with others familiar with these topics. All of however, continues to minimize the importance of
these factors may affect how eagerly an individual the digital divide, asserting that for the all intents
seeks out and consumes information on the Internet. and purposes it no longer exists.
Thus, while some call for broad-based approaches
to eliminating the global digital divide and govern-
Whose Responsibility is the ment intervention to eliminate the U.S. digital di-
vide, others argue that nothing at all needs to be done,
Digital Divide? that market forces will bridge the digital divide with-
Opinions vary about whose responsibility it is to ad- out any other action being taken. Still others believe
dress the digital divide, whether it be the global di- that access to and use of digital technologies, par-
vide, the U.S. divide, or the divide between users and ticularly the Internet, are neither necessary for every-
nonusers. At the global level, in June 2002 the United day life nor solutions to social and economic
Nations telecommunications agency argued that it problems in the United States or elsewhere.
would take concerted global action to keep the
digital divide from growing. The U.N. adopted a res- Linda A. Jackson
DIGITAL GOVERNMENT 175
See also Economics and HCI; InternetWorldwide the twelfth International World Wide Web Conference, Budapest,
Diffusion Hungary.
Lenhart, A. (2000). Whos not online: 57% of those without
Internet access say they do not plan to log on. Washington, DC:
Pew Internet & American Life Project. Retrieved July 18, 2003,
FURTHER READING from http://www.pewinternet.org/reports/pdfs/Pew_Those_
Not_Online_Report.pdf
Local Futures. (2001) Local futures research: On the movemobile and
Arrison, S. (2002, April 19). Why digital dividers are out of step. Retrieved
wireless communications. Retrieved July 18, 2003, from http://
July 17, 2003, from http://www.pacificresearch.org/press/opd/
www.localfutures.com/article.asp?aid=41
2002/opd_02-04-19sa.html
National Telecommunications and Information Administration,
Associated Press. (2002, June 22). U.N. warns on global digital di-
Economics and Statistics Administration. (n.d.) A nation online:
vide. Retrieved July 18, 2003, from http://lists.isb.sdnpk.org/
How Americans are expanding their use of the Internet. Retrieved
pipermail/comp-list/2002-June/001053.html
July 18, 2003, from http://www.ntia.doc.gov/ntiahome/dn/
BBC News. (2002, March 10). Digital divisions split town and country.
html/toc.htm
Retrieved July 18, 2003, from http://news.bbc.co.uk/2/hi/
Ombwatch. (2002, August 18). Divided over digital gains and gaps.
science/nature/1849343.stm
Retrieved July 18, 2003, from http://www.ombwatch.org/article/
Carvin, A. (2000). Mind the gap: The digital divide as the civil rights
articleview/1052/
issue of the new millenium. Multimedia Schools, 7(1), 5658.
The relevance of ICT in development. (2002, May-June) The
Retrieved July 17, 2003, from http://www.infotoday.com/mmschools/
Courier ACP-EU, 192, 3739. Retrieved 17 July 2003, from
jan00/carvin.htm
http://europa.eu.int/comm/development/body/publications/
Cattagni, A., & Farris, E. (2001). Internet access in U.S. public schools
courier/courier192/en/en_037_ni.pdf
and classrooms: 19942000 (NCES No. 2001-071). Retrieved July
Spooner, T., & Rainie, L. (2000). African-Americans and the Internet.
18, 2003, from http://nces.ed.gov/pubsearch/pubsinfo
Washington, DC: Pew Internet & American Life Project. African
.asp?pubid=2001071
Americans and the Internet. Retrieved July 18, 2003, from
Childrens Partnership. (2000). Online content for low-income and un-
http://www.pewinternet.org/reports/pdfs/PIP_African_Americans_
derserved Americans: The digital divides new frontier. Retrieved July
Report.pdf
17, 2003, from http://www.childrenspartnership.org/pub/low_income/
UCLA Center for Communication Policy. (2000). The UCLA Internet
Cooper, M. N. (2002, May 30). Does the digital divide still exist? Bush ad-
report: Surveying the digital future. Retrieved July 18, 2003, from
ministration shrugs, but evidence says yes. Retrieved July 18, 2003,
http://www.ccp.ucla.edu/UCLA-Internet-Report-2000.pdf
from http://www.consumerfed.org/DigitalDivideReport20020530
UCLA Center for Communication Policy. (2003). The UCLA Internet
.pdf
report: Surveying the digital future, year three. Retrieved July 18,
Digital Divide Network staff. (2003). Digital divide basics fact sheet.
2003, from http://www.ccp.ucla.edu/pdf/UCLA-Internet-Report-
Retrieved July 18, 2003, from http://www.digitaldividenetwork
Year-Three.pdf
.org/content/stories/index.cfm?key=168
U.S. Department of Commerce. (1995). Falling through the Net: A sur-
eEurope. (19952002). An information society for all. Retrieved July
vey of the have nots in rural and urban America. Retrieved July
18, 2003, from http://europa.eu.int/information_society/eeurope/
18, 2003, from http://www.ntia.doc.gov/ntiahome/fallingthru.html
index_en.htm
U.S. Department of Commerce. (2000). Falling through the Net: Toward
European Union. (2002, May 4). e-Government and development:
digital inclusion. Retrieved July 18, 2003, from http://search.ntia
Br idg ing the gap. Ret r ie ved July 18, 2003, from http://
.doc.gov/pdf/fttn00.pdf
europa.eu.int/rapid/start/cgi/guesten.ksh?p_action.gettxt=gt&doc=
U.S. Department of Commerce. (2002). A nation online: How Americans
SPEECH/02/157|0|RAPID&lg=EN&display=
are expanding their use of the Internet. Retrieved July 18, 2003, from
Gorski, P. (Fall, 2002). Dismantling the digital divide: A multicultural
http://www.ntia.doc.gov/ntiahome/dn/anationonline2.pdf
education framework. Multicultural Education, 10(1), 2830.
Weiser, E. B. (2002). The functions of Internet use and their social and
Hoffman, D. L., & Novak, T. P. (1998, April). Bridging the racial di-
psychological consequences. Cyberpsychology and Behavior, 4(2),
vide on the Internet. Science, 280, 390391.
723743.
Hoffman, D. L., Novak, T. P., & Schlosser, A. E. (2000). The evolu-
tion of the digital divide: How gaps in Internet access may impact
electronic commerce. Journal of Computer Mediated Commu-
nication, 5(3), 157.
Jackson, L. A., Ervin, K. S., Gardner, P. D., & Schmitt, N. (2001a).
The racial digital divide: Motivational, affective, and cognitive cor-
relates of Internet use. Journal of Applied Social Psychology, 31(10),
DIGITAL GOVERNMENT
20192046.
Jackson, L. A., Ervin, K. S., Gardner, P. D., & Schmitt, N. (2001b). Electronic government (e-government) is intimately
Gender and the Internet: Women communicating and men search- connected to human-computer interaction (HCI).
ing. Sex Roles, 44(56), 363380.
Jackson, L. A., von Eye, A., Biocca, F., Barbatsis, G., Fitzgerald, H. E.,
Critical HCI issues for e-government include technical
& Zhao, Y. (2003, May 2024). The social impact of Internet Use: and social challenges and interactions between the
Findings from the other side of the digital divide. Paper presented at two. First, at a broad, societal level, the adaptation of
government and civic engagement to increasingly public using a range of information and commu-
computerized environments raises political, orga- nication technologies (ICTs). The public includes
nizational, and social questions concerning use, the individuals, interest groups, and organizations,
appropriate contexts or environments for use, recip- including nonprofit organizations, nongovernmental
rocal adaptation mechanisms, learning and the de- organizations, firms, and consortia. The definition of
sign of government work, the design of political and e-government used here also includes e-democracy,
civic communities of interest, and the design of that is, civic engagement and public deliberation
nations themselves as well as international governance using digital technologies. Governments in indus-
bodies. Second, HCI focuses on human character- trialized and developing countries are experiment-
istics and their relationship to computing. The sig- ing with interactive systems to connect people with
nificant human characteristics of importance to government information and officials. Many ob-
e-government include cognition, motivation, lan- servers have claimed that interactive technologies
guage, social interaction, and ergonomics or hu- will revolutionize governance. We must wait to see
man factors issues. The usability and feasibility of how and to what extent individuals and groups
e-government require a deep understanding by de- will use computing to affect civic engagement and
signers of individual, group, and societal cognition how governments will use computing to influence
and behavior. political and civic spheres.
On the technological side HCI is concerned with
the outputs and processes of design and develop-
ment of systems and interfaces. Third, HCI and Development Paths of E-government
e-government intersect is the design of computer Initial efforts by government agencies to develop
systems and interface architectures. Design questions e-government entailed simply digitizing and post-
apply to input and output devices, interface archi- ing static government information and forms on the
tectures (including all types of dialogue interfaces World Wide Web using the language, displays, and
for individuals and shared spaces for multiple users), design of existing paper-based documents. Beginning
computer graphics, maps, visualization tools, and during the 1990s and continuing into the present
the effects of these systems and interface architec- many government agencies have begun to adapt op-
tures on the quality of interaction among individ- erations, work, and business processes and their
uals, groups, and government. Fourth, HCI examines interface with the public to simplify and integrate
the development process itself, ranging from how information and services in online environments.
designers and programmers work to the evaluations The federal governments of the United States,
of human-computer systems in terms of feasibil- Canada, Finland, and Singapore are among those at
ity, usability, productivity and efficiency and, more the forefront of e-government in terms of the amount
recently, their likelihood to promote and sustain dem- of information and interactivity available to the
ocratic processes. public and attention to system development and
These issues may be described separately; how- interface architecture. The country-level Web por-
ever, e-government projects require attention to sev- tal designed to help people navigate and search in-
eral of these issues simultaneously. For example, formation for entire federal governments is one of
user-friendly and socially effective applications that the key types of e-government initiatives. The U.S.
cannot be implemented in a government setting for government Web portal (www.FirstGov.gov) is an
reasons of privacy, fairness, cost, or user resistance interface with a search tool meant to serve as a sin-
prove infeasible for e-government. Multiple con- gle point of entry to U.S. government information
straints and demands therefore make this area chal- and services. The federal government of Singapore
lenging for governments. developed a single Web portal, called Singov
Electronic government is typically defined as the (www.gov.sg), to simplify access to government in-
production and delivery of information and services formation for visitors, citizens, and businesses.
inside government and between government and the Similarly, the Web portal for the government of
Canada (www.canada.gc.ca) was designed in terms of Standardization, consolidation, and integration of

three main constituents: Canadians, non-Canadians, information, operations, and interfaces with the pub-
and Canadian business. lic have been the key drivers for e-government in most
federal government efforts.
The ability to digitize visual government in-
Organizational Redesign through formation is an additional development path for
e-government. To note one example: The U.S. House
Cross-Agency Integration Committee on Government Reform Subcommittee
During the 1990s several federal agencies and state on Technology, Information Policy, Intergovernmental
governments created virtual agenciesonline Relations, and the Census Web-casts its hearings and
sources of information and services from several agen- makes testimony before the committee searchable
cies organized in terms of client groups. For exam- online. Previously, testimony was not searchable un-
ple, during the early 1990s the U.S. federal government til a transcript of a hearing was produceda process
developed the websites students.gov, seniors.gov, and that could take up to six months.
business.gov to organize and display information us- Considerable human, financial, and technical re-
ing interfaces designed specifically for these popula- sources are required to design, develop, build, and
tions with a single point of entry into a government maintain state-of-the-art e-government. For this rea-
portal focused on each populations interests. By the son, many local governments in poor economic
end of the administration of President Bill Clinton areas or in cities and towns with small and medium
approximately thirty cross-agency websites existed populations lack resources to build interactive
in the U.S. federal government. e-government unless resources are provided by fed-
Beginning in 2001, the U.S. federal government eral and state governments. In the United States some
continued this development path by undertaking of the most developed state-level e-government sites
more than twenty-five cross-agency e-government are in the states of Washington and Virginia. Municipal
projects. The development path shifted from a government websites vary dramatically in quality
loose confederation of interested designers in the gov- and level of development.
ernment to an enterprise approach to e-government
managed and controlled centrally and using lead
agencies to control projects. The desire for internal Interactivity and E-government
efficiencies, as much as desire for service to the Interactive e-government services include online tax
public, drives these projects. Several payroll systems payments, license applications and renewals, and
are being consolidated into a few payroll systems for grants applications and renewals. The city of Balti-
the entire government. Multiple and abstruse (dif- more website (http://www.ci.baltimore.md.us/) has
ficult to comprehend) requirements for finding won awards for implementation of computing tech-
and applying for government grants are being stream- nology in government. The website allows citizens
lined into one federal online grants system called to pay parking fines, property taxes, and water and
e-grants. Myriad rulemaking processes in agencies other bills. Users can search crime statistics by geo-
throughout the federal government, although not graphic area within the city and track several city
consolidated, have been captured and organized in services, including trash removal and street clean-
the interface architecture of one Web portal, called ing. The city of Baltimore has implemented an on-
e-rulemaking. The website recreation.gov uses an line version of the 311 service available in some other
architecture that organizes recreation information large U.S. cities, which allows citizens to request city
from federal, state, and local governments. System information and services over the telephone. Citizens
design and interface architecture simplify search, nav- can report and then track the status of a request
igation, and use of information concerning recre- for city services, including removal of abandoned
ational activities, recreational areas, maps, trails, vehicles, repair of potholes, removal of graffiti, and
tourism sites, and weather reports by location. requests for a change in traffic signs. These requests
not only provide interactivity but also promote gov- cultural Environment (http://www.cba.nau.edu/
ernment compliance and accountability to voters by facstaff/becker-a/Accessibility/main.html), focuses
making provision of city services more transparent on the development of tools for government agen-
to the public. cies to assess the usability of systems and sites for the
Interactivity is increasing as governments con- elderly as well as standards of measurement for eval-
tinue to develop systems and as citizens adapt to gov- uating such sites. Developers will use evaluation tools
ernment online. To note a few trends: In the United to measure a sites accessibility in terms of reading
States the number of online federal tax filings in- complexity and potential usability issues such as font
creased from 20,000 in 1999 to 47 million, or size and font style, background images, and text jus-
about 36 percent of individual filings, in 2002. The tification. Transformational tools will convert a
Environmental Protection Agency reports that it saves graphical image to one that can be seen by those users
approximately $5 million per year in printing and with color-deficiency disabilities. Developers are cre-
mailing costs by providing information digitally to ating simulation tools to model many of the prob-
the public. Public health agencies at all levels of gov- lems that elderly users experience, such as yellowing
ernment increasingly access centralized information and darkening of images. Finally, compliance tools
online through the Centers for Disease Control and will be designed to modify webpages to comply with
Protection of the U.S. Public Health Service. usability requirements for the elderly.
Other U.S. researchers are working with the Social
Security Administration, the Census Bureau, and the
Usability and E-government General Services Administration to better provide for
Usability studies in HCI examine the ease and effi- their visually impaired users in a project entitled
ciency with which users of a computer system can Open a Door to Universal Access. Project researchers
accomplish their goals as well as user satisfaction are building and prototyping key technologies for
with a system. Usability in e-government is impor- disabled employees at the partner agencies. These
tant because it is likely to affect public participation technologies will later be transferred to the private
in ways that might result in unequal access or dis- sector for wider dissemination in work settings.
crimination due to biases built into design and Usability includes all elements of accessibility,
architecture. including look and feel, readability, and naviga-
One area of usability concerns disabled people. bility. For example, usability research focused on
Many governments around the world have passed local government websites indicates that the read-
laws to ensure usability to the disabled. Section 508 ing level required to comprehend information on
of the U.S. Rehabilitation Act (29 U.S.C. 794d), as websites often exceeds that of the general popula-
amended by the Workforce Investment Act of 1998 tion, raising concerns about accessibility, compre-
(P.L. 105-220), 7 August 1998, mandates a set of re- hension, interpretation, and associated potential
quirements for U.S. federal government sites to as- for discrimination. Ongoing research regarding
sist disabled users. These requirements include e-government and usability focuses primarily on de-
standards for Web-based software and applications, velopment of tools for usability, including naviga-
operating systems, telecommunications products, bility and information representation in text, tabular,
personal computers, video, and multimedia prod- graphical, and other visual forms.
ucts. Major federal services initiatives have been
delayed and others upgraded to ensure compliance
with Section 508 requirements. Internet Voting
Disabilities increase as a population ages and One of the most important developments in e-
chiefly include visual impairment and decreases in government, with great significance for issues in HCI,
cognitive and motor skills important in an online is Internet voting. People have debated three main
environment. A research initiative, Toolset for Making possibilities for Internet voting. First, computerized
Web Sites Accessible to Aging Adults in a Multi- voting can be used at polling places in a closed
system within a secure computer local area network velopment of graphical tools to simplify complex
(LAN). Local votes would be recorded from indi- information. This project will develop and assess
vidual voting consoles and tallied at local polling sta- quality graphics for federal statistical summaries con-
tions. Second, voting consoles or kiosks can be located sidering perceptual and cognitive factors in reading,
in areas widely accessible to the general population, interaction, and interpretation of statistical graphs,
such as public libraries or shopping malls. Third, maps, and metadata (data about data). The project
Internet voting might take place from remote loca- addresses four areas: conversion of tables to graphs,
tions, such as homes or offices. representation of metadata, interaction of graphs
Many observers predicted that Internet voting and maps, and communication of the spatial and
would simplify voting processes and thereby increase temporal relationships among multiple variables.
voter participation. These predictions are far from The project uses Web-based middlewaresoftware
reality at present. Current systems and architectures which connects applicationsto enable rapid de-
lack the security and reliability required for Internet velopment of graphics for usability testing.
voting of the third type. In addition to questions Another research project, Integration of Data
of feasibility, experts are uncertain of how Internet and Interfaces to Enhance Human Understanding
voting would affect participation and the cogni- of Government Statistics: Toward the National
tive, social, and political process of voting itself. Statistical Knowledge Network (http://ils.unc.edu/
A current research study, Human Factors govstat/), takes a different HCI approach. Mem-
Research on Voting Machines and Ballot Design bers of the civically engaged public often struggle to
(http://www.capc.umd.edu/rpts/MD_EVoteHuFac access and combine the vast and increasing amount
.html), focuses on the human-machine interface in of statistical dataoften in a variety of formats
voting. Given the prominence of issues surrounding available from government agency websites.
traditional voting methods during the 2000 U.S. Researchers working in cooperation with govern-
presidential election, researchers from the University ment agencies are developing standardized data for-
of Maryland are developing a process to evaluate mats and studying social processes to facilitate
several automated voting methods and ballot de- integration of search results. In addition, the pro-
signs. The study compares technologies such as jects research team is developing a solutions archi-
optical scanning and digital recording of electronic tecture to accommodate users with a variety of
equipment and evaluates the effect of various vot- communications and hardware needs and provid-
ing methods and ballot designs on the precision with ing for broad-based usability requirements.
which voters intentions are recorded and other crit-
ical variables.
Ways Forward
The technological potential exists for individuals,
Representing Complex groups, and communities to participate in and shape
government in new ways. Some observers specu-
Government Information late that increased access to government online
Government statistics are a powerful source of in- will lead to greater interest, knowledge, and discus-
formation for policymakers and the public. Large, sion of politics. The Internet might allow citizens to
democratic governments produce and distribute a organize and mobilize resources in powerful new
vast quantity of statistical information in printed ways. The Internet enables groups and communities
and electronic form. Yet, vital statistics continue to to deliberate in new, possibly more effective ways.
be stored in databases throughout governments and Some observers have also speculated that comput-
in forms that are not easily accessible, navigable, or ing will lead to direct democracy, with individuals
usable by most citizens. A U.S. project called Quality voting on a wide range of issues. Currently, little evi-
Graphics for Federal Statistics (http://www.geovista dence shows that this potential is being realized.
.psu.edu/grants/dg-qg/intro.html) focuses on de- Those groups already civically engaged are using
computing to enhance their activities. The propen- Alvarez, R. M. (2002). Ballot design options, California Institute of
sity to simplify and distort information in public dis- Technology. Retrieved February 17, 2004, from http://www.capc
.umd.edu/rpts/MD_EVote_Alvarez.pdf
course is not abated by changes in media. Ceaparu, I. (2003). Finding governmental statistical data on the Web:
Unequal access to the Internet and a wide range A case study of FedStats. IT & Society, 1(3), 117. Retrieved February
of computerized information and communication 17, 2004, from http://www.stanford.edu/group/siqss/itandsociety/
tools, roughly divided between people with education v01i03/v01i03a01.pdf
Conrad, F. G. (n.d.). Usability and voting technology: Bureau of Labor
and people without, highly correlated with income Statistics. Retrieved February 17, 2004, from http://www.capc.umd
and political participation, creates a digital divide in .edu/rpts/MD_EVote_Conrad.pdf
e-government in spite of advances in HCI. Lack of lit- David, R. (1999). The web of politics: The Internets impact on the
American political system. New York: Oxford University Press.
eracy and lack of computer literacy worsen the digi- Dutton, W. H. (1999). Society on the line: Information politics in the
tal divide in access. Disparities among rich and poor digital age. Oxford, UK: Oxford University Press.
nations parallel digital-divide challenges within coun- Dutton, W. H., & Peltu, M. (1996). Information and communication
tries. Yet, innovations in several developing countries technologiesVisions and realities. Oxford, UK: Oxford University
Press.
and in rural areas invite some degree of optimism. Echt, K. V. (2002). Designing Web-based health information for older
Rural farmers and craftspeople are beginning to con- adults: Visual considerations and design directives. In R. W. Morrell
nect through the Internet to enhance their economic (Ed.), Older adults, health information, and the World Wide Web
well-being. Rural communities in China are using the (pp. 6188). Mahwah, NJ: Lawrence Erlbaum Associates.
Fountain, J. E. (2001). Building the virtual state: Information technol-
Internet, as yet on a modest scale, to decry local cor- ogy and institutional change. Washington, DC: Brookings Institution
ruption and in some cases have forced the federal gov- Press.
ernment to intervene in local affairs. Interfaces for Fountain, J. E. (2002). Information, institutions and governance:
Advancing a basic social science research program for digital gov-
preliterate populations are being developed. ernment. Cambridge, MA: National Center for Digital Government,
Human-computer interaction begins with the John F. Kennedy School of Government.
study of the mutual adaptation of social and tech- Fountain, J. E., & Osorio-Urzua, C. (2001). The economic impact of
nical systems. We cannot predict the path or the out- the Internet on the government sector. In R. E. Litan & A. M. Rivlin
(Eds.), The economic payoff from the Internet re volution
come of the many and varied complex adaptation (pp. 235268). Washington, DC: Brookings Institution Press.
processes now in play. One of the chief sources of Harrison, T. M., & Zappen, J. P. (2003). Methodological and theo-
learning for designers of e-government has been to retical frameworks for the design of community information sys-
focus on tools for building and sustaining democ- tems. Journal of Computer-Mediated Communication, 8(3).
Retrieved February 17, 2004, from http://www.ascusc.org/jcmc/
racy rather than to focus merely on efficiency. While vol8/issue3/harrison.html
researchers learn more about human cognition, Harrison, T. M., Zappen, J. P., & Prell, C. (2002). Transforming new
social interaction, and motivation in computer- communication technologies into community media. In N. W.
Jankowski & O. Prehn (Eds.), Community media in the informa-
mediated environments and while designers develop tion age: Perspectives and prospects (pp. 249269). Cresskill, NJ:
new tools and interfaces to encompass a wider range Hampton Press Communication Series.
of activities and discourse in online environments, Hayward, T. (1995). Info-rich, info-poor: Access and exchange in the
large-scale adaptation continues between societies, global information society. London: K. G. Saur.
Heeks, R. (Ed.). (1999). Reinventing government in the information
governments, and technology. age: International practice in IT-enabled public sector reform. London
and New York: Routledge.
Jane E. Fountain and Robin A. McKinnon Hill, K. A., & Hughes, J. E. (1998). Cyberpolitics: Citizen activism in the
age of the Internet. Lanham, MD: Rowman & Littlefield.
Holt, B. J., & Morrell, R. W. (2002). Guidelines for website design
See also Online Voting; Political Science and HCI for older adults: The ultimate influence of cognitive factors. In
R. W. Morrell (Ed.), Older adults, health information, and the World
Wide Web (pp. 109129). Mahwah, NJ: Lawrence Erlbaum
Associates.
FURTHER READING Internet Policy Institute. (2001). Report of the National Workshop on
Internet Voting: Issues and research agenda. Retrieved February
17, 2004, from http://www.netvoting.org
Abramson, M. A., & Means, G. E. (Eds.). (2001). E-government 2001
Kamarck, E. C., & Nye, J. S., Jr. (2001). Governance.com: Democracy in
(ISM Center for the Business of Government). Lanham, MD:
the information age. Washington, DC: Brookings Institution Press.
Rowman & Littlefield.
DIGITAL LIBRARIES 181
Margolis, M., & Resnick, D. (2000). Politics as usual: The cyberspace hidden facts or patterns within databases. The term
revolution. Thousand Oaks, CA: Sage. digital libraries has been defined in many ways. For
Nass, C. (1996). The media equation: How people treat computers,
televisions, and new media like real people and places. New York:
example:
Cambridge University Press.
Norris, P. (2001). Digital divide: Civic engagement, information poverty,
The Digital Library is the collection of services
and the Internet worldwide. Cambridge, UK: Cambridge University and the collection of information objects that
Press. support users in dealing with information ob-
OLooney, J. A. (2002). Wiring governments: Challenges and possibili- jects available directly or indirectly via electronic/
ties for public managers. Westport. CT: Quorum Books.
Putnam, R. (2000). Bowling alone: The collapse and revival of American
digital means (Fox and Urs 2002, 515).
community. New York: Simon & Schuster. Digital libraries are organizations that pro-
Rash, W. (1997). Politics on the nets: Wiring the political process. New vide the resources, including the specialized staff,
York: Freeman. to select, structure, offer intellectual access to,
Schwartz, E. (1996). Netactivism: How citizens use the Internet.
Sebastapol, CA: Songline Studios. interpret, distribute, preserve the integrity of,
Wilheim, A. G. (2000). Democracy in the digital age: Challenges to and ensure the persistence over time of collec-
political life in cyberspace. New York: Routledge. tions of digital works so that they are readily avail-
able for use by a defined community or set of
communities(Fox and Urs 2002, 515).
A collection of information which is both dig-
DIGITAL LIBRARIES itized and organized (Lesk 1997, 1).

Digital libraries are a set of electronic resources
and associated technical capabilities for creating,
For centuries the concept of a global repository of
searching, and using information . . . they are an
knowledge has fascinated scholars and visionaries
extension and enhancement of information stor-
alike. Yet, from the French encyclopedist Denis
age and retrieval systems that manipulate digi-
Diderots LEncylopedie to the British writer H. G.
tal data in any medium (text, images, sounds,
Wellss book World Brain to Vannevar Bushs (director
static or dynamic images) and exist in distrib-
of the U.S. Office of Scientific Research and
uted networks (Borgman et al. 1996, online).
Development) Memex (a desktop system for storing
and retrieving information) to Ted Nelsons Project Clifford Lynch, a specialist in networked infor-
Xanadu (a vision of an information retrieval system mation, made a clear distinction between content
based on hyperlinks among digital content con- that is born digital and content that is converted into
tainers), the dream of such an organized and ac- digital format and for which an analogue counter-
cessible collection of the totality of human knowledge part may or may not continue to exist.
has been elusive. However, recent technological The definitions of digital libraries have consider-
advances and their rapid deployment have brought able commonality in that they all incorporate notions
the far-reaching dream into renewed focus. The tech- of purposefully developed collections of digital in-
nologies associated with computing, networking, formation, services to help the user identify and ac-
and presentation have evolved and converged to cess content within the collections, and a supporting
facilitate the creation, capture, storage, access, re- technical infrastructure that aims to organize the col-
trieval, and distribution of vast quantities of data, lection contents as well as enable access and retrieval
information, and knowledge in multiple formats. of digital objects from within the collections.
During the late 1980s and early 1990s the term Yet, the term digital libraries may have constrained
digital libraries emerged to denote a field of inter- development in that people have tended to restrict
est to researchers, developers, and practitioners. The their view of digital libraries to a digital version of
term encompasses specific areas of development such more traditional libraries. They have tended to focus
as electronic publishing, online databases, informa- on textual content rather than on the full spectrum
tion retrieval, and data miningthe process of in- of content typesdata, audio, visual images, simu-
formation extraction with the goal of discovering lations, etc. Much of the development to date has
focused on building collections and tools for organ- development has been elusive. Because of the mul-
izing and extracting knowledge from them. Experts tidisciplinary roots of the field, the different per-
only recently have acknowledged the role of the cre- spectives, and the lack of consensus on definition,
ators and users of knowledge and the contexts in which we can have difficulty understanding the basic
they create and use. Dagobert Soergel, a specialist in constructs of digital libraries. At its simplest inter-
the organization of information, characterizes much pretation, the term digital libraries brings together
of the digital library activity to date as still at the stage the notions of digital computing, networking, and
of horseless carriage; to fulfill its fullest potential, the content with those of library collections, services,
activity needs to move on to the modern automobile and community. Researchers are giving attention to
(Soergel 2002, 1). the 5S framework developed by Edward A. Fox, di-
rector of Digital Libraries Laboratory at Virginia
Tech, Marcos Andr Gonlaves of Digital Libraries
Key Concepts Research, and Neill A. Kipp of Software Architecture.
Whereas researchers have expended considerable This framework defines streams, structures,
effort in developing digital libraries, theoretical spaces, scenarios, and societies to relate and unify
the concepts of documents, metadata (descrip-
tions of data or other forms of information content),
services, interfaces, and information warehouses that
are used to define and explain digital libraries:
Vannevar Bush on the Memex
Streams: sequences of information-carrying el-
S
cientist Vannevar Bushs highly influential essay As
We May Think (1945) introduced the idea of a device ements of all typescan carry static content and
he called the memexinspiring others to develop dynamic content
digital technologies that would find and store a vast amount Structures: specifications of how parts of a whole
of information. are arranged or organized, for example, hypertext,
The owner of the memex, let us say, is inter-
taxonomies (systems of classification), user rela-
ested in the origin and properties of the bow and tionships, data flow, work flow, and so forth
arrow. Specifically he is studying why the short Spaces: sets of objects and operations performed
Turkish bow was apparently superior to the on those objects, for example, measure, proba-
English long bow in the skirmishes of the bility, and vector spaces (a form of mathemati-
Crusades. He has dozens of possibly pertinent
cal representation of sets of vectors) used for
books and articles in his memex. First he runs
through an encyclopedia, finds an interesting but
indexing, visualizations, and so forth
sketchy article, leaves it projected. Next, in a his- Scenarios: events or actions that deliver a func-
tory, he finds another pertinent item, and ties the tional requirement, for example, the services that
two together. Thus he goes, building a trail of are offereddata mining, information retrieval,
many items. Occasionally he inserts a comment summarization, question answering, reference
of his own, either linking it into the main trail or
and referral, and so forth
joining it by a side trail to a particular item. When
Societies: understanding of the entities and their
it becomes evident that the elastic properties of
available materials had a great deal to do with the interrelationships, individual users, and user
bow, he branches off on a side trail which takes communities
him through textbooks on elasticity and tables
of physical constants. He inserts a page of long-
hand analysis of his own. Thus he builds a trail Digital Libraries Today
of his interest through the maze of materials avail- A report from the Presidents Information Tech-
able to him. nology Advisory Committee (PITAC) in 2001 ac-
Source: Bush, V. (1945, July). As we may think. The Atlantic Monthly, 176(1).
Retrieved March 25, 2004, from http://www.theatlantic.com/unbound/
knowledges the need for much more work to be
flashbks/computer/bushf.htm accomplished before we can think of digital libraries
as fully successful in the United States. The report
identifies deficiencies in digital content availability government to play a more aggressive and proactive
and accessibility: Less than 10 percent of publicly role in provision of digital content to all and to use
available information is available in digital form, and digital library technologies and content to transform
less than 1 percent of the digital content is in- the way it services its citizens.
dexed, and therefore identifiable, via Web search en- Another key area identified by the PITAC report
gines. Thus, the visible Web is still small relative to is the opportunities and challenges of digital libraries
the total potential Web. The report goes on to ac- and their long-term preservation. Experts see a slow
knowledge the need to create digital library collec- and steady leakage of digital content from the Web as
tions at a faster rate and much larger scale than are content is updated, archived, or removed. They also
currently available. The report also identifies the need see a need for both standards for digital preserva-
for improved metadata standards and mechanisms tion and archival processes for periodic transfer/trans-
for identifying and providing access to digital library formation to new formats, media, and technologies.
content and the need to advance the state of the Finally, the PITAC report says the issue of in-
art in user interfaces so that digital library users with tellectual property rights needs to be addressed for
different needs and circumstances can use interfaces digital libraries to achieve their full potential. In par-
better suited to their contexts. ticular, clarification was sought by the PITAC
The PITAC report acknowledges that much of Committee on access to information subject to copy-
the progress to date in digital libraries has resulted right, the treatment of digital content of unknown
from the federal governments investments through provenance or ownership, policies about federally
multiagency digital-library research and develop- funded digital content, and the role of the private
ment initiatives and through provision of access to sector.
libraries of medical and scientific data. In 1993 the
National Science Foundation (NSF) funded Mosaic,
the first Web browser to run on multiple plat- The Signicance for HCI
forms, thereby encouraging widescale access to dig- The first decade of digital library research and de-
ital content via the Internet and the Web. In 1994 the velopment provided ample evidence that our abil-
Digital Libraries Initiative (DLI)involving NSF, ity to generate and collect digital content far exceeds
Defense Advanced Research Projects Agency our ability to organize, manage, and effectively use
(DARPA), and National Aeronautics and Space it. We need not look further than our own experi-
Administration (NASA)funded six university-led ences with the growth of digital content and services
consortia to conduct research and development to on the Web. Although the Web may be perceived
make large distributed digital collections accessible by the majority of the using public as a vast library,
and interoperable. In 1998 the program was exit is not a library in several important aspects.
panded to include the National Institutes of Experts acknowledge the importance of under-
Health/National Library of Medicine (NIH/NLM), standing how people interact with digital libraries,
the Library of Congress, National Endowment for how their needs relate to new types of information
the Humanities (NEH), Federal Bureau of Investi- available, and the functionality that is needed by these
gation (FBI), National Archives and Records Admin- new types of information. Numerous experts have
istration (NARA), the Smithsonian Institution, and called for more user-centric approaches to the de-
the Institute for Museum and Library Services. sign and operation of digital libraries. However, these
Other federal agencies compiled some of the same calls tend to still see the user involved only in
largest publicly accessible databases, such as Earth- reaction to the development of certain collections.
observing satellite data, weather data, climate data, Thus, user-centric seems to mean user involve-
and so forth. Most recently, new forms of digital data ment rather than placement of the user and po-
library collections have been initiated, including dig- tential user at the center of digital library activity.
ital libraries of molecules, cells, genomes, proteins, For a truly user-centric approach to emerge,
and so forth. The PITAC report calls for the federal we must start by understanding user need and
A meteorologist at the console of the IBM 7090 electronic computer in the Joint Numerical Weather Prediction
Unit, Suitland, Maryland, circa 1965. This computer was used to process weather data for short and long-range
forecasts, analyses, and reseach. Photo courtesy of the U.S. Weather Bureau.
context. This understanding includes recognizing especially in digital distributed environments in

users both as individuals and as part of their so- which people question assumptions about iden-
cial context. Who the users are, what they are try- tity and provenance. Similarly, economic, business,
ing to do, and how they interact with others are all and market frameworks will complicate digital li-
meaningful areas of discovery for future digital library use and development.
brary development. Social groupsbe they fami- For people interested in human-computer in-
lies, work groups, communities of practice, learning teraction, digital libraries offer a complex, wide-
communities, geographic communities, and so on spread environment for the research, development,
grow up with, create, structure, accept, and use in- and evaluation of technologies, tools, and ser-
formation and knowledge. Digital library content, vices aimed at improving the connection of people
tools, and services are needed to support these to d i g i t a l e nv i ro n m e n t s a n d to e a ch o t h e r
groups and the individuals whom they encompass. through those environments. Effective interac-
Issues of trust, reputation, belief, consistency, and tion between human and computer is essential
uncertainty of information will continue to prevail, for successful digital libraries.
Research everyday life. Such integration requires customized

Digital-library research has made significant progress and customizable user interfaces that encompass
in demonstrating our ability to produce digital dynamic user models (with knowledge of the history,
versions of traditional library collections and ser- needs, preferences, and foibles of the users and their
vices. However, what began as an effort to create dig- individual and social roles). The fourth direction is
ital libraries has been transformed into something the reduction of data to actionable information.
much more dynamic than was originally envisioned. This reduction requires developing capabilities to re-
The idea of curated, network-accessible repositories duce human effort and provide focused, relevant, and
was (and remains) a fundamental need of schol- useful information to the user; to do this again requires
arly inquiry and communication, as was the idea that an in-depth understanding of the users and their
these repositories should support information in individual and social contexts. The fifth research direc-
multiple formats, representations, and media. tion is to improve accessibility and productivity
However, not until people made serious efforts to through developments in information retrieval, im-
build such resources, particularly for nontextual age processing, artificial intelligence, and data mining.
digital content (audio, image, and video, for exam-
ple), did they realize that this venture would
stretch the limits of existing disciplinary boundaries Realizing the Potential
and require involvement of new interdisciplinary The PITAC report offers a vision for digital li-
collaborations. braries (universally accessible collections of hu-
The NSF recently sponsored a workshop of dig- man knowledge):
ital-library scholars and researchers to frame the long- All citizens anywhere anytime can use an Internet-
term research needed to realize a new scholarly inquiry connected digital device to search all of human
and communication infrastructure that is ubiquitous knowledge. Via the Internet, they can access knowl-
in scope and intuitive and transparent in operation. edge in digital collections created by traditional
Five major research directions were recommended. libraries, museums, archives, universities, govern-
ment agencies, specialized organizations, and even
The first direction is expansion of the range of digi- individuals around the world. These new libraries
tal content beyond traditional text and multimedia to offer digital versions of traditional library, museum,
encompass all types of recorded knowledge and ar- and archive holdings, including text, documents,
tifacts (data, software, models, fossils, buildings, sculp- video, sound and images. But they also provide
tures, etc.). This content expansion requires improved powerful new technological capabilities that enable
tools for identification, linkage, manipulation, and vi- users to refine their inquiries, analyze the results,
and change the form of the information to inter-
sualization. The second research direction is the use act with it, such as by turning statistical data into a
of context for retrieving information. Such context graph and comparing it with other graphs, creat-
has two dimensions: the relationships among digital ing animated maps of wind currents over time,
information objects and the relationship between these or exploring the shapes of molecules.
objects and users needs. In particular, because of con- Very-high-speed networks enable groups of digi-
tinually accumulating volumes of digital content, such tal library users to work collaboratively, commu-
nicate with each other about their findings, and
context needs to be automatically and dynamically use simulation environments, remote scientific in-
generated to the extent possible. However, automated struments, and streaming audio and video. No mat-
tools could be supplemented with contextual addi- ter where digital information resides physically,
tions from the using community (akin to reader in- sophisticated search software can find it and pres-
put to the Amazon.com website, for example). ent it to the user. In this vision, no classroom, group
Involvement of the using community in building the or person is ever isolated from the worlds great-
est knowledge resources. (PITAC 2001, 1)
knowledge environment will also build a sense of
ownership and stewardship relative to the particular Clearly underlying this vision is the notion of en-
content and services of interest. The third research di- gaged communities of both information providers
rection is the integration of information spaces into and information users. Sometimes called knowledge
communities, these communities, defined by a shared constrain the kinds of institutional developments
interest in knowing or wanting to know about a sub- that we envision.
ject area, are in constant flux. Understanding the dy-
namics of knowledge communities, why, when, and Jos-Marie Griffiths
how they form or cease to function, will be impor-
tant to the realization of the PITAC vision. Similarly, See also Information Organization; Information
researchers need to acknowledge the social con- Retrieval
struction of knowledge and the roles of various mem-
bers in the communities over time. Traditional
publishers, libraries, museums, archives, and other
information collection and distribution entities that FURTHER READING
are bound by the physicality of their collections Atkins, D. (1999). Visions for digital libraries. In P. Schauble & A. F.
and audiences can clearly be represented in virtual Smeaton (Eds.), Summary report of the series of joint NSF-EU work-
environments. However, the real power of the emerging groups on future directions for digital libraries research
ing technologies is their unleashing of human cre- (pp. 1114). Washington, DC: National Science Foundation.
Bishop, A. P., & Starr, S. L. (1996). Social informatics of digital library
ativity, connection, and collaboration in their creation, use and infrastructure. Annual Review of Information Science
discovery, and sharing of new knowledge. Developing and Technology (ARIST), 31, 301401.
technologies that are more human-centric in their Borgman, C. L., Bates, M. J., Cloonan, M. V., Efthimiadis, E. N.,
design and function is a critical element in achieving Gilliland-Swetland, A., Kafai, Y., Leazer, G. H., & Maddox, A. B.
(1996). Social aspects of digital libraries: Final report to the National
this future. Science Foundation. Los Angeles: Graduate School of Library &
Perhaps the greatest potential change that may Information Studies, UCLA. Retrieved January 26, 2004, from
result from digital libraries of the future will be in http://dlis.gseis.ucla.edu/DL/UCLA_DL_Report.html
Bush, V. (1945). As we may think. In J. Nyce & P. Kahn (Eds.), From
the institutional framework. When collection con- Memex to hypertext: Vannevar Bush and the minds machine
tent no longer needs to be physically colocated, (pp. 85110). San Diego, CA: Academic Press.
when service providers no longer need to be phys- Diderot, D., & le Rond D Alembert, J. (Eds.). (17581776). Encyclopedie
ically close to their intended user communities, and ou dictionnaire raisonne des sciences, des arts et des mtiers, par une
societe de gens de letteres (Encyclopedia or rational dictionary of
when the roles of provider and user blend, people sciences, arts, and the professions, by a society of people of letters)
will question the continued need for physical insti- (2nd ed). Luca, Italy: Andr Le Breton.
tutions and information-professional roles. Such a Fox, E. A., Gonalves, M. A., & Kipp, N. A. (2002). Digital libraries. In
future may well see librarians, museum profession- H. Adelsberger, B. Collis, & J. Pawlowski (Eds.), Handbook on in-
formation systems (pp. 623641). Berlin: Springer-Verlag.
als, and others working within knowledge commu- Fox, E. A., & Urs, S. R. (2002). Digital libraries. Annual Review of
nities, not just as providers to those communities. Information and Science and Technology (ARIST), 46, 503589.
As digital libraries and their contents are dis- Griffiths, J.-M. (1998). Why the Web is not a library. In B. Hawkins &
P. Battin (Eds.), The mirage of continuity: Reconfiguring academic
persed across the Internet, and as permanent avail- information resources for the twenty-first century (pp. 229246).
ability and access to those contents are assured, Washington, DC: Council on Library and Information Resources,
the need for individual institutions to own and Association of American Universities.
house collections and service access points (the Lesk, M. (1997). Practical digital libraries: Books, bytes and bucks.
San Francisco: Morgan Kaufmann.
means by which individuals can request and receive Lynch, C. A. (2002). Digital collections, digital libraries, and the dig-
service, i.e. an online catalog, a physical library, a itization of cultural heritage information. First Monday, 7(5).
reference desk, or an online help desk) will dimin- National Science Foundation. (2003, June). Report of the NSF work-
ish. For institutions whose reputations have shop on digital library research directions. Chatham, MA: Wave of
the Future: NSF Post Digital Library Futures Workshop.
grown with the growth and maintenance of their Nelson, T. H. (1974). Dream machines: New freedoms through com-
scholarly library collections, how will this future puter screensA minority report (p. 144). Chicago: Nelson.
play out? Although the opportunities are significant Presidents Information Technology Advisory Committee, Panel on
Digital Libraries. (2001). Digital libraries: Universal access to hu-
and the technological developments astounding, the man knowledge, report to the president. Arlington, VA: National
abilities of institutions to change at a similar pace Coordination Office for Information Technology Research and
are not clear. Issues of trust and control are likely to Development.
DRAWING AND DESIGN 187
Soergel, D. (2002). A framework for digital library research: Broadening els, but also with design evaluation and decision
the vision. D-Lib Magazine, 8(12). Retrieved January 26, 2004 from making. Despite the almost universal adoption of
http://www.dlib.org/dlib/december02/soergel/12soergel.html
Waters, D. J. (1998). The Digital Library Federation: Program agenda.
computer-aided design software, it is typically used
Washington, DC: Digital Libraries, Council of Library and in the laterdesign developmentphases of a de-
Information Resources. sign process, after many of the basic design decisions
Wells, H. G. (1938). World brain. Garden City, NY: Doubleday, Doran. have already been made. One reason for this, and a
primary motivation for supporting sketching, dia-
gramming, and drawing interfaces in computer-aided
design, is that during the conceptual phases many de-
DRAWING AND DESIGN signers prefer to work with pencil and paper.
The history of computers and human-computer
Ever since the Sketchpad system of computer graph- interaction shows a strong tendency to favor a
ics pioneer Ivan Sutherland, designers have dreamed problem-solving approach, and computer languages
of using drawing to interact with intelligent systems. have quite properly focused on requiring pro-
Built in the early 1960s, Sketchpad anticipated mod- grammers to state problems precisely and definitely.
ern interactive graphics: The designer employed a This has, in turn, colored a great deal of our soft-
light pen to make and edit a drawing and defined its ware, including computer-aided design, which de-
behavior by applying geometric constraints such mands of its users that they be able to precisely
as parallel, perpendicular, and tangent lines. However, articulate what they are doing at all times. Yet de-
the widespread adoption of the windows-mouse signing in particular, and drawings more generally,
interface paradigm on personal computers in the seem at least sometimes ill-suited to this historical
1980s relegated pen-based interaction to a special- paradigm. Although the goal of designing is to ar-
ized domain, and for many years little research was rive at definite design decisions that make it possi-
done on computational support for freehand draw- ble to construct an artifact, during the process
ing. The re-emergence of stylus input and flat dis- designers are often quite willing to entertain (or tol-
play output hardware in the 1990s renewed erate) a great deal of uncertainty. This makes build-
interest in pen-based interfaces. Commercial soft- ing human-computer interfaces for computer-aided
ware has mostly focused on text interaction (em- design an interesting challenge, and one that may
ploying either a stylized alphabet or full-fledged ultimately demand new forms of computational
handwriting recognition), but human-computer representations.
interfaces for computer-aided design must also sup- The development of freehand interfaces for com-
port sketching, drawing, and diagramming. puter-aided design will certainly depend on tech-
Computer-aided design (CAD) is widely used in nical advances in pen-based interaction. However,
every design discipline. CAD software supports mak- successful drawing-based interfaces for design will
ing and editing drawings and three-dimensional ultimately also be informed by research on design
computer graphics models, and in most design firms, processes (how designing works and how people do
computer-aided design applications have replaced design) as well as by the efforts of cognitive psy-
the old-fashioned drawing boards and parallel rules. chologists to understand the role of drawing and vi-
Digital representations make it easier for a design sual representations in thinking. Along with the
team to share and edit drawings and to generate com- development of freehand-drawing software systems,
puter graphics renderings and animated views of a research on design and visual cognition has recently
design. The predominant use of computers in de- enjoyed a resurgence of interest. In addition to
sign is simply to make and edit drawings and mod- human-computer interaction, relevant work is be-
els, leaving it to human designers to view, evaluate, ing done in design research, artificial intelligence,
and make design decisions. However, computational and cognitive science. An increasing number of con-
design assistants are being increasingly brought in ferences, workshops, and journals are publishing
to help not only with creating drawings and mod- work in this growing research area.
Drawing as an Interface to Everything develop it through a series of sketches, ultimately

People sketch, diagram, and draw to compose, producing a precise and detailed drawing. Both di-
consider, and communicate ideas. Information about agram and sketch are typically preliminary repre-
the ideas is coded in the lines and other drawing sentations used in early design thinking to capture
marks and the spatial relationships among them. the essence of an idea or to rapidly explore a range of
Some people may think of drawings as belonging to possibilities. A diagram employs shapes and spatial
the realm of aesthetics, or as ancillary representa- relations to convey essentials concisely. A sketch,
tions to real thinking carried out in words or math- however, is often more suggestive than definitive, and
ematics, but many disciplinesfrom logic to sports, it may convey details while simultaneously avoiding
from physics to music, from mathematics and bi- specificity. A schematic drawing involves more detail
ology to designemploy drawings, sketches, and di- and complexity than a diagram and is usually intended
agrams to represent ideas and reason about them. as a more precise and definitive representation.
Many scientific and engineering disciplines use well- Features of a drawing that are potentially rele-
defined formal diagram languages such as molecu- vant include shape and geometry, topology, curva-
lar structure diagrams, analog and digital circuit ture and points of inflection of lines, absolute and
diagrams, or user-modeling language (UML) dia- relative dimensions, positions of drawing marks and
grams in human-computer interactions. Drawing is spatial relationships among them, line weights, thick-
seldom the sole vehicle for communicating infor- ness and color, speed and sequence of execution, and
mation, but in many domains it is either a primary relationships with nearby text labels. In any partic-
representation or an important auxiliary one, as a ular drawing only some of these features may be rel-
look at whiteboards in any school or company will evant. For example, a diagram of digital logic is largely
confirm. indifferent to geometry but the drawing topology
Drawing plays a special role in design. In phys- (connections among the components) is essential to
ical domains such as mechanics, structural engi- its meaning. On the other hand, in a schematic draw-
neering, and architecture, and in graphic (and ing of a mechanism or a sketch map, geometry and
graphical user interface) designs, a design drawing scale are essential.
correlates directly with the ultimate artifact: The In addition to the information that a designer
drawings geometry corresponds directly to the geom- decides deliberately to communicate, a drawing also
etry of the artifact being designed. For example, a conveys information about the designers intent, that
circle represents a wheel. In other disciplines, a di- is, metainformation about the designing process.
agram may correlate symbolically, for example a For example, the speed with which a sketch is exe-
supply-demand curve in economics. Even in domains cuted, the extent to which a designer overtraces draw-
where a graphic representation only abstractly rep- ing marks, the pressure of the pen, or the darkness
resents the artifact being designed, drawing supports of the ink all offer important information. Intense
the supposing, proposing, and disposing process of overtracing in one area of the drawing may indicate
design decision making. For these reasons, drawing that the designer is especially concerned with that
can be an interaction modality to a wide range of part of the design, or it may reveal that the draw-
computational processes and applicationsdrawing ing represents several alternative design decisions.
as an interface to everything. A quickly made sketch may reflect broad, high-level
thinking, whereas a slow drawing may reveal a
high degree of design deliberation.
Whats in a Drawing? During design brainstorming it is common to
Drawings range from conceptual diagrams to rough find several sketches and diagrams on the same sheet
sketches to precisely detailed drawings. The purposes of paper or whiteboard; they may be refinements of
of these representations differ, although designers a single design, alternative designs, designs for dif-
may employ them all in the course of designing: ferent components of the artifact, or even represen-
Beginning with a conceptual diagram of an idea, they tations of entirely unrelated ideas.
Input Issues users original drawing, even if the system has rec-
Two different approachesink based and stroke ognized sketched components and could replace
basedto building pen-based interaction systems them with precise visual representations. Many de-
are currently being followed, and each has certain ad- signers consider the imprecise, rough, and sugges-
vantages. An ink-based system registers the draw- tive nature of a sketch or diagram to be of great value
ing marks the user makes in an array of pixels and therefore prefer a hand-drawn sketch to a re-
captured by a video camera or scanner, which serves fined, geometrically precise beautified drawing.
as input for an image-processing system to parse and On the other hand, some users strongly prefer to
interpret. A stroke-based system records the mo- work with perfectly straight lines and exact right an-
tion of the users pen, usually as a sequence of x,y (and gles rather than crude-looking sketches. This de-
sometimes pressure and tilt) coordinates. To an ink- pends at least in part on the users own experience
based system any drawing previously made on paper with drawing: Novices are more likely to feel un-
can serve as scanned input, whereas one that is stroke- comfortable with their sketching ability and prefer
based must capture input as it is produced. This makes to work with beautified drawings, whereas seasoned
dynamic drawing information such as velocity, pen designers tend to see the nuances of their hand-drawn
pressure, and timing available to stroke-based sys- sketches as positive characteristics. Whether beauti-
tems. Many stroke-based systems, for example, use fication is considered helpful or harmful also de-
timing information to segment drawing input into pends in part on the drawings intended purpose.
distinct drawing elements, or glyphs.
Designers traditionally distinguish between free-
hand and hard-line drawings. Freehand drawings Recognition Issues
are typically made with only a stylus, whereas hard- A great deal of research in interactive drawing
line drawings are made using a structured interface, aims at recognizing sketches, diagrams, and draw-
previously a triangle and parallel rule, today the ings for semantic information processing by intelli-
menus and tool palettes of a conventional computer- gent systems that apply domain knowledge to reason
aided design program. The structured interface about designs. After the system extracts from the
has certain advantages: In selecting drawing elements drawing the semantics of a proposed design, then
from a tool palette the designer also identifies them, various knowledge-based design aids, such as sim-
eliminating the need for the low-level recognition of ulation programs, expert systems, and case-based
drawing marks that freehand drawing systems typ- reasoning tools, and other automated advisors can
ically require. While this helps the computer probe brought to bear. An interface that recognizes and
gram to manage its representation of the design, interprets the design semantics of sketches and di-
many designers feel that the structured interface im- agrams enables a designer to employ these programs
poses an unacceptable cognitive load and requires a in the early phases of designing. For example, a pro-
greater degree of commitment and precision than is gram that recognizes the components and connec-
appropriate, especially during the early phases of de- tions of a mechanical diagram can construct and
signing. Designers also complain that menus and execute a computer simulation of the mechanism. A
tool systems get in the way of their design flow. A program that recognizes the layout of an architec-
freehand drawing conveys more subtle nuances of tural floor plan can retrieve from a database other
line and shape than a hard-line drawing. Freehand similar or analogous floor plans. A program that rec-
drawings are often less formal and precise and more ognizes a sketched layout of a graphical user inter-
ambiguous than hard-line representations, all ar- face can generate code to construct that interface.
guably advantageous characteristics in the early A variety of recognition approaches have been
phases of design thinking. explored, including visual-language parsing and
Some computer-based drawing systems auto- statistical methods. Parsing approaches consider a
matically replace hand-drawn sketchy shapes and drawing as an expression in a visual language com-
lines with beautified ones. Other systems retain the posed of glyphs (simple drawing marks such as
arrows, circles, and rectangles) arranged in various Designers often sketch during the early stages of
spatial relations into configurations. Typically a low- design thinking, and therefore a sketch may serve the
level recognizer first identifies the glyphs. Some sys- dual purpose of (1) recording what the designer
tems restrict glyphs to a single stroke, requiring, already has decided and (2) exploring possible
for example, that a box be drawn without lifting the alternatives. Sketches in general will vary along the
pen; others allow multiple-stroke glyphs, allowing dimensions of ambiguity and precision, and even
the box to be drawn as four distinct strokes. After the within a single sketch some parts may record defi-
glyph recognizer has identified the basic drawing el- nite and precise design decisions while other parts
ements, the parser identifies legal visual expressions are vague, amorphous, and imprecise, representing
by matching the drawing against grammar rules. work-in-progress exploration. Recognition-based
Smaller visual unitsinitially glyphs, then configu- drawing systems must be able to deal with these
rations arranged in specific spatial relationsmake extremes as well as with the range of representations
up more complex visual expressions. Each design in between, and they must also be able to determine
domain has its own visual language, so parsing ap- autonomouslyfrom the drawing itselfwhat de-
proaches to general-purpose sketch recognition must gree of ambiguity and imprecision the designer in-
either be told which visual language to use or must tended to convey. For example, a recognition-based
determine this information from the context. system might be able to distinguish between its own
Statistical methods such as Bayesian networks and failure to recognize precise input and a drawing that
hidden Markov models have proved successful in other is deliberately indeterminate. The ability of a system
kinds of recognition, notably speech recognition and to sustain ambiguous and imprecise representations
natural-language understanding. Statistical techniques is for this reason especially important, and this
make it possible to build visual-language recognizers may pertain not only to the interface-recognition
without having to manually construct a grammar for algorithms, but also to any back-end processes behind
each domain-specific language. the interface that later represent or reason about
Against sketch recognition the argument is lev- the designs.
eled that people are highly sensitive to recognizer A recognizer can support imprecision and ambi-
failure and will not tolerate imperfect recognizer per- guity in several ways. Recognition-based interfaces
formance. Experience (for instance, with speech-to- can catch, resolve, or mediate potential errors and am-
text systems and early handwriting recognizers) biguities at input time, for example, by presenting the
shows that users become quite frustrated unless user with a sorted list of alternative interpretations.
recognition is extremely reliable, that is, has accu- Visual-language interpreters can employ fuzzy-logic
racy rates above 99 percent. On the other hand, un- techniques, representing match probabilities in the
like speech and character recognitionwhere it can parse, or they may allow the parse to carry multiple
be assumed that the input has only one intended in- alternative interpretations. Rather than requiring an
terpretationuncertainty in various forms may be entire drawing to represent a single visual sentence, a
more acceptable in drawing, especially when a de- recognizer may take a bottom-up approach that iden-
signer wants to preserve ambiguity. Then for tifies some parts of the drawing while allowing others
sketch recognition, the methods of sustaining am- to remain uninterpreted.
biguity and vagueness would be at least as impor-
tant as accuracy.
An intermediate approach to recognition asks Avoiding Recognition:
the user to label the elements of a sketch rather than
attempt low-level glyph recognition. In this hybrid Annotation and Multimodal Systems
approach the user enters a freehand drawing; then Another response to the problem of recognition is
after the user has labeled the elements (choosing to avoid it entirely and simply manage drawings as
from a palette of symbols) the system can reason design representations independent of their seman-
about the drawings spatial organization. tic content. This approach is taken in systems that
treat drawings as components of a collection of mul- efforts include SKETCH!, Teddy, Chateau, SketchVR,
timodal conversations. Despite a popular myth of and Stilton.
the lone creative designer, real-world design typi- Despite its name, the SKETCH! program does
cally involves a team of participants that includes ex- not interpret line drawings; rather, the designer con-
perts from a variety of design disciplines as well as trols a 3D modeler by drawing multistroke gestures,
other stakeholders, and a process that can range in for example, three lines to indicate a corner of a rec-
duration from weeks to years. The record of the tangular solid. Teddy enables a user to generate
designing process (the design history) can therefore volumes with curved surfaces (such as Teddy bears)
include successive and alternative versions over time by inflating 2D curve drawings. It uses simple
and the comments of diverse participants, along with heuristics to generate a plausible model from a sketch.
suggestions, revisions, discussions, and arguments. Chateau is a suggestive interface: It offers alter-
Sketches, diagrams, and drawings are important native 3D completions of a 2D sketch as the user
elements in the record of this design history. draws, asking in effect,Do you mean this? Or this?
Design drawings are inevitably expressions in a SketchVR generates three-dimensional models from
larger context of communication that includes spo- 2D sketches by extrusion. It identifies symbols and
ken or written information, photographs and video, configurations in the drawing in the 3D scene and
and perhaps computational expressions such as equa- replaces them with modeling elements chosen from
tions or decision trees. This gives rise to a wide range a library. In Stilton, the user draws on top of the dis-
of multimodalities. For example, a designer may (a) play of a 3D scene; the program uses heuristics about
mark up or redline a drawing, photograph, 3D likely projection angles to interpret the sketch.
model, or video to identify problems or propose
changes, or add text notes for similar reasons; (b) in-
sert a drawing to illustrate an equation or descrip- The Future
tive text or code; (c) annotate a drawing with spoken Much of the personal computer era has been domi-
comments, recording an audio (or video) track of nated by interfaces that depend on text or on inter-
a collaborative design conversation as the drawing acting with mouse-window-menu systems. A renewed
is made or attaching audio annotations to the draw- interest in sketch-based interaction has led to a new
ing subsequently. Associated text and audio/video generation of systems that manage and interpret hand-
components of the design record can then be used drawn input. Today, human-computer interaction re-
in conjunction with the drawing; for example, text search is enabling computer-aided design software to
can be indexed and used to identify the role, func- take advantage of sketching, drawing, and diagram-
tion, or intentions of the accompanying drawings. ming, which have long been essential representations
in design, as well as in many other activities. Progress
in freehand-drawing interaction research will go hand
From Sketch to 3D in hand with research in design processes and cog-
Designers in physical domains such as mechanical, nitive studies of visual and diagrammatic reasoning.
product, and industrial engineering and architecture
often sketch isometric and perspective drawings to Mark D. Gross
describe three-dimensional artifacts. Therefore,
sketch-recognition research has long sought to build See also Evolutionary Engineering; Pen and Stylus Input
systems that can generate three-dimensional mod-
els from two-dimensional sketches. Although this
goal has not yet been achieved in the general case of
arbitrary 2D sketches, a variety of approaches have FURTHER READING
been pursued, each with particular strengths and Davis, R. (2002). Sketch understanding in design: Overview of work
limitations, and each supporting specific kinds of at the MIT AI lab. In R. Davis, J. Landay & T. F. Stahovich (Eds.),
sketch-to-3D constructions. Recent representative Sketch understanding: Papers from the 2002 AAAI Symposium
(pp. 2431). Menlo Park, CA: American Association for interfaces. In Proceedings of the Human Factors in Computing
Artificial Intelligence (AAAI). (SIGCHI) Conference (pp. 368375). The Hague, Netherlands:
Do, E. Y.-L. (2002). Drawing marks, acts, and reacts: Toward a com- ACM Press.
putational sketching interface for architectural design. AIEDAM Negroponte, N. (1973). Recent advances in sketch recognition. In
(Artificial Intelligence for Engineering Design, Analysis and AFIPS (American Federation of Information Processing) National
Manufacturing), 16(3), 149171. Computer Conference, 42, 663675. Boston: American Federa-
Forbus, K., Usher, J., & Chapman, V. (2003). Sketching for military tion of Information Processing.
courses of action diagrams. In International Conference on Intelligent Oviatt, S., & Cohen, P. (2000). Multimodal interfaces that process what
User Interfaces (pp. 6168). San Francisco: ACM Press. comes naturally. Communications of the ACM, 43(3), 4553.
Goel, V. (1995). Sketches of thought. Cambridge MA: MIT Press. Pinto-Albuquerque, M., Fonseca, M. J., & Jorge, J. A. (2000). Visual
Gross, M. D., & Do, E. Y.-L. (2000). Drawing on the back of an enve- languages for sketching documents. In Proceedings, 2000 IEEE
lope: A framework for interacting with application programs by International Symposium on Visual Languages (pp. 225232). Seattle,
freehand drawing. Computers and Graphics, 24(6), 835849. WA: IEEE Press.
Igarashi, T., & Hughes, J. F. (2001). A suggestive interface for 3-D draw- Saund, E., & Moran, T. P. (1994). A perceptually supported sketch ed-
ing. In Proceedings of the ACM Symposium on User Interface Software itor. Paper presented at the ACM Symposium on User Interface
and Technology (UIST) (pp. 173181). New York: ACM Press. Software and Technology, Marina del Rey, CA.
Igarashi, T., Matsuoka, S., & Tanaka, H. (1999). Teddy: A sketching in- Sutherland, I. (1963). Sketchpad: A man-machine graphical com-
terface for 3-D freeform design. In Proceedings of the SIGGRAPH munication system. In Proceedings of the 1963 Spring Joint Computer
1999 Annual Conference on Computer Graphics (pp. 409416). New Conference (pp. 329346). Baltimore: Spartan Books.
York: ACM Press/Addison-Wesley Publishing Co. Suwa, M., & Tversky, B. (1997). What architects and students perceive
Kurtoglu, T., & Stahovich, T. F. (2002). Interpreting schematic sketches in their sketches: A protocol analysis. Design Studies, 18,
using physical reasoning. In R. Davis, J. Landay, & T. Stahovich. 385403.
(Eds.), AAAI Spring Symposium on Sketch Understanding Turner, A., Chapman, D., & Penn, A. (2000). Sketching space. Computers
(pp. 7885). Menlo Park, CA: AAAI Press. and Graphics, 24, 869876.
Landay, J. A., & Myers, B. A. (1995). Interactive sketching for the early Ullman, D., Wood, S., & Craig, D. (1990). The importance of draw-
stages of interface design. In CHI 95Human Factors in Computing ing in the mechanical design process. Computers and Graphics,
Systems (pp. 4350). Denver, CO: ACM Press. 14(2), 263274.
Larkin, J., & Simon, H. (1987). Why a diagram is (sometimes) Zeleznik, R., Herndon, K. P., & Hughes, J. F. (1996). SKETCH: An
worth 10,000 words. Cognitive Science, 11, 6599. interface for sketching 3-D scenes. In SIGGraph 96 Conference
Mankoff, J., Hudson, S. E., & Abowd, G. D. (2000). Providing inte- Proceedings (pp. 163170). New York: ACM Press.
grated toolkit-level support for ambiguity in recognition-based
E-BUSINESS
EDUCATION IN HCI
ELECTRONIC JOURNALS
ELECTRONIC PAPER TECHNOLOGY
ELIZA
E
E-MAIL
EMBEDDED SYSTEMS
ENIAC
ERGONOMICS
ERRORS IN INTERACTIVE BEHAVIOR
ETHICS
ETHNOGRAPHY
EVOLUTIONARY ENGINEERING
EXPERT SYSTEMS
EYE TRACKING
The Technological Revolution

E-BUSINESS A new landscape for conducting e-business has arisen
with the proliferation of technologies that facilitate
Although business challenges such as time and space e-business, such as information communication tech-
now matter less, new challenges arise when people nologies (ICT) (any communication device or ap-
conduct electronic business (e-business). These chal- plication encompassing radio, television, cellular
lenges arise from two fundamental sources: global phones, satellite systems, etc.); enterprise resource
customers cultural values and culturally sensitive planning (ERP) (any software system designed to
technology applications. Important cultural ques- support and automates the business processes of
tions affect e-business and global customers. Such medium and large businesses); electronic data inter-
questions include (1) why is culture important to change (EDI) (an information system or process
consider when conducting e-business? and (2) integrating all manufacturing and related applica-
how do companies leverage their information tech- tions for an entire enterprise; and manufacturing re-
nology (IT) applications in light of cultural differ- source planning (MRP) (a system for effectively
ences exhibited by global customers? Answering these managing material requirements in a manufacturing
questions can help companies that use IT in a multi- process). Agility, flexibility, speed, and change are the
cultural market. conditions for developing e-business models. In
193
addition, by using information and telecommunica- tive (Davis and Benamati 2003, 8). In essence, e-busi-
tion systems, companies are able to communicate ness means any Internet or network-enabled busi-
with their global customers where barriers such as ness, for example companies can buy parts and
time zones, currencies, languages, and legal systems supplies from each other, collaborate on sales pro-
are reduced or eliminated. As a result, global cus- motion, and conduct joint research.
tomers can be reached anywhere and at anytime. On the other hand, e-commerce is a way of do-
Services and products can be obtained whenever, ing business using purely the Internet as a means,
wherever, and by whomever. The digital economy whether the business occurs between two partners
is blazing a new path for doing business where the (business to businessB2B), between a business and
notion of value through people becomes the its customers (business to customersB2C), be-
driving force for a successful model of e-business. tween customers (C2C), between a business and em-
With the advent of the World Wide Web, busi- ployees (B2E), or between a business and government
ness is increasingly becoming an online environ- (B2G).
ment. Traditional brick-and-mortar businesses have According to Effy Oz (2002), an expert in infor-
evolved into click-and-mortar businesses. mation technology and ethics, there are three cate-
Additionally, the Internet has changed from a gories of organizations that want to incorporate the
communications tool used mostly by scientists to a Web into their e-business: (1) organizations that have
business tool used by companies to reach millions a passive presence online and focus on online ad-
of customers across the globe. As a result, the Internet vertising, (2) organizations that use the Web to
has become a powerful business resource because its improve operations, and (3) organizations that cre-
technology enables firms to conduct business ate stand-alone transaction sites as their main or only
globally (Simeon 1999). In addition, online sales eas- business.
ily penetrate global markets. Some companies treat In contrast, e-commerce is not exclusively about
Web customers as a new type of audienceso united buying and selling. Although the ultimate goal of
in their use of the Internet that national differ- business is profit generation, e-commerce is not ex-
ences no longer apply. Other companies, such as IBM, clusively about buying and selling. Instead, the real
Microsoft, and Xerox, have developed local versions goal of e-commerce is to improve efficiency by the
of their websites. These versions run off regional deployment of technologies. Factors that influence
servers, address technical issues (such as the need to the development of e-commerce are a competitive
display different character sets), and provide infor- environment, strategic commitment of the company,
mation about local services and products. Occasion- and the required competencies. Thus, the definition
ally they reflect aesthetic differencessuch as cultural of e-commerce has a more restricted application than
biases for or against certain colorsbut few com- that of e-business.
panies actively consider cultural variations that might
enhance the delivery of their products.
Understanding Cultural Concepts
Explained below are three different categories of cul-
What Are E-business ture. The first category is national culture in which
the differences of the cultural values are based on
and E-commerce? four key dimensions. First is the individualism-
The terms e-business and e-commerce have slightly collectivism dimension, which denotes a cultures
different meanings. E-business is . . . a broader term level of freedom and independence of individuals.
that encompasses electronically buying, selling, serv- Second is the power-distance dimension, which
icing customers, and interacting with business part- denotes the levels of inequality expected and accepted
ners and intermediaries over the Internet. Some by people in their jobs and lives. Third is the un-
exper ts see e-business as the objective and certainty-avoidance dimension, which denotes how
e-commerce as the means of achieving that objec- societies deal with the unknown aspects of a dif-
E-BUSINESS 195
ferent environment and how much people are of culture when developing and implementing IT
willing to accept risks. Fourth is the masculinity- applications. Companies have difficulty in under-
femininity dimension, which denotes a cultures rank- standing or even recognizing cultural factors at a
ing of values such as being dominant, assertive, tough, deeper level because the factors are complex and
and focused on material success. subtle. Companies understanding of cultural fac-
The second category of culture is related to or- tors is normally only superficial, which is why peo-
ganizational culture. According to Edgar J. Schein, ple have difficulty observing the magnitude of the
an organizational psychologist, organizational cul- impact of such factors on the success or failure of
ture is a property of a group. It arises at the level e-business companies.
of department, functional groups, and other orga- Although people have conducted an increasing
nizational units that have a common occupational amount of research in global IT, this research has
core and common experience. It also exists at been primarily limited to descriptive cross-cultural
every hierarchical level of the organizations and at studies where comparison analyses were made be-
the level of the whole organization (Schein 1999, tween technologies in different national cultures. A
1314). Thus, intense organizational culture can re- universal interface should not be mistakenly con-
sult in manifestations such as the phrases the way sidered as one interface for all customers. The con-
we do things around here, the rites and rituals of cept of universalism is somewhat misleading in
our company, our company climate, our com- this context. The most important goal is to ensure
mon practices and norms, and our core values. that customers feel at home when exploring the
A third category of culture can be called in- Internet.
formation technology culture. Information tech- Fundamentally, cultural factors have strong in-
nolog y culture often overlaps national and fluences on global customers preferences. Each cus-
organizational cultures. Indeed, IT culture is part of tomer has his or her own culturally rooted values,
the organizational culture, which determines whether beliefs, perceptions, and attitudes. When loyal cus-
the user (i.e., customer) accepts or resists the tech- tomers are satisfied with the way they have been buy-
nology to be used. IT culture can be defined as the ing goods and services, they resist changes. Making
sets of values and practices shared by those mem- purchases online is less desirable to many customers.
bers of an organization who are involved in IT-re- The fact that customers cannot touch or smell the
lated activities, such as information system products that they want makes some resistant to e-
professionals, and managers who are involved in IT- business. Customers also can be resistant because
related activities (i.e., programming, system analy- they lack the skills to use new technologies and an
sis and design, and database management). understanding of how e-business is conducted.
Different ethnic cultures demonstrate different
cognitive reactions, requiring different environmental
Global Customers: Challenges stimuli (Tannen 1998). Similarly, Web-marketing
psychology depends on different mixtures of cogni-
of Cultural Differences tive and behavioral elements (Foxall 1997). Language,
IT applications in the context of e-business have values, and infrastructure can also be barriers to e-
become more important because today companies business. For example, the preference of many
of all sizes and in all sectors are adopting the Chinese people for a cash-based payment system or
principles of cultural diversity, as opposed to cul- cash-on-delivery is the main obstacle to con-
tural convergence, when reaching out to global cus- ducting e-business in China. The phenomenon
tomers. Some questions that are worth considering can be explained by factors such as a lack of real credit
are why global customers resist new IT implemen- cards, a lack of centralized settlement systems (the
tation, how organizational culture affects new cus- ability for credit cards to be used anywhere), and a
tomers attitudes toward new IT implementation, lack of trust in conducting business via the Internet
and why many companies fail to consider the role (Bin, Chen, and Sun 2003).
When people of Malaysia and Australia were to face. Customers can negotiate better with a
asked to evaluate eight websites from their countries, seller because such a setting allows the reciprocal
the findings confirmed that the subjects had no pref- communication of interests and intentions. After
erence for one-half the websites but had a preference seller and customers establish rapport, this rapport
associated with subject nationality for the other one- creates a more trusting relationship. This trusting
half (Fink and Laupase 2000). A study of mobile- relationship could lead to repeated transactions.
phone use in Germany and China showed that people Trusting the seller or buyer is crucial in certain
accepted support information and rated it as more cultures.
effective when it was culturally localized. Similar Companies that conduct e-business need to find
studies have shown that cultural factors influence new mechanisms and strategies that overcome such
how long customers stay on the Internet, how likely cultural differences. In a situation such as the
they are to buy products online when the content study of Chinese customers, where credit cards were
is presented in their native language, and the usability not the common system of payment, a pragmatic
of Web design elements. strategy might be to buy online and pay offline. The
Researchers believe that culturally specific ele- findings in the research of customer interfaces for
ments increase international participation in con- World Wide Web transactions indicate that there are
ducting e-business more than do genre-specific significant cultural variations in why people used the
elements. Interestingly, the role of culture in user in- Internet (OKeefe et al. 2000). Their U.S. subjects
terface design can be identified as the localization el- used the Internet solely to search for information,
ements that could be considered as cultural markers. whereas their Hong Kong subjects used the Internet
These cultural markers are influenced by a specific to communicate socially. A wise e-business strat-
culture or specific genre (Barber and Badre 1998). egy for a company is thus to enhance personal com-
Examples of cultural markers are interface design el- petence for Western values and to seek out social
ements that reflect national symbols, colors, or forms relationships and shared loyalty for Eastern values.
of spatial organization. After reviewing hundreds of Another e-business strategy emphasizes lever-
websites from different countries and in different aging technology. Electronic businesses have two op-
languages, Barber and Badre posited that different tions when designing websites for customers in
cultural groups prefer different types of cultural different countriesdesign one website for all or
markers. localized websites for each country. If the audience
crosses national borders, a single website may be ap-
propriate. For instance, websites exist for Arctic re-
E-business Strategies searchers and astronomers. However, this strategy is
Because global customers come from all over the less likely to be successful when no overriding pro-
world, their demands, needs, and values are more di- fessional or occupational focus unifies the audience.
vergent than similar. Cultural context and cultural The alternative is for companies to develop local ver-
distance may have an impact on how goods and serv- sions of their websites. These local versions may be
ices can be delivered to themthat is, on marketing run off regional servers to enhance performance
channels and logistics. Hence, e-business companies or to display different character sets. They also can
must fully understand the values that affect cus- emphasize different product lines. Unfortunately,
tomers preferences. Companies need to tailor unless the company is highly decentralized, varia-
their products to customers electronic requirements. tions in the basic message or mode of presentation
Selling products electronically means that that might enhance delivery of its products to peo-
businesses must consider international channels of ple in another culture are rarely seen.
distribution that fit with customers values. The elec- Melissa Cole and Robert OKeefe (2000) believe
tronic environment can become a barrier to suc- that Amazon.com and Autobytel.com (an auto sales
cessful business endeavors. For example, in some website) have transcended global differences by em-
cultures a business transaction is best conducted face ploying a standardized transaction-oriented inter-
E-BUSINESS 197
face. Such an interface may be practical for people titudes, tastes, selection, and participation in e-busi-
who have a limited goal (such as deciding which book ness. Hence, companies need to fully understand cul-
to buy) but may not be practical for people who tural variances in order to make decisions on which
do not. Because different audiences use the Internet e-business strategies work best. Some basic questions
for different purposes, standardized features may not for future research would be: (1) What makes for
be practical for all the nuances of cultural values. universally appealing IT practices? (2) Does accept-
Designing interfaces for people who are searching ability or familiarity drive global IT use? (3) How
for social relationships, rather than seeking infor- does one successfully introduce technology applica-
mation, imposes different requirements on Web tions that are unusual or not appropriate in a coun-
retailers and designers. try? (4) How can cultural differences be considered
Culture has significant impacts on global cus- in the planning of IT practices? In a nutshell, com-
tomers and software designers. The merging con- panies and Web designers need to be sensitive to the
cepts of culture and usability have been termed different needs of global customers and to build
cultural user interfaces by Alvin Yeo (1996) and strategies and interfaces that consider cultural as-
culturability by Barber and Badre (1998). Yeo talks sumptions and characteristics. Taking advantage of
about cultures effect on overt and covert elements national differences and preferences provides re-
of interface design. Tangible, observable elements source-based competencies and competitive advan-
such as character sets and calendars are overt and tages for e-businesses. Companies need a more
easy to change, whereas metaphors, colors, and icons innovative e-business model. With new e-business
may reflect covert symbols or taboos and be difficult practices, success is centered on peoples values,
to recognize and manipulate. Barber and Badre assert agility, speed, flexibility, and change. Hence, the com-
that what is user friendly to one nation or culture mon business phrase Think globally, act locally
may suggest different meanings and understandings may not be as practical as Think locally, act glob-
to another. Therefore, efforts to build a generic global ally. Reaching out to global customers means re-
interface may not be successful. Instead, cultural flecting their local cultures, language, and currency.
markers should be programmatically changed to fa-
cilitate international interactions. Norhayati Zakaria
See also Anthropology and HCI; Ethnography;

Implications Website Design
Because of globalization and competitiveness in in-
ternational business, many multinational and local
companies have considered implementing tech- FURTHER READING
nologies to conduct e-business. Among the primary Barber, W., & Badre, A. (1998). Culturability: The merging of culture
factors that companies must consider are the effect and usability. Human Factors and the Web. Retrieved March 1, 2004,
of culture on customers technology acceptance and from http://www.research.att.com/conf/hfweb/proceedings/barber/
customers cultural values. Companies should ad- index.htm
Bin, Q., Chen, S., & Sun, S. (2003). Cultural differences in e-commerce:
dress the appropriateness of management policies A comparison between the U.S. and China. Journal of Global
and practices across countries. For example, man- Information Management, 11(2), 4856.
agers need to make decisions concerning certain sit- Cole, M., & OKeefe, R. M. (2000). Conceptualizing the dynamics of
uations such as whether a global company can globalization and culture in electronic commerce. Journal of Global
Information Technology Management, 3(1), 417.
override national cultural differences and when lo- Cooper, R. B. (1994). The inertia impact of culture on IT imple-
cal policies are best. mentation. Information and Management, 17(1), 1731.
IT provides vast opportunities for companies to Davis, W. S., & Benamati, J. (2003). E-commerce basics: Technology
foundations and e-business applications. New York: Addison-Wesley.
compete in the global and electronic arena. At the Fink, D., & Laupase, R. (2000). Perceptions of web site design char-
same time, customers from different cultures can acteristics: A Malaysian/Australian comparison. Internet Research,
differ significantly in their perceptions, beliefs, at- 10(1), 4455.
Foxall, G. R. (1997). Marketing psychology: The paradigm in the wings.

London: Macmillan.
Hofstede, G. (1980). Culture's consequences: International differences
EDUCATION IN HCI
in work-related values. Beverly Hills, CA: Sage.
Honald, P. (1999). Learning how to use a cellular phone: Comparison Education in human-computer interaction (HCI)
between German and Chinese users. Technical Communication: teaches students about the development and use of
Journal of the Society for Technical Communication, 46(2), 196205. interactive computerized systems. Development
Janssens, M., Brett, J. M., & Smith, F. J. (1995). Confirmatory cross-
cultural research: Testing the viability of a corporate-wide safety involves analysis, design, implementation, and
policy. Academy of Management Journal, 38, 364382. evaluation, while use emphasizes the interplay be-
Johnston, K, & Johal, P. (1999). The Internet as a virtual cultural tween the human users and the computerized sys-
region: Are extant cultural classifications schemes appropriate?
Internet Research: Electronic Networking Applications and Policy,
tems. The basic aim of instruction in HCI is that
9(3), 178186. students learn to develop systems that support users
Kowtha, N. R., & Choon, W. P. (2001). Determinants of website de- in their activities.
velopment: A study of electronic commerce in Singapore. Education in HCI is primarily conducted in two
Information & Management, 39(3), 227242.
O'Keefe, R., Cole, M., Chau, P., Massey, A., Montoya-Weiss, M., &
contexts: academia and industry. HCI is an impor-
Perry, M. (2000). From the user interface to the consumer inter- tant element in such diverse disciplines as computer
face: Results from a global experiment. International Journal of science, information systems, psychology, arts, and
Human Computer Studies, 53(4), 611628. design. Key elements of HCI are also taught as in-
Oz, E. (2002). Foundations of e-commerce. Upper Saddle River, NJ:
Pearson Education. dustry courses, usually with more focus on the de-
Rosenbloom, B., & Larsen, T. (2003). Communication in international sign and development of interactive systems.
business-to-business marketing channels: Does culture matter?
Industrial Marketing Management, 32(4), 309317.
Ryan, A. M., McFarland, L., Baron, H., & Page, R. (1999). An inter-
national look at selection practices: Nation and culture as expla- Development of HCI as a Discipline
nations for variability in practice. Personnel Psychology, 52, 359391. The first education programs in computer science
Sanders, M. (2000). World Net commerce approaches hypergrowth. and computer engineering were developed in the
Retrieved March 1, 2004, from http://www.forrester.com/ER/
Research/Brief/0,1317,9229,FF.html
1970s and 1980s. They dealt extensively with hard-
Schein, E. H. (1999). The corporate culture survival guide: Sense and ware and software; mathematics was the main sup-
nonsense about cultural change. San Francisco: Jossey-Bass. porting discipline. As the field of computer science
Simeon, R. (1999). Evaluating domestic and international web-sites has developed, other disciplines have been added to
strategies. Internet Research: Electronic Networking Applications and
Policy, 9(4), 297308. accommodate changes in use and technology. HCI
Straub, D., Keil, M., & Brenner, W. (1997). Testing the technology is one such discipline; it has been added to many
acceptance model across cultures: A three country study. computer science curricula during the 1990s and
Information & Management, 31(1), 111.
Tannen, R. S. (1998). Breaking the sound barrier: Designing auditory
early 2000s.
displays for global usability. Human Factors and the Web. Retrieved In order to promote a more unified and co-
March 1, 2004, from http://www.research.att.com/conf/hfweb/ herent approach to education in HCI, the Special
proceedings/tannen/index.htm Interest Group on Human-Computer Interaction
Wargin, J., & Dobiey, D. (2001). E-business and change: Managing the
change in the digital economy. Journal of Change Management,
(SIGCHI), part of the Association for Computing
2(1), 7283. Machinery (ACM), the worlds oldest and largest in-
Yeo, A. (1996). World-wide CHI: Cultural user interfaces, a silver lin- ternational computing society, decided in 1988 to
ing in cultural diversity. SIGCHI Bulletin, 28(3), 47. Retrieved initiate development of curriculum recommenda-
March 1, 2004, from http://www.acm.org/sigchi/bulletin/1996.3/
international.html. tions. The result of this work was published in
1992 under the title ACM SIGCHI Curricula for
Human-Computer Interaction. The report defined
the discipline and presented six main content areas.
ECONOMICS AND HCI The report also provided four standard courses: CS1
(User Interface Design and Development), CS2
See Denial-of-Ser v ice Attack; Digital Cash; (Phenomena and Theories of Human-Computer
E-business; Hackers Interaction), PSY1 (Psychology of Human-Computer
EDUCATION IN HCI 199
A Personal StoryBringing HCI Into the Real World
In teaching HCI concepts, I often try to make connections to interaction with the real world.
One of the classrooms in which I teach is adjacent to a chemistry lab. A solid wooden door connects the two rooms.
Until recently, a large white sign with red lettering was posted on the door, visible to all in the classroom, reading, Fire
door. Do not block. I found nothing remarkable about this arrangement until one day I noticed that the door has no knob,
no visible way of opening it. Further examination showed that the hinges are on the inside of the door, so that it opens into
the classroom. A bit of thought led to the realization that the door is for the students in the chemistry lab; if a fire breaks
out in the lab they can escape into the classroom and then out into the corridor and out of the building.
All well and good, but where does that leave students in the classroom? Imagine a fire alarm going off and the smell
of smoke in the air. My students rush to what looks to be the most appropriate exit, and find that there's no way of open-
ing the door marked Fire door, and that pushing on it is not the solution in any case. When I describe this scenario to my
HCI students in the classroom, as an example of inadequate design in our immediate surroundings, it usually gets a few
chuckles, despite the context. Still, they can learn a few lessons about design from this example.
Messages are targeted at specific audiences, and messages must be appropriate for their audience. Here we have two po-
tential audiences, the students in each of the two adjoining rooms. For the students in the chemistry lab, the sign would be
perfectly appropriate if it were visible on the other side of the door. For the students in the classroom, less information
would actually improve the message: Important: Do not block this door would be sufficient. This avoids drawing at-
tention to the function of the door, functionality that is not targeted at those reading the sign. In general, conveying an un-
ambiguous message can be difficult and requires careful thought.
The sign no longer hangs on the door, which now stands blank.
Robert A. St. Amant
Interaction), and MIS1 (Human Aspects of Informa- use in order to contribute to consumer awareness of
tion Systems). interactive systems. It emphasized the role of com-
CS1 and CS2 were designed to be offered in se- puters in organizations and evaluation of the suit-
quence in a computer science or computer engi- ability of technological solutions. Although the
neering department. CS1 focused on HCI aspects of students were not thought of as system builders, the
software, dealing primarily with practical develop- ACM SIGCHI report recommended teaching pro-
ment of interfaces. It was defined as a general course gram design and implementation as well as the use
that complemented basic programming and soft- of tools such as spreadsheets and databases that have
ware engineering courses. CS2 was for students spe- considerable prototyping and programming
cializing in HCI, and it examined HCI in a broader capability.
context, presented more-refined design and evalua- This classical curriculum has been very influen-
tion techniques, and placed more emphasis on sci- tial as a framework and source of inspiration for the
entific foundations. integration of HCI into many educational programs.
The PSY1 course was designed to be offered in a Several textbooks have been created to cover these
psychology, human factors, or industrial engineer- areas, including Prentice-Halls 1993 Human-
ing department. It stressed the theoretical and em- Computer Interaction, Addison-Wesleys 1994
pirical foundations of human-computer interaction. Human-Computer Interaction, and Addison-Wesleys
Here too the emphasis was more on design and eval- 1998 Designing the User Interface. A classical refer-
uation techniques and less on implementation. ence for the graduate level is Readings in Human-
The MIS1 course was designed to be offered in Computer Interaction: Toward the Year 2000,
an information systems department. It focused on published in 1995 by Morgan Kaufmann.
Typical Problems Since the late 1990s, small mobile computers and
HCI literature includes numerous guidelines and Web-based applications have presented new chal-
methods for analyzing users work, for implemen- lenges. The design of interfaces for such technolo-
tation, and for evaluation. There are also many dis- gies is only supported to a very limited extent by the
cussions of specific designs for interactive systems, methods and guidelines that are currently taught in
since the systematic design of a user interface is an many HCI courses. Textbooks that deal with these
essential activity in the development process. However, challenges are beginning to appear. Web Site Usability,
although courses expose students to a rich variety of published in 1999 by Morgan Kaufmann, teaches de-
systems, as soon as the students are confronted sign of Web-based applications. HCI courses have
with the task of designing a new system, they are to some extent been adjusted to include brief de-
equipped with very little in the way of methodologies. scriptions of novel systems and devices to inspire
Current education in HCI also does not pay students to use their imaginations, but most educa-
enough attention to the interplay between design tion in HCI still focuses on designing and develop-
and implementation. Design and implementation ing traditional computer systems.
can be seen as separate activities, but the tools used Guidelines for developing interactive interfaces
for implementation support certain designs and im- typically include careful analysis of the context of
pede others. When the two activities are treated sep- use, which has traditionally been work activities. Yet
arately, this fundamental relation is ignored. the new technologies are used in a multitude of other
Another weakness of many introductory courses contexts, such as entertainment, and these new con-
is that they focus solely on design and implemen- texts must be taken into consideration for future
tation and fail to stress the importance of evalua- guidelines.
tionof defining and measuring usability in a
systematic manner. Within a single course, it is im-
possible to master all the issues involved in the de- Integrating Practical Development
velopment of a user interface, but students should For students of HCI truly to understand the na-
be exposed to all the issues and understand their im- ture of the field, they must try putting their knowl-
portance and how they are related. If they only learn edge into action. There are two basically different
about design and implementation and not about ways of giving students experience with practical de-
evaluating the usability of their products, we risk velopment: through course exercises and student
ending up with systems that are attractive on the sur- projects.
face but are of no practical use to a real user. The op- The ACM SIGCHI curriculum contains pro-
posite approach is to focus primarily on evaluation posals for limited development tasks that students
from the very beginning. Students learn to evaluate can solve as exercises in a course. CS1 encourages a
the usability of an existing system through a course focus on design and implementation, using inter-
in usability engineering, which they can take in the face libraries and tools. CS2 suggests having students
first semester of an undergraduate program. Field begin from less well-defined requirements, thereby
evaluations and other, more complicated, forms of changing the focus more toward user work and task
evaluations can then be introduced in later semesters. analysis. It is suggested that the students also com-
plete design, implementation, and evaluation activ-
ities. The problem with such exercises is that they are
New Challenges limited in time and therefore tend to simplify the
HCI education continues to be challenged by new challenges of interface development. In addition, ex-
technological developments. The PC revolution that ercises are usually conducted in relation to just one
occurred in the middle of the 1990s and the wide- course. Therefore, they usually involve topics from
spread use of graphical user interfaces required more that one course only.
focus on graphical design. Many courses have adapted A more radical approach is to have students work
to these developments. on projects that involve topics from a cluster of
EDUCATION IN HCI 201
courses. There are some courses of study in which There are a growing number of cross-disciplinary
HCI is one element in a large project assignment that programs that involve development and use of com-
student teams work to complete. These courses in- puters. In several of these, HCI is becoming a key
troduce general issues and support work with the discipline among a number of scientific approaches
project assignmentfor example, an assignment to that are merged and integrated in one institutional
develop a software application for a specific organ- setting.
ization might be supported with courses in HCI, Finally, multidisciplinary education programs
analysis and design, programming, and algorithmics with an explicit and strong focus on design are be-
and data structures. This basic pedagogical approach ginning to appear. These programs handle the chal-
introduces students to theories and concepts in a lenge from emerging technologies by using an overall
context that lets the students see the practical appli- focus on design to treat such diverse disciplines as
cations of those theories and concepts. Projects computer science, architecture, industrial design,
undertaken during different semesters can be dif- communication and interaction theory, culture and
ferentiated by overall themes. Such themes might re- organization theory, art, media, and aesthetics.
flect key challenges for a practitionerfor example, The goal is to educate students to think of themselves
software development for a particular organiza- as designers who posses a rich and constructive
tion or design of software in collaboration with users. understanding of how modern information tech-
Using projects as a major building block in each nology can be used to support human interaction
semester increases an educational programs flexi- and communication. HCI will be a core subject in
bility, for while the content of a course tends to be such programs.
static and difficult to change, the focus of the proj-
ects is much easier to change and can accommodate Jan Stage
shifting trends in technology or use. Thus while
courses and general themes of the projects can be See also Classrooms
fixed for several years, the content of the projects can
be changed regularly, so that, for example, one year
students work on administrative application systems FURTHER READING
and the next on mobile devices.
Managers from organizations that hire students Baecker, R. M., Grudin, J., Buxton, W. A. S., & Greenberg, S. (Eds.).
after graduation have emphasized the importance of (1995). Readings in human-computer interaction: Toward the
year 2000 (2nd ed.). Los Altos, CA: Morgan Kaufmann.
projects. The students get experience with large Dahlbom, B. (1995). Gteborg informatics. Scandinavian Journal of
development projects that are inspired by actual real- Information Systems, 7(2), 8792.
world problems. In addition, the students learn to Denning, P. J. (1992): Educating a new engineer. Communications of
the ACM, 35(12), 8397.
work with other people on solving a task. The man- Dix, A., Finlay, J., Abowd, G., & Beale, R. (1993). Human-computer
agers often say that a student with that sort of train- interaction. Hillsdale, NJ: Prentice-Hall.
ing is able to become a productive member of a Hewett, T. T., Baecker, R., Card, S., Carey, T., Gasen, J., Mantei, M.,
project team in a very short time. et al. (1992). ACM SIGCHI curricula for human-computer inter-
action. New York: ACM. Retrieved July 24, 2003, from http://www.
acm.org/sigchi/cdg/
Kling, R. (1993): Broadening computer science. Communications of
The Future the ACM, 36(2), 1517.
In the last decades of the twentieth century, HCI was Mathiassen, L., & Stage, J. (1999). Informatics as a multi-disciplinary
education. Scandinavian Journal of Information Systems, 11(1),
integrated into many educational programs, and 1322.
there are no signs that the subject will diminish in Nielsen, J. (1993). Usability engineering. San Francisco: Morgan
importance in the years to come. On the contrary, Kaufmann.
Preece, J., Rogers, Y., Sharp, H., Benyon, D., Holland, S., & Carey, T.
one can expect that many programs that have a ba- (1995). Human-computer interaction. Reading, MA: Addison-
sic focus on computing and information systems but Wesley.
that lack courses in HCI will take up the subject. Rubin, J. (1994). Handbook of usability testing. New York: Wiley.
Shneiderman, B. (1998). Designing the user interface (3d ed.). Reading, sometimes change their titles, making it difficult to
MA: Addison-Wesley. arrive at an exact figure.)
Skov, M. B., & Stage, J. (2003). Enhancing usability testing skills of
novice testers: A longitudinal study. Proceedings of the 2nd
Beginning in the 1960s, the first attempts were
Conference on Universal Access in Computer-Human Interaction. made to convert scholarly journals or articles from
Mahwah, NJ: Lawrence-Erlbaum. journals into digital format. As information tech-
Spool, J. M., Scanlon, T., Schroeder, W., Snyder, C., & DeAngelo, T. nologies and telecommunications infrastructure de-
(1999). Web site usability. Los Altos, CA: Morgan Kaufmann.
veloped, digital, or electronic, journals have become
a viable alternative to print. As of 2003, over 80 per-
cent (approximately 12,000) of peer-reviewed jour-
nals are available in some electronic form
ELECTRONIC JOURNALS Fulltext Sources Online, published twice a year by
Information Today, Inc., lists by title the scholarly
Scholarly journals, which include substantive re- journals, magazines, newspapers, and newsletters
search articles and other materials, including letters that are available in some digital form. The num-
to the editor, book reviews, and announcements of ber of listings in Fulltext Sources Online grew from
meetings, trace their origins back to 1665, with Les about 4,400 in 1993 to over 17,000 by the end of 2002.
Journal des Scavans (trans., Journal of the experts) The formats of electronic journals (or e-journals)
in Paris and Proceedings of the Royal Society of London vary considerably, however.
in London. These journals developed to share sci-
entific discoveries among interested parties and to
establish who was first to have made a given discov- Electronic Journals: Journal
ery or to have advanced a given theory.
Peer review is an important part of publication Focused or Article Focused
in scholarly journals. It is a system whereby scholars E-journals can be categorized as either journal fo-
who are experts in the same field as the author (the cused or article focused. Journal-focused e-journals
authors peers) read, comment on, and recommend are complete replacements for print, providing an
publication or rejection of an article. This process is entire journal and, often, even more information
usually single-blind (the author does not know who than is available in any extant print alternative ver-
the reviewers are, but the reviewers know who the sions. A journal-focused e-journal generally has a
author is) or double-blind (the author does not know recognizable journal title, an editorial process, a col-
who the reviewers are and the reviewers do not know lection of articles on related topics, and may even
the identity of the author), which gives both readers have volumes and issue numbers. These complete e-
and authors increased confidence in the validity of journals often permit browsing through tables of
the published articles. Although it has been criticized contents and often feature a search engine that lets
from time to time, peer review remains one of the readers search for specific information. Complete
most valued aspects of publication in scholarly jour- electronic journals provide the same branding func-
nals, which are also referred to as peer-reviewed jour- tion that print journals provide. They are typically
nals, scholarly journals, or refereed journals. available directly from the primary journal publisher,
usually for a subscription charge.
Article-focused e-journals are just databases of
Status of Electronic Journals Today separate articles extracted from print or electronic
Today, according to Ulrichs Periodicals Directory, versions of the complete journal. Commercial data-
there are approximately 15,000 peer-reviewed jour- bases of separate articles may be available either from
nals actively published in all fields. (This number the primary publisher or from an aggregator service
should be considered approximate, as new jour- such as ProQuest, InfoTrac, or EbscoHost. Article-
nals are constantly being launched and old ones con- focused e-journals typically emphasize searching over
stantly ceasing publication. In addition, journals browsing and mix articles from many different jour-
E-JOURNALS 203
nals. In these databases it is selected articles, rather for member services. Members may receive a sub-
than complete journal titles, that are made available. scription to a print or electronic journal with their
Even within a journal-focused e-journals, there society membership or, increasingly, pay extra for it.
are many variations. The scholars Rob Kling and Ewa Society publishers' main revenue source is from sub-
Callahan describe four kinds of electronic jour- scriptions paid for by libraries.
nals: pure e-journals distributed only in digital form; Some say that for-profit companies (commer-
e-p-journals, which are primarily distributed elec- cial publishers) should not publish scholarly pub-
tronically, but are also distributed in paper form in lications because research and scholarship should be
a limited way; p-e-journals, which are primarily dis- freely available to all. A for-profit company owes
tributed in paper form, but are also distributed elec- its primary allegiance to its shareholders and the
tronically; and p- + e-journals, which have parallel bottom line rather than only to the propagation
paper and electronic editions. of knowledge. Subscription fees create a barrier that
Electronic journals may be mere replicas of a means only those who can pay or who belong to
print version, with papers presented in PDF for- an institution that can pay, have access to important
mat for handy printing, or they may provide a new research information. Still, in scholarly journal pub-
e-design with added functionality, color graphics, lishing, commercial publishers such as Elsevier
video clips, and links to data sets. Both browsing and Science, Wiley, and Springer publish the largest per-
searching may be possible, or only one or the centage of the scholarly journals, and that percent-
other. The availability of back issues also varies con- age is growing. For-profit publishers range from those
siderably. The American Astronomical Society has giants to relatively tiny publishers, and together they
an advanced electronic-journals system, with added publish approximately 40 percent of all scholarly
functions, links to other articles and to data sets, and journals. Libraries are the main subscribers to
extensive back files of old issues. both print and electronic journals and provide ac-
Aggregators of electronic-journal articles are cess to library constituents either by password or
companies that act as third parties to provide access Internet protocol address (the address, given in num-
to journal articles from a variety of publishers. The bers, that corresponds to an Internet location).
advantage of an aggregator or a publisher that offers University presses mostly publish mono-
many titles is, of course, the availability of many graphs, but universities and other educational insti-
articles from many journals in just one system. tutions also account for about 16 percent of scholarly
The system may offer articles from a wide variety of journals. Other publishers, mostly government agen-
publishers and the originals may be print, electronic, cies, contribute 21 percent of the titles published.
or both. Many scientists and social scientists prefer
electronic journals for the convenience of desktop
access and additional functions, such as the ability
Publishers of Scholarly Journals to e-mail an article to a colleague. E-journals also al-
From their early days, scholarly journals were pub- low scholars to save time locating and retrieving ar-
lished by scholarly societies, commercial publishers, ticles. Since almost all electronic journals have a
university presses, and government agencies. These subscription charge, libraries are the main customers,
main categories of publishers continue today with providing seamless access for faculty, students, staff,
both print and electronic-journal publishing. The or researchers.
number of journals published by each is not equally
distributed, however.
Societies may be the most visible to scholars, yet Article-Focused Alternatives
only approximately 23 percent of scholarly journals
are published by societies. They have a core con- to E-journals
stituency to serve, and publishing activities are al- Article-focused e-journals, being collections of
most always seen as a money-making venture to pay articles organized in subject-related databases,
are particularly good for in-depth reading over time The Impact of E-publishing
or for access to articles that come from unfamil-
iar sources. They extend, rather than replace, a li-
Alternatives
brarys journal collection and, like journals, are The fact that authors are now using a variety of pub-
provided to library constituents on a secure basis lishing venues leads to worries about duplicate
through passwords or other authentication. Article versions, as it is hard to tell which is the definitive or
databases are changing the nature of scholarship: archival version of a paper when multiple versions
In the late 1970s, scientists and social scientists read of the same paper are posted over time. Also, it
articles from an average of thirteen journal titles may be difficult to distinguish low-quality papers
each year; with electronic-journal databases they from high-quality papers when it is so easy for all
now read from an average of twenty-three journal papers to be posted. The positive impact of speedy
titles. access to research literature overshadows these fears
In addition to taking advantage of aggregators in many scholars minds, however, and so far some
article databases, readers can also choose to get in- scholars and students report being able to assess the
dividual articles from special electronic services, such definitiveness and quality of articles without too
as the Los Alamos/Cornell arXiv.org service or those much difficulty.
linked to by the Department of Energy, Office of All of the new electronic models, formats, and
Scientific and Technical Information PrePrint choices show us clearly that scholarly publishing is
Network (http://www.osti.gov/preprints/). These at a crossroads. To understand what impact these
services provide access to articles that may be new options for reading and publishing scholarly
preprints of articles that will be submitted to peer- materials may have, it is useful first to consider what
reviewed journals by the author, postprints the traditional structure and fundamental purposes
(copies of articles that are also published in jour- of scholarly publishing have been.
nals), or papers that will never be submitted to Traditionally, many people have been involved
traditional journals. in the business of moving scholarly ideas from the
Individual electronic articles may also be ac- hands of the author to the hands of the reader. If the
cessed at an authors website or at institutional repos- people and stages involved are seen as links in a chain,
itories. The Open Archives Initiative has led the way the first link is the author and the last link is the
in alternatives to traditional journal publishing and reader, but there are many intervening links
has inspired related initiatives that move the re- peer review, editing, distribution, indexing, sub-
sponsibility for distributing scholarship from pub- scription, and so forth. Each link adds value, but it
lishers to the scholars themselves or to the scholars also adds costs and time delays.
institutions. Institutional repositories are now at the Some of the links are by-products of a print
early planning and development stage, but ideally distribution system and reflect the limitations
will include the entire intellectual capital of a uni- of print access. Electronic distribution may be one
versity faculty, including papers, data, graphics, and way to cut out the intervening links, so an arti-
other materials. The Open Archives Initiative pro- c l e m o ve s d i r e c t l y f ro m t h e a u t h o r t o t h e
motes software standards for establishing institu- reader. But it is important to remember the func-
tional or individual e-print services (access to digital tions of those links and the value they add. Peer
preprints or postprints) so many institutions are review, for example, adds authority; editing adds
establishing OAI-compliant sites. E-print services quality; distribution adds accessibility; and archiv-
are well established in some academic disciplines, ing adds longevity. Online alternatives that pro-
in particular high-energy physics and astro- tect these functions to some degree will be the
physics. They are not as common in disciplines such most successful in the long run, although the rel-
as medicine and chemistry, which rely heavily on ative value versus cost of these functions is
peer review. hotly debated.
ELECTRONIC PAPER TECHNOLOGY 205
The Future Pullinger, D., & Baldwin, C. (2002). Electronic journals and user
Online journals today range from simplistic (and quite behaviour. Cambridge, UK: Deedot Press.
Rusch-Feja, D. (2002). The Open Archives Initiative and the OAI pro-
old-fashioned-looking) ASCII texts (texts that rely on tocol for metadata harvesting: Rapidly forming a new tier in the
the American Standard Code for Information scholarly communication infrastructure. Learned Publishing, 15(3),
Interchange, or ASCII, for data transmission) of in- 179186.
dividual articles available from aggregator services Schauder, D. (1994). Electronic publishing of professional articles:
Attitudes of academics and implications for the scholarly com-
such as Lexis-Nexis to complex multimedia and in- munication industry. Journal of the American Society for Information
teractive electronic journals available on the publishers Science, 45(2), 73100.
website. Fully electronic journals without print equiv- Tenopir, C., King, D. W., Boyce, P., Grayson, M., Zhang, Y., & Ebuen,
M. (2003). Patterns of journal use by scientists through three evo-
alents are still rare, but they are expected to become lutionary phases. D-Lib Magazine, 9(5). Retrieved July 29, 2003,
more common in many disciplines. Fully electronic from http://www.dlib.org/dlib/may03/king/05king.html
journals can be highly interactive and can include mul- Tenopir, C., & King, D. W. (2000). Towards electronic journals: Realities
timedia, links to data sets, and links to other articles; for scientists, librarians, and publishers. Washington, DC: Special
Libraries Association.
they can also encourage a sense of community among Weller, A. C. (2001). Editorial peer review: Its strengths and weaknesses.
their readers. Therefore their impact on scholarship Medford, NJ: Information Today.
in the future is likely to continue to grow.
Carol Tenopir
See also Digital Libraries ELECTRONIC PAPER

TECHNOLOGY
FURTHER READING
For nearly two thousand years, ink on paper has been
Borgman, C. L. (2000). From Gutenberg to the Global Information the near-universal way to display text and images on
Infrastructure: Access to information in the networked world.
a flexible, portable, and inexpensive medium. Paper
Harnad, S. (2001). For whom the gate tolls? How and why to free the does not require any external power supply, and im-
refereed research literature online through author/institution self- ages and text can be preserved for hundreds of years.
archiving, now. Retrieved July 28, 2003, from http://cogprints However, paper is not without limitations. Paper can-
.soton.ac.uk/documents/disk0/00/00/16/39/index.html
King, D. W., & Tenopir, C. (2001). Using and reading scholarly liter- not be readily updated with new images or text se-
ature. In M. E. Williams (Ed.), Annual review of information sci- quences, nor does it remain lightweight when dealing
ence and technology: Vol. 34. 19992000 (pp. 423477). Medford, with large quantities of information (for example,
NJ: Information Today.
Fjallbrant, N. (1997). Scholarly communication: Historical development
books). Nevertheless, although laptop computers
and new possibilities. Retrieved July 28, 2003, from http://inter- have enabled people to carry around literally thou-
net.unib.ktu.lt/physics/texts/schoolarly/scolcom.htm sands of documents and images in a portable way,
Ginsparg, P. (2001). Creating a global knowledge network. Retrieved they still have not replaced ink on paper.
July 28, 2003, from http://arxiv.org/blurb/pg01unesco.html
Kling, R., & Callahan, E. (2003). Electronic journals, the Internet, and
Imagine a thin film that possesses the look and
scholarly communication. In B. Cronin (Ed.), Annual review of in- feel of paper, but whose text and images could be
formation science and technology: Vol. 37. 2003 (pp. 127177). readily changed with the press of a button. Imag-
Medford, NJ: Information Today. ine downloading an entire book or newspaper from
Meadows,A. J. (1998). Communicating research. New York: Academic Press.
Nature Webdebates. (2001). Future e-access to the primary literature. the web onto this thin medium, rolling it up, and
Retrieved July 28, 2003, from http://www.nature.com/nature/ taking it to work with you. The technology to
debates/e-access/ make this and similar concepts possible is cur-
Page, G., Campbell, R., & Meadows, A. J. (1997). Journal publishing
(2nd ed.). Cambridge, UK: Cambridge University Press.
rently being developed. There are several different
Peek, R. P., & Newby, G. B. (1996). Scholarly publishing: The electronic approaches to creating what has become known as
frontier. Cambridge, MA: MIT Press. electronic ink or electronic paper.
Ink on paper is a very powerful medium for sev- all the way in one direction or the other, in which
eral reasons. Not only is it thin, lightweight, and in- case the color viewed will be one of the contrast-
expensive, but ink on paper reflects ambient light, ing colors or the other, or they may rotate par-
has extraordinary contrast and brightness, retains its tially, in which case the color viewed will be a shade
text and images indefinitely, has essentially a 180 between the two. For example, if the contrasting col-
viewing angle (a viewing angle is the angle at which ors are black and white, then complete rotation in
something can be seen correctly) is flexible, bend- one direction will mean that black shows, complete
able, and foldable, and perhaps most importantly, rotation in the other will mean white shows, and
consumes no power. Objectively speaking, paper is partial rotation will mean a shade of gray. The im-
an extraordinary technology. Creating a new elec- age that is formed by this process remains stable with
tronic technology that will serve as a successful pa- no additional electrical addressing on the sheet a
per surrogate and match all the positive attributes of long time (even for days). This innovative technol-
paper is no easy task. In fact, it is one of the biggest ogy was pioneered at Xeroxs Palo Alto Research
challenges facing technologists today. Center and is currently being commercialized by
Broadly defined, electronic display materials that Gyricon Media.
can be used in electronic paper applications can be Given contrasting colors of black and white, the
made from a number of different substances, reflect white side of each bead has a diffuse white reflecting
ambient light, have a broad viewing angle, have a pa- appearance that mimics the look and effect of paper,
per-like appearance and most importantly, have while the other side of the ball is black to create
bistable memory. Bistable memorya highly sought- optical contrast. Gyricon displays are typically made
after propertyis the ability of an electrically cre- with 100-micrometer balls.
ated image to remain indefinitely without the An important factor in this technologys success
application of any additional electrical power. There is the fact that the many millions of bichromal beads
are currently three types of display technologies that that are necessary can be inexpensively fabricated.
may make electronic paper or ink applications Molten white and black (or other contrasting col-
possible. These technologies are bichromal rotat- ors) waxlike plastics are introduced on opposite sides
ing ball dispersions, electrophoretic devices, and cho- of a spinning disk, which forces the material to flow
lesteric liquid crystals. to the edges of the disk, where they form a large
number of ligaments (small strands) protruding
past the edge of the disk. The jets are black on one
Rotating Ball Technology: side and white on the other, and quickly break up
into balls as they travel through the air and solidify.
Gyricon Sheets The speed of the spinning disk controls the balls
A Gyricon sheet is a thin layer of transparent plas- diameter.
tic in which millions of small beads or balls, anal- There are many applications envisioned for
ogous to the toner particles in a photocopier this type of display technology. As a paper substitute
cartridge, are randomly dispersed in an elastomer (electronic paper), it can be recycled several thou-
sheet. The beads are held within oil-filled cavities sand times; it could be fed through a copy machine
within the sheet; they can rotate freely in those cav- such that its old image is erased and the new one is
ities. The beads are also bichromal in nature; that is, presented, or a wand can be pulled across the pa-
the hemispheres are of two contrasting colors (black perlike surface to create an image. If the wand is given
on one hemisphere and white on the other hemi- a built-in input scanner, it becomes multifunctional:
sphere). Because the beads are charged, they move It can be a printer, copier, fax, and scanner all in one.
when voltage is applied to the surface of the sheet, This technology is very cheap because the materi-
turning one of their colored faces toward the side of als used and the manufacturing techniques are in-
the sheet that will be viewed. The beads may rotate expensive.
ELECTRONIC PAPER TECHNOLOGY 207
Electrophoretic Technology the capsule and the white particles to the bottom,
Electrophoretic materials are particles that move which now makes the surface appear dark at that spot.
through a medium in response to electrical stimu-
lation. Researchers at the Massachusetts Institute
of Technology pioneered a technique to create mi- Cholesteric Liquid Crystals
crocapsules with diameters of 30300 micrometers Cholesteric liquid crystal materials also have many
that encase the electrophoretic materials, which may of the positive attributes of paper, and they have the
be white particles in a dark dye fluid or black and added advantage of being amenable to full color. The
white particles in a clear fluid. They have coined the optical and electrical properties of a cholesteric
name electronic ink (or e-ink) to identify their tech- liquid crystal material allow it to form two stable tex-
nology. Material containing these microcapsules is tures when sandwiched between conducting elec-
then coated onto any conducting surface. By encap- trodes. The first is a reflective planar texture with a
sulating the particles, the researchers solved the long- helical twist whose pitch, p, can be tuned to reject
standing problem of electrophoretic materials a portion of visible light: When the material is placed
instability. (Electrophoretic materials have tenden- on a black background, the viewer sees a brilliant
cies toward particle clustering, agglomeration, and color reflection. The second is a focal conic texture
lateral migration.) By having the particles encap- that is relatively transparent. The reflection band-
sulated in discrete capsules, the particles cannot dif- width (Dl) in the perfect planar texture is approxi-
fuse or agglomerate on any scale larger than the mately 100 nanometers (100 billionths of a meter).
capsule size. This narrow selected reflection band is different from
In the technology using white particles in a dark the broadband white reflection of Gyricon and elec-
dye, when a voltage of one polarity is applied to a tronic ink reflective display renditions. Upon the ap-
surface that has been coated with this material, the plication of an applied voltage, V 1 , the planar
tiny white encapsulated particles are attracted to the structure transforms into the focal conic state that
top electrode surface so that the viewer observes a is nearly transparent to all wavelengths in the visible-
diffuse white appearance. By changing the polarity light range. The black background is then visible,
of the applied voltage, the white particles then mi- and an optical contrast is created between reflecting
grate back to the rear electrode where they are color pixels and black pixels. In this state, the volt-
concealed by the dye and the pixel appears dark to age can be removed and the focal conic state will re-
the viewer. After migration occurs in both states the main indefinitely, creating a bistable memory between
white particles stay in their location indefinitely even the reflecting planar state and the transparent focal
after the voltage is removed. Gray scale is possible by conic state. In order to revert from the focal conic
controlling the degree of particle migration with ap- state back to the planar reflecting texture, the mol-
plied voltage. This innovative technology is currently ecules must transition through a highly aligned state,
being commercialized by E Ink. which requires the application of voltage V2, which
In the system using black and white particles in is slightly higher than V1. Abruptly turning off the
a clear fluid, each microcapsule contains positively voltage after the aligned state results in the planar
charged white particles and negatively charged black texture. There are ways in which the planar texture
particles suspended in a transparent fluid.When one can be altered to make it more paperlike in its re-
polarity of the voltage is applied, the white particles flectivity. Gray scale is inherent in cholesteric liq-
move to the top of the microcapsule where they be- uid crystals technology since the focal conic domains
come visible to the user (this part appears white). At can be controlled with different levels of voltage.
the same time, an opposite polarity pulls the black Since cholesteric liquid crystal materials are trans-
particles to the bottom of the microcapsules where parent, they can be vertically integrated to create a
they are no longer visible to the viewer. By reversing true color addition scheme. Although stacking cre-
this process, the black particles migrate to the top of ates more complicated driving circuitry, it preserves
resolution and brightness levels since the pixels are so convincing that many people believed that they
vertically integrated rather than spatially arranged were talking with a human psychotherapist.
across the substrate plane, as is the case with con-
ventional liquid crystal displays. The technology was
developed at Kent State University and is now being Eliza as Psychotherapist
commercialized by Kent Displays. Cholesteric liquid In client-centered sessions a psychotherapist reflects
crystal materials are being developed for docu- back what the client says to invite further re-
ment viewers, electronic newspapers and books, and sponses instead of offering interpretations. If a client
information signs. reports a dream about a long boat ride, Eliza might
respond with Tell me about boats. Most users would
Gregory Philip Crawford not immediately assume that the program is igno-
rant of even the basic facts about boats. Weizenbaum
See also Cathode Ray Tubes; Liquid Crystal Display designed Eliza to take advantage of the user's pro-
jected illusion of understanding as a way of mask-
ing the program's profound lack of real-world
FURTHER READING knowledge. He also carefully noted that the as-
sumption of a program understanding what the user
Comiskey, B., Albert, J. D., Yoshizawa, H., & Jacobson, J. (1998). An says is one made by the user.
electrophoretic ink for all-printed reflective electronic displays.
Nature, 394(6690), 253255.
In 1966 the popular understanding of mainframe
Crawford, G. P. (2000). A bright new page in portable displays. IEEE computers as electronic brains superior to human
Spectra, 37(10), 4046. capabilities was so strong that most people did in-
Sheridon, N. K.; Richley, E. A.; Mikkelsen, J. C.; Tsuda, D.; Crowley, deed project vast knowledge and understanding onto
J. C.; Oraha, K. A., et al. (1999). The gyricon rotating ball dis-
play. Journal for the Society for Information Display, 7(2), 141. any computer. So, despite flaws and limitations
that later users would immediately notice, the first
users attributed so much credibility to Eliza's re-
sponses that some subjects have been very hard to
convince that ELIZA is not human (Weizenbaum
ELIZA 1966, 42).
Scientists were also impressed with Elizas po-
The computer program Eliza (also known as tentials. Nobel Prize winner Herbert Simon (econ-
Doctor) was created by the U.S. computer scien- omist, computer scientist, psychologist, and
tist Joseph Weizenbaum (b. 1923) as an artificial in- philosopher of science) (19162001) suggested that
telligence application for natural language the U.S. Congress fund the development of Eliza as
conversation. Considered a breakthrough when pub- a substitute for human psychotherapists, and some
lished, Eliza was named after the character Eliza psychoanalysts attempted to develop more sophis-
Doolittle, who learned how to speak proper English ticated scripts for the program. Such efforts inevitably
in G. B. Shaw's play Pygmalion. Weizenbaum devel- failed. In fact, Weizenbaum critiques Eliza's recep-
oped this program in the 1960s while a computer tion and popularity and asserts that such pro-
scientist at MIT (19631988). Eliza is actually only grams should not be used for sensitive interactions.
one specialized script running on a general conver-
sational shell program that could have various scripts
with different content. The Eliza script presents the Eliza as Conversational Interface
computer's conversational role as a mock Roger- Nevertheless, Eliza's computerized techniques for
ian (referring to the U.S. psychologist Carl Rogers) conversational interaction were a significant im-
client-centered psychotherapist while the user plays provement in human-computer interfaces. Eliza was
the role of a client. At the time the program was among the first artificial intelligence programs that
ELIZA 209
Talking with ELIZA
WHO ELSE IN YOUR FAMILY TAKES CARE
J
oseph Weizenbaum reports the following as a
OF YOU
typical conversation with ELIZA (Weizenbaum
My father.
1966).
YOUR FATHER
CLIENT: Men are all alike.
You are like my father in some ways.
ELIZA: IN WHAT WAY
WHAT RESEMBLANCE DO YOU SEE
They're always bugging us about something
or other. You are not very aggressive but I think you
don't want me to notice that.
CAN YOU THINK OF A SPECIFIC EXAM-
PLE WHAT MAKES YOU THINK I AM NOT
VERY AGGRESSIVE
Well, my boyfriend made me come here.
You don't argue with me.
YOUR BOYFRIEND MADE YOU COME
HERE WHY DO YOU THINK I DON'T ARGUE
WITH YOU
He says I'm depressed much of the time.
You are afraid of me.
I AM SORRY TO HEAR YOU ARE DE-
PRESSED DOES IT PLEASE YOU TO BELIEVE I AM
AFRAID OF YOU
It's true. I am unhappy.
My father is afraid of everybody.
DO YOU THINK COMING HERE WILL
HELP YOU NOT TO BE UNHAPPY WHAT ELSE COMES TO MIND WHEN
YOU THINK OF YOUR FATHER
I need some help, that much seems certain.
Bullies.
WHAT WOULD IT MEAN TO YOU IF YOU
GOT SOME HELP DOES THAT HAVE ANYTHING TO DO
W I T H T H E FAC T T H AT YO U R
Perhaps I could learn to get along with my
BOYFRIEND MADE YOU COME HERE
mother.
Source: Weizenbaum, Joseph. (1966, January). ELIZAA computer
TELL ME MORE ABOUT YOUR FAMILY program for the study of natural language communication between man
and machine. Communications of the ACM, 9(1). Retrieved March 22,
My mother takes care of me. 2004, from http://i5.nyu.edu/~mm64/x52.9265/january1966.html
allowed for a conversation between people and com- Several features of the interactive environment
puters. Eliza opened up the possibility of immedi- and conversational interface contributed to Eliza's
ate user input and computer response in ways that credibility. An environmental feature easily taken for
were much closer to natural human interactions. A granted today was MIT's MAC (Multi-Access
crucial property of human-computer interfaces dis- Computer) time-sharing operating system, which
closed by Weizenbaum's program is that they must allowed multiple users to have quick response times
maintain the user's sense of being heard and un- to their individual input. Eliza appeared to speak
derstood (Weizenbaum 1966, 42). In that sense Eliza back to the user the way another person would. A
satisfied one aspect of Alan M. Turing's (mathe- user could generate input spontaneously at the tele-
matician and founder of computer science and cryp- type machine and have the program respond to that
tology) test for artificial intelligence. specific input conversationally at the same teletype
not unlike today's Internet chat rooms, only with re- easily see how merely reusing input words by put-
sponses generated by a bot (robot). Compared to ting them into canned sentences leads to a loss of
submitting a stack of punched cards and waiting a meaning.
day for a printout, Eliza's interface was positively
friendly.
Achievements and
Interface Problems and Continued Inuence
The program's real achievement was as an example
How Eliza Solves Them of a conversational interface for some useful content.
Weizenbaum's program dealt with several specific This kind of interface is successful for a narrow, the-
interface problems: identifying keywords, discov- oretically well-defined, or foreseeable field of inter-
ering minimal context, choosing and calculating ap- actions such as solving simple arithmetic problems.
propriate responses (transformations), generating Eliza quickly entered into intellectual and pop-
responses for input without any keywords, and most ular culture and continues to be discussed and cited
importantly, allowing for designing separate, change- forty years later. The program has many variants, in-
able scripts that encode the content, that is, the par- cluding psychiatrists Kenneth Colbys Parry (short
ticular keywords and transformations for a given for paranoid schizophrenic), the program Racter, de-
conversational role. Thus, the shell program that scribed as artificially insane, and many more so-
computes responses and a script provide an inter- phisticated descendents.
face to the content encoded in that script.
The program first scans the user's input sentence William H. Sterner
to see if any of the words are in its dictionary of key-
words. If a keyword is found, then the sentence is See also Dialog Systems; Natural-Language Processing
decomposed by matching it to a list of possible
templates. The design of the templates is what dis-
covers some minimal context for the user's input. In FURTHER READING
one of Weizenbaum's examples, the sentence It
seems that you hate me is matched to a template for Bobrow, D. G. (1965). Natural language input for a computer prob-
lem solving system (Doctoral dissertation, MIT, 1965), source num-
the keywords YOU and ME: ber ADD X1965.
(0 YOU 0 ME) Colby, K. M., Watt, J. B., & Gilbert, J. P. (1966). A computer method
The 0 in the template stands for any number of psychotherapy: Preliminary communication. The Journal of
Nervous and Mental Disease, 142(2), 148152.
of filler words. The template is used to break up Lai, J. (Ed.). (2000). Conversational interfaces. Communications of the
the input sentence into four groups: ACM, 43(9), 2473.
(1) It seems that (2) YOU (3) hate (4) ME. Raskin, J. (2000). The humane interfaceNew directions for designing
This decomposition is then matched to one of interactive systems. New York: Addison-Wesley.
Rogers, C. (1951). Client centered therapy: Current practice, implica-
several possible reassembly rules that can be used tions and theory. Boston: Houghton Mifflin.
to generate a response. In this case the one chosen is: Turing, A. M. (1981). Computing machinery and intelligence. In D. R.
(WHAT MAKES YOU THINK I : 3 : YOU). Hofstadter & D. C. Dennett (Eds.), The mind's IFantasies and
The response then substitutes the third part of reflections on self and soul (pp. 5368). New York: Bantam Books.
(Reprinted from Mind, 49[236], 433460)
the input sentence, hate, into the response What Turkle, S. (1984). The second selfComputers and the human spirit.
makes you think I hate you (Weizenbaum 1966, 38). New York: Simon & Schuster.
That is the basic operation of Eliza, although the Weizenbaum, J. (1966). ELIZAA computer program for the study
of natural language communication between man and machine.
program has many more technical nuances. The real Communications of the ACM, 9(1), 3645.
ingenuity comes from designing the decomposition Weizenbaum, J. (1967). Contextual understanding by computers.
and reassembly rules that make up the script. We can Communications of the ACM, 10(8), 474480.
E-MAIL 211
Weizenbaum, J. (1976). Computer power and human reasonFrom location, and information on the left would indi-
judgment to calculation. San Francisco: W. H. Freeman. cate the user, so a message for SallySmith@TechU
Winograd, T. (1972). Understanding natural language. New York:
Academic Press.
would arrive at the right place. The @ symbol was
an obvious choice, according to Tomlinson, be-
cause it was a character that never appeared in names,
and already had the meaningat,so was appropriate
E-MAIL for addressing. All e-mail addresses still include this
symbol.
Electronic mail, also callede-mailor simplyemail, E-mail has grown exponentially for three decades
is a system for exchanging text messages between com- since. In the 1970s and 1980s it grew until it was a
puters. First invented in 1971, e-mail came into standard throughout American universities. Starting
very widespread usage in the 1990s, and is considered in 1988 it moved out into the nonuniversity popu-
by many to be the most important innovation in per- lation, promoted by private companies such as
sonal communications since the telephone. E-mail CompuServe, Prodigy, and America Online. A study
has changed the way businesses, social groups, and of e-mail growth between 19921994 showed traf-
many other kinds of groups communicate. fic doubling about every twelve months279 mil-
lion messages sent in November of 1992, 508 million
the next year, and topping the 1 billion messages/
History of E-mail month mark for the first time in November of
E-mail was invented in 1971 by Ray Tomlinson, who 1994 (Lyman and Varian 2004). Not only were more
was a scientist at BBN in Cambridge, Massachusetts. people getting e-mail accounts, but the people
(The first-ever e-mail message, probably QWERTY who had them were sending more and more mes-
UIOP, was sent as a test between two computers on sages. For more and more groups, there was enough
Tomlinsons desk. Many, but not all e-mail messages critical mass that e-mail became the preferred way
sent since then have been more informative.) This of communicating. By the early twenty-first century
was not the first text message sent via computer, but e-mail was no longer a novelty, but a standard way
the first-ever sent between computers using the now- of communicating throughout the world between
standard addressing scheme. The Internet, or Arpanet all kinds of people.
as it was then called, had come into existence a few
years earlier, and was used by scientists at a few lo-
cations. Users of the Arpanet system already used Format of E-mail Messages
messaging, but one could only send messages to other At its most basic, e-mail is simply a text message with
users at the same location (e.g. user TomJones at a valid address marked by To: Imagine that
State U might easily leave a message for SallySmith [email protected] now wants to send an e-mail
at the same location). Tomlinson was working on a address to [email protected]. The part of the mes-
way to send files between mainframes using file- sage after the @ sign refers to an Internet Domain
transfer program called CPYNET. He decided to also Name. If the e-mail is to be delivered correctly, this
extend the messaging system this so that users could domain must be registered on the Internet Domain
send messages to other users anywhere in the Arpanet Name Server (DNS) system, just as Web pages
system. must be. Likely, TomJoness university keeps a con-
One of the problems facing Tomlinson was ad- stantly updated list of DNS entries (a DNS lookup
dressing. How would TomJones at State U indicate service) so that it knows where to sent Toms out-
that he wanted to send a message to SallySmith at going mail.
TechU, not State U? Tomlinson chose the @ symbol The computer receiving Toms message must have
as the centerpoint for his new addressing system. an e-mail server or know how to forward to one, and
Information on the right of the @ would indicate the must have an account listed forjoe.If either of these
A Personal StoryThe Generation Gap
When I first went off to college, e-mail was something your university offered only to those savvy enough to take advan-
tage of it. It was an exclusive club: People who could send mail and have it arrive in seconds, rather than the usual two or
three days that the U.S. Postal Service required. And so, a freshman at college, I talked with my parents almost every day
for free, via e-mail, while my friends racked up large phone bills calling home. The formality of a written letter, or even a
phone call, was a treat saved only for a special occasion. But it took some time for my mother to warm to this interac-
tion; to her, e-mail was only on the computer, not personal like a letter could be. Even today, it is more like a second lan-
guage to her.
By the time I graduated from college, e-mail was commonplace and ubiquitous. Despite the diaspora of my college
friends across the country, my phone bill remained small, my e-mail rate high, until suddenly a new technology burst onto
the scene. In 1997 I started using Instant Messenger (IM), leaving a small window open on the corner of my screen. As
my friends slowly opted in we gravitated toward the peripheral contact of the buddy list and away from the more for-
mal interaction of e-mail. Gradually, I realized that long e-mail threads had been replaced by quick, frequent IM interac-
tion: a brief question from a friend, a flurry of activity to plan a night out.
But Ive become a bit of a fuddy-duddy; the technology has passed me by. Recently I added a young acquaintance to
my buddy list. He mystified me by sending brief messages: "Hi!" To this I would reply, "What's up? Did you have a ques-
tion?" This would confuse himwhy would he have a question? I finally realized that we used the medium in different
ways. To me, IM was a path for getting work done, a substitute for a quick phone call or a short e-mail. To him, the pres-
ence of friends on his buddy list was simply the warmth of contact, the quick hello of a friend passing by on the Web.
Observing his use is fascinating; he has well over a hundred friends on his list, and generally keeps a dozen or more con-
versations occurring simultaneously. No wonder I rated no more than a quick hello in his busy world! I tried to keep up
once, but found I could not match his style of use of the medium.
As new technologies arise, their new users will no doubt take to them with a gusto and facility that we cannot fully com-
prehend. It is our job as designers to ensure that we offer these users the flexibility and control to make of these new me-
dia what they will, and not limit them by the boundaries of our own imagination.
Alex Feinman
is not correct, the message will be bounced back to Standard E-mail messages also contain other,
the original sender. E-mail can be sent to multiple re- nonessential fields usually including a From field
cipients by putting multiple e-mail addresses in the identifying the sender and a Subject field summa-
To field separated by commas or by using the cc rizing the content. Other optional fields are:
field or bcc field. CC stands for Carbon Copy, and
is a convention taken from office communications Mime type: Describes the file format for attach-
long predating e-mail. If you receive an e-mail where ments
you are listed under the CC field, this means that you HTML formatting: Indicates that the message
are not the primary intended recipient of the mes- contains formatting, graphics, or other elements
sage, but are being copied as a courtesy. Recipients described in the standard Web html format
listed in the CC field are visible to all recipients. BCC Reply-To: Can list a reply to address that may
in contrast stands for Blind Carbon Copy, and con- be different from the sender. This is useful for
tents of this field are not visible to message recipi- lists that want to avoid individual replies being
ents. If you receive a BCC message, other recipients accidentally sent to the entire group.
will not see that you were copied on the message, and SMS: Indicates that the e-mail can be sent to a
you will not see other BCC recipients. device using the Simple Messaging System pro-
E-MAIL 213
tocol used by cell phones and other handheld groups are administered through buttons and
devices links on the group web page, not text commands.
Priority: Can be interpreted by some Web browser These groups may also include other features such
to indicate different priority statuses as online calendars or chatrooms.
E-mail lists, like most other groups, have certain
These are only a few of the more common op-
group norms that they follow, and newcomers should
tional fields that may be included in an e-mail. When
take note of them. Some of these are options that are
an e-mail is sent using these optional features, the
set by the list administrator:
sender cannot be sure that the recipients e-mail soft-
ware will be able to interpret them properly. No Is the list moderated or unmoderated? In
organization enforces these as standards, so it is up moderated lists, an administrator screens all in-
to developers of e-mail server software and e-mail coming messages before they are sent to the
client software to include or not include these. group. In unmoderated lists, messages are im-
Companies such as Microsoft and IBM may also add mediately posted.
specialized features that work only within their Does the list by default Reply to all? When users
systems. E-mail with specialized features that are sent hit the Reply button to respond to a list mes-
outside of the intended system doesnt usually cause sage, will they by default be writing to the indi-
undue problems, howeverthere will just be extra vidual who sent the message, or to the entire
text included in the e-mail header that can be disre- group? Not all lists are the same, and many
garded by the recipient. embarrassments have resulted in failure to no-
tice the differences. Users can always manually
override these defaults, simply by changing the
recipient of their messages in the To line.
E-mail Lists
An important technological development in the his- Lists also have group norms that are not im-
tory of e-mail was the e-mail list. Lists are one-to- plemented as features of the software, but are im-
many distributions. A message sent to an e-mail list portant nonetheless. How strictly are list members
address (e.g., [email protected]) is sent expected to stick to the topic? Is the purpose of the
by an individual and received by everyone subscribed list social or purely informational? Are commercial
to the list. One popular way of administering lists posts welcome or not? Listserv software can be con-
is using ListServ software, which was first developed figured to send an automatic Welcome message to
in 1986 for use on IBM mainframes, and currently new members explaining the formal and informal
marketed by Lsoft (www.lsoft.com). ListServ soft- rules of the road.
ware has the advantage that membership is self-
administeredyou dont need a moderators help
to subscribe, unsubscribe, or change membership Social Characteristics
options, these are done by sending messages that are
interpreted and carried out automatically by the of E-mail
server. For example, Tom Jones could subscribe him- Academic researchers in the fields of communica-
self to an open list by sending the message SUB- tions, psychology, and human-computer interac-
SCRIBE dogtalk-l to the appropriate listserv address. tion were quick to recognize that this radical new
And, just as important, he could unsubscribe him- communications method could have effects on
self later by sending the e-mail UNSUBSCRIBE both individuals and organizations. This re-
dogtalk-l. There are also a wide variety of options search area, which encompasses the study of e-mail
for list subscriptions, such as receiving daily di- and other online media, is referred to as the
gests or subscribing anonymously. study of Computer-Mediated Communications,
Another popular way of administering groups is abbreviated CMC. Some well-established charac-
through online services such as Yahoogroups. These teristics of e-mail are:
Casual Style and share problems that they would be extremely re-
Electronic mail was very quickly recognized to luctant to discuss with anyone face-to-face. Online
have some unique effects on communication style, dating services often arrange e-mail exchanges prior
and possibly have long-term effects on the groups to phone or face-to-face meetings.
that use it. Casual style is one common marker of e- Lack of social cues may sometimes promotes an
mail communication. Many people use the verb artificial feeling of closeness that Joseph Walther calls
talk rather than write, as in Ill talk to you on e- a Hyperpersonal effect (Walther, 1996). Individuals
mail rather than Ill write you an e-mail. E-mail may imagine that other are much closer to them-
never developed the formal salutations and bene- selves in attitudes than they really are, and this
dictions of lettersfew e-mails begin with Dear may lead to highly personal revelations being shared
Mr. Jones or end with Sincerely, Sally Smith. In online that would rarely be communicated face-
1978 one early e-mail user observed: One could to-face.
write tersely and type imperfectly, even to an older
person in a superior position and even to a person Egalitarianism
one did not know very well, and the recipient took Text-only communication does not convey status
no offense. The formality and perfection that most cues, or other information that tends to reinforce so-
people expect in a typed letter did not become as- cial differences between individuals. E-mail is be-
sociated with network messages, probably because lieved to promote egalitarian communication
the network was so much faster, so much more like (Dubrovsky, Kiesler, and Sethna 1991). Lower-level
the telephone (J.C.R. Licklider, quoted in Vezza employees can easily send e-mails to executives that
1978). they would never think to phone or visit, loosen-
The casual style is partly a result of the unique ing restraints on corporate communication and po-
early-Internet hacker culture, but also partly a re- tentially flattening corporate hierarchies. It has
sult of the medium itself. E-mail messages are of- also been observed that students who rarely con-
ten delivered in a few seconds, lending a feeling of tribute verbally in classes will contribute more via
immediacy. The fact that e-mail is easily deleted e-mail or other online discussion, probably because
and not printed on paper lends a feeling of im- of the increased social distance and reduced inhi-
permanence (although this is illusory, as many le- bition (Harasim 1990).
gal defendants are now finding!) While in some
settings, such as when conducting corporate or Negative Effects: Flaming and Distrust
legal business, e-mails are now expected to be The social distance and lack of inhibition can have
formal and guarded in the manner of a letter, in negative effects as well. E-mail writers more easily
general the literary genre of e-mail remains one give in to displays of temper than they would in per-
of casualness and informality. son. In person, blunt verbal messages are often
E-mail, along with other means of Computer- presented with body language and tone of voice to
Mediated Communications, also lends a feeling of alleviate anger, but in e-mail these forms of com-
social distance. Individuals feel less close, and less in- munication are not present. Recipients of rude e-
hibited via e-mail compared to being face-to-face mails may more easily feel insulted, and respond in
with message recipients. The social distance of e-mail kind. Insulting, angry, or obscene e-mail is called
has a number of good and bad effects. flaming. In one early experimental study of com-
paring e-mail and face-to-face discussions, researchers
Self-Disclosure via E-mail counted 34 instances of swearing, insults and name-
Online communication with strangers also leads calling, which were behaviors that never occurred in
to a feeling of safety, because the relationship can be a face-to-face group performing the same task (Siegel
more easily controlled. Many support groups for et al. 1986). For similar reasons, it is often harder
highly personal issues thrive as e-mail lists. Indi- to build trust through e-mail. Rocco (1998) found
viduals may use an online forum to disclose feelings that groups using e-mail could not solve a social
E-MAIL 215
dilemma that required trust building via e-mail ways. Companies are experimenting with more vir-
but groups working face-to-face could do so easily. tual teams, and allowing workers to telecommute
Beyond these interpersonal difficulties that can more often, because electronic communications
occur online, there are some practical limitations of make it easier to stay in touch. Universities offer more
e-mail as well. The asynchronous nature of e-mail off-campus class options than ever before for the
makes it difficult to come to group decisions (see same reason. Organizations may take on more dem-
Kiesler and Sproull 1991). Anyone who has tried to ocratic decision-making practices, perhaps polling
use e-mail to set up a meeting time among a large employees as to their cafeteria preferences or park-
group of busy people has experienced this difficulty. ing issues, because collecting opinions by e-mail is
far easier than previous methods of many-to-
many communication.
Culture Adapts to E-mail
These observations about the effects of e-mail were
made relatively early in its history, before it had be- Future of E-mail
come as widespread as it currently is. As with all new Electronic mail has been such a successful medium
technologies, however, culture rapidly adapts. It has of communication that it is in danger of being
not taken long, for example, for high-level business ex- swamped by its own success. People receive more
ecutives to assign assistants to screen e-mails the way electronic mail than they can keep up with, and strug-
they have long done for phone calls. It is probably still gle to filter out unwanted e-mail and process the rel-
the case that employees are more likely to exchange e- evant information without overlooking important
mail with top executives than to have a phone or details. Researchers have found that e-mail for many
personal meeting with them, but the non-hierarchi- people has become much more than a communica-
cal utopia envisioned by some has not yet arrived. tion medium (Whittaker and Sidner 1996). For ex-
A simple and entertaining development helps e- ample, many people do not keep a separate address
mail senders convey emotion a little better than book to manage their personal contacts, but instead
plain text alone. Emoticons are sideways drawings search through their old e-mail to find colleagues
made with ASCII symbols (letters,numbers and punc- addresses when needed. People also use their over-
tuation) that punctuate texts. The first emoticon was crowded e-mail inboxes as makeshift calendars,
probably : ) which, when viewed sideways, looks like a to-do lists, and filing systems. Designers of high-
smiley face. This emoticon is used to alert a recipient end e-mail client software are trying to accommo-
that comments are meant as a joke, or in fun, which date these demands by incorporating new features
can take the edge off of blunt or harsh statements. such as better searching capability, advanced filters
Most experienced e-mail users also develop per- and threading to help users manage documents
sonal awareness and practices that aid communi- (Rohall and Gruen 2002). E-mail software is often
cation. Writers learn to reread even short messages integrated with electronic calendars and address
for material that is overly blunt, overly personal, or books to make it easy to track appointments and
otherwise ill conceived. If harsh words are exchanged contacts. And e-mail is increasingly integrated
via e-mail, wise coworkers arrange a time to meet with synchronous media such as cell phones, instant
face-to-face or on the phone to work out differences. messaging, or pagers to facilitate decisionmaking and
If a group needs to make a decision over e-mail, such other tasks that are difficult to accomplish asyn-
as setting a meeting time, they adopt practices chronously.
such as having the first sender propose multiple-
choice options (should we meet Tuesday at 1 or
Wednesday at 3?) or assigning one person to col- The Spam Problem
lect all scheduling constraints. A larger, more insidious threat to e-mail comes in
Groups also take advantage of e-mails good char- the form of spam or junk e-mail. Spam refers
acteristics to transform themselves in interesting to unwanted e-mail sent to many recipients. The
term was first used to describe rude but fairly in- 2004), designed to limit spamming. This bill would,
nocuous e-mailings, such as off-topic comments among other things, mandate that commercial e-
sent to group lists, or personal messages acciden- mailers provide opt-out options to recipients
tally sent to a group. But spam has taken on a more and prohibit false e-mail return addresses and false
problematic form, with unscrupulous mass-mar- subject headings. This bill will not eliminate the prob-
keters sending unsolicited messages to thousands lem, because most spam currently originates outside
or even millions of e-mail addresses. These spam- of the United States. Similar multinational efforts
mers are often marketing shady products (video spy may eventually have an effect, however.
cameras, pornographic websites) or worse, solicit- Individuals can also purchase antispam software
ing funds in e-mail scams. These professional spam- or antispam services that will delete some (but not
mers take advantage of two of the characteristics of all) unwanted e-mails. The best way to avoid receiving
e-mail that have made it so popular: its flexibility spam is never to list your e-mail address on your
and inexpensiveness. Spammer usually forge the website in machine-readable text. Many spam lists
from line of the e-mails they send, so that their are assembled by automatic spider software that
whereabouts cannot be easily blocked. (Messages combs through webpages looking for the telltale @
usually include Web addresses hosted in nations sign. If you still want your e-mail to be available
where it would be difficult to shut them down.) on the Web, two simple ways around this are to re-
Spammers also take advantage of the fact that e- place the @ symbol in your e-mail address with the
mail is essentially free for senders. The only sig- word at or create a graphic of your e-mail ad-
nificant cost of e-mail is borne by recipients, who dress and use it as a substitute for the text.
must pay to store e-mail until it can be read or Despite these challenges, electronic mail has
deleted. Low sending cost means that spammers can carved itself an essential place in the social world
afford to send out advertisements that get only a of the twenty-first century and should continue to
miniscule fraction of responses. The effect of this grow in importance and usefulness for many years
spamming is that users are often inundated with to come.
hundreds of unwanted e-mails, storage requirements
for service providers are greatly increased, and the Nathan Bos
marvelously free and open world of international
e-mail exchange is threatened. See also Internet in Everyday Life; Spamming
What is the solution to spam? Many different
groups are working on solutions, some primarily
technical, some legal, and some economic or so- FURTHER READING
cial. Software companies are working on spam fil-
Bordia, P. (1997). Face-to-face versus computer-mediated communi-
ters that can identify and delete spam messages cation. Journal of Business Communication, 34, 99120.
before they appear in a users inbox. The simplest C A N - S PA M l e g i s l a t i o n . Re t r i e ve d Ma rch 3 1 , 2 0 0 4 , f ro m
ones work on the basis of keywords, but spammers http://www.spamlaws.com/federal/108s877.html
quickly developed means around these with clever Crocker, D. E-mail history. Retrieved March 31, 2004, from www.
livinginternet.com
misspellings. Other filters only let through e-mails Dubrovsky, V. J., Kiesler, S., & Sethna, B. N. (1991). The equalization
from known friends and colleagues. But most phenomenon: Status effects in computer-mediated and face-to-
users find this idea distastefulisnt the possibility face decision-making groups. Human-Computer Interaction, 6,
of finding new and unexpected friends and colleagues 119146.
Garton, L. & Wellman, B. (1995). Social impacts of electronic mail
one of the great features of the Internet? Research in organizations: a review of the research literature. In B. R. Burleson
continues on filters that use more sophisticated al- (Ed.), Communications Yearbook, 18. Thousand Oaks, CA: Sage.
gorithms, such as Bayesian filtering, to screen out a Harasim, L. M. (Ed.). (1990). Online education: perspectives on a
new environment (pp. 3964). New York: Praeger.
high percentage of unwanted e-mail. There are Hardy, I. R. (1996). The evolution of ARPANET e-mail. History Thesis,
also attempts afoot to outlaw spam. In December University of California at Berkeley. Retrieved March 31, 2004,
2003 the U.S. Congress passed a bill (CAN-SPAM, from http://www.ifla.org/documents/internet/hari1.txt
EMBEDDED SYSTEMS 217
Kiesler, S., & Sproull, L. S. (1992). Group decision-making and The embedding of computers in larger systems
communication technology. Organizational Behavior and Human enables the implementation of almost unlimited ap-
Decision Processes, 52, 96123.
Lyman, P., & Varian, H. R. (2000). How much information. Retrieved
proaches to control and signal processing. A com-
March 31, 2004, from http://www.sims.berkeley.edu/how-much- puter can implement complex control algorithms
info that can adapt to the changing operation of a
Rocco, E. (1998). Trust breaks down in electronic contexts but can be larger system. Once a computer has been embedded
repaired by some initial face-to-face contact. In Proceedings of
Human Factors in Computing Systems, CHI 1998 (pp. 496502). in a larger system, it can also be used to provide
Rohall, S. L., & Gruen, D. (2002). Re-mail: A reinvented e-mail pro- additional functionality, such as communications
totype. In Proceedings of Computer-Supported Cooperative Work with other computers within or outside the larger
2002. New York: Association for Computer Machinery.
Siegel, J., Dubrovsky, V., Kiesler, S., & McGuire, T. W. (1986). Group
system that it serves. It can also be used to support
processes in computer-mediated communication. Organizational improved interfaces between machines and human
Behavior and Human Decision Processes, 37, 157186. operators. In addition, an embedded computing sys-
Sproull, L., & Kiesler, S. (1991). Connections: New ways of working in tem can be updated or altered through the loading
the networked organization. Cambridge, MA: The MIT Press.
Vezza, A. (1978). Applications of information networks. In Proceedings
of new software, a much simpler process than is
of the IEEE, 66(11). required for changes to a dedicated mechanism or
Walther, J. B. (1996). Computer-mediated communication: Impersonal, analog circuit.
interpersonal, and hyperpersonal interaction. Communication People living in modern technological societies
Research, 23, 343.
Whittaker, S. & Sidner, C. (1996). E-mail overload: Exploring personal come into contact with many embedded systems
information management of e-mail. In Proceedings of Computer- each day. The modern automobile alone presents
Human Interaction. New York: ACM Press. several examples of embedded systems. Computer-
Zakon, R. H. (1993). Hobbes Internet timeline. Retrieved March
31, 2004, from http://www.zakon.org/robert/internet/timeline/
based engine control has increased fuel efficiency,
reduced harmful emissions, and improved automo-
bile starting and running characteristics. Computer-
based control of automotive braking systems has
enhanced safety through antilock brakes. Embedded
EMBEDDED SYSTEMS computers in cellular telephones control system man-
agement and signal processing, and multiple com-
Embedded systems use computers to accomplish spe- puters in a single handset handle the human interface.
cific and relatively invariant tasks as part of a Similar control and signal-processing functions
larger system functionas when, for example, a com- are provided by computers in consumer entertain-
puter in a car controls engine conditions. Computers ment products such as digital audio and video play-
are embedded in larger systems because of the ca- ers and games. Embedded computing is at the core
pability and flexibility that is available only through of high-definition television. In health care, many
digital systems. Computers are used to control other people owe their lives to medical equipment and ap-
elements of the system, to manipulate signals directly pliances that could only be implemented using
and in sophisticated ways, and to take increasing re- embedded systems.
sponsibility for the interface between humans and
machines in machine-human interactions.
Prior to this embedding of computers in larger Dening Constraints
systems, any nontrivial system control required the The implementation and operation constraints on
design and implementation of complex mechanisms embedded systems differentiate these systems from
or analog circuitry. These special-purpose dedicated general-purpose computers. Many embedded sys-
mechanisms and circuits were often difficult to de- tems require that results be produced on a strict sched-
sign, implement, adjust, and maintain. Once im- ule or in real time. Not all embedded systems face this
plemented, any significant changes to them were constraint, but it is imposed much more on em-
impractical. Further, there were severe limits on the bedded systems than on general-purpose computers.
types of control that were feasible using this approach. Those familiar with personal computers rarely think
of the time required for the computer to accomplish not rely on human intervention to address failures
a task because results are typically returned very that might arise. Users of personal computers accept
quickly, from the average users point of view. Some that software often includes bugs, but the same users
personal-computing operations, such as very large expect that their hardwarein this context, devices
spreadsheet calculations or the editing of large, high- such as household appliances, automobiles, and tele-
resolution photographs, may take the computer a no- phoneswill operate without problems. Tradition-
ticeable amount of time, but even these delays are ally, such devices have been very robust because they
rarely more than an inconvenience. In contrast, were relatively simple. The embedding of comput-
embedded systems can operate at unimaginable ing into these sorts of devices offers potential for
speeds, but if an embedded system violates a real- greater functionality and better performance, but
time constraint, the results can be catastrophic. For the consumer still expects the familiar robustness.
example, an automobile engine controller may need Further, embedded computing is often found in
to order the injection of fuel into a cylinder and the systems that are critical to the preservation of hu-
firing of a sparkplug at a rate of thousands of injec- man life. Examples include railroad signaling devices
tions and sparks each second, and timing that devi- and medical diagnostic and assistive technology such
ates by less than one-thousandth of a second may as imaging systems and pacemakers. These systems
cause the engine to stall. Systems that involve control must be robust when first placed in service and must
or signal processing are equally intolerant of results either continue to operate properly or fail only in
that come early or late: Both flaws are disruptive. ways that are unlikely to cause harm. Further, as men-
Limited electrical power and the need to remove tioned above, these systems must operate without
heat are challenges faced by the designers of many human intervention for extended periods of time.
embedded systems because many embedded appli- Most current embedded systems operate in
cations must run in environments where power is isolation, but some perform their functions with lim-
scarce and the removal of heat is inconvenient. ited monitoring and direction from other comput-
Devices that operate on batteries must strike a bal- ers. As with general-purpose computing, there
ance between demand for power, battery capacity, appears to be a trend toward increasing the inter-
and operation time between charges. Heat removal operability of embedded systems. While increasing
is a related problem because heat production goes the interaction among embedded systems offers the
up as more power is used. Also, embedded systems potential for new functionality, networking of em-
must often fit within a small space to improve porta- bedded computing devices also increases security
bility or simply to comply with space constraints im- concerns.
posed by a larger system. Such space constraints
exacerbate the problem of heat removal and thus
further favor designs that limit power consumption. An Illustrative Example
A cellular telephone, for example, features embed- People tend instead to think of embedded systems
ded systems that are hampered by significant power in conjunction with cutting-edge technology, such
and space constraints. A less obvious example is the as the various spacecraft developed and deployed by
avionics package for a general-aviation aircraft. Such NASA. The first embedded computer used by NASA
a system must not draw excessive power from the in a manned spacecraft was developed for the Gemini
aircrafts electrical system, and there may be little program in the early 1960s. That computer was used
space available for it in the aircraft cockpit. for guidance and navigation. (The Mercury program
Users of older personal computers learned to ex- preceding Gemini involved manned space flight, but
pect frequent computer failures requiring that the the flights were simple enough to be controlled from
users restart the computer by pressing a combina- the ground.) The NASA programs following Gemini
tion of keys or a reset button. Newer personal placed increasing reliance on embedded comput-
computers are more robust, but many embedded ers to accomplish a range of tasks required for the
systems demand even greater robustness and can- successful completion of manned space missions.
EMBEDDED SYSTEMS 219
Unmanned space flights have needed embedded wipers and power windows may be controlled by
computers to provide flight control for spacecraft embedded systems. In some instances late-model
too far away to tolerate control from Earth. automobiles that have been recalled to the factory
Closer to Earth, however, the modern auto- have had the required repair accomplished entirely
mobile may contain a hundred embedded com- through an embedded computer software change.
puters, each with greater computational capabilities Embedded communication and navigation systems
than the single computer that traveled on the for automobiles are now available, and these systems
Gemini space flights. Embedded computer engine are more complex than those used in the early space
control was introduced in the late 1970s to satisfy program. In addition, the human interface between
emissions requirements while maintaining good per- the automobile and its driver is now managed by
formance. Those who have operated automobiles one or more embedded systems. In the early
from before the days of embedded systems will 1980s, several automakers replaced analog human
recall that those automobiles were more difficult to interfaces with computer-based interfaces. Some
start when the weather was too cold or too hot or of those interfaces were not well received. Later-
too wet. Automobiles of that era were also less model automobiles retained the computer control
fuel efficient, emitted more pollution, and had per- of the interfaces, but returned to the more familiar
formance characteristics that varied with driving analog appearance. For example, many drivers
and environmental conditions more than is the case prefer the dial speedometer to a digital display, so
today. Embedded computer engine control addresses even though the speedometer is actually con-
these variations by adapting the engine control in trolled by a computer, auto designers reverted from
response to sensed environmental and engine op- digital to analog display.
eration data.
The next element in the automobile drive train
is the transmission. The first cars with auto- Increasing Dependence
matic transmissions typically suffered from poorer Embedded computers can be used to implement far
performance and fuel economy than cars with more sophisticated and adaptive control for com-
manual transmissions. Modern automatic trans- plex systems than would be feasible with mechani-
missions controlled by embedded computers, by cal devices or analog controllers. Embedded systems
comparison, compare favorably with manual trans- permit the human user to interact with technology
missions in both performance and economy. as the supervisor of the task rather than as the
The computer control supports the selection of controller of the task. For example, in an automo-
different shifting strategies depending on whether bile, the engine controller frees the driver from hav-
the driver prefers sports driving or economy driving to recall a particular sequence of actions to start
ing. Further, manufacturers can match a single a car in cold weather. Similarly, the automatic trans-
transmission to a wide range of engines by chang- mission controller releases the driver from track-
ing the software in the transmission controller. ing engine speed, load, and gear, leaving the driver
The embedded transmission system can also be free to concentrate on other important driving tasks.
configured to communicate with the embedded The computers ability to manage mundane tasks ef-
engine system to generate better performance and ficiently is one of the great assets of embedded sys-
economy than each system could achieve oper- tems. Unfortunately, the increased complexity that
ating independently. embedded systems make possible and the increased
Other familiar automotive capabilities provided separation between the user and the machine also
through embedded systems include cruise control, introduce new potential dangers.
control of antilock brakes, traction control, active Embedded systems make previously impractical
control of vehicle suspension, and control of steer- applications practical. Prior to the late 1970s, mo-
ing for variable power assist or four-wheel steering. bile telephone service was cumbersome and ex-
Automobile interior climate and accessories such as pensive because of limited capabilities to manage the
Jeffrey, K. (2001). Machines in our hearts: The cardiac pacemaker, the

available radio communication channels. Embedded implantable defibrillator, and American health care. Baltimore: Johns
computing initially made the modern cellular tele- Hopkins University Press.
phone industry feasible because computers embed- Jurgen, R. (1995). Automotive electronics handbook. New York: McGraw-
Hill.
ded within the cellular telephone base stations
Leveson, N. (1995). Safeware: System safety and computers. Reading,
provided radio channel management and efficient MA: Addison-Wesley.
hand-offs as mobile users moved from cell to cell. Shaw, A. ( 2001). Real-time systems and software. New York: John Wiley
Newer digital cell phones include computers em- & Sons.
Stajano, R. (2002). Security for ubiquitous computing. West Sussex, UK:
bedded within the handsets to improve communi- John Wiley & Sons.
cation and power efficiency. Without embedded Vahid, F., & Givargis, T. (2002). Embedded systems design: A unified
computing, the explosive expansion of the cell phone hardware/software introduction. New York: John Wiley & Sons.
industry could not have occurred. Implanted pace- Wolf, W. (2001). Computers as components. San Francisco: Morgan
Kaufmann Publishers.
makers help maintain the human hearts proper
pumping rhythm and have improved the quality and
duration of life for many people, but those people
are now dependent upon this embedded system.
ENIAC
The Future The Electronic Numerical Integrator and Computer
Embedded systems will certainly expand in func- (ENIAC), built at the University of Pennsylvania
tionality, influence, and diversity for the foreseeable between 1943 and 1946, was the first electronic dig-
future. The digital technology required to implement ital computer that did useful work. Large analog
embedded systems continues to improve, and com- computers had existed since Vannevar Bush and
puter hardware is becoming more powerful, less ex- his team had built the differential analyzer in 1930.
pensive, smaller, and more capable of addressing Depending on one's definition, the first digital
electrical power considerations. In parallel, tech- computer may have been the exper imental
niques for the production of robust real-time soft- Atanasoff-Berry machine in 1940. Unlike its pred-
ware are steadily improving. Digital communication ecessors, ENIAC possessed many of the features of
capability and access is also expanding, and thus fu- later digital computers, with the notable exceptions
ture embedded systems are more likely to exhibit of a central memory and fully automatic stored
connectivity outside of their larger systems. Lessons programs.
learned from early interfaces between humans and In terms of its goals and function, ENIAC was
embedded systems coupled with the improve- the first digital supercomputer, and subsequent su-
ments in embedded computing should yield better percomputers have continued in the tradition of hu-
interfaces for these systems. Embedded systems are man-computer interaction that it established.
likely to become so woven into everyday experi- They tend to be difficult to program, and a techni-
ence that we will be unaware of their presence. cally adept team is required to operate them. Built
from state-of-the-art components, they involve a de-
Ronald D. Williams manding trade-off between performance and reli-
ability. Their chief purpose is to carry out large
See also Fly-by-Wire; Ubiquitous Computing numbers of repetitious numerical calculations, so
they emphasize speed of internal operation and tend
to have relatively cumbersome methods of data input
FURTHER READING and output. Remarkably, ENIAC solved problems of
kinds that continue to challenge supercomputers
Graybill, R., & Melhwm, R. (2002). Power aware computing. New York:
Kluwer Academic Press/Plenum.
more than a half-century later: calculating ballistic
Hacker, B. (1978). On the shoulders of Titans: A history of Project Gemini. trajectories, simulating nuclear explosions, and pre-
Washington, DC: NASA Scientific and Technical Information Office. dicting the weather.
ENIAC 221
A technician changes a tube in the ENIAC computer during the mid-1940s. Replacing a faulty tube required
checking through some 19,000 possibilities. Photo courtesy of the U.S. Army.
Historians debate the relative importance of var- The problem that motivated the U.S. Army to in-
ious members of the ENIAC team, but the leaders vest in ENIAC was the need for accurate firing tables
were the physicist John W. Mauchly, who dreamed of for aiming artillery during World War II. Many new
a computer to do weather forecasting, and the engi- models of guns were being produced, and working
neer J. Presper Eckert. After building ENIAC for the out detailed instructions for hitting targets at vari-
U.S. Army, they founded a company to manufac- ous distances empirically by actually shooting the
ture computers for use in business as well as in gov- guns repeatedly on test firing ranges was costly in
ernment research. Although the company was time and money. With data from a few test firings,
unprofitable, their UNIVAC computer successfully one can predict a vast number of specific trajecto-
transferred the ENIAC technology to the civilian sec- ries mathematically, varying such parameters as gun
tor when they sold out to Remington Rand in 1950. angle and initial shell velocity. The friction of air re-
Both development and commercialization of digi- sistance slows the projectile second by second as it
tal computing would have been significantly delayed flies, but air resistance depends on such factors as the
had it not been for the efforts of Mauchly and Eckert. momentary speed of the projectile and its altitude.
Thus, the accuracy of calculations is improved by di- paper instructions in one hand and used the other
viding the trajectory into many short intervals of hand to turn rotary switches on the tall function ta-
time and figuring the movement of the projectile in bles, one 09 switch for each digit. Arranged in rows
each interval on the basis of the output of the pre- from head to ankle height, these switches had a sim-
ceding intervals and changing parameters. ple color coding to reduce errors: Every fifth row
At the dedication ceremony for ENIAC in of knobs was red and the others black; the plates be-
1946, the thirty-second trajectory of an artillery shell hind the knobs alternated shinny with black, three
was calculated to demonstrate the machine's effec- columns at a time. Multiplication, division, and
tiveness. Using desk calculators, people would take square-root calculation were handled by specially
three days to complete the job, compared with thirty built components that could be plugged in as needed.
minutes on the best ballistics analog computer, the A master programmer unit handled conditional (if-
differential analyzer. ENIAC did the calculation ac- then) procedures.
curately in twenty secondsless than the time the Programmers might require a month to write a
shell would be in the air. Because World War II had program for ENIAC and from a day to a week to
ended by the time ENIAC was ready, the first real job set up and run one, but this was not as inefficient
it did was evaluating the original design for the ther- as it seems because after the machine was ready to
monuclear (hydrogen) bomb, finding that the de- do a particular job, a large number of runs could
sign was flawed and causing the atomic scientists be cranked out rapidly with slight changes in the pa-
to develop a better approach. rameters. ENIAC continued to do useful work for
Filling 167 square meters in a large room, 27- the military until October 1955. Parts of this pio-
metric ton ENIAC was constructed in a U shape, neering machine are on display at the National
with the panels and controls facing inward toward Museum of American History in Washington, D.C.,
an area where the operators worked. ENIAC was built along with videos of Presper Eckert explaining
with about eighteen thousand vacuum tubes, con- how it was operated.
suming 174 kilowatts of electric power and keep-
ing the room quite hot. Many experts had been William Sims Bainbridge
skeptical that the machine could work because
vacuum tubes frequently burned out, but taking great See also Atanasoff-Berry Computer; Supercomputers
care in testing the tubes and running them below
their specifications kept failures in use to about six
hundred a year. FURTHER READING
ENIAC had both an IBM card reader and an au-
tomatic card punch, used chiefly for output and McCartney, S. (1999). ENIAC: The triumphs and tragedies of the world's
first computer. New York: Walker.
input of data calculated during one run that would Metropolis, N., Howlett, J., & Rota, G.-C. (Eds.). (1980). A history of
be used later in another run; the cards were not used computing in the twentieth century. New York: Academic Press.
to enter programs. The computer was programmed Stern, N. (1981). From ENIAC to UNIVAC: An appraisal of the Eckert-
largely by plugging in equipment and connecting by Mauchly computers. Bedford, MA: Digital Press.
Weik, M. H. (1961, January/February). The ENIAC story. Ordnance,
means of cables the twenty accumulators (electronic 37.
adders) that performed the calculations. Hundreds
of flashing lights on the accumulators gave the op-
erators clues about how the work was progressing.
The calculations were done in the decimal system,
rather than binary, and parameters were input man- ERGONOMICS
ually by setting rotary switches. Switches also con-
trolled local program-control circuits. To set The field of human factors and ergonomics plays an
parameters for a given run, the programmers held important and continuing role in the design of
ERGONOMICS 223
human-computer interfaces and interaction. Re- An Ergonomics Approach to

searchers in human factors may specialize in prob-
lems of human-computer interaction and system
Human-Computer Interaction
design, or practitioners with human factors creden- When it comes to designing computer systems,
tials may be involved in the design, testing, and human factors and ergonomics takes the view that
implementation of computer-based information dis- people do not simply use computers; rather, they
plays and systems. perform tasks. Those tasks are as various as con-
Human factors and ergonomics can be defined trolling aircraft, creating documents, and monitor-
as the study, analysis, and design of systems in which ing hospital patients. Thus, when analyzing a
humans and machines interact. The goal of hu- computer system, it is best to focus not on how
man factors and ergonomics is safe, efficient, effec- well people interact with it (that is, not on how
tive, and error-free performance. Human factors well they select a menu option, type in a com-
researchers and practitioners are trained to create mand, and so forth), but how well the system allows
systems that effectively support human performance: them to accomplish their task-related goals. The qual-
Such systems allow work to be performed efficiently, ity of the interface affects the usability of the system,
without harm to the worker, and prevent the worker and the more congruent the computer system is with
from making errors that could adversely affect the users task- and goal-related needs, the more suc-
productivity, or (more importantly), have adverse cessful it will be. David Woods and Emilie Roth (re-
affects on him or herself or others. Research and prac- searchers in cognitive engineering) describe a triad
tice in the field involve the design of workplaces, sys- of factors that contribute to the complexity of prob-
tems, and tasks to match human capabilities and lem solving: the world to be acted on, the agents (au-
limitations (cognitive, perceptual, and physical), as tomated or human), and how the task is represented.
well as the empirical and theoretical analysis of In human-computer systems, the elements of the
humans, tasks, and systems to gain a better under- triad are, first, aspects of the task and situation for
standing of human-system interaction. Methodolo- which the system is being employed; second, the hu-
gies include controlled laboratory experimentation, man operator or user; and third, the manner in which
field and observational studies, and modeling and information relevant to the task is represented or
computer simulation. Human factors and er- displayed. The role of the computer interface is to
gonomics traces its roots to the formal work deserve as a means of representing the world to the hu-
scriptions and requirements of the engineer and man operator.
inventor Frederick W. Taylor (18561915) and the Research related to human factors and er-
detailed systems of motion analysis created by the gonomics has addressed numerous topics related to
engineers Frank and Lillian Gilbreth (18681924, human-computer interaction, including the design
18781972). As human work has come to require of input devices and interaction styles (for example,
more cognitive than physical activities, and as peo- menus or graphical interfaces), computer use and
ple have come to rely on increasingly sophisticated training for older adults, characteristics of textual
computer systems and automated technologies, hu- displays, and design for users with perceptual or phys-
man factors researchers and practitioners have ical limitations. Areas within human factors and er-
naturally moved into the design of computerized, as gonomics that have direct applicability to the design
well as mechanized, systems. Human factors engi- of human-computer systems include those focused
neering in the twenty-first century focuses on the on appropriate methodologies and modeling tech-
design and evaluation of information displays, on niques for representing task demands, those that deal
advanced automation systems with which human with issues of function allocation and the design of
operators interact, and on the appropriate role of human-centered automation, and those concerned
human operators in supervising and controlling com- with the design of display elements that are relevant
puterized systems. to particular tasks.
Methodologies and Modeling Another task-analytic framework that features

goal decomposition is GOMS, which stands for goals,
Frameworks for Task And operators, methods, and selection rules, has been de-
Work Analysis veloped by Card, Moran, and Newell (1983) and
applied in subsequent research. Tasks are decom-
John Gould and Clayton Lewis (1985) claimed posed into task goals and the methods, or sequences
that a fundamental component of successful com- of operators (or actions) that can accomplish them.
puter system design requires an early and contin- Selection rules describe the means by which the op-
ual focus on system users and the tasks that they need erator selects a particular method. GOMS models
to perform. In addition, Donald Norman (1988) has have been used to describe, predict, and analyze hu-
suggested that for a computer system to be suc- man interactions with computing systems. For in-
cessful, users must be given an appropriate model of stance, GOMS models that have been constructed
what the system does, and how; information on sys- to describe user interactions with computer systems
tem operation must be visible or available; and users have actions such as keystrokes or mouse move-
must have timely and meaningful feedback regard- ments, along with more cognitive actions such as
ing the results of their actions. Users should never reading the screen. The psychologist David Kieras
be unable to identify or interpret the state of the com- (1997) has described how GOMS models can be used
puter system, nor should they be unable to identify to predict learning and task execution times for soft-
or execute desired actions. Therefore, human factors ware systems; GOMS (along with a variation sup-
and ergonomics research that focuses on human- porting the modeling of parallel activities) has
computer system design devotes considerable ener- also been used to model large-scale human-com-
gies to analysis of system components, operator puter systems in order to predict task times for a sys-
characteristics, and task requirements, using task and tem under design.
work analysis methods. Christine Mitchell (1987) has developed a
A hierarchical task analysis (HTA), as described third modeling framework, operator function mod-
by Annett and Duncan (1967), decomposes a task eling, which has been applied to the design of in-
into a hierarchical chain of goals, subgoals, and ac- formation displays. In this framework, systems goals,
tions. Plans are associated with goals to specify subgoals, and activities are represented as a set of in-
how and when related subgoals and activities are car- terconnected nodes; each node (corresponding to
ried out, and can take on a variety of structures. a goal, subgoal, or activity) has the potential to change
For instance, plans can have an iterative structure, its state in response to external inputs or the states
describing activities that are repeated until some cri- of higher-level goals. This technique makes it possi-
terion is met. Plans may also describe a set of strictly ble for researchers to model human operators ac-
sequential activities, or they may describe activities tions in real time, which in turn helps them be
that can be done in parallel. In HTA, as in other forms sure that the necessary information is displayed at
of task analysis, one can specify task demands, cri- the appropriate time.
teria for successful completion, knowledge or skills Other forms of analyses and modeling within
required, information needed, and likely errors as- human factors, particularly in the subdiscipline of
sociated with each step. In human-computer system cognitive engineering, have focused specifically on
design HTA is used to help ensure that display in- the cognitive challenges associated with complex hu-
formation content corresponds to identified activi- man-computer systems. By modeling the complex-
ties, that the organization of displays and control ities that cause decision-making and problem-solving
activities matches task requirements, and to provide difficulties for human operators, researchers can de-
additional support (in the form of better informa- velop solutions, such as more-effective interfaces,
tion displays, training, or task requirements) to re- that mitigate those difficulties. Data collection meth-
duce the likelihood of error. ods frequently used in cognitive engineering work
ERGONOMICS 225
and task analyses include interviews with experts ately was able to select a course of action based on
in the area under consideration and observation of recognition of a situation or state.
practitioners. Methods in cognitive task and work analysis have
Some methods in cognitive task analysis focus contributed to the design of information displays for
on the identification and explication of real-world numerous types of complex systems, including
decisions made by experts. Other methods in- process control systems, military command and con-
clude cognitive work analysis, an iterative set of trol systems, and information systems.
analyses and modeling efforts that address goals and
resources relating to the tasks that must be per-
formed, strategies for performing the tasks, the Function Allocation and
influence of the sociotechnical environment on sys-
tem performance, and the knowledge and skills Automation Design
required of operators. (These methods have been Another key research area within human factors and
developed and described in monographs by Jens ergonomics that has direct application to the design of
Rasmussen, Annelise Mark Pejitsen, and L. P. human-computer systems is the appropriate alloca-
Goodstein; and Kim Vicente.) An important com- tion of functions between human operators and au-
ponent is identifying the complexities and con- tomated systems. While early efforts in function
straints that adversely affect the behavior of actual allocation tended to rely on fixed lists of functions bet-
users of the system; those constraints are often rep- ter suited to humans or machines and an either-or ap-
resented using abstraction-hierarchy models. proach, more recent allocation schemes have
Abstraction hierarchies are multilevel system mod- focused on a more human-centered approach.Within
els in which each level of the model corresponds to these schemes, allocations can range from complete
a description of the system at a different level of ab- human control to complete automation,and there can
straction. Higher levels of abstraction represent the be intermediate stages in which humans can over-
system in terms of its purpose and functions, ride automated actions or choose from and implement
whereas lower levels represent the system in terms actions recommended by the automated system.Other
of its physical implementation. Abstraction hier- models have focused on the differing roles of human
archy models lay out the purposes of the overall sys- operation and automation at different stages in the de-
tem, the functions and systems available to cision process (the information gathering stage, the
achieve those purposes, and the constraints on their analysis stage, the actual decision making, and the im-
use or implementation. The cognitive work analy- plementation stage). Selection of an appropriate
sis paradigm also uses decision-ladder descriptions level of automation may also be dynamic, changing
of tasks to be performed. A decision-ladder de- based on external task demands and circumstances.
scription represents stages of processing and re- Related to problem of function allocation are
sultant knowledge states of either human or considerations such as the degree of trust and re-
automated agents, starting with the initial in- liance operators place on automated systems, the ex-
stance of a need for response, moving through the tent to which operators are aware of and understand
observation and classification of information, and the functioning of the automation, and the degree
ending with the selection and execution of an ac- to which the use of information displays can miti-
tion. (The actual stages are activation, observation, gate any difficulties in these areas. For instance, re-
state recognition, goal selection, task and procedure search has found that information displays showing
selection, and implementation.) The model provides how well an automated decision-making element
explicitly for shortcuts between knowledge states functions can improve human operators judg-
and information-processing stages, allowing inter- ment with regard to using the aid. Other research
mediary steps to be bypassed. Such shortcuts might has studied how automation may also affect oper-
be appropriate, for example, if an expert immedi- ators degree of awareness regarding system function
V1
by keeping them effectively out of the loop, par-
ticularly for more complex systems. V8 V2
Ergonomic Studies of
Display Elements
A third way in which human factors and ergonom- V7 V3
ics have made specific contributions to the design of
human-computer interfaces is in the area of display
elements and their ability to convey task and goal-
related information to human operators. For instance,
one angle of human factors and ergonomics display
research has focused on how to represent uncertainty V6 V4
(an important contributor to task complexity) in
graphical form. Researchers have investigated the use V5
of shapes such as ellipses or rings, linguistic phrases, FIGURE 1. Example of an object display. When the
color, or sound to convey positional uncertainty and system is operating normally (solid line), based on
blurred icons to convey state uncertainty. the state of variables V1V8, a regular polygon ap-
A display-design methodology called ecological pears. Deviations (shown as a dashed line) are easily
interface design, developed by Kim Vicente, ap- perceptible.
plies outcomes from cognitive work analysis to the
design of information displays for complex sys-
tem. This approach aims to design displays that sup- Case Study: The Application
port activities such as fault identification and
diagnosis, as well as normal system monitoring and
of Cognitive Analysis Methods
control, by making goal-relevant constraints and to Display Design
properties in the system visible through the inter-
face. Principles in ecological interface design have As noted above, methods in cognitive work and task
been applied and tested in a variety of applica- analysis have been employed in the design of nu-
tions, including process control and aviation. Military merous complex human-machine systems, impact-
command and control systems make use of inter- ing the design of information displays, allocation of
faces that offer similar goal- and function-related functions between humans and automated compo-
information. nents, and the design of tasks and training require-
Researchers have also studied properties of so- ments. Cognitive analyses were used in the design of
called object displays, which integrate multiple a new naval surface vessel, as described by Ann W.
pieces of information into one graphical form. Bisantz (2003). Information sources for the analy-
Holistic properties, or emergent features, defined by ses included interviews with domain experts, design
the values or configurations of individual elements documents and written operational requirements,
of the graphical form, are used to convey informa- and other experts on the subject matter. The
tion related to higher-level goals or system states. For analyses were part of a multiyear design effort and
instance, a star (or polygon) display will graph the were performed early in the design of the vessel. The
values of system state variables graphed on individ- effort focused on information displays, manning re-
ual axes. When the system is operating under nor- quirements, and human-automation function al-
mal conditions, a symmetric polygon is formed; location for the command-and-control center of the
deviations from normal are easily identified (as ship. However, at the point of the design process
shown in Figure 1). when the analyses took place, choices regarding man-
ERGONOMICS 227
ning (the number of personnel that would be domain models can be used to identify potential
available to run the ship), the use of automation, and conflicts, such as when a system might need to be
subsequent tasks that would be assigned to per- utilized to accomplish multiple goals simultaneously,
sonnel were still to be determined. Thus, it was not or when the use of a system to accomplish one
possible for the designs to focus on detailed plans goal might negatively impact another goal. These
and specifications for the command-and-control conflicts and interactions suggest requirements for
workstations. Instead, models and research findings information displays that would let system con-
from cognitive engineering were used to make rec- trollers recognize the conflicts and make appropri-
ommendations regarding display areas and con- ate decisions.
tent as follows. For instance, Figure 2 shows a portion of a
work domain model of the naval environment. Notice
Work Domain Models that the gun system (a physical system) may be re-
As described above, functional abstractions of the quired to accomplish multiple high-level goals: It may
work domain can help significantly with the design be required for self defense against mines as well as
of information displays, because such abstractions for support of on-shore missions (land attack).
make the goals of the work domain, as well as the Because different people may be responsible for co-
functional and physical resources available to ac- ordinating and implementing these two goals, this
complish those goals, explicit. Additionally, work potential conflict indicates the need for specific
Self Defense Against

Undersea Threats Self Defense Against
Mines Neutralize Mines
Land Attack Missions Naval Surface Fire

Suppor t
Counterbattery
Deliver Ordinance for Mine Neutralization

Land Attack Systems
Short Range Deliver y
Gun System
FIGURE 2. Work domain example showing how multiple goals can rely on one physical system, indicating re-
quirements for the contents of information displays for those selecting among goals or utilizing the gun system.
Figure reprinted from Bisantz et al. (2003)
Defense of Systems Achieve

and Personnel Assigned Missions
Signature Battlespace
Maintenance Awareness
Use of
Use of some sensors
sensors makes necessary for
ship detectable offensive
Maneuverin g Battlespace Environmental functions
Systems Sensors Sensors
Moving may
disambiguate Knowledge of environmental
sensor data conditions can impact choice
of sensor settings
FIGURE 3. Portion of a work domain analysis indicating potential goal interactions involved in sensor man-
agement that would need to be displayed to operators. Source: Bisantz et al. (2003)
information display content, such as alerts or mes- defense and offensive support), display indications
sages to subsystem controllers if the gun system is un- regarding mission areas outside a controllers pri-
available or if there are plans for its use, as well as alerts mary area of responsibility, and displays for higher-
to higher-level commanders that if both the land- level commanders regarding mission goals and
attack and mine-defense goals are active, there may priorities (Bisantz et al. 2003).
be a constraint in the availability of the gun system.
Another example involves the management, use, Decision Ladder Descriptions
and configuration of sensor systems, as shown in The application of the decision ladder formalism to
Figure 3. The configuration of certain types of sen- describe operators tasks also led to recommenda-
sors depends on environmental factors such as ocean tions for display design. As noted above, the decision
conditions. In some circumstances it may be nec- ladder describes tasks in terms of the stages of acti-
essary to maneuver the ship to disambiguate sen- vation, observation, state recognition, goal selection,
sor data. This movement may in turn make the ship task and procedure selection, and implementation
detectable to enemy forces, thus creating a conflict and explicitly allows shortcuts and non-sequential
between offensive and defensive ship goals. Again, paths through this sequence. Application of this
such conflicts indicate the need for information to method to the description of tasks in the undersea
be displayed to sensor operators as well as to higher- warfare domain indicated that many of the tasks
level commanders, who may need to prioritize these comprised primarily observation and state recogni-
potentially conflicting goals. tion activities (rather than intensive goal selection
More generally, the work domain analyses led to or task-planning activities), thus suggesting that in-
recommendations regarding display areas that formation displays that highlighted potential pat-
supported communication among controllers with terns and supported training or practice with pattern
different areas of responsibility (for example, ship recognition would be valuable.
ERGONOMICS 229
Cross-linked Functional Matrices ergonomics will no doubt continue to increase its fo-
A third form of analysis, cross-linked functional ma- cus on research and methodologies appropriate for
trices, also led directly to display-design recommen- the design of complex human-computer systems. In
dations. As part of the ongoing design effort, systems the early twenty-first century, common themes within
engineers were developing detailed functional de- this work are the identification of system and task
compositions of the functions and tasks that the demands on the operator, concern for overall hu-
ship would be required to perform. These breakdowns man-system effectiveness, and optimal application
were utilized to make recommendations regarding au- of methodologies and models to answer information
tomation and display requirements for each function, requirements with supportive information displays.
as well as to document the cognitive tasks associated
with the function and to make recommendations on Ann M. Bisantz
the contents of workstation display areas that would
support those tasks and functions. For instance, one See also Task Analysis
ship function is to filter tracks, that is, to apply filter-
ing techniques to tracks (unknown contacts picked up
by radar or other sensing systems) and remove FURTHER READING
tracks that are nonthreatening from the display. To fil-
ter tracks successfully, the primary display supporting Annett, J., & Duncan, K. D. (1967). Task analysis and training design.
Occupational Psychology, 41, 211221.
monitoring and supervisory control activities should Bainbridge, L. (1983). Ironies of Automation. Automatica, 19(6),
include an alert or indication that tracks are being 775779.
filtered. The display also must provide access to de- Bennett, K. B., Toms, M. L., & Woods, D. D. (1993). Emergent features
and configural elements: Designing more effective configural dis-
tailed information, including, for example, the filter- plays. Human Computer Interaction, 35, 7197.
ing algorithms that have been employed to determine Billings, D. E. (1997). Aviation automation: The search for a human-
the significance of the track. These requirements, as centered approach. Mahwah, NJ: Lawrence Erlbaum Associates.
well those of ship functions that support the task of Bisantz, A. M., Roth, E. M., Brickman, B., Lin, L., Hettinger, L.,
& McKinney, J. (2003). Integrating cognitive analysis in a large
supervising the classification and identification of scale system design process. International Journal of Human-
tracks, were specified. Finally, when the display Computer Studies, 58, 177206.
needs of all the ships functions had been specified, a Card, S. K., Moran, T. P., & Newell, A. (1983). The psychology of human-
set of workstation display areas and their content computer interaction. Hillsdale, NJ: Lawrence Erlbaum.
Dix, A., Finlay, J., Abowd, G., & Beale, R. (1998). Human-computer
was specified. Overall, eleven display areas were interaction. London: Prentice Hall.
identified; these included the local-area picture, the Endsley, M., & Kaber, D. B. (1999). Level of automation effects on per-
task scheduling and status areas, the tactical picture formance, situation awareness and workload in a dynamic control
task. Ergonomics, 42(3), 462492.
area,the communications area,and the goals and high- Finger, R., & Bisantz, A. M. (2002). Utilizing graphical formats to con-
level constraints area, among others. Interface proto- vey uncertainty in a decision-making task. Theoretical Issues in
types were then implemented and tested,and provided Ergonomics Science, 3(1), 124.
early validation of the utility of the display area con- Gilbreth, F., & Gilbreth, L. (1919). Applied motion study. London:
Sturgis and Walton.
cept. Importantly, this study focused on identifying Gould, J. L., & Lewis, C. (1985). Designing for usability: Key princi-
appropriate information content for the displays,rather ples and what designers think. Communications of the ACM, 28(3),
than on aspects such as interaction style, hardware re- 300311.
quirements, or screen organization or design. Gray, W. D., John, D. E., & Atwood, M. E. (1993). Project Ernestine:
Validating a GOMS analysis for predicting and explaining real-
world task performance. Human Computer Interaction, 8(3),
237309.
The Future Helander, M., Landauer, T. K., & Prabhu, P. V. (1997). Handbook of
human-computer interaction (2nd ed.). Amsterdam: Elsevier
As work continues to become more and more cog- Science-North Holland.
nitive in nature, and as workplaces become more and Hoffman, R. R., Crandall, B., & Shadbolt, N. (1998). Use of the criti-
more computerized, the field of human factors and cal decision method to elicit expert knowledge: A case study in the
methodology of cognitive task analysis. Human Factors, 40(2), is an everyday term with different meanings for dif-
254276. ferent communities of practitioners and researchers,
Hutchins, E. L., Hollan, J. D., & Norman, D. A. (1986). Direct manip-
ulation interfaces. In D. A. Norman & S. W. Draper (Eds.), User cen-
but the fact that different users of the same term may
tered system design (pp. 87124). Hillsdale, NJ: Lawrence Erlbaum. refer to very different phenomena does not seem
Kirschenbaum, S. S., & Arruda, J. E. (1994). Effects of graphic and ver- to be widely recognized. A further difficulty is that
bal probability information on command decision-making. Human although many researchers collect error data, there
Factors, 36(3), 406418.
Lee, J. D., & Moray, N. (1994). Trust, self-confidence, and operators' is no established research tradition to experimen-
adaptation to automation. International Journal of Human- tally manipulate and study error as a phenome-
Computer Studies, 40, 153184. non. In short, the subject of errors is broad, but
Martinez, S. G., Bennett, K. B., & Shattuck, L. (2001). Cognitive sys-
tems engineering analyses for army tactical operations. In
the study of errors is shallow. After reviewing the
Proceedings of the human factors and ergonomics society 44th an- state of the art in human error research, this article
nual meeting (pp. 523526). Santa Monica, CA: Human Factors examines studies of the cognitive processes that lead
and Ergonomics Society. to errors in interactive behavior.
Mitchell, C. M., & Miller, R. A. (1986). A discrete control model of op-
erator function: A methodology for information display design.
IEEE Transactions on Systems, Man, and Cybernetics, SMC-16(3),
343357. Assessment of Current Research
Muir, B. M. (1987). Trust between humans and machines, and the de- Errors may affect productivity, user satisfaction, and
sign of decision aids. International Journal of Man-Machine Studies,
27, 527539. the safety of property and lives. Errors may be diag-
Norman, D. A. (1988). The psychology of everyday things. New York: nosed as training problems, design problems, sys-
Basic Books. tem problems, or organizational problems, and the
Parasuraman, R., Sheridan, T., & Wickens, C. D. (2000). A model for
types and levels of human interaction with automation. IEEE
diagnosis will determine the remediation. Training
Transacctions on Systems, Man, and Cybernetics, 30(3), 286297. problems are remediated by better documentation
Rasmussen, J., Pejtersen, A. M., & Goodstein, L. P. (1994). Cognitive or user training. Design problems are fixed by re-
systems engineering. New York: Wiley and Sons. design. Systems problems require a wider look at the
Roth, E. M., Patterson, E. S., & Mumaw, R. J. (2002). Cognitive
Engineering: Issues in User-centered System Design. In J. J.
complete task environment of the user to determine
Marciniak (Ed.), Encyclopedia of software engineering (2nd ed., pp. incompatibilities or conflicts between the design
163179). New York: Wiley Interscience, John Wiley and Sons. of multiple devices used by the same worker or in
Salvendy, G. (Ed.). Handbook of human factors. New York: Wiley the functions and responsibilities of the worker.
and Sons.
Stammers, R. B., & Shephard, A. (1995). Task analysis. In J. R. Finally, organizational problems (for example, strong
Wilson & E. N. Corlett (Eds.), Evaluation of human work (pp. pressures to reduce costs at the expense of all else,
144168). London: Taylor and Francis. including safety) may require remediations as di-
Taylor, F. W. (1911). The principles of scientific management. New York:
Norton and Company.
verse as the adoption of new procedures, changes in
Vicente, K. J. (1999). Cognitive work analysis. Mahwah, NJ: Erlbaum. organizational structure, or the replacement of top
Vicente, K. J. (2002). Ecological interface design: Progress and chal- management.
lenges. Human Factors, 44(1), 6278. Although legitimate, the breadth of phenome-
Woods, D. D., & Roth, E. M. (1988). Cognitive systems engineering.
In M. Helander (Ed.), Handbook of human computer interaction
non covered by the term human error has
(pp. 335). Amsterdam: Elsevier. tended to get in the way of understanding its nature,
detection, and correction. To some degree, these dif-
ferent meanings have caused communication diffi-
culties and have occasionally resulted in turf
ERRORS IN INTERACTIVE battles in which different communities argue for the
primacy of their level of analysis (for instance,
BEHAVIOR cognitive, systems, or organizational). More gener-
ally, these various meanings muddy the waters be-
Designing interactive systems to reduce error and cause distinctions that are important within one level
increase error detection and recovery is an impor- of analysis are lost or blurred by attempts to cast
tantand often frustratinggoal. Human error all error phenomena within the same framework.
ERRORS IN INTERACTIVE BEHAVIOR 231
To Err Is Technological
CHICAGO (ANS)Humans may err, but computers are look at the human-computer interaction and be more
supposed to be accurate all the time. Except, of course, sensitive to the human side.
they're not. And as humans rely more and more on tech- The second common mistake, which researchers clas-
nology, they have to make allowances for the error factor sified as an error of omission, takes place when a com-
of both the hardware and the people operating it, re- puter fails to detect a mishap and human operators
searchers are finding. miss it too because they haven't run through a manual
In recent studies conducted in flight simulators, pi- checklist.
lots who relied solely on automated decision aidsde- It was an error of omission that led to the crash of a
signed to reduce human erroroften found themselves Korean Air jet in 1983 after being shot down over
the victims of unintended consequences that might have Soviet airspace, Skitka said. The pilot allegedly never dou-
proved deadly in an actual flight. ble-checked the autopilot program to make sure it was
According to University of Illinois at Chicago psy- following the correct flight path. It wasn't, she said.
chologist Linda Skitka, who has been studying the phe- Indeed, in studying anonymous near-accident reports
nomenon with a teammate for five years, people working filed with the airlines by pilots, Skitka found that many
with computerized systems are prone to two kinds of mistakes involved pilots programming the flight com-
errors. puter to do specific tasks but not bothering to check that
First, when they are told by a computer to do a task, it was performing those tasks.
many do it without double-checking the machine's ac- The studies were conducted at the NASA Ames
curacy, despite the fact they've been told the system is not Research Center in California and at the University of
fail-safe. The researchers dubbed this an error of com- Illinois and have left Skitka suspicious of any task that in-
mission. volves highly technical systems that monitor events. That
For example, the test pilots were told to go through includes work in the nuclear energy and shipping indus-
a five-step checklist to determine whether or not an entries and even hospital intensive care units, where mon-
gine was on fire. One of the elements was a computerized itors are relied on for life-and-death decisions, she said.
warning signal. When they received the signal, the pi- Better technical design and operator training are po-
lots all turned off the defective enginewithout running tential solutions, she said. Perhaps the biggest problem is
through the other four steps. that many of the tasks that need to be performed in au-
It turned out that a completely different engine had tomated situations are dull. Those tasks need somehow
been on fire. When asked about their decision, all the to be made more interesting so humans don't go into
pilots said they had run through the entire checklist when autopilot themselves, she said.
in fact they had not. I'm still a fan of automation but now we've intro-
Most of these systems are being designed by engi- duced new possibilities for human error, said Skitka.
neers who think the way to get rid of human error is to (Computers) are never going to be able to be programmed
engineer the human out of the equation, said Skitka.To for every possible contingency. We have to make sure we
some extent, that's right. But to the extent that we still keep that human factor in our equation.
have human operators in the system, we need to take a Source: To err is technological, new research finds. American News
Service, October 5, 2000.
This situation holds even among the communi- one of the most influential thinkers on the topic com-
ties of researchers and practitioners interested in hu- plained, the need for human error data for vari-
man factors and human-computer interaction. As ous purposes has been discussed for decades, yet no
Approaches to the Study of Error

To err is humanand to blame it on a computer is even
more so. in Interactive Behavior
Robert Orben An analysis of errors at the cognitive level avoids nei-
ther the confusion nor the shallowness endemic to
the study of human error. Arguably the dominant
acceptable human error data bank has emerged cognitive account of errors distinguishes among
(Rasmussen 1987, 23). However, even within the hu- knowledge-based, rule-based, and skill-based errors.
man factors and human-computer interaction com- Knowledge-based errors occur when a user lacks the
munities, these various purposes are somewhat requisite knowledgefor example, if the only route
independent and attempts to shoehorn them into you know from your home to work is habitually
one approach may not be the path to progress, but crowded during rush hour, you will undoubtedly
an obstacle to progress. For example, major re- spend a lot of time waiting in traffic. If you do not
views of human errors discuss the cognitive, systems, know an alternative route, then obviously you will
and organizational perspectives on human error. not be able to take it.
Although each perspective is important, it is diffi- Rule-based errors result from learning a mal-rule
cult to know how progress in understanding the roots or applying the correct rule but on the wrong oc-
of error at, for instance, the organizational level, will casion. For example, if you know two routes to work
either provide or be aided by insights into, for in- and you also know that one is the fastest during rush
stance, the errors encountered in the course of hour and the other is the fastest on the off hours,
routine interactive behavior. taking the wrong route becomes a rule-based er-
Although a confusion of level of analysis is a ror. You have the correct knowledge, but you picked
major obstacle to understanding human error, it is the wrong rule.
not the only one. Perhaps equally damaging is the way An error is skill-based when knowledge is avail-
in which errors are collected and classified. Following able, the correct rule is selected, but a slip is made in
a tradition that goes back at least to William James, executing the rule. For example, you intend to take
the most famous error taxonomies simply amass a your rush-hour route to work, but at the critical
large number of naturally occurring slips as reported intersection you take the turn for the route that is
anecdotally by friends, colleagues, the current re- fastest during the off hours.
searchers, and prior researchers. These errors are then The same behavior, e.g., taking the wrong
subjected to a largely informal analysis that sorts the route during rush hour, can result from lack of
errors into the taxonomic categories favored by the knowledge, misapplication of a rule, or a slip. Hence,
researchers. Attempts to compare how any given er- the knowledge-based, rule-based, and slip-based ap-
ror is classified within or between taxonomies brings proach to errors is neither as neat and clean nor as
to mind the complaint that cognitive theory is rad- theory-based as it may first appear. Whether an error
ically underdetermined by data (Newell 1992, 426). is classified as skill-based, rule-based, or knowledge-
Although some of these taxonomies rely on cog- based may depend more on the level of analysis than
nitive theory as the basis of their classifications, all on its ontogeny.
lack the mechanisms to predict errors. Hence, their Unfortunately, the view that errors in routine in-
explanatory power is only post hoc and incidental. teractive behavior are stochastic is reinforced by the
Indeed, a 1997 survey of the literature on errors difficulties of systematically studying such errors.
led the researchers to conclude that errors in routine Indeed, it is almost a tautology to assert that errors
interactive behavior are regarded primarily as the rein routine interactive behavior are rare. This rarity
sult of some stochastic process. Such a view dis- may have encouraged the naturalistic approach in
courages the systematic study of the nature and origin which researchers and their confederates carry
of this class of errors. around notebooks with the goal of noting and
ERRORS IN INTERACTIVE BEHAVIOR 233
recording the occasional error. Naturalistic ap- method and cognitive, perceptual, and movement;
proaches have an important role to play in docu- GOMSgoals, operators, methods, and selection
menting the importance and frequency of error. rules).
However, they have not been particularly productive Despite a strong push from the cognitive HCI
in understanding the cognitive mechanisms that de- community, within the larger cognitive community
termine the nature, detection, and correction of the emphasis on an embodied cognition interacting
errors in interactive behavior. with a task environment to accomplish a task has
been a minority position. Fortunately, its status seems
New Directions to have changed as we now have six approaches to
The rarity of errors requires research programs that embodied cognition and at least two mechanistic ap-
will capture and document errors, not retrospec- proaches capable of modeling the control of inter-
tively, but as they occur. The systematic study of such active behavior. The components of interactive
errors of interactive behavior involves three interre- behavior can be studied by focusing on the mix-
lated paths. The first path entails creating a task ture of cognition, perception, and action that takes
environment designed to elicit a particular type of approximately 13 of a sec to occur. As human ra-
error. The second involves collecting errorful and er- tionality is bounded by limits to working memory,
ror-free behaviors and subjecting both to a fine- attention, and other cognitive functions the exact
grained analysis. A cost of these two approaches is mix of operations depends on the task being per-
that they require collecting vast amounts of cor- formed and the task environment. Understanding
rect behavior to amass a small database of errorful how the task environment influences the mix of op-
behavior. For example, in 2000 cognitive re- erations is the key to understanding human error in
searcher Wayne Gray reported that out of 2,118 goal interactive behavior, as the following four examples
events (either initiating or terminating a goal) only show:
76 or 3.6 percent could be classified as errors. The
third path entails building integrated models of cog- A GOAL STRUCTURE ANALYSIS OF THE NATURE,
nition that predict the full range of behavior, in- DETECTION, AND CORRECTION OF ERRORS In 2000
cluding reaction time, correct performance, and Gray provided a goal structure analysis of errors
errors. made programming a VCR. A cognitive model was
The study of the cognitive mechanisms that pro- written that used the same goal structure as humans
duce errors has been hampered by the long tradi- with the goals and subgoals analyzed down to those
tion in psychology of attempting to understand the that take approximately 1-s to occur (three times
mind by studying each mental function in isola- higher than required for the analysis of embodied
tion. Fortunately, contrasting trends exist. For ex- cognition). This level of analysis allowed second-by-
ample, the pioneering researchers Stewart Card, second comparisons of human behavior with model
Thomas Moran, and Allen Newell are credited with behavior (that is, model tracing).
bringing to HCI the attempt to understand in de- Places in which the model and a given human
tail the involvement of cognitive, perceptual, and on a given trial diverged were considered potential
motor components in the moment-by-moment in- errors. Each potential error was inspected to deter-
teraction a person encounters when working at a mine whether it represented a true error or a failure
computer (Olson and Olson 2003, 493). Indeed, of the model to capture the richness and diversity of
building on this work, the noted HCI investigator human goal structures. True errors were cataloged
Bonnie John developed a task analysis notation that according to the actions that the model would
captures the ways in which embodied cognition have had to take to duplicate the error. This taxon-
(cognitive, perceptual, and action) is responsive to omy avoided the use of more ambiguous terms such
small changes in the task environment. This ap- as knowledge-based,rule-based, and slip-based,
proach is called CPM-GOMS (CPMcritical path or capture errors, description errors, and mode
errors. Thus, model tracing was used to provide a models of cognitive processing is the key to under-
rigorous and objective taxonomy with which to char- standing, detecting, and correcting human errors.
acterize the nature, detection, and correction of
errors.
Applying a Bounded Rationality
A MEMORY ACTIVATION ANALYSIS OF POSTCOMPLETION
ERROR Postcompletion errors are device-specific Framework
errors made after the target task has been accom- Errors are infrequent, but not rare. Their infrequency
plished. A classic postcompletion error is making has discouraged many from studying errors within
copies of a paper but forgetting to remove the orig- the experimental laboratory and may have dis-
inal. Computational modelers Mike Byrne and Susan couraged a rigorous, theory-based approach to
Bovair showed that a model that was sensitive to understanding how cognitive processes interact with
the working memory demands of the task environ- the task environment to produce errors. The natu-
ment could duplicate the pattern of human post- ralistic approach to errors is enticing, but a hundred
completion errors. years of this approach has not yielded much progress.
Although the importance of errors must be judged
LEAST-EFFORT TRADEOFFS BETWEEN KNOWLEDGE by their effect on everyday life, the study of the na-
IN-THE-WORLD AND KNOWLEDGE IN-THE-HEAD ture, detection, and correction of errors must be pur-
Researchers Wayne Gray and Wai-Tat Fu were able sued in the laboratory.
to show an increase of errors in interactive behavior For those concerned with human errors in HCI,
due to least-effort tradeoffs between reliance on a fruitful path is to pursue the errors that emerge
knowledge in-the-world and knowledge in-the-head. from the interaction of embodied cognition with a
Subjects in two conditions of a VCR programming task being performed in a given task environment.
task could acquire show information either by look- This bounded rationality framework focuses on the
ing at a show information window (Free Access) or mixture of cognition, perception, and action that
by moving the mouse and clicking on the gray box takes approximately 1/3 of a sec to occur. The goal
that covered a field of the window (Gray Box). of this work is the creation of powerful theories that
Subjects in a third condition were required to mem- would allow researchers and practitioners to predict
orize the show information before they began pro- the nature and probable occurrence of errors within
gramming (Memory Test). Results showed that the a given task environment.
Gray Box condition made the most errors, followed
by Free Access, and then the Memory Test. The re- Wayne D. Gray
sults were interpreted to mean that the increased per-
ceptual-motor costs of information acquisition led See also Cognitive Walkthrough; User Modeling
the Free Access and Gray Box groups to an increased
reliance on error-prone memory.
FURTHER READING
I N T E G R A T E D M O D E L O F C O G N I T I O N In a 2002 paper
Allwood, C. M. (1984). Error detection processes in statistical prob-
researchers Erik Altmann and Gregory Trafton pro- lem solving. Cognitive Science, 8, 413437.
posed a goal-activation model of how people re- Allwood, C. M., & Bjorhag, C. G. (1990). Novices debugging when
member the states of the world they want to achieve. programming in Pascal. International Journal of Man-Machine
Studies, 33(6), 707724.
In subsequent work, this model was applied to Allwood, C. M., & Bjorhag, C. G. (1991). Training of Pascal novices
yield predictions about the cognitive effects of inter- error handling ability. Acta Psychologica, 78(13), 137150.
ruptions on task performance (for instance, being in- Altmann, E. M., & Trafton, J. G. (2002). Memory for goals: an acti-
vation-based model. Cognitive Science, 26(1), 3983.
terrupted by the phone while writing a paper). For Anderson, J. R., Bothell, D., Byrne, M. D., & Lebiere, C. (2002). An in-
the cognitive level of analysis, this work demonstrates tegrated theory of the mind. Retrieved October 17, 2002, from
that the basic research agenda of producing integrated http://act-r.psy.cmu.edu/papers/403/IntegratedTheory.pdf
ETHICS 235
Anderson, J. R., & Lebiere, C. (Eds.). (1998). Atomic components of Newell, A. (1992). Precis of unified theories of cognition. Behavioral
thought. Hillsdale, NJ: Erlbaum. and Brain Sciences, 15(3), 425437.
Ballard, D. H., Hayhoe, M. M., Pook, P. K., & Rao, R. P. N. (1997). Nooteboom, S. G. (1980). Speaking and unspeaking: Detection and cor-
Deictic codes for the embodiment of cognition. Behavioral and rection of phonological and lexical errors in spontaneous speech. In
Brain Sciences, 20(4), 723742. V.A. Fromkin (Ed.), Errors in linguistic performance: Slips of the tongue,
Benzeev, T. (1995). The nature and origin of rational errors in arith- ear, pen, and hand (pp. 8795). San Francisco: Academic Press.
metic thinking: Induction from examples and prior knowledge. Norman, D. A. (1981). Categorization of action slips. Psychological
Cognitive Science, 19(3), 341376. Review, 88(1), 115.
Berry, D. C. (1993). Slips and errors in learning complex tasks. In G. Ohlsson, S. (1996a). Learning from error and the design of task en-
M. Davies & L. R. H. (Eds.), Memory in everyday life. Advances in vironments. International Journal of Educational Research, 25(5),
psychology (pp. 137159). Amsterdam: North-Holland/Elsevier. 419448.
Byrne, M. D. (2001). ACT-R/PM and menu selection: applying a cog- Ohlsson, S. (1996b). Learning from performance errors. Psychological
nitive architecture to HCI. International Journal of Human- Review, 103(2), 241262.
Computer Studies, 55(1), 4184. Olson, G. M., & Olson, J. S. (2003). Human-computer interaction:
Byrne, M. D., & Bovair, S. (1997). A working memory model of a com- Psychological aspects of the human use of computing. Annual
mon procedural error. Cognitive Science, 21(1), 3161. Review of Psychology, 54, 491516.
Card, S. K., Moran, T. P., & Newell, A. (1983). The psychology of hu- Payne, S. J., & Squibb, H. R. (1990). Algebra mal-rules and cognitive
man-computer interaction. Hillsdale, NJ: Erlbaum. accounts of error. Cognitive Science, 14(3), 445481.
Elkerton, J., & Palmiter, S. L. (1991). Designing help using a GOMS Rasmussen, J. (1987). The definition of human error and a taxonomy
model: An information retrieval evaluation. Human Factors, 33(2), for technical system design. In J. Rasmussen, K. Duncan, & J. Leplat
185204. (Eds.), New technology and human error (pp. 2330). New York:
Gray, W. D. (1995). VCR-as-paradigm: A study and taxonomy of er- Wiley.
rors in an interactive task. In K. Nordby, P. Helmersen, D. J. Gilmore, Reason, J. (1990). Human error. New York: Cambridge University Press.
& S. A. Arnesen (Eds.), Human-Computer InteractionInteract'95 Simon, H. A. (1956). Rational choice and the structure of the envi-
(pp. 265270). New York: Chapman & Hall. ronment. Psychological Review, 63, 129138.
Gray, W. D. (2000). The nature and processing of errors in interactive Trafton, J. G., Altmann, E. M., Brock, D. P., & Mintz, F. E. (2003).
behavior. Cognitive Science, 24(2), 205248. Preparing to resume an interrupted task: Effects of prospective
Gray, W. D., & Boehm-Davis, D. A. (2000). Milliseconds matter: An goal encoding and retrospective rehearsal. International Journal of
introduction to microstrategies and to their use in describing and Human-Computer Studies, 58(5), 583603.
predicting interactive behavior. Journal of Experimental Psychology: VanLehn, K. A. (1990). Mind bugs: The origins of procedural miscon-
Applied, 6(4), 322335. ceptions. Cambridge, MA: MIT Press.
Gray, W. D., & Fu, W.-t. (2001). Ignoring perfect knowledge in-the- Vicente, K. J. (2002). Ecological interface design: Progress and chal-
world for imperfect knowledge in-the-head: Implications of ra- lenges. Human Factors, 44(1), 6278.
tional analysis for interface design. CHI Letters, 3(1), 112119.
Gray, W. D., & Fu, W.-t. (in press). Soft constraints in interactive be-
havior: The case of ignoring perfect knowledge in-the-world for
ETHICS
imperfect knowledge in-the-head. Cognitive Science.
Gray, W. D., John, B. E., & Atwood, M. E. (1993). Project Ernestine:
Validating a GOMS analysis for predicting and explaining
realworld performance. Human-Computer Interaction, 8(3),
237309. Philosophical interest in the ethical implications of
Gray, W. D., Palanque, P., & Patern, F. (1999). Introduction to the the development and application of computer
special issue on: interface issues and designs for safety-critical technology emerged during the 1980s, pioneered by,
interactive systems. ACM Transactions on Computer-Human
Interaction, 6(4), 309310. among others, Terrell Ward Bynum, Deborah
Heckhausen, H., & Beckmann, J. (1990). Intentional action and ac- Johnson, Walter Manerusually credited with coin-
tion slips. Psychological Review, 97(1), 3648. ing the phrase computer ethicsand James Moor.
James, W. (1985). Psychology: The briefer course. Notre Dame, IN:
University of Nortre Dame Press. (Original work published 1892.)
These philosophers and others laid the foundations
John, B. E. (1990). Extensions of GOMS analyses to expert perform- for a field of study that, for a number of years, en-
ance requiring perception of dynamic visual and auditory infor- compassed three central lines of inquiry: (1) ethical
mation. In J. C. Chew & J. Whiteside (Eds.), ACM CHI'90 Conference questions and challenges to social, moral, and po-
on Human Factors in Computing Systems (pp. 107115). New York:
ACM Press.
litical values raised by changes in society and indi-
John, B. E. (1996). TYPIST: A theory of performance in skilled typ- vidual lives, (2) the nature of computer ethics itself,
ing. Human-Computer Interaction, 11(4), 321355. and (3) ethical obligations of professional experts in
Kieras, D. E., & Meyer, D. E. (1997). An overview of the EPIC archi-
computer and information technologies and engi-
tecture for cognition and performance with application to human-
computer interaction. Human-Computer Interaction, 12(4), neering. More recently the field has broadened to in-
391438. clude strands from neighboring disciplines.
Ethics, Values, and the Impacts Computer technology raised questions about at-
tributing moral responsibility for harmful conse-
of Computer and Information quences of action as philosophers and others
Technologies noted the increasing use of computer systems in con-
trol functions, sometimes replacing human con-
Incorporating most of the work in the field, this trollers, sometimes mediating human action,
line of inquiry focuses on the impacts of comput- sometimes automating complex sequences of tasks.
ing and information technologies that raise ethical Ethical concerns went hand in hand with technical
questions as well as questions about moral, political, concerns. Where computer scientists and engi-
and social values in societies and in individuals lives. neers worried about correctness, reliability, safety
Many of the issues that emerged early on, such as in- and dependability, philosophers asked whether in-
tellectual property, responsibility, crime, privacy, au- creasing reliance on computer-controlled automa-
tonomy, free speech, and quality of life, have remained tion is warranted and whether, secondarily, it leads
important and have evolved alongside developments to a diminishment of accountability for malfunc-
in the technologies themselves. Philosophers engaged tions, dangers, and harms due to computerization.
in the study of impacts have approached their sub- Another intriguing line of questions was taken up
ject from at least two perspectives. In one they have by philosophers such as Kari Coleman, Arthur Kuflik,
asked about the nature of moral obligations in light James Moor, and John Snapper. This line concerned
of particular changes, thus being concerned with right responsibility and was spurred by actual and pre-
and wrong actions of people. In the other they have dicted advances in artificial intelligence. It asked
been concerned with the status of particular values whether aspects of human agency, such as life-
in society and how these are affected by technol- and-death decisions, should ever be delegated to
ogy-induced changes. computers no matter what the relative competency
In the case of intellectual property, philosophi- levels. A twist in this line of questions is whether a
cal interest focused on moral obligations owed to the time will come when humans will have moral obli-
creators and owners of software. Philosophers, like gations to intelligent machines.
their colleagues in law, recognized key metaphysical An issue related to that of responsibility is the
(relating to a branch of philosophy that is concerned nature and severity of computer crime and the va-
with the fundamental nature of reality and being) riety of harms wrought on others in the context of
differences between computer software and tradi- computer-mediated communications and transac-
tional forms of intellectual property and sought to tions. Philosophers participated in early debates over
understand whether and in what ways these differ- whether actions such as gaining unauthorized access
ences affect the extent and nature of property pro- to computer systems and networks should be judged
tection that software deserves. By the mid-1990s and as crimes or whether such judgment should be re-
into the present, as the Internet and World Wide Web served for cases where clear damage results, as in the
developed and increased in popularity, most of the cases of transmitting computer viruses and worms
attention given to intellectual property has been fo- and posting obscene or threatening materials.
cused on controversial questions concerning digital Privacy has been one of the most enduring issues
representations of a wide range of intellectual and in this category. Philosophers have focused attention
cultural works (including text, images, music, and on privacy as a social, political, and individual
video), peer-to-peer file sharing, and even Web-link- value threatened by developments and applications
ing (the use of Web hyperlinks to move from one of computer and information technologies.
web page to another). From the perspective of val- Philosophers have participated in the chorus of voices,
ues, philosophers have questioned social and legal which also includes scholars of law, policy, and social
decisions that have shaped the relative strength science and privacy advocates, that has denounced
and standing of intellectual property in the face of many of these developments and applications as dan-
other values, such as freedom to share. gerously erosive of privacy. As with other issues, the
ETHICS 237
nature of the activities that raise concern shifts and digital networking technologies. Many re-
through time as a result of evolving technologies and searchers have pointed to the enormous positive po-
their applications. The earliest applications to take tential of collaborating online and building
the limelight were large government and corporate community and accessing vast troves of informa-
databases. Cries of Big Brother resulted in various tion. However, some philosophers have asked whether
legal constraints, including, most importantly, the the intrusion of digital technologies debases these
U.S. Privacy Act of 1974. Through time dramatic re- spheres of lifereplacing the actual with the virtual,
ductions in the cost of hardware and improve- replacing face-to-face communication with medi-
ments in the capacities to collect, store, communicate, ated communication, replacing family and intimate
retrieve, analyze, manipulate, aggregate, match, and interactions with chat rooms and online games, and
mine data led to a proliferation in information gath- replacing human teachers and mentors with com-
ering throughout most sectors of society and an am- puterized instructionand deprives them of their
plification of early concerns. In parallel with these essentially human character and consequently de-
developments, we experienced an upsurge in identi- prives us of meaningful opportunities for emotional,
fication and surveillance technologies, from video spiritual, and social growth. The influence of
surveillance cameras to biometric (relating to the sta- Continental philosophers, including Edmund Husserl
tistical analysis of biological observations and phe- and Emmanuel Levinas, is more apparent here than
nomena) identification to techniques (such as Web in previously mentioned areas where Anglo-
cookies) that monitor online activities. Each of these American, analytical thought tends to dominate.
developments has attracted concern of a broad
constituency of scholars, practitioners, and activists
who have applied their areas of knowledge to par- Metaethics of Computer and
ticular dimensions of the developments. Philosophers,
such as Judith DeCew, Jeroen van den Hoven, Information Technology
James Moor, Anton Vedder, and Helen Nissenbaum, Many philosophers leading the inquiry of ethics and
have taken up particularly two challenges: (1) im- information technology have raised questions about
proving conceptual understanding of privacy and the the nature of the inquiry itself, asking whether
right to privacy and (2) refining theoretical under- anything is unique, or uniquely interesting, about
pinnings and providing a systematic rationale for pro- the moral and political issues raised by information
tecting the right to privacy. technology. The continuum of responses is fairly
Finally, a category of questions concerning qual- clear, from a view that nothing is philosophically
ity of life asks, more generally, about the ways com- unique about the issues to the view that settings and
p u t e r a n d i n f o r m a t i o n t e c h n o l o g i e s h ave capacities generated by computer and information
impinged on core human values. We could include technologies are so novel and so distinctive that they
in this category a variety of concerns, starting with demand new theoretical approaches to ethics. The
the digital dividethe possibility that computer tech- more conservative approaches assume that we can
nology has increased the socioeconomic gap between reduce the problems in computer ethics (for ex-
those groups of people with power and wealth and ample, any of those mentioned earlier) to the
historically disadvantaged socialeconomic, racial, more familiar terms of ethics and applied ethics, gen-
and gender groups. Such questions concerning so- erally. From there the problems are accessible to stan-
cial justice within societies have been extended to dard ethical theories. For example, although
the global sphere and the vastly different levels of ac- transmitting computer viruses is a novel phenom-
cess available in countries around the globe. enon, after we cast it as simply a new form of harm-
Another element in the category of quality of ing others property, it can be treated in those familiar
life concerns the impacts on relationships, such as terms. Other philosophers, such as Luciano
those among friends, romantic partners, family mem- Floridi, have suggested that because these technolo-
bers, and teachers and students, made by computers gies create new forms of agency or new loci of value
itself, new ethical theories are required to resolve and Electronicshave developed codes of profes-
problems. James Moor, in an essay entitled What Is sional ethics. Two issues remain controversial. One
Computer Ethics?, offers something in between. issue deals with the nature and limits of professional
Computer ethics deserves attention because it raises codes. The philosopher Michael Davis has provided
not only policy questions that are new (such as a thoughtful account of the role of codes of conduct
Should we allow computer programs to be privately in encouraging ethical professional practice, in con-
owned?) but also novel conceptual questions about trast to John Ladd, who has challenged the very pos-
the very nature of a computer program, whether sibility that professional codes of conduct can rightly
more like an idea, a process, or a piece of writing. be thought of as codes of ethics. The other issue, spe-
These conceptual puzzles, particularly acute in the cific to the professions within computer technolo-
case of privacy, explain why we continue to struggle gies, asks whether they are sufficiently similar to
to resolve so many of the controversial questions that traditional professions of law and medicine to
privacy raises. warrant the label of professions.
Computer Ethics as Porous Borders

Although the philosophical community pursuing
Professional Ethics inquiry into ethical implications of information
Some contributors to the field of computer ethics technology remains relatively small, its intellectual
have seen its greatest potential as a guide for com- borders are fluid. Since the decades of its emergence,
puter scientists, engineers, and other experts in the it has been enriched by developments in the liter-
technologies of computing and information, thus atures and methods of neighboring fields. In turn,
placing it in the general area of professional ethics. many of the works produced within those fields have
Early proponents of this idea, Donald Gotterbarn been influenced by the work of ethicists. A few ex-
and Keith Miller, added their voices to those of so- amples, where cross-disciplinary flow has been par-
cially concerned computer scientists and engineers ticularly active, bear mentioning. One example is
whostarting with Norbert Wiener and Joseph information law, which emerged into prominence
Weizenbaum, followed by Terry Winograd, Peter roughly a decade after philosophical issues of pri-
Neumann, and Alan Borningexhorted their col- vacy, intellectual property, free speech, and gover-
leagues to participate actively in steering social de- nance spurred many of its core works by legal
liberation, decision, and investment toward socially, scholars such as Lawrence Lessig, Yochai Benkler,
politically, and morally positive ends and also to warn James Boyle, Pamela Samuelson, Jerry Kang, and
of dangers and possible misuse of the powerful tech- Niva Elkin-Koren. As a result of these works, philo-
nologies of computation and information. In this sophical studies have paid greater attention to is-
area, as in other areas of professional ethics, such sues of public values, the direct effects of policy
as legal and medical ethics, key questions included on values, and the meaning for society of key
the duties accruing to computer scientists and engi- court rulings.
neers as a consequence of their specialized knowl- A second prominent influence has come from
edge and training. In the area of system reliability, the areas of philosophy and social study of science
for example, computer engineers such as Nancy and technology where theoretical writings and
Leveson have focused enormous energies to artic- empirical studies of scholars such as Langdon
ulate the duty to produce, above all, safe systems, Winner, Albert Borgman, Bruno Latour, Wiebe Bijker,
particularly in life-critical areas. Andrew Feenberg, and Donald MacKenzie have in-
Responding to calls for greater focus on pro- spired novel approaches to many of the substan-
fessional duties, at least two major professional or- tive issues of ethics and information technology. Ideas
ganizationsthe Association for Computing such as the social shaping of technical systems and
Machinery and the Institute of Electrical Engineering the values embodied in system design have focused
ETHNOGRAPHY 239
philosophical attention on design and development

details of specific systems and devices, opening a line FURTHER READING
of work that views the design of systems and devices
not as a given but rather as a dependent variable. Adam, A. (2002). Cyberstalking and Internet pornography: Gender
and the gaze. Ethics and Information Technology, 2(2), 133142.
Although influenced by such ideas, ethicists approach Brey, P. (1997). Philosophy of technology meets social constructivism.
them with a different goal, not only seeking to de- Techne: Journal of the Society for Philosophy and Technology, 2(34).
scribe but also to evaluate systems in terms of moral, Retrieved March 24, 2004, from http://scholar.lib.vt.edu/ejour-
political, and social values. Philosophers who have nals/SPT/v2n3n4/brey.html
Bynum, T. W. (2001). Computer ethics: Its birth and its future.
pursued these lines of inquiry include Deborah Ethics and Information Technology, 3(2), 109112.
Johnson, Jeroen van den Hoven, and Philip Brey, who Dreyfus, H. L. (1999). Anonymity versus commitment: The dangers
has interpreted many key works in social construc- of education on the Internet. Ethics and Information Technology,
1(1), 1521.
tivism (an approach to the social and humanistic Elkin-Koren, N. (1996). Cyberlaw and social change: A democratic
study of technology that cites social factors as the approach to copyright law in cyberspace. Cardozo Arts &
primary determinants of technical development) for Entertainment Law Journal, 14(2), 215end.
philosophical audiences and developed a concept of Floridi, L. (1999). Information ethics: On the philosophical founda-
tions of computer ethics. Ethics and Information Technology, 1(1),
disclosive ethics (a philosophical approach which 3756.
holds that system design may disclose ethical im- Gotterbarn, D. (1995). Computer ethics: Responsibility regained. In
plications). D. G. Johnson & H. Nissenbaum (Eds.), Computers, ethics, and so-
Interest in design as a dependent variable has cial values (pp. 1824). Englewood Cliffs, NJ: Prentice Hall.
Grodzinsky, F. S. (1999). The practitioner from within: Revisiting the
also led to collaborations among philosophers, com- virtues. Computers and Society, 29(1), 915.
puter scientists, and researchers and designers of Introna, L. D. (2001). Virtuality and morality: On (not) being dis-
human-computer interfaces who have been inspired turbed by the other. Philosophy in the Contemporary World, 8(1),
3139.
by the complex interplay between computer systems Introna, L. D., & Nissenbaum, H. (2000). Shaping the Web: Why the
and human values. These collaborations are im- politics of search engines matters. The Information Society,
portant test beds of the idea that a rich evaluation 16(3), 169185.
of technology can benefit from simultaneous con- Johnson, D. G. (2001). Computer ethics (3rd ed.). Upper Saddle River,
NJ: Prentice Hall.
sideration of several dimensions: not only techni- Johnson, D. G. (2004). Computer ethics. In H. Nissenbaum & M. E.
cal design, for example, but also, ideally, empirical Price (Eds.), Academy and the Internet (pp. 143167). New York:
effects on people and an understanding of the val- Peter Lang.
ues involved. For example, Lucas Introna and Helen Moor, J. H. (1985). What is computer ethics? Metaphilosophy, 16(4),
266275.
Nissenbaum studied a search-engine design from Nissenbaum, H. (2004). Privacy as contextual integrity. Washington
the point of view of social and political values. Their Law Review, 79, 119158.
study tried to achieve an in-depth grasp of the work- Spinello, R. A., & Tavani, H. T. (Eds.). (2001). Readings in cy-
berethics. Sudbury, MA: Jones and Bartlett.
ings of the search engines system, which they van den Hoven, J. (1994). Towards principles for designing politi-
evaluated in terms of fairness, equality of access to cal-administrative information systems. Van den Hoven, J. (1994).
the Web, and distribution of political power within Towards principles for designing political-administrative infor-
the new medium of the Web. Other researchers who mation systems. Information and Public Sector, 3, 353373.
reach across disciplines include Jean Camp and
Lorrie Cranor, as well as Batya Friedman, Peter Kahn,
and Alan Borning, who have developed value-sen-
sitive design as a methodology for developing com- ETHNOGRAPHY
puter systems that take multiple factors into
consideration. Ethnography has several meanings. It means the
study and recording of human culture; it can also
Helen Nissenbaum mean the work produced as a result of that
studya picture of a people. Today, however, some-
See also Law and HCI; Privacy; Value Sensitive Design one interested in human-computer interaction (HCI)
is more likely to encounter ethnography as a term At the same time, ethnographers generally ac-
that describes a research method. As a research cept the existence of a reality external to the con-
method, ethnography is widely used in a multi- structs of the knower. In accepting this external
tude of ways, but unfortunately also misused. reality, and therefore also acknowledging the value
of talk regarding what is knowable about it, ethnog-
raphy differs from, for example, postmodernism,
Principles of Ethnography which tends to limit discussion to human appre-
Ethnography is most properly understood as a re- hensions rather than the effect of these apprehen-
search methodology. This methodology, which can s i o n s o n t h e wor l d . In te l l e c t u a l l y b e t we e n
be described as participant observation, or, more in- postmodernism and positivism, ethnography is an
formally, fieldwork, is rooted in the social science of empirical research strategy that strives to account
anthropology. Anthropology is the discipline that at- satisfactorily for the dynamics of different cul-
tempts to develop a holistic account of everything tures, or of human culture in general, without try-
having to do with human beings and their activi- ing to construct transcendent laws of human action.
tieshumankind in space and time.
Ethnography developed specifically as the chief
approach of cultural anthropology, one branch of History of Ethnography
anthropology. Cultural and social anthropologists Historically, ethnography emerged out of late-nine-
use long-term participation and observation to de- teenth-century British anthropology. Historians of
velop deeply contextualized accounts of contempo- anthropology credit the British anthropologist W.
rary ways of life. This distinguishes them from, for H. R. Rivers (18641922) as the first full practitioner
example, archaeological anthropologists, who aim of ethnography. Before Rivers, anthropologists gen-
to construct similarly holistic understandings of past erally relied entirely on data collected by missionar-
cultures. ies or colonial officials; at most, they had only brief
What is distinctive about ethnography is its par- personal experience of the cultures they analyzed.
ticular way of knowing, what a philosopher would Rivers stressed the importance of long-term, depth
call its epistemological approach. (Epistemology is exposure to the culture of interest, so that the con-
the philosophical study of how we come to know texts of cultural facts such as kin terms or magical
what we know, or how we justify our belief that we practices could be more fully grasped.
know.) Ethnographers pay as much attention to the The Polish expatriate Franz Boas (18581942)
ways in which people (including ethnographers brought ethnography to anthropology in the United
themselves) perceive what they know as they do to States, both by his own fieldwork and through teach-
how people act because ethnographers believe that ing his students, notably Ruth Benedict (18871948)
the ways in which people act are affected by their cul- and Margaret Mead (19011978). Partly because
tural milieu. Even when focused on actions, ethno- of the great variety characteristic of the Native
graphers pay close attention to what those actions American cultures that were its principle foci, U.S.
reveal about what people take as known. That is, ethnography came to emphasis the particularity of
ethnographic methodology accepts the impor- each culture.
tance of cultural construction to both ideas and The anthropologist usually credited with in-
action. Further, ethnography attempts to incorpo- venting ethnography, however, is the Polish no-
rate this acceptance into both the generation and bleman Branislaw Malinowski (18841942). His long
representation of anthropological knowledge. This 1925 introduction to Argonauts of the Western Pacific
approach differs from modernist or positivist apis still presented as the classic statement of ethnog-
proaches to knowledge, such as those informing sta- raphy. Of particular note in Malinowskis ap-
tistical reasoning, which attempt to isolate facts proach to ethnography was his stress on the need for
from their knowers and from knowers culturally de- the ethnographer to develop an emotional de-
pendent acts of knowing. pendency upon the native informants. Only if the
ETHNOGRAPHY 241
ethnographer were cut off from regular contact with ethnographic representation (including more lit-
his or her culture of origin could he or she hope to erary forms, such as poetry), they also fed directly
develop the insider perspective still taken as the into subaltern studiesthe study of oppressed peo-
hallmark of good ethnographic writing or film. ples. The trend, part of the critique of anthropol-
Under Malinowskis influence, the participa- ogys role in colonialism, toward privileging the
tive aspect of ethnographic observation developed natives cultural understanding over the outsiders
a particular quality. Even though the ethnographer cultural understanding, is influential in contempo-
knows that his or her knowledge of the culture is not rary cultural studies.
yet complete, he or she tries to participate in cultural Despite the internal anthropological critiques,
activity, performing an indigenous role as well as the cachet of ethnography grew considerably during
possible. The hope is that informants will critique the 1980s in other fields. It was frequently featured
ethnographers performance and thereby accelerate as the methodology of preference in feminist cri-
the pace of the ethnographers learning. This very tiques of social science and was drawn on by those
active form of research differs markedly from the ap- advocating field studies in social psychology. Today,
proach of social scientists focused on minimizing most cultural anthropologists and many other schol-
their impact on the society under observation. ars continue to practice ethnography, drawing on
Ethnographic perspectives blended with sociol- this long and rich tradition of practice and critique.
ogys tradition of field research, especially in the
founding of new subfields such as community stud-
ies. After World War II it became popular to model Ethnography Meets Human-
sociological studies as closely as possible on labo-
ratory science, but nonetheless ethnography main- Computer Interaction
tained a vigorous and continuous presence in In research on computing, ethnography has been as
important theoretical and empirical fields. For ex- much a basis for critique as a methodology. The con-
ample, ethnographic studies were a key part of the structs of early HCIfor example, the notion of the
1960s displacement of industrial sociology (with its (individual) man [sic](individual) machine in-
focus on bureaucracy), by the sociology of work terfacewere critiqued, and social scientists began
(which focused on the actual labor process). undertaking fuller, contextual ethnographic studies
Sociological ethnography did not retain all aspects of computing. In 1991, for example, the anthro-
of anthropological ethnographys epistemology, how- pologist David Hakken argued that, instead of striv-
ever. For example, Malinoswki claimed that an ing after human-centered computing, a better goal
ethnographers position as an outsider gave special would be culture-centered. That is, rather than try-
insights, as he or she would be able to see aspects ing to design systems to meet universal character-
of a culture less visible to insiders, who took the cul- istics, they should be oriented toward specific cultural
ture for granted. Sociological ethnographers who frameworks.
were studying their own culture obviously could not Field study of actual computing encouraged some
claim to have an outsiders insights. computer scientists to broaden their conception of
Interestingly, while Malinowski stressed how cul- the nature of interfaces by developing systems for
ture influenced the ways of knowing of the targets computer-supported collaborative (or collective)
of ethnographic investigation, he believed the ethno- work. Terry Winograd and Fernando Floress 1986
graphic knowledge of the professional anthropolo- theoretical critique of positivism in computer sci-
gist to be fully scientific in the positivist sense. He ence (Understanding Computers and Cognition: A
did not in general address the cultural biases and as- New Foundation for Design) drew on ethnography.
sumptions the anthropologist brought to the field. Geoffrey Bowker is among the many computer-
Such contradictions made ethnography vulnerable science ethnographers who now refer to the an-
to critiques in the 1980s. Not only did those critiques tiformalist trends in computer science as social
spawn an interest in new, experimental forms of informatics.
Anthropological ethnographers such as Henrick intelligence: Beginning in the mid-1980s, the former
Sinding-Larsen and Lucy Suchman were central to moved from being an object of derision for those
perhaps the single most influential tradition of so- trained in rigorous natural science to a privileged
cial informatics, variously referred to in Europe as technique, albeit distorted from its anthropologi-
the Scandinavian approach or as user participation cal form. In the process of domesticating ethnog-
in systems development and called participatory de- raphy for informatics, some informaticians turned
sign in the United States. For a variety of practical as themselves into self-trained ethnographers, while
well as political reasons, Nordic systems developers other social scientists (e.g., Crabtree) developed a
wanted to broaden the influence of users over sys- quick and dirty ethnography. To build effective ob-
tems development, and ethnography seemed like a ject-oriented as opposed to relational databases, one
good way to gain entry into the users world. Kristen needs a good understanding of the notional things
Nygaard, arguably the computer scientist with the relevant to a work process. After some initial ef-
most sustained influence on this trend, under- forts to identify these themselves, informaticians
stood the potential value of ethnography. He returned the task over to social scientists who, after a
cruited Sinding-Larsen onto an Oslo hospital project week or so hanging out in a work site, could gen-
called Florence, one of the three projects recognized erate a list of apparently relevant notions. Such lists,
as foundational to the Scandinavian approach. however, are likely to lack the depth of interrelational
T h ro u g h t h e i r p a r t i c i p a t i o n i n t h e a n nu a l and contextual understanding that would come from
ksyen and IRIS (Information Research in Scan- longer, more intense participant observation.
dinavia) and decennial Computers in Context In quick and dirty appropriations of fieldwork,
conferences, ethnographers like Suchman, Jeanette ethnography ceases to be an epistemology and is re-
Blomberg, and Julian Orr brought ethnography into duced to a technique, one of several qualitative re-
a continuing dialogue with Nordic and, later, U.S. search tools. In the late 1990s, before her untimely
systems development. death, Forsythe despaired of these developments. The
From its inception in projects such as Florence, computer scientist Jonas Lwgren and the cultural
however, the relationship between ethnography and anthropologist James Nyce have criticized HCI re-
HCI studies has been complex. The relationship has searchers interested in ethnography for wanting to do
spawned a wide variety of approaches as well as mis- ethnography but only managing an ethnographic gaze.
understandings. For example, the frequent failure of This gaze, not ethnography in the full sense, has been
a computerized system to perform in the manner in- incorporated into practices as diverse as program eval-
tended might be a consequence of either its design uation and educational research, with uneven conse-
or of something in the context of its use. To figure quences. When their limitations are understood, the
out which, one has to investigate the use context. One various appropriations of the ethnographic gaze
could do this by making the laboratory more like the can be of substantial value, but it should not be
use context (one approach to usability), or one confused with ethnography in the full sense.
could examine how the system is actually used in the It is ironic that just when ethnography was un-
real world, through ethnographic studies of use. der attack in its home base of cultural anthropology,
Because usability studies and use studies sound its general popularity as a way of knowing was
similar and have some things in common, they were spreading very widely. Too often, however, re-
sometimes glossed together, even though they have search self-labeled as ethnographic violates one or
very different epistemological underpinnings. What more of the epistemological premises at the core of
one learned, or concluded could not be learned, de- anthropological ethnography. As a result, the term
pended upon training and professionally preferred is used to cover such a broad array of approaches
reading style (Allwood and Hakken 2001). The med- as to have lost some of its meaning.
ical anthropologist Diana Forsythe has chronicled Its methods are often qualitative, but ethnog-
how a different misunderstanding emerged from the raphy is not just qualitative methods. Indeed, ethno-
bizarre courting dance of ethnography and artificial graphers often also deploy quantitative methods in
ETHNOGRAPHY 243
their data collection and invoke numbers in their graphic gaze. In evaluating especially these latter, it
analyses. Good ethnography integrates various kinds is useful to keep in mind that, to the cultural an-
of information, but particularly information derived thropologist, ethnography is more than an array of
from active participation, which is at the center of methods; it is a way of knowing.
the idea of ethnography.
David Hakken
Current Ethnography of HCI See also Anthropology and HCI; Sociology and HCI
Fortunately, today there is a rich body of good ethno-
graphic HCI research. One type of study focuses
specifically on the work of people developing com- FURTHER READING
puter systems. Forsythes 2001 Studying Those Who
Study Us: An Anthropologist in the World of Artificial Allwood, C. M., & Hakken, D. (2001). Deconstructing use: Diverse
Intelligence includes several ethnographic studies of discourses on users and usability in information system devel-
computer scientists in the process of developing opment and reconstructing a viable use discourse. AI & Society,
15, 169199.
artificial intelligence and knowledge-engineering sys- Blank, J. (2001). Mullahs on the mainframe: Islam and modernity among
tems. The cultural anthropologist Gary Downey uses the Daudi Borhas. Chicago: University of Chicago Press.
ethnography to study the education of computer en- Blomberg, J. (1998). Knowledge discourses and document practices:
gineers, while the linguist and computer specialist Negotiating meaning in organizational settings. Paper presented
at the annual meeting of the American Anthropological Association,
Stephen Helmreich focuses on computer scientists Philadelphia, PA.
who see themselves as developing artificial forms of Bowker, G., Star, S. L., Turner, W., & Gasser, L. (1997). Introduction.
life in code. Hakkens Cyborgs@cyberspace? reports In G. Bowker, S. L. Star, W. Turner, & L. Glasser (Eds.), Social sci-
ence, technical systems, and cooperative work: Beyond the great di-
on the results of his ethnographic study of Nordic vide (pp. xxxiii). Mahwah, NJ: Lawrence Erlbaum Associates.
systems developers. Clifford, J., & Marcus, G. (Eds.). (1986). Writing culture: The poetics
Another body of HCI ethnography looks at com- and politics of ethnography. Berkeley and Los Angeles: University
puting among actual users in the general population. of California Press.
Crabtree, A. (1998). Ethnography in participatory design. Paper pre-
The sociologist Susan Leigh Stars The Cultures of sented at the Participatory Design Conference, Seattle, WA.
Computing (1995) contains several good examples. Downey, G. (1998). The machine in me: An anthropologist sits among
Hakkens The Knowledge Landscapes of Cyberspace computer engineers. New York: Routledge.
(2003) deals with knowledge management in com- Ehn, P. (1988). Work-oriented design of computer artifacts. Stockholm:
Almqvist & Wiksell.
mercial organizations, social services, and schools. Forsythe, D. (2001). Studying those who study us: An anthropologist in
A new group of design anthropologists are trying to the world of artificial intelligence. Palo Alto, CA: Stanford University
put this knowledge to work in product development. Press.
Hakken, D. (1991). Culture-centered computing: Social policy and
A final body of HCI ethnography reaches back development of new information technology in England and the
to an earlier tradition by placing computing prac- United States. Human Organization, 50(4), 406423.
tices in broad social contexts. Computing Myths, Class Hakken, D. (1999). Cyborgs@cyberspace?: An ethnographer looks to the
Realities (1993) is one example, as is The Internet: An future. New York: Routledge.
Hakken, D. (2003). The knowledge landscapes of cyberspace. New York:
Ethnographic Approach (1999), which examines Routledge.
Internet use in Trinidad. As computing becomes Hakken, D., & Andrews, B. (1993). Computing myths, class realities.
more densely integrated into non-Western social for- Boulder, CO: Westview Press.
mations, studies such as Jonah Blanks Mullahs on Helmreich, S. (1999). Silicon second nature: Culturing artificial life in
a digital world. Berkeley and Los Angeles: University of California
the Mainframe (2001) will provide well-considered Press.
insights to those who seek to integrate technical and Miller, D., & Slater, D. (1999). The Internet: An ethnographic approach.
economic development work. Oxford, UK: Berg.
Nyce, J., & Lwgren, J. (1995). Toward foundational analysis in hu-
In sum, there is reason to expect continuing man-computer interaction. In P. J. Thomas (Ed.), The social and
expansion of rich ethnography in the study of interactional dimensions of human-computer interfaces (pp. 3746).
HCI, as well as new appropriations of the ethno- Cambridge, UK: Cambridge University Press
Orr, J. (1996). Talking about machines: An ethnography of a modern Technological Evolution

job. Ithaca, NY: Cornell University Press. People like to think that new technology is invented
Sinding-Larsen, H (1987). Information technology and management
of knowledge. AI & Society, 1, 93101.
by heroic geniuses such as Leonardo da Vinci and
Star, S. L. (Ed.). (1995). The cultures of computing. Oxford, UK: Thomas Alva Edison, but many scholars argue that
Blackwell. innovation is more like a process of biological evo-
Suchman, L. (1987). Plans and situated actions. Cambridge, UK: lution in which individuals play only minor roles. In
Cambridge University Press.
Van Mannen, J. (1983). The fact of fiction in organizational ethnog- his classic 1922 study of social change, the sociolo-
raphy. In J. Van Maanen (Ed.), Qualitative methodology (pp. 3755). gist William F. Ogburn acknowledged that inventors
Newbury Park, CA: Sage. tend to be more intelligent than the average. But, he
Whyte, W. F. (1991). Participatory action research. Newbury Park, CA:
Sage.
explained, inventions are collective additions to an
Winograd, T., & Flores, F. (1986). Understanding computers and cog- existing cultural base that cannot occur unless the
nition: A new foundation for design. Norwood, NJ: Ablex. society has already gained a certain level of expert-
ise in the particular area. For example, the telegraph
was created by combining many existing elements,
such as electricity, coils, batteries, signaling, and al-
EVOLUTIONARY phabet codes. A personal computer connected to the
World Wide Web combines a television set (moni-
ENGINEERING tor), telephone (Internet connection), and typewriter
(keyboard), plus a myriad of small electronic and
How can people make computers solve problems programming innovations, each of which has its own
without having to program them with every detail heritage reaching back in time to earlier inventions
needed for a solution? One answer is to endow the such as the telegraph and the electronic calculator.
computers with some form of artificial intelligence Ogburn said that technical evolution takes place
that allows them to make judgments independently. in four interrelated steps: invention, accumulation,
Another answer makes computers assist humans diffusion, and adjustment. He devoted an entire chap-
in breeding solutions rather the way farmers breed ter of his influential book to a long list of inventions
animals and plants. This is called genetic program- and discoveries that were made independently by
ming, or evolutionary programming, and it opens two or more people, illustrating the fundamental
the possibility of very different relationships between role of culture in preparing the way for such inno-
people and their machines. vations. These inventions accumulate, such that an
Since the dawn of the agricultural age thou- advancing culture possesses more of them each year,
sands of years ago, people have understood that and they diffuse from their original application and
artificial selection can gradually transform pop- geographic area to others. Finally, society must ad-
ulations of living creatures to make them more use- just to new technologies, often by innovating in other
ful for human beings. For example, people who fields, as happened in recent history when the spread
herd cattle might slaughter the cows that give lit- of the Web made new kinds of e-businesses possi-
tle milk, while allowing the best dairy cows to sur- ble, which in turn required both programming
vive and have offspring. Over generations, this will and economic innovations.
produce cows that give more milk. Genesis 3031 From this perspective, invention is very much like
reports that Jacob fully understood how to breed genetic mutation in biology. Many mutations are dis-
sheep and goats for desirable characteristics in bib- advantageous and die out, and the same is true for
lical times. In the mid-nineteenth century, Charles inventions. From the vast menagerie of inventions
Darwin developed the concept of natural selection, produced, only some survive, spread, and combine
which has become a key analytic principle in biol- with other successful inventions in a manner anal-
ogy. In the twentieth century, historians, social sci- ogous to sexual reproduction, producing offspring
entists, and computer scientists began to apply that in turn may combine with other inventions if
evolutionary ideas to technology. they survive in the harsh environment of the market-
EVOLUTIONARY ENGINEERING 245
place. Like living creatures, technological innovations good paths through the maze, perhaps even many
inherit characteristics from their predecessors, ex- copies of the one best solution.
hibit great diversity caused by both mutation and re- Selection, reproduction, and mutation are
combination, and undergo selection that allows some enough for evolution to take place as it does in mi-
to survive while others become extinct. croorganisms, but humans and many other complex
life forms have one more important process: sexu-
ality. From an evolutionary perspective, sex has
Genetic Programming one great advantage. It combines genes from dif-
In 1975 the computer scientist John Holland offered ferent lineages.
an evolutionary approach to programming that he Imagine that you are breeding dairy cows. Per-
called "genetic plans," but that today is known as ge- haps Bossie the cow produces lots of milk but is mean
netic algorithms. In biology, the genetic code is carried and kicks you whenever you approach. Bertha is
by DNA molecules, which are long strings of nucleotide sweet tempered but produces little milk. Assuming
bases denoted by four letters of the alphabet: A (ade- these characteristics are genetically determined, you
nine), C (cytosine), G (guanine), and T (thymine). By let both Bossie and Bertha have many babies. Then
analogy, genetic algorithms typically employ strings you breed the offspring of one with the offspring
of letters or numbers that can be read by the soft- of the other, hoping to get a variety of mixtures of
ware system as if they were short programs specifying their characteristics. Perhaps one of Bossie's and
a series of actions. For example, URDL might instruct Bertha's granddaughters, Bessie, combines their
the cursor of the computer to go through a series of virtues of abundant milk and sweet disposition. She
moves on the screen: up, right, down, and left. becomes the mother of your herd of ideal dairy cows.
A genetic algorithm system contains many of Another granddaughter, Bortha, is mean and un-
these strings, perhaps thousands of them. Each can productive. She becomes roast beef.
be interpreted as different instructions for solving The equivalent of sexual reproduction in genetic
the same well-specified problem. The URDL code algorithms is called crossover. After checking the fit-
might be instructions for moving the cursor through ness of each string in the system, the program makes
a maze on the computer screen. (URDL would not some new strings by adding part of one high-fitness
be a very good solution to the maze problem, how- string to part of another. In the maze example, UUDUU
ever, because it merely returns the cursor to where and RLRRL could produce a string of higher fitness,
it started.) All the strings currently in the system are UURRU, through crossover. Adding crossover to a
tested to see how far each one takes the cursor genetic algorithm generally allows it to find better so-
from the start of the maze toward the goal. That is, lutions quicker, especially if the problem is difficult.
the system evaluates the fitness of each string. Each of the strings evolving through a genetic al-
Then it performs selection, removing strings like gorithm is a kind of computer program, and this
URDL that have very poor fitness and copying strings method can actually be used to write software. In re-
that have high fitness. cent years, experiments have shown the approach
Selection cannot be too harsh, or it will elimi- can solve a very wide range of engineering problems,
nate all variation among the strings, and variation is in hardware as well as software.
essential to evolution. One way to increase variation
is through mutationrandomly adding, subtract-
ing, or substituting a letter in a string. Suppose the Engineering Applications
first four moves in the maze are UURR, and the path A team led by the computer scientist John Koza at
then branches both up and down. Then both Stanford University has been using genetic program-
UURRU and UURRD will have high fitness, and both ming to design electronic circuits, including filters for
UURRL and UURRR will have lower fitness. After audio systems optimized to block some frequencies
repeating these steps many times, this process will and pass others, controllers such as automobile cruise
result in a population of long strings that represent control devices, and circuit generators that produce
desired outputs. In a way, designing electronic circuits improved over time, the better to serve the needs
is the opposite of our maze example, because it in- of human beings.
volves assembling electronic components into a maze In the GOLEM (Genetically Organized Lifelike
of circuits having the desired characteristics, rather Electro Machines) project, the DEMO team has used
than finding the best path through a pre-existing maze. genetic programming to design populations of sim-
A number of researchers have used genetic program- ple robots that are then actually built, so that their
ming to design the details of neural networks, a characteristics can be compared in the real world,
kind of artificial intelligence approach that can be built rather than just in computer simulations. These very
as analog electronic circuits but are usually expressed simple robots have one task only: to creep across the
digitally in software instead. floor. They are constructed out of piston-like actu-
One of the classic challenges for genetic pro- ators and bars that form a structure that can hinge
gramming was designing truss structure for use in at the connection. In the computer, a population
large machines and architectural buildings. The prob- of simulated robots evolves by means of genetic al-
lem is to design an open structure of beams and gird- gorithms through hundreds of generations. Mutation
ers that can reach horizontally out from a wall a given occasionally adds or subtracts a bar or actuator at
distance and support a given weight, and to do so a r a n d o m l y ch o s e n p o i n t i n t h e s t r u c t u re .
with the minimum amount of material. This prob- Selection tests how far each of the simulated ro-
lem involves many variables, because beams and gird- bots could creep in a given period of time, and those
ers of various sizes can be connected in many that could go the farthest have more offspring in the
different ways to form the truss, so it is a nice com- next generation. A number of the robots have ac-
putational challenge. It is also relatively straight- tually been fabricated and can indeed crawl
forward to test the quality of the solutions produced around the laboratory. The DEMO team set up the
by the computer, either through mathematical analy- computer and fabrication system, and established its
sis or through actually building a design and see- goal, but the robots' design was carried out auto-
ing how much weight it can support before breaking. matically by evolution.
The same approach can be useful in chemical engi-
neering, for example to determine an optimum com-
bination of raw materials and temperature to produce Challenges and Opportunities
high quality plastics reliably at low cost. While evolutionary computing can solve many en-
A research team at Brandeis University, called gineering problems, it is used relatively seldom out-
DEMO (Dynamical and Evolutionary Machine side research studies. One reason is that the
Organization) and including the scientists Jordan calculations are time consuming, especially in fig-
Pollack and Gregory Hornby, has used genetic pro- uring the relative fitness of all the strings. Parallel
gramming to design robots and virtual creatures. processing helps here, because it will allow all the fit-
DEMO believes that evolutionary engineering will ness evaluations to happen simultaneously, and John
be greatly facilitated by two related technological de- Koza's team works with a cluster of a thousand com-
velopments: improvements in the quality of computers. But there is also a human-computer inter-
puter aided mechanical design, including simulations, action barrier, because we do not yet have convenient
and development of inexpensive methods for systems that professional engineers can use to tell
rapid prototyping and manufacture of single items. the genetic algorithm what kind of solution it is sup-
Together, these two innovations could achieve posed to evolve. A related problem concerns how to
both the cost-effectiveness of mass production and give the computer the real-world information it can
the customer satisfaction of high-quality skilled use to evaluate the fitness of competing designs.
craftsmanship with unique designs. Combined with In the near future we can imagine that engineers
genetic programming, these methods could produce will be able to define problems through a comfort-
an evolutionary system in which machines constantly able multimedia computer interface, and highly
EXPERT SYSTEMS 247
intelligible output will describe solutions as they Ogburn, W. F. (1922). Social change. New York: Huebsch.
evolve. Manufacturing companies could unobtru- Pollack, J. B., Lipson, H., Hornby, G., & Funes, P. (2001). Three gen-
erations of automatically designed robots. Artificial Life, 7(3),
sively use their customers to evaluate the fitness of 215223.
designs by abandoning mass production of standard
models in favor of near infinite variety. The result
could be an approach that radically enhances human
creativity at the same time it inspires people to think
about engineering design in a fresh way.
EXPERT SYSTEMS
Expert systems (ES) are computer systems that cap-
See also Artificial Intelligence ture and store human problem-solving knowledge
(expertise) so that it can be utilized by less knowl-
edgeable people. An alternate term is knowledge-based
FURTHER READING expert systems. Expert systems imitate human ex-
perts reasoning processes in solving specific prob-
Bainbridge, W. S. (in press). The evolution of semantic systems. Annals lems and disseminate scarce knowledge resources,
of the New York Academy of Science. leading to improved, consistent results. As the knowl-
Basalla, G. (1988). The evolution of technology. Cambridge, UK:
Cambridge University Press. edge in an expert system is improved and becomes
Deb, K., & Gulati, S. (2001). Design of truss-structures for minimum more accurate, the system may eventually function
weight using genetic algorithms. Finite Elements in Analysis and at a higher level than any single human expert can
Design, 37(5), 447465.
Dennett, D. C. (1995). Darwin's dangerous idea. New York: Simon &
in making judgments in a specific, usually narrow,
Schuster. area of expertise (domain).
Gallagher, J. C., & Vigraham, S. (2002). A modified compact genetic Expert systems are part of artificial intelligence
algorithm for the intrinsic evolution of continuous time recurrent (the subfield of computer science that is concerned
neural networks. In W. B. Langdon, E. Cant-Paz, K. Mathias, R.
Roy, D. Davis, R. Poli et al. (Eds.), GECCO 2002: Proceedings of the
with symbolic reasoning and problem solving).
Genetic and Evolutionar y Computation Conference (pp. They use a symbolic approach to representing
163170). San Francisco: Morgan-Kaufmann. knowledge and simulate the process that experts
Goodman, E. D., Seo, K., Rosenberg, R. C., Fan, Z., Hu, J., & Zhang, use when solving problems. Knowledge, once
B. (2002). Automated design methodology for mechatronic sys-
tems using bond graphs and genetic programming. In 2002 NSF captured through the knowledge acquisition
Design, Service, and Manufacturing Grantees and Research process, must be represented, typically as produc-
Conference (pp. 206221). Arlington, VA: National Science tion rules (knowledge representation methods
Foundation.
Holland, J. H. (1975). Adaptation in natural and artificial systems.
in which knowledge is formalized into rules con-
Cambridge, MA: MIT Press. taining an IF part and a THEN part and,
Hornby, G. S., & Pollack, J. B. (2002). Creating high-level components optionally, an ELSE part). However, additional
with a generative representation for body-brain evolution. Artificial knowledge representations (formalisms for repre-
Life, 8(3), 223246.
Hornby, G. S., & Pollack, J. B. (20012). Evolving L-systems to gener-
senting facts and rules about a subject or a spe-
ate virtual creatures. Computers and Graphics, 25(6), 10411048. cialty) exist; each problem has a natural fit with
Koza, J. R. (1992). Genetic programming. Cambridge, MA: MIT Press. one or more knowledge representations. To be use-
Koza, J. R., Keane, M. A., & Streeter, M. J. (2003). Evolving inventions. ful, knowledge must be utilized through a reason-
Scientific American, 288(2), 5259.
Li, Y., Rangaiah, G. P., & Ray, A. K. (2003). Optimization of styrene re- ing process implemented in the inference engine
actor design for two objectives using a genetic algorithm. (the expert system component that performs
International Journal of Chemical Reactor Engineering, 1, A13. reasoning [thinking]). The structure of expert sys-
Miller, G. (2000). Technological evolution as self-fulfilling
prophecy. In J. Ziman (Ed.), Technological innovation as an evo-
tems is important, as are the application areas to
lutionary process (pp. 203215). Cambridge, UK: Cambridge which expert systems have been successfully
University Press. applied.
Expertise, Experts, and edge engineer interviews expert(s) and develops

an understanding of a problem. Then he or she iden-
Knowledge Acquisition tifies an appropriate knowledge representation and
Expertise is the extensive, task-specific knowledge inferencing approach.
acquired from training, reading, and experience. Acquired knowledge must be organized and
Through their expertise, experts make better and stored in a knowledge base. A good knowledge
faster decisions than nonexperts in solving complex representation naturally represents the problem do-
problems. A person requires much time to become main. People have developed many useful knowl-
an expert. Novices become experts only incremen- edge representation schemes through the years. The
tally. Experts have a high skill level in solving specific most common are production rules and frames.
problem types. Experts generally are good at rec- Rules are used most frequently. Any number of rep-
ognizing and formulating problems, solving prob- resentations may be combined into a hybrid knowl-
lems quickly and properly, explaining their solutions, edge representation.
learning from experience, restructuring knowledge, Most commercial (called production) ES are rule
breaking rules when necessary, and determining rel- based. Knowledge is stored as rules, as are the
evant factors. When faced with new problems, problem-solving procedures. Knowledge is presented
their solutions tend to be pretty good. as production rules in the form of condition-action
To mimic a human expert, we can implement an pairs: IF this condition occurs, THEN some ac-
ES that directly incorporates human expertise. tion will (or should) occur. This action may include
Typically, an ES can explain how it obtains solutions establishing a fact to be true, false, or true to some
and why it asks for specific information and estimate degree, which is known as a confidence level. For
its measure of confidence in its solutions, as would example, a rule may state: IF the light bulb is off,
an expert. THEN move the switch to the ON position.
The objective of an expert system is to transfer Each rule is an autonomous chunk of expert-
expertise from an expert to a computer system and ise. When utilized by the inference engine, the
then to other nonexpert humans. This transfer in- rules behave synergistically (relating to combined
volves knowledge acquisition, knowledge repre- action or operation). Rules are the most common
sentation, knowledge inferencing, and, finally, form of knowledge representation because they
knowledge transfer to the user. Knowledge is are easily understood and naturally describe many
stored in a knowledge base and reasoned with by an real situations.
inference engine. A frame includes all the knowledge about a par-
Knowledge acquisition is the process of extract- ticular object. This knowledge is organized in a hi-
ing, structuring, and organizing knowledge and erarchical structure that permits a diagnosis of
transferring it to the knowledge base and sometimes knowledge independence. Frames are used exten-
to the inference engine. Formally, knowledge is a col- sively in ES. Frames are like index cards that describe
lection of specialized facts, procedures, and judg- conditions and solutions. Given a real problem, a
ment rules. case-based reasoning (CBR) inference engine searches
A knowledge engineer applies AI methods to ap- the frames to identify a closest match to solve a prob-
plications requiring expertise, including developing lem. This procedure is similar to an expert remem-
expert systems. Knowledge engineering involves bering a specific situation that is like or close to the
knowledge acquisition, representation, validation, new one encountered.
inferencing, explanation, and maintenance.
Knowledge engineering also involves the coopera-
tion of human experts in explicitly codifying the Inferencing and Explanation
methods used to solve real problems. An expert system literally can reason (e.g., think).
The most common method for eliciting After the knowledge in the knowledge base is at a
knowledge from an expert is interviews. The knowl- sufficiently high level of accuracy, it is ready to be
EXPERT SYSTEMS 249
used. The inference engine is a computer program (when the ES asks the user for some information);
that accesses the knowledge and controls one or more advanced systems include the how? question (how
reasoning processes. a certain conclusion was reached).
The inference engine directs a search through A key issue in expert systems is the fuzziness of
the knowledge base. It asks for facts (for the IF part the decision-making process. Typical problems have
of rules) that it needs in order to fire rules and reach many qualitative aspects (the engine sounds funny),
their conclusions (the THEN part). The program de- and often when a rule reaches a conclusion, the ex-
cides which rule to investigate, which alternative to pert may feel it is right only about seven times out
eliminate, and which attribute to match. The most of ten. This level of confidence must be consid-
common inferencing methods for rule-based sys- e re d . Ce r t a i n t y t h e o r y p e r f o r m s t h i s t a s k .
tems are backward and forward chaining. Certainty factors (CF) express belief in an event (fact
Backward chaining is a goal-driven approach to or hypothesis) based on evidence (or the experts as-
problem solving. One starts from an expectation sessment) along a scale, for example, anywhere from
of what is to happen (hypothesis), then seeks evi- 0 (completely false) to 1 (completely true). The cer-
dence (facts) to support (or contradict) the expec- tainty factors are not probabilities but rather indi-
tation. An ES starts with a goal to be verified as either cate how true a particular conclusion is.
true or false. Then it looks for a rule that has that
goal in its conclusion. It then checks the premise of
that rule in an attempt to satisfy this rule. When nec- Expert System Components and Shells
essary, the ES asks the user for facts that it needs to The three major components in every expert system
know. If the search for a specific solution fails, the are the knowledge base, inference engine, and user in-
ES repeats by looking for another rule whose con- terface. An expert system may contain the following
clusion is the same as before. The process continues additional components: knowledge acquisition sub-
until all the possibilities that apply are checked or system, blackboard (workplace), explanation subsys-
until the initially checked rule (with the goal) is sat- tem (justifier), knowledge refining system, and user(s).
isfied. If the goal is proven false, then the next goal Most expert systems do not contain the knowl-
is tried. edge refinement component. When a system
Forward chaining is a data-driven approach. It makes an error, the error is captured and examined,
starts with all available information (facts) and tries and the knowledge base is updated through the
to reach conclusions. The ES analyzes the problem knowledge acquisition subsystem.
by looking for the facts that match the IF portion of The knowledge base contains the relevant knowl-
its IF-THEN rules. As each rule is tested, the program edge necessary for understanding, formulating, and
works its way toward one or more conclusions. solving problems. It includes facts such as the prob-
Typically, backward chaining is utilized in di- lem situation and theory of the problem area and
agnostic systems such as those for the medical or special heuristics (rules that direct the use of knowl-
equipment repair areas, whereas forward chaining edge to solve specific problems). The inference en-
is utilized in financial and accounting applications. gine may include general-purpose problem solving
Automatic control applications, such as those that and decision-making rules.
run steel mills and clay processing plants, typically The inference engine is the brain of the ES. It is
use forward chaining. also known as the control structure or rule in-
Human experts are often asked to explain their terpreter (in rule-based ES). The inference engine
decisions. Likewise, ESs must also be able to explain provides a methodology for reasoning about infor-
their actions. An ES must clarify its reasoning, rec- mation in the knowledge base to reach conclusions.
ommendations, or other actions. The explanation It provides directions about how to use the sys-
facility does this. Rule-based ES explanation traces tems knowledge by developing the agenda that or-
rules that are fired as a problem is solved. Most ES ganizes and controls the steps taken whenever
explanation facilities include the why? question consultation is performed.
Expert systems contain a language processor for Successful expert systems examples include
friendly communication between the user and the MYCIN (medical diagnosis, Stanford University),
computer. This communication can best be per- XCOM (computer system configuration, Digital
formed through menus in a graphical user interface. Equipment Corporation), Expert Tax (tax planning,
Some ESs use natural language processors. Coopers & Lybrand), Loan Probe (loan evaluation,
Knowledge engineers generally use a software Peat Marwick), La-Courtier (financial planning,
tool called a shell. Expert system shells (com- Cognitive Systems), LMOS (network manage-
puter programs that facilitate the relatively easy im- ment, Pacific Bell), LMS (production planning,
plementation of a specific expert system, similar to scheduling, and management, IBM), and Fish-Expert
the concept of a Decision Support System genera- (disease diagnosis, north China). The Nestle Foods
tor) include the major components of the expert sys- Corporation developed an expert system to provide
tems (except for the knowledge itself). Generally a accurate information and advice on employee
shell can represent knowledge in only one or two pension funds. America Online utilizes an expert sys-
ways (for example, rules and frames) and manipu- tem to assist its help desk personnel. Many help desks
late them in a limited number of ways. Using a shell, incorporate expert systems that may be accessed
the knowledge engineer can focus on the knowledge either by the organizations personnel or customers.
because the shell manages the knowledge, the inter- Expert systems enable organizations to capture
face, the inferencing method(s), and the inferencing the scarce resource of expertise and make it available
rules. Only the knowledge need be added to build an to others. Expert systems affect an organizations bot-
expert system. Examples of some commercial rule- tom line by providing expertise to nonexperts. Expert
based shells are Corvid Exsys, Ginesys K-Vision, systems often provide intelligent capabilities to other
CLIPS, and JESS. Most shells run directly on information systems. This powerful technology will
World Wide Web servers. Users and knowledge en- continue to have a major impact on knowledge de-
gineers access them with Web browsers. ployment for improving decision-making.
Jay E. Aronson
Expert Systems Application Areas
Expert systems can be classified by their general prob-
See also Artificial Intelligence; Information
lem areas. Expert systems classes include:
Organization
Interpreting: Inferring situation descriptions
from observations
Predicting: Inferring likely consequences of given
situations FURTHER READING
Diagnosing: Inferring system malfunctions from
Allen, B. P. (1994). Case-based reasoning: Business applications.
observations Communications of the ACM, 37(3), 4042.
Designing: Configuring objects under constraints Aronson, J. E., Turban, E., & Liang, T. P. (2004). Decision support
Planning: Developing plans to achieve goals systems and intelligent systems (7th ed). Upper Saddle River, NJ:
Prentice Hall.
Monitoring: Comparing observations to plans,
Awad, E. M. (1996). Building expert systems: Principles, procedures, and
flagging exceptions applications. Minneapolis/St. Paul, MN: West Publishing.
Debugging: Prescribing remedies for malfunc- Dean, T., Allen, J., & Aloimonos, Y. (2002). Artificial intelligence: Theory
tions and practice. Upper Saddle River, NJ: Pearson Education POD.
Feigenbaum, E., & McCorduck, P. (1983). The fifth generation. Reading,
Repairing: Executing a plan to administer a pre-
MA: Addison-Wesley.
scribed remedy Giarratano, J. C. (1998). Expert systems: Principles and program-
Instructing: Diagnosing, debugging, and cor- ming. Pacific Grove, CA: Brooks Cole.
Hart, A. (1992). Knowledge acquisition for expert systems. New York:
recting student performance McGraw-Hill.
Controlling: Interpreting, predicting, repairing, Jackson, P. (1999). Introduction to expert systems (3rd ed.). Reading,
and monitoring system behaviors MA: Pierson Addison-Wesley.
EYE TRACKING 251
Kolonder, J. (1993). Case-based reasoning. Mountain View, CA: Morgan frame-by-frame study of the captured video stream
Kaufmann. of the eye. Relatively recent advances in eye-tracking
Russell, S. J., & Norvig, P. (2003). Artificial intelligence: A modern ap-
proach (2nd ed.). Upper Saddle River, NJ: Prentice Hall.
technology, which have resulted in increasingly ac-
Winston, A. (1992). Artificial intelligence (3rd ed.). Reading, MA: curate, comfortable, unobtrusive, and inexpensive
Addison-Wesley. eye trackers, have made possible the adoption of
Zahedi, F. (1993). Intelligent systems for business. Belmont, CA: eye trackers in interactive applications. Todays eye
Wadsworth.
trackers typically employ a video camera to capture
an image of both the eyes pupil and the corneal re-
flection of a nearby on- or off-axis light source (usu-
ally infrared because infrared light is invisible to
EYE TRACKING the naked eye). Computer and video-processing hard-
ware are then employed to calculate the eyes point
Eye-movement-based, or gaze-based, interaction has of regard (defined as the point on the screen or other
been suggested as a possible easy, natural, and fast stimulus display being looked at by the viewer) based
noncommand style of interaction with computers. on these two optical features. Such video-based
Indeed eye typing, in which the user gazes at letters corneal-reflection eye trackers have become the de-
to type, has often been cited as the prototypical ex- vice of choice in interactive applications.
ample of an eye-based, noncommand, interactive
application. Such gaze-based interaction relies on Interactive Control: Early Application Examples
the interfaces ability to track the position and move- Demonstrations of two early interactive uses of
ment of the users eyes. eye trackers were published in 1990 by Robert Jacob
In general, the purpose of eye tracking can be di- and by India Starker and Richard Bolt. Starker and
agnostic or interactive. In diagnostic applications, Bolt used eye movements to navigate in, and inter-
eye movements are recorded so that scene ele- act with objects within, a graphical fantasy world.
ments that the user looked at and possibly paid at- Jacob used eye movements to enable the user to se-
tention to can be analyzed and evaluated later. Used lect objects on a desktop display (in this case Navy
diagnostically, eye trackers help evaluate the users ships that could be selected by gaze to obtain in-
attentive behavior. Interface usability studies can use formation about them). Both examples of interac-
eye movement data to test the visibility of some fea- tive eye-tracking applications exposed a key difficulty
ture of an interface display, such as desktop or with the use of eye movements for interaction, which
webpage elements. For example, the researcher Joseph Jacob called the Midas-touch problem. When eye
H. Goldberg and his colleagues reported in 2002 on movements are used for selection of objects shown
the use of eye movements to evaluate the design of on a computer display, simulating a mouse-based
webpages, including the positioning of webpage por- interface, selection confirmation may be problem-
tals. In 1998 the psychologist Keith Rayner reviewed atic. With a mouse, the user typically confirms se-
a host of similar diagnostic eye-movement appli- lection of interface features by clicking mouse
cations. This article concentrates on interactive, real- buttons. The eyes, however, cannot register button
time applications of eye tracking. clicks. The result is that, just as everything Midas
touched indiscriminately turned to gold, anything
gazed at can potentially be selected. To allow the user
Eye-Tracking Devices to select objects via eye movements, Jacob proposed
The movements of the eyes have been measured and a measure of dwell time to activate selection; that is,
studied for more than a hundred years. Early tech- the eyes would have to rest on an object for a certain
niques, however, were either invasive (for example, amount of timefor example, 500 milliseconds.
involving the embedding of a scleral coil in a contact Starker and Bolt used a similar thresholding tech-
lens worn by the user), or not particularly precise, nique to zoom in on objects of interest in their vir-
such as early video-oculography, which relied on tual 3D world.
The Midas-touch problem is related to a more and earlier, fertilizing research in a variety of inter-
general concern associated with eye movements, active scenarios. Eye typing is still an exemplary
namely, eye movement analysis. During normal vi- interactive application, particularly since it provides
sual perception, the fovea (the retinal region of high- computer access and a means of communication for
est acuity) is moved about the visual field through certain disabled users, such as quadriplegics. Among
very fast eye movements known as saccades. Saccadic the research teams that have examined interactive
eye movements are so fast, and as a result the scene gaze-based object selection in virtual reality are
is so blurred during a saccade, that the viewer is ren- Vildan Tanriverdi (working with Robert Jacob) and
dered effectively blind during these brief periods of Nathan Cournia and his colleagues. In this interac-
foveal repositioning. Areas or regions of interest are tive situation, eye-based pointing, based on a ray-
inspected with more scrutiny during a relatively stead- casting approach (selecting objects in the virtual
ier form of gaze known as a fixation. Characteristic world by shooting an invisible ray along the line of
eye movements include pursuit of moving objects sight) was found to be more efficient than arm-based
(pursuit movements) and involuntary eye rotation pointing when arm pointing relied on the arm ex-
counteracting head movement (vestibular move- tension approach (lifting the arm to point as if a laser
ments), among others. For interactive uses, fixations were attached to ones fingertip).
are usually of greatest interest, because fixated loca- Research on general gaze-based selective eye-
tions are highly correlated with locations to which tracking applications has also continued, with ad-
the user is devoting the greatest visual attention. Eye- vances in applications in which gaze is used as an
movement analysis is used to characterize the type of indirect indicator of the users intent. Arguing that
eye movement recorded by the eye tracker. This is a gaze-based pointing is an interactive strategy that
particularly important component of eye tracking inappropriately loads the visual channel with a mo-
since w ithout analysis it may not be know n tor control task, Shumin Zhai and his colleagues con-
whether regions in the visual field are simply being tend that gaze-based functionality is fundamentally
glanced at, fixated, or passed over by saccades. at odds with users natural mental model, in which
Several different techniques for eye movement the eye searches for and takes in information while
analysis are available; the two most popular are based coordinating with the hand for manipulation of ex-
on the eyes position or its velocity. Both seek to iden- ternal objects. Zhai and colleagues provide an alter-
tify fixations in the time course of eye movements. native to direct gaze-based selection of user interface
Position-based methods typically classify fixations a objects by using gaze as an indirect accelerator of the
priori by measuring the deviation of sampled eye mouse pointer.
movements from some recently calculated central lo- Other gaze-based interaction styles use gaze
cation. Velocity-based techniques, on the other hand, indirectly to manipulate interface objects without
start by defining saccades, since these are often eas- necessarily requiring the users awareness of the
ier to locate in the eye movement time course. eye tracker. An example of this class of eye-track-
Fixations are generally then assumed to be composed ing modality is gaze-contingent graphical rendering
of eye movement points that lie outside the brief sac- of complex scenes. By exploiting knowledge of the
cadic periods. Just as in position-based techniques, users gaze and the limited capacity of human pe-
a threshold, this time of velocity, is used to locate sac- ripheral vision, the system is able to focus its limited
cades in an effort to classify fixations. The impor- resources on the display regions projected onto the
tant point here is that regardless of which technique users fovea. This technique is particularly well suited
is used, some form of eye movement analysis is usu- to applications burdened with rendering complex
ally required to make sense of eye movement patterns. graphical data. David Luebke and his colleagues have
developed a technique for gaze-contingent render-
Interactive Control: Recent Advances ing in which 3D graphical objects are rendered at
Eye-tracking equipment has improved greatly high resolution only when the user is focusing on
from the early interactive applications of the 1990s them directly. These view-dependent level-of-detail
EYE TRACKING 253
techniques thus aim to degrade the spatial resolu- of eye-tracking technology research is to eliminate
tion of 3D graphical objects imperceptibly. In a re- the necessity for calibration altogether. An autocal-
lated method focused on degrading temporal ibrating eye tracker, possibly based on multiple cam-
resolution, Carol OSullivan and her colleagues in- era input, may soon be available.
teractively manipulated the precision of 3D object Anticipating improved sensing technologies, such
collisions in peripheral areas so that the gap between as autocalibrating eye trackers, an emerging hu-
colliding objects in the periphery was larger due to man-computer interaction strategy is concerned with
lower precision. designing attentive user interfaces, or AUIs, in
Indirect gaze-based pointing approaches have which the interface is aware of the direction of the
also been developed to support computer-mediated users (visual) attention. By tracking the users eyes,
communication systems. A prevalent problem in AUIs attempt to match the characteristics of computer
such systems is the lack of eye contact between displays to certain characteristics of human vision,
participants and the lack of visual deictic (look at such as the distinction in human vision between foveal
this) reference over shared media. In multiparty and peripheral vision. Such AUIs make better use of
teleconferencing systems, these deficiencies lead to limited resources by tailoring display content to that
confusion over who is talking to whom and what which is useful to human vision (e.g., a small fraction
others are talking about. Roel Vertegaal and his fel- of the display at the point of regard is displayed at high
low researchers offer a solution to both problems by resolution while peripheral regions are displayed at
tracking each participants gaze. Their gaze-aware lower resolution, matching the resolvability of pe-
multiparty communication system provides still im- ripheral human vision while simultaneously con-
ages or video of remotely located participants in a serving computational resources). As the technology
virtual teleconferencing and document-sharing sys- matures, eye trackers will endure as a significant com-
tem. Containing pictures or video of participants ponent in the design of interactive systems.
faces, 3D boxes rotate to depict a participants gaze
direction, alleviating the problem of taking turns Andrew T. Duchowski
during communication. Furthermore, a gaze-directed
spot of light is shown over a shared document to in-
dicate the users fixated regions and thereby provide FURTHER READING
a deictic reference.
Baudisch, P., DeCarlo, D., Duchowski, A. T., & Geisler, W. S. (2003,
March). Focusing on the essential: Considering attention in dis-
play design. Communications of the ACM, 46(3), 6066.
Future Directions Bowman, D. A., & Hodges, L. F. (1997). An evaluation of techniques
While many interactive eye-tracking applications for grabbing and manipulating remote objects in immersive vir-
tual environments. In Proceedings of the Symposium on Interactive
have successfully been developed, current state-of- 3D Graphics (pp. 3538). Providence, RI: ACM.
the-art eye trackers suffer from the requirement of Cournia, N., Smith, J. D., & Duchowski, A. T. (2003, April). Gaze-
calibration. In most cases, especially those requiring vs. hand-based pointing in virtual environments. In Proceedings
high levels of accuracy, the eye tracker must be cal- of CHI 03 (Short Talks and Interactive Posters (pp. 772773).
Fort Lauderdale, FL: ACM.
ibrated to each user before the device can be oper- Duchowski, A. T. (2003). Eye tracking methodology: Theory & practice.
ated. Even more problematic, some eye trackers suffer London: Springer-Verlag.
from calibration drift; that is, during interactive ses- Goldberg, J. H., & Kotval, X. P. (1999). Computer interface evaluation
sions the eye-tracker accuracy degrades, requiring using eye movements: Methods and constructs. International
Journal of Industrial Ergonomics, 24, 631645.
recalibration. Eye trackers are steadily improving, Goldberg, J. H., Stimson, M. J., Lewenstein, M., Scott, N., & Wichansky,
and not all eye trackers require extensive calibration. A. M. (2002). Eye tracking in web search tasks: Design implica-
Operating at a coarse resolution and utilizing fa- tions. In Proceedings of Eye Tracking Research & Applications (ETRA)
(pp. 5158). New Orleans, LA: ACM.
cial recognition subsystems, some eye trackers Jacob, R. J. (1990). What you look at is what you get: Eye move-
provide general gaze direction with a substantially ment-based interaction techniques. In Proceedings of CHI 90 (pp.
reduced calibration requirement. The current goal 1118). Seattle, WA: ACM.
Luebke, D., Hallen, B., Newfield, D., & Watson, B. (2000). Perceptually Rayner, K. (1998). Eye movements in reading and information pro-
driven simplification using gaze-directed rendering (Technical Report cessing: 20 years of research. Psychological Bulletin, 124(3), 372422.
CS-2000-04). Charlottesville: University of Virginia. Salvucci, D. D., & Goldberg, J. H. (2000). Identifying fixations and sac-
Majaranta, P., & Raiha, K.-J. (2002). Twenty years of eye typing: Systems cades in eye-tracking protocols. In Proceedings of Eye Tracking
and design issues. In Proceedings of Eye Tracking Research & Research & Applications (ETRA) (pp. 7178). Palm Beach Gardens,
Applications (ETRA) (pp. 1522). New Orleans, LA: ACM. FL: ACM.
Nielsen, J. (1993, April). The next generation GUIs: Non-command Starker, I., & Bolt, R. A. (1990). A gaze-responsive self-disclosing
user interfaces. Communications of the ACM, 36(4), 8399. display. In Proceedings of CHI 90 (pp. 39). Seattle, WA: ACM.
OSullivan, C., Dingliana, J., & Howlett, S. (2002). Gaze-contingent Tanriverdi, V., & Jacob, R. J. K. (2000). In Proceedings of CHI 00 (pp.
algorithms for interactive graphics. In J. Hyn, R. Radach, & H. 265272). The Hague, Netherlands: ACM.
Duebel (Eds.), The minds eye: cognitive and applied aspects of eye Vertegaal, R. (1999). The GAZE groupware system: Mediating joint
movement research (pp. 555571). Oxford, UK: Elsevier Science. attention in multiparty communication and collaboration. In
Proceeding of CHI 99 (pp.294301). Pittsburgh, PA: ACM.
FACIAL EXPRESSIONS
FLY-BY-WIRE
FONTS
computer will be a significant step towards devel-

FACIAL EXPRESSIONS oping an effective human-machine interface. To-
wards this end, it is important to understand the ways
The human face provides information that regulates a in which a system with the ability to understand
variety of aspects of our social life.Verbal and non-ver- facial gestures (analysis), and the means of au-
bal communication are both aided by our perception tomating this interpretation and/or production (syn-
of facial motion; visual speech effectively compliments thesis) might enhance human-computer interaction.
verbal speech. Facial expressions help us identify our Also essential is understanding how meaning is de-
companions and inform us of their emotional state.In rived from the complex rigid and non-rigid motions
short,faces are windows into the mechanisms that gov- associated with the face; this is perhaps key to the
ern our emotional and social lives.To fully understand machine perception of human expression.
the subtlety and informativeness of the face and the
complexity of its movements,face perception and face
processing are becoming major topics of research by Psychological Importance
cognitive scientists,sociologists,and most recently,re-
searchers in human computer interaction (HCI),com- of Facial Expressions
puter vision, and computer graphics. The face is a multi-signal, multi-message response
The automation of human face processing by a system capable of tremendous flexibility and
255
specificity. It is the site for sensory inputs and the on quantitative and qualitative analysis of facial
communicative outputs. In 1978 Paul Ekman, a psy- expressions was undertaken by Paul Ekman and
chologist at the University of California medical W. Friesen, who produced a widely used system for
school, described how faces convey information describing visually distinguishable facial movements
via four general classes of signals: in 1978. This system, called the Facial Action Cod-

ing System, or FACS, is based on the enumeration of
static facial signals: permanent features of the
all action units that cause facial movements; the
face like the bony structure and soft tissue masses
combination of these action units results in a large
contributing to facial appearance;
slow facial signals: changes in facial appearance
set of possible facial expressions.
FACS coding is done by individuals trained to
over time, wrinkles, texture, and so on;
artificial signals: exogenously determined fea-
categorize facial motion based on the anatomy of fa-
cial activitythat is, how muscles singly and in com-
tures, such as eyeglasses, and cosmetics; and
rapid facial signals: phasic changes on neuro-
bination change facial appearance. Some muscles
give rise to more than one action unit, and the cor-
muscular activity leading to visible changes in
respondence between action units and muscle units
facial appearance.
is approximate. A FACS coder, after undergoing
All four of these classes contribute to facial recog- extensive training, codes expressions by dissecting
nition; however,only the rapid signals convey messages an expression, decomposing it into the specifics that
via emotions in a social context. The neuropsychology produced the motion. This coding is done by ana-
of facial expressions supports the view that facial move- lyzing the relationship between components of the
ments express emotional states, and that the two expressions and judgments made by the coders from
cerebral hemispheres are differently involved in the static photographs and more recently from videos.
control and interpretation of facial expressions. In 1978 John Bassili, a psychologist at the Uni-
Some of the initial work studying the relationship be- versity of Toronto, argued that because facial mus-
tween facial expressions and emotions was undertaken cles are fixed in certain spatial arrangements, the
in the early nineteenth century by Charles Darwin and deformations of the elastic surface of the face to
C.B.Duchenne du Boulogne,a neurophysiologist.Their which they give rise during facial expressions may
work still has a strong influence on the research tech- be informative in the recognition of facial expres-
niques used to examine expression perception. sions. To verify his claim, Bassili conducted experi-
Until recently the majority of studies on facial ments by covering the faces of actors with black
expressions have continued to examine the percep- makeup and painting white spots in at random lo-
tion of posed expressions in static photographs. Most cations. Faces were divided into upper and lower re-
of these studies suggest that seven universal cate- gions (to correlate with FACS data for upper and
gories of expressions can be recognized by members lower regions) and recognition studies were con-
of all cultures, both literate and preliterate. Re- ducted. This study showed that in addition to the
searchers are now beginning to study facial expres- spatial arrangement of facial features, the movement
sions in spontaneous and dynamic settings to avoid of the surface of the face does serve as a source of in-
the potential drawbacks of using static expressions formation for facial recognition. This significant ob-
and to acquire more realistic samples. The problem, servation is that we use muscles and connected skin
of course, is how to categorize active and sponta- to model facial action dynamically.
neous facial expressions in order to extract infor- Recently the video-coding community has
mation about the underlying emotional states. proposed a new standard for the synthesis of facial
action called Facial Action Parameters (FAPs). FAPs
is a basic parameterization of facial movement
Representations of Facial Motion that resembles the efforts of the earlier facial ani-
After the seminal nineteenth century work of mation systems. Underlying this parameterization
Duchenne and Darwin, the most significant work are implicit hooks into the FACS system. The FAPs
FACIAL EXPRESSIONS 257
model does extend upon the earlier representations cated method that tracked linear facial features to
by defining a set of sixty-eight parameters. Addi- estimate the corresponding parameters of a 3-D wire
tionally, these parameters are lumped into two frame face model, which made it possible to repro-
separate levels, visemes (representing mouth pos- duce and recognize facial expressions. They used con-
ture correlated to a phoneme) and expressions (rep- tour tracking, which required that facial features
resenting expressions like joy, sadness, anger, disgust, be highlighted with make-up for robust tracking. In
fear, and surprise with a single parameter). All the 1996 Irfan Essa and his colleagues at the Massa-
FAPs are constructed by combining lower-level fa- chusetts Institute of Technology and the Georgia In-
cial actions and are widely used to synthesize facial stitute of Technology, developed a similar method.
motions, primarily for low-bandwidth telecommu- At the cost of significantly higher computation, their
nication applications. approach incorporated a more detailed model of the
face using finite element techniques and very fine
pixel-by-pixel measurements of image motion, cou-
Tracking Facial Motion pled with feature tracking to measure facial action.
To support natural human-computer interaction, To aid in the tracking and analysis of facial motions,
there have been several attempts to track facial ex- their approach combined the measurement of mo-
pressions from video sequences. Kenji Mase, while a tion from a video with a model. Both approaches
researcher at NTT and the ATR Media Integration generate very good interpretations and animations
and Communications Research Laboratories in Ky- of facial expressions, and researchers are currently
oto, Japan, was the first to introduce an analysis of working on other methods that rely on simple fea-
video sequences and to present a method to track ture tracking and color/motion tracking for track-
action units using optical flow. The results of this ing faces.
1991 approach showed the usefulness of motion es-
timation using optical flow for observing facial mo-
t i o n . A f e w ye a r s l a te r, Ir f a n E s s a a n d A l ex Recognition of Facial Motion
Pentland at the Massachusetts Institute of Tech- Recognition of facial expressions can be achieved by
nology, employed a similar approach, but added a categorizing a set of predetermined facial motions
muscle-based face model to extract finer facial move- as in FACS, rather than determining the motion of
ments. Around the same time, M. J. Black from Xe- each facial point independently. Black and Yacoob
rox PARC, working with Yaser Yacoob and Larry use local parameterized models of image motion
Davis at the University of Maryland, extended this to measure FACS and related parameters for
approach to include the use of local parameterized recognition. These methods show an 86 percent over-
models of image motion to track faces; their method all accuracy in correctly recognizing expressions over
required an affine model of the different regions of their database of 105 expressions (which included
the face for tracking. In 1996 D. DeCarlo and D. data from live subjects and television shows). Mase,
Metaxes added a deformable model of the face to es- using a smaller set of data (thirty test cases), obtained
timate motion and shape for tracking, which served an accuracy of 80 percent. Both methods rely on
as a constraint in the measurement of motion. All FACS combinations to recognize expressions.
these approaches show that a more robust tracking A 1999 research project by Jeffrey Cohn and
and estimation of facial movements is improved his colleagues at Carnegie Mellon University reported
by combining information about the model of the 91, 88, and 81 percent agreements for brow, eye, and
face and very detailed measurements of facial move- mouth movements, between manual FACS codes
ment. They also show that the more detailed the and their automated system. Their system uses hi-
model (resulting in higher computational cost), the erarchical motion estimation coupled with feature
better the tracking. point tracking to measure facial movements. Their
In 1993 Demetri Terzopoulos and Keith Waters database contains about a hundred university stu-
at the University of Toronto introduced a sophisti- dents making expressions as instructed. In a 1999
paper, Marian Stewart Bartlett and her colleagues re- audio for affect) and to develop systems that allow
ported an accuracy of 91 percent over eighty se- for extended interactions, so that a system can ob-
quences (each sequence is six frames) containing six serve users for an extended time to get to know
upper-face FACS actions. They used a hybrid ap- them. Significant progress has been made in build-
proach that combined spatial variation, feature tracking systems that can recognize expressions in a non-
ing, and motion estimation within a neural network invasive manner. With the growth of robots and
framework. The last two methods are, however, interactive environments, it is easy to predict that
not recognizing expressions per se, but are compar- technology will soon be developed in which the real-
ing them to FACS codes that were validated by hu- time interpretation of expressions will be key.
man experts.
In 1991 Paul Ekman suggested that the timing Irfan Essa
of expressions was an important cue in detecting the
difference between true and fake facial expressions See also Affective Computing; Gesture Recognition
and emotions, and in the mid 1990s, Essa and his
colleagues proposed an approach based on a grow-
ing body of psychological research that argued that FURTHER READING
it was the dynamics of the expression, rather than
detailed spatial deformations, that was important in Bassili, J. (1978). Facial motion in the perception of faces and of emo-
expression recognition. They moved away from a tional expression. Journal of Experimental Psyschology, 4, 373
379.
static analysis of expressions (which is how the FACS Bartlett, M. S., Hager, J. C., Ekman, P., & Sejnowski, T. J. (1999). Mea-
model was developed) towards a whole-face analy- suring facial expressions by computer image analysis. Psycho-
sis of facial dynamics in motion sequences, which physiology, 36(2), 253263.
Black, M. J., & Yacoob, Y. (1997). Recognizing facial expressions in im-
could only be achieved by an automated percep- age sequences using local parameterized models of image motion.
tion of facial motion in image sequences within a International Journal of Computer Vision, 25(1), 2348.
dynamic estimation framework. Using over fifty-two Bruce, V., & Burton, M. (1992) Processing images of faces. Norwood,
video sequences of twelve subjects, they were able to NJ: Ablex Publishing.
Cohn, J., Zlochower, A., Lien, J., & Kanade, T. (1999). Automated face
achieve a recognition accuracy of 98 percent in analysis by feature point tracking has high concurrent validity with
recognizing expressions of happiness, anger, disgust, manual faces coding. Psychophysiology, 36, 3543.
and surprise. Darrell, T., Essa, I., & Pentland, A. (1996). Task-specific gesture mod-
Several new methods for extracting FAPs (as op- eling using interpolated views. IEEE Transaction on Pattern Analy-
sis and Machine Intelligence, 18(12).
posed to FACs) from video have also been been in- Darwin, C. (1965). The expression of the emotions in man and animals.
troduced to support automatic extraction of FAPS Chicago: University of Chicago Press.
parameters, which in turn could aid in the recog- DeCarlo, D., & Metaxas, D. (2000). Optical flow constraints on de-
formable models with applications to face tracking. Interna-
nition of facial expressions. An important aspect tional Journal of Computer Vision, 38(2), 99127.
of this new work is to aid in bandwidth limited com- Duchenne, G.-B. (1990). The mechanism of human facial expression:
munication with 3D faces. Studies in emotion and social interaction. Cambridge University
Press; Editions de la Maison des Sciences de lHomme.
Eisert, P., & Girod, B. (1998). Analyzing facial expression for virtual
conferencing. IEEE Computer Graphics & Applications, 18(5).
Expression and Emotions Ekman, P. (1978). Facial signs: Facts, fantasies and possibilities. In T.
While the relationship between emotions and ex- Sebeok, (Ed.), Sight, Sound and Sense. Bloomington: Indiana Uni-
pressions has been well studied in the psychological versity Press.
Ekman, P. (1991). Telling lies: Clues to deceit in the marketplace, pol-
literature, quantifiable representations that connect itics, and marriage. New York: Norton. Ekman, P., & Friesen, W.
what can be observed and modeled to what can be (1978). Facial action coding system. Palo Alto, CA: Consulting Psy-
inferred with knowledge from other sources (like chologists Press.
Essa, I., Basu, S., Darrell, T., & Pentland, A. (1996). Modeling, track-
context and familiarity with the person being inter- ing and interactive animation of faces and heads using input from
acted with) also play a major role. To this end, it is video. In Proceedings of Computer Animation Conference 1996 (pp.
important to rely on some of the other sensors (like 6879). New York: IEEE Computer Society Press.
FLY-BY-WIRE 259
Essa, I., & Pentland, A. (1997). Coding, analysis, interpretation, and Center in Edwards, California, in 1972: The test plat-
recognition of facial expressions. IEEE Transaction on Pattern Analy- form was an F-8C Crusader aircraft. Since then,
sis and Machine Intelligence, 19(7), 757763.
Mase, K. (1991). Recognition of facial expressions for optical flow.
fly-by-wire has been used widely in military aircraft;
IEICE Transactions, Special Issue on Computer Vision and its Ap- it is currently also used in many civilian passenger
plications, E 74(10). aircraft, including most Airbus models (starting with
MPEG.(1999).Overview of the MPEG-4 standard.Technical Report ISO/ the Airbus A320 in 1988) and the Boeing 777.
IEC JTC1/SC29/WG11 N2725. International Organisation for Stan-
dardization (ISO), Seoul, South Korea. Retrieved April 8, 2004, from In the 1990s computers also came to be em-
http://drogo.cselt.stet.it/mpeg/standards/mpeg-4/mpeg-4.htm. bedded into the control of automobiles. Antilock
Parke, F., & Waters, K. (1996). Computer facial animation. Wellesley, braking, in which the computer derives inputs from
MA: AK Peters.
Pelachaud, C., Badler, N., & Viaud, M. (1994). Final report to NSF of
the drivers brake pedal and rapidly pulses brakes
the standards for facial animation workshop. Philadelphia: National where necessary, is an example. Other drive-by-wire
Science Foundation, University of Pennsylvania. Retrieved April approaches are in development, including mecha-
8, 2004, from http://www.cis.upenn.edu/hms/pelachaud/workshop nisms that will allow cars to sense the distance to the
_face/workshop_face.html.
Picard, R. (1997). Affective computing. Cambridge, MA: MIT Press.
car in front and slow down to keep this spacing
Tao, H., Chen, H., Wu, W., & Huang, T. (1999). Compression of mpeg- sufficient for safety.
4 facial animation parameters for transmission of talking heads. There are several advantages to using fly-by-wire
IEEE Transactions on Circuits and Systems for Video Technology, technology. First, the weight of a fly-by-wire control
9(2), 264.
Waters, K., & Terzopoulos, D. (1992a).The computer synthesis of ex- system is generally much less than that of traditional
pressive faces. Philosophical Transactions of the Royal Society of controls. In aircraft, this is no small advantage.
London, B, 335(1273), 8793. Second, since computers are so much faster than hu-
Waters, K., & Terzopoulos, D. (1992b).Modelling and animating faces
using scanned data. The Journal of Visualization and Computer
mans, fly-by-wire makes quicker reactions to rapid
Animation, 2(4), 123128. changes in the controlled system or the operating
Yacoob, Y., & Davis, L. (1994). Computing spatio-temporal repre- environment possible. This is especially useful in air-
sentations of human faces. In Proceedings of the Computer Vision craft applications where, to improve maneuver-
and Pattern Recognition Conference (pp. 7075). New York: IEEE
Computer Society.
ability, changes are made to the airframe (structure
of the aircraft) that render it intrinsically less stable.
In such cases, the control must be extremely agile to
respond to the onset of instability before the insta-
FLY-BY-WIRE bility becomes irrecoverable and the airplane crashes.
Another advantage to fast reaction is the potential
Fly-by-wire is a phrase used to describe situations in for reduced drag, as a result of an improved trim set-
which computers are used as an indispensable me- ting of the controls. Third, a well-designed system
diating agent between a human operator and a can reduce both the physical and mental workload
medium of transportation. As the term suggests, it of the pilot. Removing the direct mechanical-hy-
first saw the light of day in aircraft applications; now, draulic linkage from the cockpit controls to the con-
however, the term is also often taken to include trol surfaces reduces the physical effort required to
systems in which computers are used in the con- handle them, and having a mediating system that
trol of automobiles. displays information in an appropriate way re-
The traditional means of controlling an air- duces the mental strain on the pilot.
craft is to connect the pilot controls to the control Lives can be lost if fly-by-wire systems fail. For
surfaces by means of mechanical or hydraulic link- this reason, a great deal of effort is spent in ensuring
ages. By contrast, in a fly-by-wire system, the oper- that they are highly reliable. A commonly quoted re-
ators commands are wholly or partially fed into a quirement for civilian aircraft is that the system
computer, which then determines the appropriate should not fail catastrophically at a rate of more than
control settings. once every billion ten-hour flights. This number is
Perhaps the first complete fly-by-wire system was so low that fault-tolerant techniques must be used
built at what is now the NASA Dryden Flight Research to ensure that even if some components of the system
fail, there is enough reserve capacity for the system appropriately mediated by the system; the ques-
to continue operating. The system must also have tion is where the line should be drawn. For example,
the intelligence to switch the more critical part of the one significant source of danger to aircraft is pilot-
workload out of malfunctioning units to those induced oscillations (PIOs). A PIO occurs when the
that are still running. (See the discussion of fault tol- pilot and the aircraft dynamically couple in such a
erance below.) way as to induce instability. For example, if the pilot
applies too great an input to attempt to correct a de-
viation from the desired-state trajectory, the system
Basics of Fly-By-Wire Operation can deviate in the other direction. The pilot might
When a pilot applies an input to a control, this is rec- then overcorrect for this second deviation; the sys-
ognized and translated into a digital signal, which is tem can deviate some more, and so on. PIOs are a
sent to the processors at the heart of the fly-by- known potential cause of loss of aircraft. One im-
wire system. These processors then compute the ap- portant application of fly-by-wire systems is to
propriate control output to be sent to the actuators recognize when such PIOs might occur and appro-
(for example, the ailerons, rudder, spoilers, and flaps). priately modulate the pilots commands to the
The bulk of the complexity lies in the computation control surfaces.
of the control output: A fly-by-wire system does not
merely translate the pilot commands to actuator sig-
nals; it also takes the current aircraft state and the Response Time: The Key Parameter
safe-operating envelope into account as it makes The response time of a fly-by-wire system is key to
its calculations. From this arises one of the major the quality of control provided by the system. Fly-
controversies in fly-by-wire control: To what ex- by-wire systems may fail not only by providing the
tent must the fly-by-wire system protect the aircraft wrong output (or no output at all), but also by
from potentially incorrect or harmful inputs from providing the correct output too late. The tasks that
the pilot? One school of thought requires that the run on the system must therefore be scheduled in
pilot be empowered to override the system at will in such a way that the real-time deadlines are met. Task-
an emergency, to almost fly the wings off the air- scheduling algorithms are a very active area of re-
plane if that is what is required to maintain safety search in real-time computing. Typically, one assigns
and save lives. The argument in favor of this approach tasks to processors and then the operating system of
is that the actual envelope of safe (or at least sur- each processor executes a scheduling algorithm to
vivable) operation is somewhat broader than the for- schedule them (that is, to decide what task is to
mal safety envelope. The pilot of a commercial jetliner run at what time). One example of a scheduling al-
can be expected to have, through years of experience gorithm is the earliest deadline first policy, in which,
and training, enough experience to make judgments as the term implies, the task with the earliest dead-
about what risks are worth taking in a grave emer- line is picked to run. The designer must ensure
gency, one that may not have been foreseen in its en- that there are enough computational resources to
tirety by the designers of the fly-by-wire system. meet all task deadlines: To do this requires a care-
The other school of thought would have the fly- ful analysis that takes into account how long exe-
by-wire system be the ultimate arbiter of what cution of each task will take in a worst-case scenario,
control inputs are safe, and modulate the pilots com- which tasks need to be run and at what rate, the ef-
mands when this is deemed appropriate. Proponents ficiency of the scheduling algorithm, and the capa-
of this viewpoint may argue that the probability of bilities of the individual processor.
the pilots making a mistake is sufficiently large that
the system should be able to winnow out commands
that it believes to be dangerous. Fault Tolerance
Both schools would probably agree that when Since lives depend on the successful operation of fly-
the pilots input is obviously harmful, it should be by-wire systems, the systems must exhibit extraor-
FLY-BY-WIRE 261
dinarily low failure rates. We have already mentioned Acceptance tests are not perfect. There is always
the effective failure rate of one in ten billion per fly- the chance that an erroneous output will fall within
ing hour. These failure rates are far lower than those the acceptable range of output or that an unusual
of individual processors. For this reason, it is im- set of circumstances will cause a correct output to
portant that fly-by-wire systems be fault-tolerant. fall outside what is designated as the acceptable
Fault tolerance means that the system can toler- range. If the acceptable range is too narrow, the num-
ate a certain number of faults and still function ac- ber of correct outputs falsely tagged as incorrect
ceptably. This requires the presence of redundancy, tends to increase; if it is too wide, a larger number
that is, extra capacity in the system that can be uti- of faulty outputs will slip through. Devising an ac-
lized whenever components fail. Redundancy can be ceptable compromise is a difficult and important
classified into four broad categories: hardware, soft- problem.
ware, time, and information. The designer of hardware redundancy must en-
sure that correlated failures are kept to a minimum.
Hardware Redundancy For example, the control lines that connect the pilot
Hardware redundancy is the presence of addi- inputs to the elevator and rudder controls must be
tional hardware and the associated controls required spaced as far apart as possible, to make it less likely
to manage it. Broadly speaking, there are two ways that a single event, such as the collapse of a por-
in which faults can be recovered through hardware tion of the floor, severs all of them. Also, the design
redundancy: forward and backward. In forward must be such that faults are prevented from spread-
recovery, the system is able to mask the effects of failing. For example, a short circuit in one part of the
ure so that no time is lost. Triple-modular redun- system should not trigger a chain reaction that burns
dancy (TMR), in which there are three processors out a substantial number of units. Likewise, erro-
and a voting circuit, is an example of hardware re- neous data must be kept from spreading. Fault- and
dundancy that makes forward recovery possible. The error-containment zones are generally established
three processors execute the same code, and their to suitably isolate each subset of hardware from
output is voted on. The majority result of the vote the others.
that is, the output produced by the majority of
processorsis the output of the TMR module. It Software Redundancy
is easy to see that if one of the three processors were Software redundancy is crucial because software in
to fail, the two remaining processors would consti- high-performance systems tends to be far more com-
tute the majority output and would mask the in- plex than hardware, and software faults can pose a
correct output of the faulty processor. Another considerable risk to correct functioning. Modern
example of hardware redundancy is the use of software-engineering techniques go a long way to-
multiple actuators for the same control surface: If ward reducing the number of faults per thousand
one actuator fails and pushes in the wrong direction, lines of code; however, this is still not sufficiently low
the remaining actuators should have enough phys- to meet the stringent requirements of fly-by-wire,
ical capacity to overwhelm it and manage to config- and redundancy techniques are often used. Software
ure the surface correctly. redundancy consists of using multiple versions of
In backward recovery, the system recognizes a software to do the same function and then vote on
faulty output and then reruns the program on a func- the results. One might, for example, have three in-
tioning processor. The faulty output is typically rec- dependent teams of software developers write an en-
ognized by means of an acceptance test. An acceptance gine controller module, run each version on a
test alerts the system to a fault when it detects out- different member of a TMR cluster, and vote on
put outside an acceptable range. For example, if the the results. The hope is that the number of coinci-
pressure sensor in a submersible reports that the pres- dent failures (i.e., multiple software versions fail-
sure at a depth of 200 meters is 1 atmosphere, the ac- ing on the same input) is small enough for the system
ceptance test would flag an error. to be sufficiently reliable.
Designers go to great lengths to reduce the pilot controls are received in analog form by four ac-
chances of coincident failure. The teams developing tuator control electronics (ACE) units, converted to
the various versions are generally kept isolated from digital form, and sent to the primary flight computers
one another to avoid mistakes being copied from one (PFC). The PFC complex (described in more detail
team to another. Different programming lan- below), which also receives sensor data on altitude,
guages may be used to reduce the chances of com- temperature, speed, and so forth, executes the flight
mon errors being caused by a commonality of control algorithms and sends its outputs back to the
language. They may also attempt to use different al- ACE units. These units then convert this digital out-
gorithms to do the same function, to reduce the put to analogue outputs to specify the control sur-
chance of numerical instability striking multiple ver- face settings.
sions for the same input. The PFC complex has three PFC channels: left,
center, and right. Each channel consists of three lanes,
Time and Information Redundancy which are full-fledged computational systems,
Time redundancy means having enough time to rebuilt around the AMD 29050, Motorola 68040,
run failed computations. The vast majority of hard- and the Intel 80486 processors, respectively. Thus,
ware failures are transient: They go away after each channel uses a diversity of hardware to reduce
some time. A powerfuland simpleapproach is the chances of coincident failures. By contrast,
simply to wait for a while and then retry the same software diversity is not used: The same code (actu-
computation on the same processor. If the failure ally, with slight modifications to suit the three dif-
is transient and the waiting time has been long ferent processors), in the Ada programming language,
enough for the transient effect to die away, this ap- runs on each of the three channels.
proach will work. Time redundancy is obviously also Communication is over three bidirectional buses
necessary to effect backward recovery. of about 2-megabits-per-second capacity. Each PFC
Information redundancy that relies on codes that channel can transmit on exactly one of these buses:
detect and correct errors is useful primarily in deal- These buses are labeled left, center, and right, in-
ing with faults in memory or in communication. dicating which channel is entitled to transmit on
Memory is often subject to transient upsets. For ex- that bus. Every lane on every channel monitors each
ample, when energetic charged particles (like al- of the buses. Similarly, each ACE unit can broad-
pha particles) pass through a memory cell, they cast on exactly one bus: two are permitted to use
can affect its contents. While the charged particles the left bus, and one each on the center and right
do not usually cause permanent harm to the phys- buses.
ical structure, the content has suffered a spurious The three lanes of each channel can be in one of
change. Communication is also subject to error three states: command, standby, or monitor. Only
because of noisy channels. Noisy channels pose the the command lane of each channel is permitted to
greatest problem in the case of wireless communi- transmit on the bus. To reduce the chance of coin-
cation, less of a problem when electric cables are cident failures, the system tries to ensure that the
used, and the least problem when optical fibers are command lanes of each channel use a different
used. Codes render data more resilient by setting up processor. For example, the command lanes may
internal correlations that can be exploited to detect be chosen such that the Intel lane of the first chan-
(and possibly correct) erroneous bits. nel, the Motorola lane of the second channel, and
the AMD lane of the third channel are the designated
command lanes. Of course, failures among the lanes
Example: Boeing 777 may make it impossible to ensure such hardware di-
versity among the command lanes. If the command
Fly-By-Wire System lane processor is detected (by the other lanes in its
The Boeing 777 fly-by-wire system makes exten- channel) to have failed four times, it is replaced by
sive use of redundancy to ensure high reliability. The another lane from its channel.
FLY-BY-WIRE 263
In normal functioning, therefore, each of the on Dependable Systems and Networks (DSN02) (pp. 210218).
three buses carries the output of its corresponding Cupertino, CA: IEEE CS Press.
Briere, D., & Traverse, P. (1993). AIRBUS A320/A330/A340 electri-
channel. The input values used by each channel cal flight controls: A family of fault-tolerant systems. Proceedings
are obtained by selecting the middle value of the of the Fault-Tolerant Computing Symposium (FTCS-23) (pp.
copies read on the buses. Each ACE has its own set 616623). Cupertino, CA: IEEE CS Press.
of actuators to which it relays the commands received Carter, J., & Stephenson, M. (1999). Initial flight test of the produc-
tion support flight control computers at NASA Dryden Flight Research
from the PFCs. For the sake of fault-tolerance, an in- Center (NASA Technical Memorandum TM-1999-206581). Wash-
dividual control surface may be driven by more than ington, DC: NASA.
one actuator. Cataldo, A., Liu, X., & Chen, Z. (2002). Soft walls: Modifying flight
control systems to limit the flight space of commercial aircraft. Re-
trieved July 18, 2003, from http://buffy.eecs.berkeley.edu/Research
Summary/03abstracts/acataldo.1.html
Analysis deGroot, A., Hooman, J., Kordon, F., Paviot-Adet, E., Iounier, Lemoine,
The increasing capability and decreasing cost of mod- M., et al. (2001). A survey: Applying formal methods to a software-
intensive system. In 6th IEEE International Symposium on High As-
ern processors has made fly-by-wire systems more surance Systems Engineering (HASE 01) (pp. 5564). Cupertino,
practical. Their original use was in military aircraft. CA: IEEE CS Press.
However, starting in the late 1980s, fly-by-wire has Droste, C. S., & Walker, J. E. (2001). The general dynamics case study
been used in civilian aircraft as well. Sophisticated on the F-16 fly-by-wire flight control system. Reston, VA: American
Institute of Aeronautics and Astronautics.
fault-tolerant algorithms have to be used to meet the Fielding, C. (2001). The design of fly-by-wire flight control systems.
stringent reliability requirements, and a suitable task Retrieved July 18, 2003, from http://www.shef.ac.uk/acse/ukacc/
assignment and scheduling algorithm are used to en- activities/flybywire.pdf
Johnson, B. (1989). The design and analysis of fault-tolerant digital sys-
sure that critical tasks meet their deadlines. tems. New York: Addison-Wesley.
The very high reliability requirements of fly-by- Knight, J. C. (2002). Safety-critical systems: Challenges and directions.
wire systems pose a challenge to the process of es- In 24th International Conference on Software Engineering (ICSE
timating their reliability and certifying them as fit to 02) (pp. 547550). Cupertino, CA: IEEE CS Press.
Krishna, C. M., & Shin, K. G. (1997). Real-time systems. New York:
fly. Although many reliability models for software McGraw-Hill.
exist in the literature, none of them is completely Leveson, N. (1995). Safeware: System safety in the computer age. New
convincing, and much more research needs to be York: Addison-Wesley.
done in this area. Littlewood, D., Popov, P., & Strigini, L. (2001). Modeling software
design diversity: A review. ACM Computing Surveys, 33(2), 177
Fly-by-wire can provide control capability far 208.
superior to traditional control techniques. One Perrow, C. (1999). Normal accidents. Princeton, NJ: Princeton Uni-
intriguing application, in the wake of the 2001 ter- versity Press.
Peterson, I. (1996). Fatal defect: Chasing killer computer bugs. New
rorist incidents involving aircraft, is using fly-by- York: Vintage Books.
wire to implement no-fly zones around important Riter, R. (1995). Modeling and testing a critical fault-tolerant multi-
buildings. These zones can be made part of the process system. In Proceedings of the Fault-Tolerant Computing
specified flight constraints, and the system pro- Symposium (pp. 516521). Cupertino, CA: IEEE CS Press.
Schmitt, V., Morris, J. W., & Jenney, G. (1998). Fly-by-wire: A histori-
grammed to ensure that pilot commands to breach cal and design perspective. Warrendale, PA: Society of Automo-
these zones are countermanded, while still retain- tive Engineers.
ing flight safety. Storey, N. (1996). Safety-critical computer systems. New York: Addi-
son-Wesley.
Thomas, M., & Ormsby, B. (1994). On the design of side-stick con-
C. M. Krishna trollers in fly-by-wire aircraft. ACM SIGAPP Applied Computing
Review, 2(1), 1520.
Voas, J. (1999). A world without risks: Let me out! In 4th IEEE Inter-
national Symposium on High Assurance Systems Engineering (HASE
FURTHER READING 99) (p. 274). Cupertino, CA: IEEE CS Press.
Yeh, Y.C. (1998). Design considerations in Boeing 777 fly-by-
wire computers. In 3rd IEEE High Assurance Systems Engineering
Aidemark, J., Vinter, J., Folkesson, P., & Karlsson, J. (2002). Experi-
Conference (HASE) (pp. 6473). Cupertino, CA: IEEE CS
mental evaluation of a time-redundant execution for a brake-
Press.
by-wire application. Proceedings of the International Conference
Johannes Gutenberg (c. 13951468) invented the

FONTS first reliable means of casting individual, move-
able, and reusable metal printing type. The first print-
Font terminology derives from letterpress printers intended printed books to rival scribes
ing, and comprises the elements of printed language manuscripts, and so the printed roman type repli-
(the letters, numbers, punctuation marks, and other cated the humanistic minuscule and the humanistic
items) called a character set. The traditional use of cursive writing styles.
the term font changed with digital technology. Font Printed books replaced manuscripts, and de-
once meant a character set at one size, in one style, signed fonts replaced pen-based replications. Fonts
of one typeface, within one family. Helvetica bold that evolved in Italy, France, and England from about
italic 12, for example, is the twelve-point font of the 1470 to about 1815 weathered critical evaluation over
bold style of the italic typeface from the Helvetica time and remain standards. Fonts designed after
family. about 1815 generally were extensions, revivals, or
Font in todays common use generalizes the variations of these earlier models.
more specific terms typeface (a character sets in-
trinsic design and visual properties), type style (a
typeface variation), and typeface family (font char- Style Properties
acter sets, components, and variations that com- Style properties of fonts are width (the character
prise a group through shared identifying width excluding serifs, serifs being the horizontal ter-
characteristics). This article uses font in the general mination to vertical strokes), weight (the vertical
common usage. stroke thickness), contrast (the horizontal stroke
thickness compared to the weight), serif form (brack-
eted, unbracketed, or none), x-height (the height
Historical Background of the small letters), and curve axis (either inclined
Language is the unified and systematic means of com- or vertical). Table 1 gives the style properties of four
munication by standardized sounds and signs that common fonts.
function through convention. For European culture Font size is the body height measured in points
and its offshoots in the New World, this means al- and is not the letter size. In digital type, one point
phabetic sign groups, written and read horizon- is one seventy-second of an inch. A metal type piece
tally from left to right. is an elongated rectangle. The body end shows the
History identifies a three-stage development to- character in high relief for making the stamp-like
ward the alphabet. At first there are forms for ob- ink transfer onto paper. Most languages require a
jects, called pictograms; then there are forms for uniform character line, and so require a uniform
ideas, called ideograms; finally there come to be forms height system.
for sounds, called phonograms. These signs and sym- Type characters and type bodies vary in width
bols evolved from nonstandard forms without an but not in height, and do not take up the entire body
organizational system to standard forms and systems surface. Letter size varies among fonts in the same
and thence to the alphabet. point size. Garamond, for example has a smaller x
The visual form of Western European written height than Helvetica, and so appears smaller at
languages derives from the Roman monumental cap- the same font size. Digital font size retains the defi-
itals, or majuscules, that date from about 100 CE, and nition that font size is body size, although the font
the Carolingian minuscules that dates from about bodyalso called the bounding boxis no longer
800 CE. These writing styles evolved and varied widely physical.
from about 1350 to 1500, when Renaissance Normative text sizes range from eight points to
scribes formalized them as the humanistic minus- twelve points and headings usually start at fourteen
cule, or roman style, and the humanistic cursive or points but can be larger. Most current software can set
italic style. font sizes in point decimals, enabling fine adjustments.
FONTS 265
TABLE 1. Style Properties in Approximate Percentages for the Respective Letter
property
width 78.0% h 83.5% h 80.0% h 76.5% h
weight 19.0% h 12.5% h 13.5% h 15.3% h
contrast 18.5% v 48.0% v 88.9% v 42.5% v
serif form unbracketed bracketed none (sans serif) bracketed
x height 59.3% h 60.0% h 73.0% h 68.0% h
curve axis vertical rotated left vertical rotated left
h=capital letter height; v = vertical stroke thickness (weight)
Font Design the mid-1980s, Adobe Systems, Apple Computer,

Knowledge of font style evolution and the visual and Monotype Corporation collaborated to offer
properties of the written and printed archetypes pro- PostScript fonts on the LaserWriter printer. When
vides a formal understanding of font design. A sense combined with the Macintosh WYSIWYG (what
of font design aesthetics differs with the intellect and you see is what you get) display and Aldus Page-
sensibility of the designer, and the design and visual Maker, which was the first desktop publishing pro-
challenges of creating a new font are the same as they gram, digital graphics and typesetting became widely
always were. Only technological factors have changed. accessible.
Drawing is the traditional approach to visual in- Though PostScript fonts, also known as Type 1
vestigation and the design process, and font design fonts, are widespread, OpenType and Unicode are
oftenbeginswithdrawingsonpaper.Designdrawings emerging industry standards. OpenType is a one-file
resolve intercharacter and intracharacter visual prob- format developed in the mid-1990s jointly by Adobe
lems.Whilenosinglecharacter,orglyph,isperfect,the and Microsoft. It has two main advantages: cross-
designer strives for balanced compromise that yields platform compatibility, and extended Latin and non-
anaestheticandfunctionalwhole.Designersmanually Latin language application. OpenType combines
markfinaloutlinedrawingsforanchorpointsandspac- existing formats with additional extensions com-
ing side bearings. Technicians then plot anchor patible with operating systems for Mac, Windows,
points and side bearings into a font-specific applica- and Unix (with FreeType an open source font en-
tion designed to compile digital font formats and de- gine). OpenType also replaces ISO-Latin encoding
scriptor files. Testing and editing follow and continue with Unicode encoding (the mapping of character
until the font is approved for release and distribution. codes to glyphs). (Unicode is an international en-
coding standard that assigns a unique code number
to each glyph. A glyph is a visual representation of a
Digital Font Formats character.) Unicode enables OpenType to have mul-
Originally, digital typesetting required markup lan- tiple and multilingual character sets, some with more
guage to control expensive raster image proces- than 65,000 characters, and OpenType is compati-
sors (RIP) and photo-based output equipment. In ble with PostScript Type 1 fonts. Other font formats
A Personal StoryOur Most Memorable Nightmare
Many participants in a digital workflow experience unforeseen problems, and most clients, writers, editors, designers,
developers, and publishers have a most memorable project storytheir worst nightmare. Here is an anecdote to exemplify
such a situation concerning fonts.
Several years ago, a client in Asia contracted us to design an English-language publication produced in Europe. We
exchanged the files by e-mail without incident between the client in Asia and our studio in America. We were careful about
cross-platform issues such as: Mac to PC conversion, different software packages and versions, even different font pub-
lishers and versions. Throughout the project stages, we had no content or file problems, and the client approved the
project for production in Europe.
After several weeks we got a call from the frantic client that the project was on hold due to massive typographic er-
rors! Text letters had changed to different letters, letters had transposed, and so on. We could not imagine how such a sit-
uation could happen. However, it was true.
Everyone set aside finding fault, and worked together for a solution.
In review, we determined that the project began on a PC for writing and editing, changed to a Mac for design, changed
back to PC for client approval, and then was e-mailed to a European service provider who probably ran a language
translation program on the project. Still, cross-platform projects can work. While we do not know the true cause of the
problem, we suspect changes in text encoding caused the changes in letter assignments.
We learned that human-to-human interaction together with human-computer interaction is good business. We learned
to plan for technical problems, and that early and frequent communication with everyone throughout the workflow
does reduce risk and increases resolution when problems do happen.
Thomas Detrie and Arnold Holland
serve specific computing platforms, purposes, and (outline file). The two files are necessary because the
devices. These font formats include ClearType, technology used for video screen display is different
that improves font legibility on liquid crystal display from the technology used for a printed reproduc-
(LCD) screens; GX fonts, that improve font handling tion. The screen font is a low-resolution screen pixel,
in interactive graphics; and TrueType fonts, that con- or picture element representation of the printer font.
tain complete font information in one file. Despite a technique that uses a grayscale to make the
letter edges appear smooth, the limited digital in-
PostScript Fonts formation of a screen font prevents high-resolution
In 1984 Adobe Systems introduced the programming output.
language PostScript to define shapes in outline us- The screen-font file contains screen font sizes
ing Bezier curves (curves defined by two anchor- and styles. Standard sizes are nine point, ten point,
ing end points and two middle points that can be twelve point, fourteen point, eighteen point, and
moved to change the shape of the curve). PostScript twenty-four point. Standard styles are roman, italic,
fonts have no specific size. A PostScript output de- bold, and bold italic.
vice renders the characters in specific sizes as des-
ignated. Because mathematical terms specify outline Multiple-Master Fonts
fonts, they require less memory than visual bitmap- Artisans adjusted early fonts to be optimal for each
font data. Outline fonts retain smooth contours when point size. Giambattista Bodoni (17401815) dis-
slanted, rotated, or scaled to any size. plays 144 cuts of his roman type in the first vol-
Two main components comprise PostScript fonts, ume of Manuale Tipografico. Creating optimal
a screen font (bitmap file) and PostScript printer font designs for each font declined early for economic
FONTS 267
reasons, despite the development of machines like flow. Common font problems are font organization,
Bentons pantograph punch cutter (1885), which font omission, and pseudo fonts.
made it easier to cut master patterns. A font filing system makes it easier to collect and
Standard digital fonts have traditional style vari- send the fonts associated with a document, or to re-
ationthe attributes of weight and width. In 1989 construct a document if problems occur. A font
Adobe Systems introduced multiple-master tech- filing system also helps prevent errors caused by us-
nology, which allowed for the manipulation of fonts ing the wrong font, mixing font formats, or mixing
along multiple axes (in this case the term axis repre- font publishers.
sents qualities such as the weight, width, and optical The Mac operating system (OS) and Microsofts
size of a character). The master designs determine Windows operating system handle fonts. On ma-
the dynamic range for each axis in a font, and Post- chines using a UNIX operating system, however, fonts
Script technology enables interpolation between the are part of a Windows application installed widely
master designs. The dynamic range of a multiple- on UNIX machines. Many fonts available in Win-
master font with two axes, weight and width, cov- dows are not available for UNIX, so cross-platform
ers permutations from light condensed through bold documents need font formats accessible to both
expanded. Multiple-master fonts with an optical-size systems.
axis can improve legibility in smaller-sized fonts Generally, it is good to keep system folder fonts
by, for example, opening closed letter parts, reduc- to a minimum. Each font added to your system is ac-
ing contrast, strengthening serifs, and increasing tive and takes up random-access memory (RAM),
width. For larger-sized fonts, the optical-size axis can the short-term memory the computer uses to store
add refinements such as increased contrast, thin- information in process. (This is as opposed to using
ner serifs, and decreased width. storage memory, called read-only memory, or ROM).
When there are many fonts using up RAM, there is
TrueType Fonts less RAM available for applications.
TrueType is outline font technology that combines Many digital font publishers supply standard
the screen font and the printer font into one file. fonts. However, standard fonts from different pub-
TrueType fonts have no specific sizes and can have lishers are not identical. It is best to avoid mixing
benefits in a mixed-platform environment. However, publishers within the same font family. Computers
TrueType fonts seldom print well on PostScript out- do not distinguish fonts with the same name from
put equipment. TrueType technology is different different publishers, or always substitute one pub-
from PostScript technology, so a PostScript printer lisher for another. This can alter document ap-
must interpret a TrueType font or substitute a Post- pearance and increase processing times. Therefore,
Script font. This often increases raster image proces- computer users are advised to create separate font
sor time. folders and to file each font by name. Then one
Because of their incompatibility, TrueType and can have subfolders for different publishers versions
PostScript fonts are best not used in the same doc- of the same font. (For example, in a folder for Gara-
ument. Some fonts may print as bitmaps or not at mond typefaces, one might have subfolders labeled
all. Even when a PostScript font and a TrueType font Adobe Garamond and Agfa Garamond for ver-
have the same name, their metrics (the size and spac- sions of that font from those two publishers.)
ing limits) are different. This can create type ID con- Missing fonts, missing font components, and
flicts, and if the font does print, it may reflow and pseudo fonts are common font problems. Consider,
cause different line breaks. for example, the calligraphy font Apple Chancery.
Users can use the toolbar style buttons B (bold)
and I (italic) to create Apple Chancery Bold Italic.
Font Management However, Apple Chancery Bold Italic is only a screen
Font management reduces font problems to improve representation. Apple Chancery Bold Italic does not
computer performance, usability, and project work- exist as a printer font. A non-PostScript printer may
emulate the font for printing but a PostScript out- Baecker, R. M., & Marcus, A. (1990). Human factors and typography
put device will substitute a default font or not print for more readable programs. Reading, MA: Addison-Wesley.
Berry, W. T., & Poole H. E. (1966). Annals of printing. London: Bland-
the font. Only active PostScript fonts will print on ford Press.
PostScript equipment. Bigmore, E. C., & Wyman, C. W. H., (Eds.). (1978). A bibliography of
Font utility programs add convenience and pro- printing. New Castle, DE: Oak Knoll Books.
ductivity to font tasks and can help reduce font man- Bringhurst, R. (1992). The elements of typographic style. Vancouver,
Canada: Hartley and Maerks.
agement problems. In addition to handling the Carter, R., Day, B., & Meggs, P. (2002). Typographic design: Form and
problems mentioned above, font utility programs communication. Hoboken, NJ: John Wiley & Sons.
can create bitmapped fonts from PostScript out- Mller-Brockman, J. (1985). Grid systems in graphic design. New York:
Hastings House.
line fonts for accurate display. They can interpo- Dowding, G. (1961). An introduction to the history of printing types.
late missing font sizes and help improve fonts printed London: Wace.
on non-PostScript output devices. Frutiger, A. (1980). Type, sign, symbol. Zrich, Switzerland: ABC
Font utility programs can list the font style vari- Verlag.
Frutiger, A. (1989). Signs and symbols: Their design and meaning. New
ations together on a menu. Normally, the applica- York: Van Nostrand Reinhold.
tion font menu displays active fonts alphabetically Gerstner, K. (1974). Compendium for literates: A system of writing.
by attribute, not alphabetically by name. This in- Cambridge, MA: MIT Press.
creases the tendency to make style variations with Jaspert, W., Pincus, B., Turner, W., & Johnson, A. F. (1970). The en-
cyclopaedia of type faces. New York, NY: Barnes & Noble.
the font style buttons located on the application tool Johnson, A. F. (1966). Type designs: Their history and development.
bar. A font family list helps avoid pseudo fonts, London: Deutsch.
and makes it easier to select active fonts. Karow, P. (1994). Digital typefaces: Description and formats. New York:
Springer-Verlag.
Font utility programs also enable easier font ac- Karow, P. (1994). Font technology: Methods and tools. New York:
tivation or deactivation, as well as the designation of Springer-Verlag.
font sets. Font sets provide job-specific font lists. Lawson, A. S. (1971). Printing types: An introduction. Boston: Bea-
When a supplier requests the font list for a docu- con Press.
McLean, R. (1980). The Thames & Hudson manual of typography. New
ment, that list is the same as the font set. York: Thames & Hudson.
Although we have discussed fonts from a techni- McGrew, M. (1993). American metal typefaces of the twentieth century.
cal angle and, for the purposes of this encyclopedia, New Castle, DE: Oak Knoll Books.
focused on digital fonts, fonts can also be discussed Morison, S. (1999). A tally of types. Boston: David R. Godine.
Moxon, J. (1958). Mechanick exercises on the whole art of printing,
from an aesthetic, cultural, or linguistic perspec- 168384. London: Oxford University Press.
tive. Typography, a subject closely connected to fonts, Muir, P. H., & Carter, J. (Eds.). (1983). Printing and the mind of man.
places heavy emphasis on readability issues, usually Munich, Germany: Karl Pressler.
Prestianni, J. (Ed.). (2002). Calligraphic type design in the digital age:
in the context of traditional printing technology. Font An exhibition in honor of the contributions of Hermann and Gudrun
work associated with computer printing and display Zapf. Corte Madera, CA: Gingko Press.
focuses mainly on legibility issues in the generation Prust, Z. A. (1997). Graphic communications: The printed image.
of new letter designs. Tinley Park, IL: Goodheart-Willcox.
Ruegg, R., & Frlich, G. (1972). Basic typography. Zrich, Switzerland:
ABC Verlag.
Thomas Detrie and Arnold Holland Ruder, E. (1981). Typographie: A manual of design. New York: Hast-
ings House.
See also Laser Printer; Unicode Shneiderman, B. (1998). Designing the user interface: Strategies for
effective human-computer interaction. Reading, MA: Addison-
Wesley.
Steinberg, S. H. (1996). Five hundred years of printing. New Castle, DE:
Oak Knoll Books.
FURTHER READING Sutton, J., & Bartram, A. (1968). An atlas of typeforms. New York: Hast-
ings House.
Adobe Systems. (1999). PostScript language reference, 3E. Reading, MA: Updike, D. B. (1980). Printing types: Their history, forms, and use. New
Addison-Wesley. York: Dover.
GAMES
GENDER AND COMPUTING
GEOGRAPHIC INFORMATION SYSTEMS
GESTURE RECOGNITION
GRAPHICAL USER INTERFACE
G
GRID COMPUTING
GROUPWARE
Purpose: Games must have a point, no matter

GAMES how banal. The point is to some extent a goal
in itself.
Games are activities that are designed to entertain
players and that are governed by rules. Games are Modern types of games are computer or video
competitive, usually pitting players against each other games. They are a type of interactive entertainment
or against tasks set by the rules. Typically the rules in which the player controls electronically generated
define ways to win and lose. images that appear on a video display screen. Such
The elements common to all games are: games include video games played in the home on
special machines or home computers and those played
Competition: Games involve competition be- in arcades.
tween two or more players or at least a challenge A computer game is not always a video game or
to a single player. vice versa. The usual distinction today is subtle; a game
Rules: Games must have rules; after all, a game is a computer game if it is played on a general-
is its rules. purpose computer; game is a video game if it is
Goals: Players must have a drive to play a played on a computer that is specialized for game play.
game, be it for a high score, personal gain, phys- Computer games feature a large collection of direct
ical survival, exploration, or just entertainment. controls exploiting the full computer keyboard,
269
whereas video games tend to use more layers of menus That year also brought the release of the first home
or activity sequences via a game controller. games console, the Magnavox Odyssey, developed by
One difference between computer games and Ralph Baer based on the idea hed had twenty-one
video games arises from the fact that computers have years previously at Loral Electronics. The Odyssey
high-resolution monitors, optimized for one person connected to a television set and came with twelve
watching at close range, whereas video game con- games and two hand controls. More than 100,000
soles use a much lower-resolution commercial tel- units were sold during its first year, but its success was
evision as their output device, optimized for watching short lived.
at a greater distance by more than one person. As a In 1974 Atari released Home Pong. It used a new
result, most computer games are intended for sin- technology, the microchip. Ataris approach meant
gle-player or networked multiplayer play, whereas that a single chip could be used to perform all the
many video games are intended for local multiplayer operations required to play the game; this game
play, with all players viewing the same TV set. became known as Pong on a chip. The Odyssey
used separate discrete circuits for each operation
(collision detection, on-screen scoring, etc.), which
Games History meant that it was much more expensive to produce.
To understand the concept of games and its relation The Odyssey is the first example of a killer tech-
to computer technologies we should know how nology (an application for the technology so inno-
games have shaped the games industry so far. vative and fascinating that a large number of people
In 1951 Ralph Baer, senior engineer at Loral would be compelled to buy it), but it was soon priced
Electronics, suggested creating an interactive game out of the market as a succession of games using
that people could play on their television sets. His large-scale integrated (LSI) chips similar to Ataris
idea was not developed at the time, but it is the first was released by numerous competing companies.
example of anyone considering new technology as The number of Pong-type games on the market
a medium for playing games. In fact, ten years passed continued to increase until the arrival of the next killer
before the first real computer game was created. In technology. In 1976 the Fairchild company released
1961 students at the Massachusetts Institute of its Video Entertainment System (later renamed
Technology (MIT) designed and programmed the Channel-F). This system was a programmable con-
game Spacewar. Spacewar ran on the universitys pro- sole, meaning that it was not limited to a specific game
grammed data processor (PDP-1), a computer that or set of games preprogrammed into it. Atari released
took up the floor space of a small house. The track- its own programmable console, the Atari 2600, in 1977.
ball was invented to control the game. For a further Both consoles ran games from plug-in cartridges.
decade Spacewar was the limit of video gaming, Other consoles followed based on this model.
and therefore video gaming was available to a only The first single-circuit board computer, the
select few people with access to large-scale comput- Apple I, also was released in 1976. This computer
ing resources such as the PDP series machines. began a surge in the popularity and availability of
In 1971 Nutting Associates released Computer home computers, and by the early 1980s a number
Space, a game based on Spacewar and developed of competing systems had emerged, including the
by Nolan Bushnell. It was the first arcade video game, now-ubiquitous IBM PC. This was a time when
but it was not very popular, and only fifteen hundred many companies were trying to predict what the
units were sold. Bushnell attributed this failure to next killer technology would be. Some believed that
the game being too complicated, noting that people the answer was a fusion of games console and
werent willing to read instructions (Winter 2004). personal computer. Mattel and Coleco released up-
He decided to go into business for himself, calling grade packs to convert their consoles into personal
his new company Atari. In 1972, only a year after computers. Neither product was successful.
Computer Space, Atari released Pong, Bushnells sec- The belief that personal computer technology
ond arcade video game. would kill games consoles led most manufacturers to
GAMES 271
abandon games consoles. Atari switched from games occurred. Most PCs and games consoles can now
to manufacturing its own line of personal comput- play movies and music as well as video games.
ers, and Mattel and Coleco bowed out of the com-
puter industry altogether in order to manufacture
their own games console. These actions turned out Game Design
to be in poor judgment because the relatively high In interactive entertainment all games are de-
cost of personal computers led many who wanted signed to create an enjoyable, involving, challenging
just a games console to look elsewhere. In 1985 the experience for the player. The interactive enter-
Nintendo Entertainment System (NES) provided ex- tainment industry classifies its game titles by
actly what games players wanteda low-cost dedi- genre. Genres are important concepts for game de-
cated games console. It was quickly followed in 1986 sign and can often influence the technologies used
by the rival Sega Master System. in a game. Some of the common genres are action,
In 1989 Nintendo released the Game Boy, a hand- adventure, strategy, role-playing, and simulation.
held games console. By 1991 the Game Boy had been Other taxonomies (systems of classification) exist.
joined by competition from the Atari Lynx and the Game design is a broad process that involves every-
Sega Game Gear, both technically superior to the thing from the target audience to game play mechanics
Game Boy. The Game Boy had an eight-bit proces- to the atmosphere exhibited by a games content. It is
sor (a computer circuit able to process data of eight- a complex process that requires a complete under-
bit length that can be addressed and moved between standing of technology, game theory, storytelling, mar-
storage and the computer processor) and a mono- keting, team leadership, and project management.
chrome LCD screen with no built-in lighting. Its ri- The objective is for a design to be as complete as
vals both featured full-color backlit screens, and the possible but also flexible to unexpected changes
Lynx featured even a sixteen-bit processor at a designers make in the game specification. Designers
time when sixteen-bit games consoles were new. lay out the design in a primary document called the
However, both the Lynx and the Game Gear fell by game specification, which is used as the main
the wayside, and the Game Boy remained popular reference in the development process.
for twelve years until its successor, the Game Boy
Advance, was released in 2001. The Game Boys real
killer technology was its portability. Its rivals were Game Development Process
heavy and cumbersome by comparison, and their When computer games first became mainstream dur-
backlit screens used up so much power that they had ing the 1980s one person usually designed and pro-
a limited battery life. grammed a game, and that process is still the case for
Also in 1989 Sega released its sixteen-bit games simple games for low-end platforms such as mobile,
console, the Genesis (known as the MegaDrive in personal digital assistant (PDA), or even simple two-
Japan and Europe). Nintendo replied in 1991 with dimensional games for PCs. However, to create a
the Super Nintendo Entertainment System (SNES). competitive game with a three-dimensional engine,
By the mid-1990s these sixteen-bit consoles, com- an involving story, and hundreds of megabytes of
bined with the growing success of multimedia PCs media content, a game studio employing many mul-
(driven by the widespread adoption of the CD-ROM titalented people must work for four or five years.
drive), combined to kill off the midrange personal Teams work simultaneously to create the game
computers such as the Atari ST and Commodore according to the game specification. An average game
Amiga, which could not compete with the games studio will have three main teams: programming,
consoles in price and could no longer compete with art, and level design (the creation of environment,
IBM PCs and Apple Macs in performance. stages, or missions playable by a gamer in any type
Since then progress has been driven by the move of computer or video game).
toward faster processors and larger storage space, but Programmers work on coding the game engine
a move toward technological convergence also has (the core software component of a computer
game) and level-designing tools. Level designers Microsoft Windows, namely the Win32 program-
use these tools to lay out the structure of the game, ming environment. The main reason for this ac-
including content from the artists and storytelling ceptance is the availability of the Microsoft
from the game specification. The art team creates DirectX game development API, which together with
game objects (the components of games which can C++ is the industry standard.
be rendered and with which the player can react; On console platforms such as the Playstation 2,
today they mostly occur in three dimensions3D) the GameCube, or even the GBA, C++ is again used
from concept sketches. as the programming language. However, specialized
The nature of game programming most often is console hardware devices, such as the Playstation 2s
governed by the current standards in game technol- vertex processing unit (a graphics processing unit able
ogy. This scenario is common to such a point that to apply programs that process pixels by applying shad-
game engines might undergo upgrades even during ing, lighting, or other rendering effects to every ver-
the development of the games. Game implementa- tex to be processed), are often accessed by low-level
tion is as dependent on the technological standards assembly (a symbolic language that is converted by
of the games industry as it is on the requirements of a computer into executable machine-language pro-
a game specification. grams) calls rather than by higher level APIs.
During the past decade interactive entertainment
has made huge advances in computing. The best
known of these advances are in 3D rendering Video Technology and 3D Graphics
hardware and software. Equally important are other Computer games are best known for graphics be-
advances in game technology, such as artificial in- cause such visuals are the most important aspect
telligence, algorithms (detailed sequences of actions of human-computer interaction, and computer
or procedures to perform or accomplish some task games succeed largely in their exploitation of a com-
or solve a problem) for multiplayer networked games, puter interface.
software and hardware for sound media, and input From the early days of the PC until the Super VGA
peripherals such as controllers. Computer games (Video Graphics Array) era, video cards were meas-
have had a profound effect on computing and hu- ured by their 2D visual quality, the number of colors
man-computer interaction (HCI). Computer games they could display, and the resolutions they supported.
have been sources of innovations in several areas, in- 2D games render animated sprites (small bitmap
cluding platforms (computers and consoles), appli- images often used in animated games, but also used as
cation programming interfaces (API), graphics a synonym for icon) and a background directly to a
and sound hardware and software, peripherals, game video cards screen buffer (an area in the computers
logic, and artificial intelligence. memory (RAM) where a small amount of data is stored
for a short amount of time before it is used). Early 3D
games, such as Flight Simulator, Descent, and iD
Platform and Language Softwares Wolfenstein 3D and Doom, used software-
To date more than sixty-five video game systems have rendered raytracing (a technique for producing views
been marketed. Recent and well-known systems of a virtual three-dimensional scene on a computer)
include Dreamcast, GBA, GameCube, Nintendo 64, to calculate the pixel (the smallest picture element of
Playstation, Playstation 2, Xbox, PSP, and NGage. a digital image) values rendered to the screen buffer.
New systems are introduced almost each year. Unfortunately, software 3D rendering could not
The most popular programming languages for handle a scene with more than a few hundred poly-
computer game implementation are C and C++ lan- gons, limiting games to relatively low screen resolu-
guages. They are fast, compiled (an executable pro- tions, and was not powerful enough to remove many
gram is created from source code), high-level visual flaws from the raytracing process.
languages that work on practically all platforms. The 3D games soon evolved to the point that they
most accepted platform for computer games is needed more complex scenes with higher numbers
GAMES 273
of polygons. Although software raytracing engines Articial Intelligence

worked, they were not fast enough. Thus, engi- Artificial intelligence (AI) in games generally differs
neers became interested in implementing common from artificial intelligence in academic computer sci-
3D algorithms in hardware. During the mid-1990s ence. Much of this difference has to do with differ-
several 3D APIs were candidates for hardware im- ent goals. In games AI typically needs to exhibit only
plementation. enough realism that its actions are consistent with
OpenGL (based on SGIs (Silicon Graphics the game world and rules. AI as a computer sci-
Incorporated) IRIX GL) was a successful standard ence discipline has a broader goalto pass the Turing
API because of its ease of use and extensive indus- test (a behavioral test conceived by Alan Turing in
try support. Microsofts Direct3D was new to graph- 1950 designed to test whether or not a system is
ics but was supported by Windows 95. Hardware intelligent), in other words, to behave indistin-
vendors also proposed their own APIs, such as NECs guishably from a human. Sometimes game logic (the
PowerVR (used in the Sega Dreamcast) and 3Dfx internal mechanism of a game that performs all
Glide. The game that revolutionized 3D graphics was the tasks needed for it to work) and AI share this
iD Softwares Quake, the first 3D shooter (a game goal. However, games are less strict about what is ac-
that presents a lot of violent fighting and shooting cepted as artificial intelligence.
and depicts the action from the perspective of the Artificial intelligence in games is limited by the
characters-players) to feature entirely 3D objects us- constraint to perform as quickly as possible, leaving
ing polygon-modeled characters instead of sprite time for other game logic (e.g., collision detection
players. Although Quake ran on most computers and response) and rendering while maintaining a
with its excellent software renderer, iD Software also smooth frame rate. This constraint does not allow
developed a version called GLQuake that rendered for complex AI systems, such as a neural network (a
using OpenGL. This development occurred around general-purpose program that has applications out-
the same time that Microsoft released a hardware- side of potential fields, including almost any prob-
OpenGL driver for Windows 95 and 3Dfx shipped lem that can be regarded as pattern recognition in
its first Voodoo 3D accelerator card (a special printed some form), genetic algorithms, or even advanced
circuit board, usually plugged into one of the com- knowledge-based systems. As a result, behavior for
puters expansion slots, that makes the computer AI game agents tends to be determined by relatively
work faster). For the first time consumer-level hard- simple finite state machines (FSM) (abstract ma-
ware ran a 3D software API that could be used for chines that have only a finite, constant amount of
real-time 3D graphics. memory) with cleverly designed, predefined rules.
When graphics accelerators first hit the mar- Regardless of the differences, many concepts from
ket, the main visual improvements were smooth tex- academic AI are used in games. Chief among the con-
t u re s a n d l i g h t i n g e f f e c t s . To d ay h a rdw a re cepts is the rules-based system, a fairly generic
supports many more features. concept considering that all AI characters, as play-
As hardware progresses the capabilities of real- ers in a game, must adhere to game rules and formu-
time 3D graphics are quickly approaching those of late their own rules by which to play. AI players are
prerendered raytracing software as opposed to real- equally goals based; however, often their goals are
time raytracing software. Although real-time 3D is different from those of the human player, depend-
still raytraced, a variety of texturing and rendering ing on whether the game is competitive or cooper-
methods is used to produce lifelike, surreal, and im- ative. AI systems can be implemented using FSM,
pressive effects. Many advanced rendering techniques which analyze scenarios on a per-context basis. FSMs
devised to improve realism, and allowing photo- are often paired with fuzzy logic (an approach to
realistic quality, are now standard in most graphics computing based on degrees of truth rather than
engines. the usual true or false Boolean logic on which
OpenGL and Direct3D are the only APIs left on the modern computer is based), which quantizes fac-
the playing field of PC graphics. tors in decision making.
The most sophisticated aspects of game AI are that as many as 114 million people could be play-
often better termed game logic. Specifically, ing online games by 2006.
pathfinding (the problem of determining a path in Online games fall into a number of games gen-
a complex environment) is a challenging problem, res. At its simplest level an online game could be an
particularly for enemy bots in first-person shoot- interactive website that allows visitors to play simple
ers and movable units in real-time strategy games. games against each other, but many video games now
The most commonly used algorithms in pathfind- come with an online competitive element to them,
ing are Dijkstras algorithm for simple pathfinding allowing a player to connect to the Internet and chal-
in static mazelike environments, and A* for pathfind- lenge friends or complete strangers. Online com-
ing in more open areas, particularly with obstacles munities have grown up around such games, and
moving dynamically. developers often expend significant resources
In role-playing games involving nonplayer char- catering to them.
acters (NPCs), the human player interacts with AI
whose rules are defined by a script. These NPCs use
prewritten dialogues. A well-scripted NPC can often Controller Interaction with
act more appropriately, if not realistically, in the con-
text of a game than can a simple artificial life agent (a Games and Health Issues
virtual creature controlled by software, such as a crea- Since the early years of computer games and console
ture in Black & White), which can understand the raw gaming systems, the game controller has played an
logic of game theory but can only poorly grasp the important role in human-computer interaction by
players relationship with the game world. NPC script- allowing users to directly interact with video games.
ing plays an even greater role in defining AIs rules- A game controller is a computer input device that
based system. For example, NPCs in Biowares takes data from the user and transfers it to the com-
role-playing game Baldurs Gate use a rules-based puter or gaming console, where the software inter-
script to determine which weapons to use in which prets the data and performs the action that the
combat situations. user wanted. Examples of controllers are the typical
gamepad, the flight simulator controller, the ana-
logue joystick, the light gun (a pointing device for
Networking computers that is similar to a light pen), and
Overall, games have not changed networking nearly Sonys EyeToy.
as much as networking has changed games. The mul- These devices all work in the same way: They take
tiplayer aspect, particularly on the PC, is essential to movement and action cues from the user, translate
modern computer gaming in practically all genres. them into tiny electrical pulses that a computer
The most prominent game network software is can understand, and send them to the machine on
Microsofts DirectPlay, part of the DirectX API. It which the game is being played. Controllers mostly
builds a wrapper (a library of software programs which have been designed to provide only input. However,
hide low-level details of a programming language) now companies are designing controllers that pro-
on top of TCP/IP and other network protocols, allow- vide feedback or output in the form of force feed-
ing programmers to quickly connect game boxes on back, also known as haptic feedback, which is a
a network and exchange packets (one unit of binary rumble or vibration of the controller. For example,
data capable of being routed through a computer players can now feel when they are being tackled in
network) without having to code a sophisticated Madden NFL or feel the force of being shot in UT.
communication system from scratch. This feature adds to the fun of the game.
With broadband Internet finally starting to spread Game controllers were originally developed to
to the consumer mainstream, online gaming is start- bring the arcade experience to the home, either
ing to really take off. Market research has suggested through video games on the computer or through
GAMES 275
console gaming systems. When designing arcade cab- adapted for use in training simulations for the mili-
inets, companies could create custom sticks and pads tary, law enforcement, air traffic controllers, and op-
for each individual game, but building a new con- erators of all kinds of physical equipment.
troller for every game that is designed for home con-
soles or computers would not be feasible. When the Abdennour El Rhalibi
Atari 2600one of the first gaming consoleswas
released in 1977 it was bundled with a couple of See also Artificial Intelligence; Multiagent Systems;
square-based broomstick-like joysticks, each with an Three-Dimensional Graphics
eight-position lever and an action button. These joy-
sticks were some of the first game controllers. The
Atari joystick was built with economy and durability FURTHER READING
in mind but not ergonomics. A major problem
with such a joystick is the way in which it is operated. Adams, J. (2002). Programming role playing games with DirectX.
The constant movement of the players wrist neces- Indianapolis, IN: Premier Press.
Anderton, C. (1998). Digital home recording. San Francisco: Miller
sary to operate such a joystick can cause painful irri- Freeman Books.
tation of overused muscles and tendons, leading to Barron, T., & LostLogic. (2002). Multiplayer game programming. Rose-
RSI (repetitive strain injury) and other pathologies ville, CA: Prima Tech.
such as carpal tunnel syndrome. To avoid this health Bates, B. (2001). Game design: The art and business of creating games.
Roseville, CA: Prima Tech.
issue, developers began to design controllers with Binmore, K. (1997). Fun and games: A text on game theory. Lexington,
smaller, thumb-operated joysticks. Before thumb- MA: D. C. Heath.
operated joysticks were developed, developers used Chapman, N. P. (2003). Digital media tools. Chichester, UK; Hoboken,
NJ: Wiley Ed.
directional pads on game controllers. These direc- Crooks, C. E., & Crooks, I. (2002). 3D game programming with Direct
tional pads were used on such controllers as the orig- X 8.0. Hingham, MA: Charles River Media.
inal Nintendo controller. A directional pad had Danby, J. M. A. (1997). Computer modeling: From sports to space flight,
four directions of movement on it. A person could from order to chaos. Richmond, VA: William-Bell.
DeLoura, M. (Ed.). (2000). Game programming gems. Hingham, MA:
move up or down, left or right. Today controllers use Charles River Media.
combined efforts of directional pads, thumb-con- Deloura, M. (Ed.). (2001). Game programming gems II. Hingham, MA:
trolled joysticks, and many combinations of buttons Charles River Media.
to allow the user to play games safely. Dempski, K. (2002). Real-time rendering tricks and techniques in
DirectX. Indianapolis, IN: Premier Press.
Engel, W. F. (Ed.). (2002). Direct3D ShaderX: Vertex and pixel shader
tips and tricks. Plano, TX: Wordware Publishing.
Games as Educational Tools Hallford, N., & Hallford, J. (2001). Swords and circuitry: A design-
ers guide to computer role-playing games. Roseville, CA: Prima
People increasingly use games for education in busi- Tech.
ness, management, marketing, medicine, schools; Mulholland, A., & Hakal, T. (2001). Developer's guide to multiplayer
they even use detective games. Many video and com- games. Plano, TX: Wordware Publishing.
puter games seek to provide a realistic 3D visual expe- Preece, R. S. (2002). Interaction design: Beyond human-computer inter-
action. New York: John Wiley & Sons.
rience. Training simulators for people such as pilots, Rasmusen, E. (2001). Games and information: An introduction to game
tanker captains, soldiers, and law enforcement offi- theory. Malden, MA: Blackwell Publishing.
cers similarly seek to provide a realistic experience. Rollings, A., & Morris, D. (2000). Game architecture and design. Scotts-
The U.S. Army puts soldiers through simulated dale, AZ: Coriolis Group.
Rouse, R. (2001). Game design, theory and practice. Plano, TX: Word-
training and missions, teaching not only tactics, but ware Publishing.
also teamwork and military principles. The U.S. Navy Thomas, L. C. (1984). Games, theory, and applications. New York:
has found that use of Microsofts Flight Simulator Halsted Press.
Walsh, P. (2001). The Zen of Direct3D game programming. Roseville,
game improves the performance of flight cadets. CA: Prima Tech.
Games and training simulations are converging Watt, A., & Policarpo, F. (2001). 3D games: Real-time rendering and soft-
as techniques and technology from games are being ware technology, 1. New York: ACM Press.
gender and computing. Some explanations focus on

GENDER AND gender role socialization, which deters women from
choosing careers that society deems atypical for
COMPUTING females, such as computing. Other explanations
focus on stereotypes about computing and how
There is ample evidence of relationships between gen- the social construction of computing as masculine
der and computer attitudes, computer self-efficacy, may be keeping women away. Still others consider
computer anxiety, and computer use. In general, females how gender differences in values may be implicated
have less favorable attitudes, lower self-efficacy, greater in women's decisions not to engage in computing
anxiety, and lower levels of use than do males. These activities or computing careers. Discrimination in
gender differences are pervasive and persistent. They the classroom and workplace, either blatant or in-
have been observed in the United States, the United direct, is yet another explanation for the under-rep-
Kingdom, Israel, Romania, Hong Kong, and other resentation of women in computing.
countries and in children and adults of all ages. They
have been observed for at least three decades and us-
ing a variety of measuring instruments. Gender Role Socialization
Given persistent and pervasive gender differences Research indicates that women and girls are sys-
in attitudes, feelings, and behavior regarding com- tematically steered away from technical fields through
puters, it is no surprise that women are underrepre- school culture, traditional gender roles, and other
sented in computing majors and computing societal pressures. Gender role socialization, partic-
careers. Of the 24,768 bachelor of science degrees in ularly during the high school years, works against
computer and information sciences awarded in the young womens interest in computing. Females en-
United States in 19961997, less than 7,000 went to counter traditional attitudes about computing among
women. Fewer than one in six doctorates in computer their peers, parents, and teachers. Computing envi-
science go to women. Moreover, the gender gap has ronments, such as computer labs and clubs, are typ-
been increasing despite considerable efforts to reduce ically female-unfriendly. Not only are women in the
it. In Europe, for example, female enrollment in com- minority in these settings, they are also less likely to
puter science courses declined from 28 percent in 1978 be included in discussions there, especially informal
to 13 percent in 1985 and to 9 percent in 1998. In con- conversations. When girls receive career advice, coun-
trast, women's representation in other previously male- selors may not mention or may actively discourage
dominated fields, such as medicine and veterinary them from pursuing nontraditional career choices.
medicine, has increased dramatically. Parents contribute to gender role socialization in
Women who have entered computing tend to be the technology they make available to their children.
concentrated in the soft areas such as marketing In the United States, five times as many boys as girls
and sales and support functions such as help desk have computers to use at home, with parents spend-
and customer serviceareas requiring good inter- ing twice as much money on technology products for
personal skills, which are deemed more likely in their sons as for their daughters.
women than men. Men dominate in technical areas, Video games, an important pathway to interest in
such as systems analysis and programming. computing, are played and made primarily by males.
In addition emphasizing competition and violence,
which run counter to female socialization, video games
Explaining the Relationship between typically relegate female characters to limited roles as
damsels in distress or sideshow prostitutes. When
Gender and Computing women are the main characters in video games,
Scholars from various disciplines and practitioners they are usually provocatively dressed in ways that em-
in various computer science industries have offered phasize their sexuality (as, for example, with Lara Croft
a variety of explanations for the relationship between in the video game Tomb Raider.
GENDER AND COMPUTING 277
It is therefore unsurprising that gender differ-

Computer Girl Site Offers ences in computer attitudes, self-efficacy, anxiety,
and use are smaller in Eastern European countries
Support for Young Women than in the West. What is surprising is that gender
differences exist at all, and in the same direction as
C
alling itself a bridge from high school to the com-
puter world, the Computer Girl website (www.com- in the West. Although the evidence is by no means
putergirl.us) is sponsored by ACM-W (the Association as overwhelming as it is for the United States and
for Computing Machinerys Committee on Women in Western European countries, females in Eastern
Computing). Moderated by Amy Wu, a high school stu- European countries appear to have less computer
dent with a keen interest in computer science, the site of- self-efficacy and more computer anxiety than their
fers advice from women in computing and connects girls male counterparts, and use computers less often than
with mentors and big sisters in the field. There are also do their male counterparts.
links to relevant articles and information on scholarships
for women in computing. Below is a typical question and Stereotypes and the Social Construction
answer exchange from the site: of Computing
Q: What kind of pre-college experience should I There exists a stereotype of computer users as my-
have? opically obsessed with computing to the exclusion
A: Before college, make sure you have a computer
of everything else, including people and relation-
at home, even if it means saving for months or ships. Although both male and female computer users
buying a used one. Unless you are using a com- consider this stereotype to be not me, the image
puter frequently (constantly), you do get rusty. is more threatening for females than males because
Join a computer club or, better yet, start one of of the importance of people and relationships in the
your own, particularly a girl-only club for both gender role socialization of females.
fun and for helpful support. Try to meet successful
women in computing. Write letters, send emails,
The scholars Sherry Turkle and Seymour Papert
visit officeswhatever you can do to get advice argued in a 1990 paper that women and men may have
and gain support for your interests. Take an intern- different computer styles, and that the male style is
ship with a technology company and even try to more strongly supported by the computer culture.
target a woman-owned tech company. Do your Females computer style is relational and character-
research, do your homework and don't give up! ized by efforts to connect and interact with objects
to have a relationship with objects and with the
computer itself (for example, by anthropomorphiz-
The situation appears somewhat better in Eastern ing the computer, giving it a personality). Males com-
Europe than it does in Western Europe and the puter style is impersonal and characterized by efforts
United States. The former Communist countries to distance oneself from objects, to command and
of Eastern Europe have historically produced pro- conquer, to avoid relationships, either between ob-
portionately more female technologists, engineers, jects or between self and objects, including the com-
and physicists than Western Europe or the United puter. The social construction of computing favors
States. Soviet industrialization efforts, which em- impersonal-style users.
phasized both gender equality and the importance Studies of elementary and high school children
of technology, created a relatively gender-neutral reveal that girls have accepted the masculine view of
view of technology. Thus, as late as the 1980s, as many computing. Girls are more likely than boys to believe
if not more females than males were studying to be they will not be good at computing, that they are
engineers in those countries. The majority of chil- not qualified for computing, that they will not enjoy
dren in math and computing in Bulgaria and computing, and that they will not be able to obtain
Romania are female, comprising more than double a good position in computing. Girls believe that com-
the proportion of females in similar courses in the puting work is done in isolation, that it involves sit-
United Kingdom. ting at a computer screen all day, and that the primary
Narrowing the Gap
(ANS)Two journal entries by 13-year-old Liliana TechGYRLS and similar programs. They strive to expose
Guzman capture the idea behind TechGYRLS, a national girls to technologyand professional women who use it
program designed to narrow the knowledge and skills gap in an all-girl setting.
between girls and boys in the realm of technology. []
I like TechGYRLs because when you make a mistake Pamela Cox, an elementary school teacher and
there are no boys to laugh at you, Guzman wrote in a TechGYRLS leader, said the 14-week, after-school program
journal recording her experiences in the program in Dallas for 9- to 13-year-olds focuses on using a computer ani-
last year, when she was in sixth grade. Her second entry mation program as well as robotics construction kits.
said: I think if you try you may get it, but if you give The animation program requires them to do some
up, you won't get it. basic programming, Cox said.They start by creating geo-
Girls are not keeping up with boys when it comes to metric figures, then they can add a picture to it. It's really
computers and other technologies, say experts in the field. neat. What's great is when they get to the point where they
If technology is going to play a significant role in our can help each other. All you have to do is get them started.
future, something most experts agree is inevitable, and The Dallas program is one of seven around the coun-
women are to be equal partners in shaping that future, we try, all affiliated with the YWCA of the USA. . . .
must find ways to capture and maintain girls' interest in We're working with girls coming into a phase of
computers and technology, said Marla Williams, exec- life where their decision may be to paint their nails or
utive director of the Women's Foundation of Colorado. hang out in a mall, said Khristina Lew, a YWCA
Williams' group recently released a report reviewing spokeswoman. Or they can focus on something that will
various computer games and software marketed to boys have a positive effect on their lives. Girls can be shy or em-
and girls and found a significant gap in their educational barrassed about pursuing traditionally masculine careers.
value. Boys' products develop their computer literacy and TechGYRLS is a program that can make them more com-
programming skills, it found, while products for girls fortable with that idea and hopefully will stay with
emphasize fashion and hairdos. The report encouraged cre- them as they grow up.
ation of higher-quality software titles of interest to girls. Karen Pirozzi
Capturing and maintaining girls' interest in computers Source: Girls get help in keeping up with boys in computer skills.
American News Service, 2000.
is a goal shared by the founders of the YWCA-affiliated
activity is either programming or office administra- genders believed that a computing career would re-
tion. They also believe that computing requires math- quire a continual updating of skills, women believed
ematics, and girls tend to feel less competent in this more strongly than did men. Other findings from
mathematics than do boys. this study indicated that both genders assigned
Similar findings regarding the nature of com- gendered preferences in computing subdisciplines.
puting work were obtained in a 1998 study of For example, both believed that women preferred
adults in Australia, Hong Kong, the United Kingdom, multimedia and documentation to programming.
and the United States. Women were more likely than In discussing the social construction of computing,
men to believe that computing careers involved soli- feminist scholars have suggested three reasons why
tary work in front of a computer screen and would women are staying away from computing: commu-
not involve teamwork. They were more likely than nicative processes that handicap women (e.g., exclud-
men to believe that computer work would not involve ing them from informal communications), social
travel, and that the profession required technical, networks that favor men, and male claims to knowledge
mathematical, and management skills. Though both overlaid by gendered power relations (men having more
GENDER AND COMPUTING 279
power than women). Feminists assert that it is this mas- their contributions are less likely to be acknowledged
culine discourse, embedded in a masculine comput- and valued, and they are more likely to be discour-
ing culture, that is the major deterrent to women. aged and even ridiculed for engaging in computing
Some feminist scholars have suggested that more activities. While the downplaying of women's contri-
sophisticated theorization about computing is needed butions is frequently unintentional, the cumulative
to encourage women into the field. They point out that effect is lower self-esteem and self-confidence, espe-
computing is a new and amorphous field consisting cially about the likelihood of success in computing. By
of a set of disparate and complex practices and tech- adolescence females express less self-confidence about
nologies. Academic computing does not accurately re- computing than do males, even when faced with ob-
flect the field. It relies too heavily on mathematical jective evidence of equivalent performance.
formalism and ignores the creative approaches to com- Women who do enter computing careers face sub-
puting that are needed in the workplace. Moreover, the tle and sometimes not-so-subtle discrimination in
image of the computer scientist as antisocial and un- the workplace. They are often assigned less-chal-
interested in the end users is actually contrary to the lenging work, passed over for promotion, and less of-
requirements of most computing work for strong ten acknowledged for their contributions than are
interpersonal skills and attention to the end user. men with similar qualifications and job experience.
Gender Differences in Values

The gender gap in computing may reflect gender differ- The Internet: A Special Case?
ences in work values. Women place greater value than The Internet, once the almost exclusive domain of
do men on interpersonal relationships at work and on males, is now an equal-opportunity technology, but
work that helps others. The masculine construction of only in the United States. Females comprise just over
computing may lead women to perceive computing half of the U.S. Internet users, a dramatic change
careers as incompatible with their work values. Research from their 5 percent representation in 1995. However,
supports this view. Women state that they are ac- males are the predominant Internet users in all other
tively rejecting computing careers in favor of careers countries. For example, 75 percent of Brazilian, 84
that better satisfy their values. For example, previously percent of Russian, 93 percent of Chinese, and 96
male-dominated careers such as medicine and law have percent of Arab Internet users are male. Within the
experienced a dramatic increase in the number of fe- U.S. evidence it appears that although women are as
male participants, due at least in part to the con- likely as men to access the Internet, they do so less
strual of these professions as compatible with women's often, for shorter periods of time, visit fewer Web
valuing of relationships and service to others. domains, and engage in fewer and different Internet
Other work values that have been implicated in activities than males. For example, women are more
women's decisions not to enter computing fields are likely than men to use the Internet to communi-
the value of pay and job status, particularly when these cate with family and friends, consistent with womens
values conflict with the family values. Women value stronger interpersonal orientation.
high pay and job status less than do men, and value Some researchers have argued that aspects of
work flexibility to accommodate family more than do the online world discourage women from using the
males. Thus, the promise of high-paying, high-status Internet. Pornography and sexual predators can make
computing jobs will be less attractive to women the Internet seem dangerous. Flamingsending or
who are making career decisions than it will be to men. posting abusive, hostile remarks electronicallyis a
form of direct aggression of the sort that women are
Discrimination socially conditioned (if not biologically predisposed)
From elementary school to the workplace, comput- to avoid. Gender differences in communication dy-
ing environments have been characterized as unfriendly namics in the real world may be replicated online.
if not outright hostile to women. Females receive less Women get less respect and acknowledgement for
encouragement to engage in computing activities, their contributions; men dominate online discussions.
Is the Relationship between Gender cine, for example). In contrast, men who enter comput-
ing tend to be focused on computing itself. Women's
and Computing a Problem? interest in computing appears to develop more slowly
There are at least two reasons for believing that and much later than men's interest: Men's interest
the declining number of women in computing is a often develops in early childhood as a result of play-
problem. First, for women, the decision to refuse to ing video games, while women's interest often devel-
consider computing careers closes off potentially re- ops in high school as a result of using computers
warding futures for them. Women typically work in in as a means to an end (for example, as a tool for com-
jobs that pay less, have fewer benefits, and offer fewer pleting school projects such as science reports).Women
opportunities for advancement than the jobs typi- are more likely than men to begin their computing
cally held by men, including computing jobs. careers at a later stage (generally after age thirty), often
Excluding such careers from consideration is par- following a first career or a career break. Women in
ticularly troublesome at a time when women are computing often regard themselves as having high
more likely to be the main or sole providers for their ability in math and science. They enjoy logical think-
families. Computing careers generally pay well, pro- ing and problem solving, and see computing careers
vide good benefits for families, and offer opportu- as a good fit with their abilities.
nities for personal growth and advancement. For
women to dismiss computing, especially if dismissal External Factors
is based on misperceptions or misinformation about Women in computing report having family, friends,
computing, puts them at a disadvantage in terms of and teachers who supported their interest and deci-
their own needs as workers and family providers. sion to enter computing. Institutional support also ap-
Second, the computing field needs women work- pears to be a factor: The availability of high-quality,
ers. Currently, and into the foreseeable future, affordable education at a desired location influenced
there is a strong and unmet need for qualified work- some women's decisions to enter computing. In par-
ers in computing. One way to meet this need is to ticular, recruiting strategies that emphasize the high
encourage more women to enter the field. quality and low cost of the institution's computing
programs and that emphasize geographic factors such
as nearness to family have been more successful than
Characteristics of Women strategies that do not explicitly mention these factors.
Finally, the high pay, favorable job prospects, and
in Computing opportunities for challenging work and advancement
There are of course some women who have overcome that careers in computing offer are all reasons that
gender socialization, gender stereotypes, the mascu- women cited as important in their decision to enter
line construction of computing, and overt and subtle computing.
discrimination to find satisfying careers in comput-
ing. A considerable amount of research has focused
on identifying what characteristics set these women Increasing the Representation
apart, both from men in computing and from women
in other careers. When asked their reasons for choos- of Women in Computing
ing computing as a career, these women gave responses A number of recommendations are available on how
that in some cases reflected their personal attributes best to increase the representation of women in com-
and in some cases reflected external factors. puting. First, the image of computing needs to be re-
constructed to reflect the diversity of skills and
Personal Attributes approaches that are desirable in the field today more
Women who enter computing often do so because they accurately. In both curriculum design and pedagogy,
perceive it as related to their other interests, which greater emphasis should be placed on interpersonal,
are often (but not always) people-oriented (medi- business, and multitasking management skills, all of
GEOGRAPHIC INFORMATION SYSTEMS 281
which are needed in today's computing careers, along Brosnan, M. J., & Lee, W. (1998). A cross-cultural comparison of gen-
with technical skills. der differences in computer attitudes and anxiety: The U.K. and
Hong Kong. Computers in Human Behavior, 14(4), 359377.
Second, both boys and girls need to be provided Cassell, J., & Jenkins, H (Eds.). (1998). From Barbie to Mortal Kombat:
with positive computing experiences early on (in el- Gender and computer games. Cambridge: MIT.
ementary school or earlier). Computing environments Cone, C. (2001). Technically speaking: Girls and computers. In P. O'Reilly,
should be made more welcoming to girls. For exam- E. M. Penn, & K. de Marrais (Eds.), Educating young adolescent girls
(pp. 171187). Mahwah, NJ: Lawrence Erlbaum Associates.
ple, environments could focus more on cooperation Durndell, A., & Haag, Z. (2002). Computer self-efficacy, computer
as a means of problem solving than on competition anxiety, attitudes toward the Internet and reported experience with
to determine who has the best solution. Computing the Internet, by gender, in an East European sample. Computers in
Human Behavior, 18, 521535.
activities should take into account female socializa- Gorriz, C., & Medina, C. (2000). Engaging girls with computers through
tion and stereotypic influences on interests and pref- software games. Communications of the Association of Computing
erences. For example, multimedia computing activities Machinery, 43(1), 4249.
or activities that involve cooperation and collabora- Jackson, L. A., Ervin, K. S., Gardner, P. D., & Schmitt, N. (2001). Gender
and the Internet: Women communicating and men searching. Sex
tion, particularly among same-gender others, have Roles, 44(56), 363380.
been shown to be well received by girls. Less clear is Kirkpatrick, H., & Cuban, I. (1998). Should we be worried? What the
whether video games targeting girls are helping to research says about gender differences in access, use, attitudes and
achieve gender equity in computing. achievement with computers. Educational Technology, 38(4), 5661.
Margolis, J. & Fisher, A. (2002). Unlocking the clubhouse: Women in com-
Third, young women need more complete and puting. Cambridge, MA: MIT.
accurate information about computing careers. The Morahan-Martin, J. (1998). Males, females and the Internet. In
diversity of skills and approaches needed in comput- J. Gackenbach (Ed.), Psychology and the Internet (pp. 169198).
San Diego, CA: Academic Press.
ing and connections between computing and Panteli, N., Stack, J., & Ramsay, H. (2001). Gendered patterns in com-
other fields should be highlighted when educating puting work in the late 1990s. New Technology, Work and Employ-
students about computing careers. It should also be ment, 16(1), 317.
made clear how computing careers can satisfy a di- Robertson, M., Newell, S., Swan, J., Mathiassen, L., & Bjerknes, G.
(2001). The issue of gender within computing: Reflections from the
verse set of values. UK and Scandinavia. Information Systems Journal, 11(2), 111126.
Fourth, role models and mentors need to be as Sanger, J., Wilson, J., Davies, B., & Whittaker, R. (1997). Young chil-
available to girls and women as they are to boys and dren, videos and computer games. London: Falmer.
men. Research on gender and computing has stead- Schott, G., & Selwyn, N. (2000). Examining the male, antisocial
stereotype of high computer users. Journal of Educational Comput-
fastly and uniformly advocated the use of mentors and ing Research, 23(3), 291303.
role models for recruiting and retaining women. Taken Tapscott, D. (1998). Growing up digital: The rise of the Net generation.
together, these recommendations may help more New York: McGraw-Hill.
Teague, G. J. (2002). Women in Computing: What brings them to it,
women discover satisfying careers in computing. what keeps them in it? SIGCSE Bulletin, 34(2), 147158.
Turkle, S., & Papert, S. (1990). Epistemological pluralism: Styles and
Linda A. Jackson cultures within the computer culture. Signs: Journal of Women in
Culture and Society, 16(1), 128148.
See also Digital Divide; Sociology and HCI
FURTHER READING GEOGRAPHIC

American Association of University Women (AAUW). (2000). Tech-
savvy: Educating girls in the new computer age. Washington, DC:
INFORMATION SYSTEMS
AAUW Educational Foundation.
Balka, E., & Smith, R. (Eds.) (2000). Women, work and computeriza- A geographic information system (GIS; the abbre-
tion. Boston, Kluwer. viation is used for the plural, geographic information
Blair, K., & Takayoshi, P. (Eds.). (1999). Feminist cyberspace: Mapping
gendered academic spaces. Stamford, CT: Albed.
systems, as well) is capable of performing just about
Brosnan, M. J. (1998). Technophobia: The psychological impact of in- any conceivable operation on geographic infor-
formation technology. London: Routledge. mation, whether editing and compilation, analysis,
mining, summarizing, or visualization and display. lutions (cell sizes) ranging down to 1 meter or less.
Geographic information is a particularly well-defined Raster data is also the preferred format for digital el-
type of information, since it refers specifically to the evation models (DEMs), which represent Earth's
surface and near-surface of Earth and links observa- topographic surface through measurements at reg-
tions and measurements to specific locations (for all ular intervals. DEM data is available for most of
intents and purposes the term geospatial is synony- the United States at 30-meter resolution, and for the
mous with the term geographic). Maps are the most world at 1-kilometer resolution.
familiar form of geographic information, so a GIS can In vector representation, phenomena are repre-
be considered simplistically as a computerized col- sented as collections of points, lines, or areas, with
lection of maps, but a far wider assortment of types associated attributes. Vector representation is widely
of information can be included in a GIS than are in- used to disseminate data from the U.S. Census, for
cluded on a map, including customer records (e.g., example, providing summary statistics for states,
records that a mail order company might keep on counties, cities, or census tracts, and representing each
its customers) that are tagged with geographic loca- reporting zone as an area. Lines and areas are most
tions such as street addresses, or images of Earth's sur- often represented as sequences of straight-line seg-
face from remote sensing satellites, or information ments connecting points, and as such are termed poly-
gathered using the Global Positioning System (GPS). lines and polygons respectively. Vector representation
Today GIS development is a major application of is also used for the street centerline databases that de-
computing technology, with an annual market for soft- scribe the locations of streets, roads, and highways,
ware, data, and services totaling on the order of $10 bil- and are widely used to support way finding.
lion. The general public is likely to encounter GIS A vector GIS is also capable of representing rela-
through Web-based services such as MapQuest that of- tionships between objectsfor example, between points
fer maps and driving directions computed from digi- representing incidents of crime and the neighborhoods
tal maps. Most municipal governments will use GIS to in which the crimes occurred. This capability allows
track, manage, and plan the use of their geographically places of work to be linked to workers' home locations,
based assets and activities, as will utility and telecom- or connections to be made between bus routes. Because
munication companies, resource management agen- relationships are in general unaffected by stretching or
cies, package delivery companies, and departments of distortion of the geographic space, they are generally
transportation. GIS is extensively used in the military termed topological data, to distinguish them from geo-
for tasks such as targeting missile systems, planning bat- metric data about object positions and shapes.
tlefield tactics, and gathering intelligence. A GIS database makes use of both raster and vec-
tor formats, and typically will contain several distinct
layers, or representations of different phenomena
GIS Representations over the same geographic area.For example,layers might
Two major forms of data representation are used include representations of maps of topography, soils,
in GIS: raster and vector. In raster form, an area is roads,rivers and lakes,and bedrock geology.By including
represented as an array of rectangular cells, and vari- all of these layers in a single database, it is possible to use
ation of some phenomenon of interest over the area the GIS to explore relationships and correlations, for ex-
is expressed through values (e.g., degrees of light or ample between soils and bedrock geology, and to com-
darkness, colors, or numeric values representing such bine layers into measures of suitability for various types
properties as annual rainfall) assigned to the cells. of land use or vulnerability to pollution.
This form is used for remotely sensed images from
satellites, and today GIS users have easy access to a
wide range of such images, from government sources GIS Functions
such as NASA and the U.S. Geological Survey, to The most important parts of a GIS are those that sup-
commercial sources such as IKONOS and Quickbird. port its basic functions, allowing users to compile,
Images of interest to GIS users will have spatial reso- edit, store, and display the various forms of geographic
information. Defining location on Earth's surface can by hand, given the technical issues involved in mov-
be a complex task, given the large number of altering or deleting hand-drawn lines on paper maps. The
native coordinate systems and projections available other was the measurement of the area of arbitrar-
to mapmakers. A GIS thus needs the ability not only ily shaped zones on maps, as required, for exam-
to convert between raster and vector representations, ple, in the task of inventorying land use, or the
but also to overcome differences between coordi- planning of new subdivisions. This use of GIS as a
nate systems (such as the latitude-longitude coordi- personal assistant that can perform tasks on geo-
nate system), between map projections (such as the graphic data that the user finds too tedious, expen-
Mercator projection or the Lambert Conformal Conic sive, inaccurate, or time-consuming to perform by
projection), and between the various mathematical hand, drove almost all of the first thirty years of GIS
figures used to approximate the shape of Earth. In the development. It required modes of human-computer
United States, for example, geographic data may use interaction (HCI) that were suited to the task, pro-
either of two mathematical figures: the North viding a comparatively skilled user with easy access
American Datum of 1927, based on the Clarke el- to information and the results of analysis.
lipsoid of 1886, or the newer North American Datum More recently, however, a number of other re-
of 1983, based on a unified global geodetic system. quirements have come to dominate developments
Many distinct coordinate systems are in use, rang- in GIS-driven HCI. A GIS that is used in a vehicle to
ing from the high-accuracy State-Plane Coordinate provide the driver with instructions on reaching his
systems defined by each U.S. state to the lower-ac- or her destination must convey the information with-
curacy Universal Transverse Mercator system origi- out distracting the driver from the driving task. Many
nally devised for military applications. of the in-vehicle navigation systems that are now be-
Once the foundation for such basic operations has ing installed in cars, either as original equipment or
been built, a GIS developer can quickly add a vast ar- as optional accessories, provide for auditory in-
ray of functions and capabilities. These may include structions as well as visual output, and may in-
sophisticated algorithms for designing and printing clude voice-recognition functions for input as
hard-copy maps, algorithms to identify optimum well. HCI issues also arise when GIS must be de-
routes for vehicles through street networks or opti- signed for use by people with visual impairment, and
mum locations for new retail outlets, methods for com- there have been several interesting developments
puting correlations between data in different layers or along these lines in the past ten years.
for combining layers into measures of suitability, Advanced GIS use requires a high level of skill
and methods for evaluating potential land use deci- and training on the part of its user. Courses in GIS
sions. All these uses are termed spatial analysis; when often include advanced work in map projections and
applied to extremely large data sets in an exploratory in spatial statistics. Thus another set of HCI issues
mode they are termed data mining. arise in GIS applications that are designed for use by
The list of supported forms of spatial analysis is children, or by other groups whose knowledge of ad-
huge, and an industrial-strength GIS will offer liter- vanced GIS concepts is limited. Encarta, Microsofts
ally thousands. In addition, there is an active market CD-ROM encyclopedia, for example, offers a
in extensions to basic GIS products, offered by third number of functions associated with geographic
parties and designed to be compatible with a spe- information, including simple mapmaking, and
cific vendor's base product. retrieval of information using maps as organizing
frameworks. A child cannot be expected to under-
stand map projections, so it is common for such sys-
GIS as Human-Computer Interaction tems to display information on a three-dimensional
The original motivation for the development of GIS globe rather than on a flattened or projected
in the 1960s came from the need to automate cer- Earth. A child cannot be expected to understand the
tain basic operations. One was map editing, which cartographer's concept of scale or representative frac-
is very difficult and time-consuming if performed tion, so such systems resort to clever metaphors as
284 BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION;;
Geographic Information Systems Aid Land Conservation
(ANS)Technology that helps a firm like Sears, Roebuck the impact of political issues on their neighborhoods. For
and Co. find a good department store location or steers a example, it can project graphic overlays on maps that il-
developer to an ideal housing site is also proving useful to lustrate land use issues affecting a neighborhood and
nonprofit organizations concerned with land stewardship sources of campaign dollars in a city election.
issues such as conservation, environmental justice and We were able to use this technology to depict the trends
sustainable development. in pesticide use in the state, said Steven Romalewski, the
The technology is known as Geographic Information program's director, as well as (mapping) where the ma-
Systems, and computer giant Hewlett Packard and soft- jority of Mayor Rudolph Giuliani's campaign contribu-
ware maker Environmental Systems Research Institute Inc. tions were coming from. He said many New York City
have awarded grants totaling $6 million to put it in the voters seemed interested to learn it didn't come from one
hands of land preservation groups. of the city's five boroughs.
While GIS analyzes large amounts of complicated in- The Southern Appalachian Forest Coalition, an alli-
formation about an area of terrain and turns it into easily ance in six states from Virginia to Alabama, used GIS tech-
understood maps and graphics, it is costly and requires nology to raise awareness of old-growth forest preservation
trained users and hefty computer hardware. and to identify remaining wild areas. The California Wild-
According to Hewlett Packard executive Forrest Whitt, lands Project is using its GIS grant to create a statewide
some people in the computer industry wanted to even map for habitat conservation.
the playing field by putting GIS technology employed by John Maggio
mineral exploiters and private developers into the hands Source: Land Conservation Groups Benefit from Development
Technology. American News Service, March 9, 2000.
of nonprofit groups.
The New York Public Interest Research Group (NYPIRG)
used grant money awarded in 1998 to help voters track
the basis for specifying level of detail, for example approximation, or sampling of the real thing that
by allowing the user to raise or lower the view- it purports to represent. Only in certain very limited
point relative to Earth's surface, revealing less and circumstances, such as the representation of objects
more detail respectively. that are truly mathematical, including the straight
lines of land surveys, is it possible to achieve close to
perfect representation.
Virtual Reality and Uncertainty This fundamental principle of GIS representa-
A GIS contains a representation of selected aspects tions has led to great interest in the topic of uncer-
of Earth's surface, combining raster and vector tainty, which can be defined as the difference between
formats to achieve a representation using the binary what the database tells the user about the real world
alphabet of digital systems. When used at the office and what the real world would reveal to the user if
desk, as is typical of the vast majority of GIS applica- visited directly. In some cases uncertainties can be
tions, the representation in effect replaces the real resolved from the user's own memory, particularly
world, limiting its user's perception of reality to if the user has visited the area that is represented in
the information contained in the database. The the database. GIS use is thus always most success-
real geographic world is infinitely complex, reveal- ful if combined with personal knowledge.
ing more detail the closer one looks apparently ad There are many sources of uncertainty, includ-
infinitum, so it follows that any database represen- ing measurement error (it is impossible to meas-
tation must be at best a generalization, abstraction, ure location on Earth's surface exactly), generalization
(the omission of local detail from representations in Mobile GIS is already in use in many applications.
the interests of simplicity or limiting data vol- Utility company workers, surveyors, and emergency
ume), vagueness in the definitions of key terms (for incident managers already routinely have access to
example, there are variations in the way soil and land GIS capabilities through suitably configured PDAs,
cover are classified, and confusion on the part of the although these devices are more likely to be used for
user about what the data are intended to represent. data input than for analysis. Information technolo-
This last type of uncertainty commonly arises when gies are routinely used by scientific workers in the field
the user of a certain collection of data misunder- to record observations, and GPS transponders are used
stands the intent of the creator of the data, perhaps to track animals to develop models of habitat use.
because of poor documentation. For those uses of mobile GIS, the user is in con-
Uncertainty has been studied within several the- tact both with the database and with the reality
oretical frameworks, including geostatistics, spatial represented by the database. The term augmented re-
statistics, and fuzzy-set theory. Each has its benefits, ality (AR) is often used to describe those uses, since
and each is suited to particular settings. Statistical ap- sensory reality is being extended through informa-
proaches are most appropriate when uncertainty tion technology, allowing the user to see things that
arises because of measurement error, or when it are for one reason or another beyond the senses.
can be characterized using probabilistic models. Fuzzy AR can be used to see under the surface of the
sets, on the other hand, appear to be more appro- street when digging to install new pipes, allowing the
priate when dealing with imperfect definitions, or construction crew to avoid accidentally damaging ex-
when experts are uncomfortable making precise clas- isting facilities. AR can be used to address the inability
sifications of geographic phenomena. of visually impaired people to see their surroundings,
and exciting developments have occurred recently in
the development of systems to aid such personal way
Augmented Reality and Mobile GIS finding. AR can be used to superimpose historic views
In recent years the development of wireless networks on the field of view, creating interesting opportuni-
and miniaturized devices has raised the possibility ties in tourism. The ability of a cell phone user to see
of a fully mobile GIS, no longer confined to the desk- the locations of nearby businesses displayed in map
top. Laptop computers now have virtually the form on the cell phone screen is also a form of AR.
same computational power and storage capacity as The long-term implications of AR are profound,
desktop workstations, and wireless networks can pro- since they give people the ability to sense aspects of
vide bandwidths approaching those available via their surroundings that are beyond their senses. AR
Ethernet and other local-area networks. Laptops are also presents concrete problems for HCI. The dis-
relatively cumbersome, however, with heavy battery plays provided by laptop computers and PDAs are
consumption, and personal data assistants (PDAs) adversely affected by the strong light conditions typ-
offer better mobility with some sacrifice in com- ical of the outdoors. Displays that clip on eyeglasses
putational power and storage capacity. Wearable offer comparatively high resolution (similar to a
computers are increasingly practical; with wear- PDA), but make it very difficult to implement point-
able computers a central processing unit and stor- and-click interaction. Displays on cell phones are too
age devices are packed into a cigar-box-sized package small for many GIS applications, which tend to re-
to be worn on the belt, and visual output is provided quire large display areas for suitable resolution (it is
through devices clipped to eyeglasses. In summary, difficult, for example, to annotate street maps with
then, we are approaching a point at which it will names on a cell phone screen). Reference has already
be possible to use GIS anywhere, at any time. This been made to the problems of visual display for driv-
clearly has the most significant implications when ers. Input devices, such as one-handed keyboards, are
GIS is used in the subject area, allowing the user to also difficult to use. Finally, heads-up display, in which
be in direct sensory contact with the phenomena be- information from the GIS is superimposed directly
ing studied and analyzed. on the field of view, requires head-mounted displays
that are much more cumbersome and difficult to use provides the format standards to ensure sufficient uni-
than a display clipped to the eyeglasses. formity in these data sets as well as the metadata stan-
dards to allow parts of the patchwork to be described
effectively.
GIS and Data Sharing Since 1993 the United States has made enormous
Traditional paper maps are very efficient reposito- investments in information technology infrastruc-
ries of geographic information, and as such repre- ture for data sharing. In accordance with U.S. law, the
sent major investments. The typical national vast bulk of geographic information produced by the
topographic map sheet of the U.S. Geological Survey, federal government is in the public domain, free of
for example, covering an area approximately 15 kilo- copyright restrictions, and available for no more than
meters on a side at a scale of 1:24,000 costs on the the cost of reproduction. Today the amount of such
order of $100,000 to create, and must be regularly data available from websites is on the order of
updated with new information if it is to remain cur- petabytes (quadrillions of bytes), and growing rap-
rent. It takes some fifty thousand such sheets to cover idly. This free resource has in turn stimulated the de-
the forty-eight contiguous states, and if the entire velopment of a wide range of applications and an
series were to be recreated today the total investment industry dedicated to adding value to data by mak-
would be in excess of $5 billion. Remote sensing satel- ing it easier to use, more current, or more accurate.
lite programs require investments in the hundreds The term geolibrary has been coined to de-
of millions; a 1993 study by the U.S. Office of Man- scribe the websites that provide geographic infor-
agement and Budget found total annual investment mation. By definition a geolibrary is a library that can
in geographic information by federal agencies to ex- be searched based on geographic locationthat is
ceed $4 billion. Not surprisingly, then, society has capable of answering queries of the form What have
traditionally relied on national governments to make you got about there? The National Research Council
these kinds of investments, through national map- has explored the concept and status of geolibraries
ping agencies and national space programs. Only na- in one of a series of reports pertaining to the NSDI.
tional governments have been able to afford the cost Geolibraries present interesting issues of user inter-
of the complex systems needed to create maps. face design. All allow users to display a world map
Today, this picture is changing radically. Anyone and to zoom in to an area of interest, refining the
with $100 can purchase a GPS receiver capable of search criteria with additional requirements. But the
determining location to better than 5 meters and can area of interest for many users is defined not by a loca-
use it to make digital maps of local streets or prop- tion on a map or by coordinates, but by a place-name,
erty boundaries. Mapping software is available for and many users will not be able to easily locate that
the average PC, with the result that today virtually place-name on a map. This issue is solved through
anyone can be a cartographer, making and pub- the use of a gazetteer, an index that converts place-
lishing maps on the Internet. names to coordinates. But most gazetteer entries pro-
Moreover, national governments find it increas- vide only a point reference, which is problematic for
ingly difficult to justify the kinds of annual expendi- extended features of complex shape, such as rivers or
tures needed to maintain mapping programs. In 1993 mountain ranges.
the U.S. National Research Council began to advocate The U.S. National Geospatial Data Clearinghouse
the concept of a National Spatial Data Infrastructure is an example of a geolibrary that allows search
(NSDI), a set of institutional arrangements and stan- and retrieval across a distributed archive, in effect
dards that would coordinate a new form of decen- allowing its users to visit and search several libraries
tralized production of geographic information. The simultaneously and with minimal effort. Such ca-
NSDI is intended to support a patchwork approach, pabilities are made possible by metadata, the infor-
replacing uniform, government-produced series of mation that describes the contents of data sets in
maps with coverage at varying scales produced as ap- standard form, allowing vast catalogs of data to be
propriate by local, state, or federal agencies. The NSDI searched quickly and easily. The dominant metadata
standard for geographic information is the Federal contributors and would return suitable hits. The user
Geographic Data Committee's Content Standard for would then be able to use a chosen data set, but rather
Digital Geospatial Metadata. It provides hundreds than copying it to the users client GIS, the data set
of potential fields for the description of the contents, would be accessed transparently over the Internet.
lineage, quality, and production details of a data set. Most GIS vendors now offer software to support
the development of GIS services and Web-based map-
ping. Some services are available free, and others on
Web-Based GIS a subscription basis. But it remains to be seen whether
Early efforts to build geolibraries, beginning in the the provision of services is capable of providing suf-
mid 1990s, focused on the need to distribute data ficient cash flow to a company, and whether Web-based
sets, by analogy to the traditional library whose re- GIS is a viable long-term commercial proposition.
sponsibility ends when the book is placed in the
hands of the reader. Under this model each user was
required to maintain a full GIS capability, since all The IT Mainstream
transformation and analysis occurred at the users The history of GIS has been one of specialized appli-
end. Since the advent of the Web and the widespread cation of information technology. The development
popularity of Web browsers, more and more serv- of the first GIS in the 1960s required many original
ices have been developed by servers, with a conse- developments and inventions, including the first map
quent reduction in the complexity of the software scanner, the first topological data structure, and the
the user must have to use the data sets. Today, a user first algorithm for map overlay. Today, however, the
of a standard Web browser can access services for majority of the software in a GIS is industry standard,
many basic GIS operations. The task of geocoding, implementing mainstream solutions for operating
for example, which consists of converting street mail- systems, application development, object-oriented
ing addresses to coordinates, is now available from database design, and graphic interfaces. Undoubtedly
a number of sites, including MapQuest. Similarly GIS has become closer over the years to the IT main-
it is possible to access remote services for converting stream, and today it is common for records in large
place-names to coordinates, and it is expected that database solutions to be tagged with geographic loca-
more and more GIS services will be available in tion. For example, the locations of credit card transac-
this form by 2010. tions are routinely tagged with location in space and
To be automatic and transparent, such Web serv- time to support mining for evidence of misuse through
ices require adherence to standards, and many the detection of anomalous behavior. Mainstream
such standards have been developed in recent solutions are attractive to software developers be-
years by such organizations as the Open GIS cause they allow massive economies of scale through
Consortium. They allow a user's client GIS to request the adoption of standard technologies that can serve
data from a geolibrary or a geocoding service from many disparate applications.
a provider, taking care of such issues as coordinate On the other hand it is clear that GIS applica-
system transformation and clipping of data to match tions will always be to some degree distinct from
a user's study area. A vast range of mapping, geoli- the mainstream. Cell phone mapping applications,
brary, and other services are now available and fully for example, push the limits of available screen area
interoperable with popular GIS software. For ex- and resolution. GIS database applications raise dif-
ample, a user of ArcGIS, a family of GIS software ficulties when the phenomena to be represented are
products created by ESRI (Environmental Systems fundamentally continuous rather than discrete: for
Research Institute), might determine that a needed example, roads and rivers are continuous features,
data set is not available on the desktop computer's not easily broken into the discrete chunks of data-
hard drive, and might search ESRI's Geography base records. Topography deals with continuous sur-
Network website for suitable data. The search would faces not easily broken into squares or triangles for
be initiated over a distributed archive of registered discrete representation. In all of these cases the need
for discrete representation causes downstream is- Raper, J. (2000). Multidimensional geographic information science.
sues for applications (for example, representing New York: Taylor and Francis.
Snyder, J. P. (1997). Flattening the earth: Two thousand years of
continuously curved streets as straight lines with map projections. Chicago: University of Chicago Press.
sharp bends leads to difficulties in simulating driver Worboys, M. F. (1995). GIS: A computing perspective. New York: Taylor
behavior). and Francis.
Moreover, GIS is about the representation of Zhang, J. X., & Goodchild, M. F. (2002). Uncertainty in geographi-
cal information. New York: Taylor and Francis.
an infinitely complex real world, and its effective use
will always require an understanding of the nature
of that world, and the consequences of the inevitable
generalization, approximation, and sampling that
occur in digital representation. GESTURE RECOGNITION
Michael F. Goodchild The use of gesture, particularly hand gesture, as a
means of communicating with computers and
See also Navigation machines is attractive for several reasons. First, many
researchers observe that humans possess great facil-
ity in performing gestures and appear to do so spon-
FURTHER READING taneously. Second, the hands and arms always comes
attached to the human end of the humancomputer
Chrisman, N. R. (1997). Exploring geographic information systems.
New York: Wiley. interaction exchange. There is no need to hunt for the
Clarke, K. C. (1999). Getting started with geographic information sys- missing remote control or to equip the human with a
tems (2nd ed.). Upper Saddle River, NJ: Prentice-Hall. communications device if the computer could observe
DeMers, M. N. (2000). Fundamentals of geographic information sys-
tems (2nd ed.). New York: Wiley.
the user and react accordingly. Third, as the space in
Duckham, M., Goodchild, M. F., & Worboys, M. F. (2003). which we interact extends from one screen to many,
Fundamentals of geographic information science. New York: Taylor from small screen to large, and from the confines
and Francis. of the two-dimensional panel surface into the three-
Hearnshaw, H. M., & Unwin, D. J. (Eds.). (1994). Visualization in
geographical information systems. New York: Wiley.
dimensional space beyond it, gestures present the
Kennedy, M. (1996). The Global Positioning System and GIS: An promise of natural interaction that is able to
introduction. Chelsea, MI: Ann Arbor Press. match both the added expanse and dimensionality.
Leick, A. (1995). GPS satellite surveying. New York: Wiley.
Longley, P. A., Goodchild, M. F., Maguire, D. J., & Rhind, D. W. (Eds.).
(1999). Geographical information systems: Principles, techniques,
management and applications. New York: Wiley. Organizing Concepts
Longley, P. A., Goodchild, M. F., Maguire, D. J., & Rhind, D. W. (2001). One may think of the human end of the interactive
Geographic information systems and science. New York: Wiley.
MacEachren, A. M. (1995). How maps work: Representation, visu-
chain as being able to produce three key interactive
alization, and design. New York: Guilford Press. signals: things that can be heard, seen, and felt (ignor-
Medyckyj-Scott, D., & Hearnshaw, H. M. (Eds.). (1993). Human ing taste and smell as currently far-fetched for HCI).
factors in geographical information systems. London: Belhaven In this sense the computers input devices can be
Press.
National Research Council. (1993). Toward a coordinated spatial data
thought of as the sensory organs detecting the signals
infrastructure for the nation. Washington, DC: National Academy sent by its human partner. Under this formulation
Press. speech interfaces require auditory computer input,
National Research Council. (1999). Distributed geolibraries: Spatial and the plethora of input devices by which the user
information resources. Washington, DC: National Academy Press.
O'Sullivan, D., & Unwin, D. J. (2002). Geographic information analy- moves a mouse or joystick or depresses keys would
sis. Hoboken, NJ: Wiley. constitute the computers tactile sense. The sensory
Peng, Z. R., & Tsou, M. H. (2003). Internet GIS: Distributed geographic receptor for gesture is vision. One might relax this
information services for the Internet and wireless networks.
Hoboken, NJ: Wiley.
vision requirement to allow the use of various glove
Peuquet, D. J. (2002). Representations of space and time. New York: and magnetic, acoustic, or marker-based tracking
Guilford. technologies. For this discussion we shall include
GESTURE RECOGNITION 289
these approaches with the caveat that they are in- manipulative gesture systems typically use the shape
tended as waypoints toward the goal of vision-based of the hand to determine the mode of action (e.g.,
gesture understanding. to navigate, pick something up, point, etc.), while the
To move beyond promise to practice, one needs hand motion indicates the path or extent of the con-
to understand what the space of gestures is and what trolled motion.
it can afford in interaction. We organize our discus- When used in a manipulative fashion, gesture in-
sion around a purpose taxonomy. Interactive ges- terfaces have a lot in common with other direct
ture systems may be divided into three classes: (1) manipulation interfaces, the only distinction being
manipulative, (2) semaphoric, and (3) conversational. the device that is used. As such, many of the same
The human hands and arms are the ultimate multi- design principles apply in building manipulative ges-
purpose tools. We use them to modify objects around ture interfaces. These include ensuring rapid enough
us (moving, shaping, hitting, etc.) to signal one an- visual feedback for the control, size of, and distance
other and in the general service of language. While to targets of manipulation and considerations for
the psychology and psycholinguistics of gesture is a fatigue and repetitive stress order (as when one has
very involved field our tripartite segmentation ade- to maintain hand positions, poses, and attitudes by
quately covers the use of gesture in HCI. These dis- maintaining muscle tension).
tinctions are not perfunctorythey have great Gestures used in communication/conversation
significance for the vision-based processing strat- differ from manipulative gestures in several signifi-
egy employed as well as the design of the interac- cant ways. First, because the intent of the latter is ma-
tive system that utilizes the gesture. nipulation, there is no guarantee that the salient
features of the hands are visible. Second, the dynamics
Manipulative Gesture Systems of hand movement in manipulative gestures differ
Manipulative gesture systems follow the tradition of significantly from those in conversational gestures.
Richard Bolts Put-That-There system, which per- Third, manipulative gestures may typically be aided
mits direct manipulation. The user interacted with by visual, tactile, or force feedback from the object
a large wall-size display moving objects around the (virtual or real) being manipulated, while conversa-
screen, with the movements tracked by an electro- tional gestures are typically performed without such
magnetic device. As will be seen later, this work may constraints. Gesture and manipulation are clearly
also be classified as conversational since cotemporal different entities sharing between them possibly the
speech is utilized for object manipulation. feature that both may utilize the same body parts.
We extend the concept to cover all systems of di-
rect control. The essential characteristic of manipu- Semaphoric Gesture Systems
lative systems is the tight feedback between the Semaphores are signaling systems in which the bodys
gesture and the entity being controlled. Since poses and movements are precisely defined to desig-
Bolts seminal work there has been a plethora of sys- nate specific symbols within some alphabet. Tradi-
tems that implement finger tracking/pointing, a va- tionally, semaphores may involve the use of the
riety of finger flyingstyle navigation in virtual human body and limbs, light flashes, flags, and the
spaces or direct-manipulation interfaces, such as con- like. Although semaphore use inhabits a miniscule
trol of appliances, computer games, and robot portion of the space of human gestures, it has attracted
control. Other manipulative applications include in- a large portion of vision-based gesture research and
teraction with wind tunnel simulations, voice syn- systems. Semaphore gesture systems predefine some
thesizers, and an optical flowbased system that universe of whole gestures g i G. Taking a cate-
detects one of six gross full-body gestures (jumping, gorial approach,gesture recognition boils down to
waving, clapping, drumming, flapping, marching) determining if some presentation pj is a manifesta-
for controlling a musical instrument. Some of these tion of some g i . Such semaphores may be either static
approaches use special gloves or trackers, while oth- gesture poses or predefined stylized movements. Note
ers employ only camera-based visual tracking. Such that such systems are patently not sign language
recognition systems in that only isolated symbols are space-time gestures that treat time as a physical
entertained. Sign languages include syntax, gram- third dimension.
mar, and all the dynamics of spoken language sys- One of the most common approaches for the
tems. Some attempts have been made to recognize recognition of dynamic semaphoric gestures is based
isolated sign language symbols (e.g., finger spelling), on the Hidden Markov Model (HMM). First applied
but the distance between this and sign language un- by Yamato, Ohya, and Ishii in 1992 to the recogni-
derstanding is as far as that between optical char- tion of tennis strokes, it has been applied in a myr-
acter recognition and natural language understanding. iad of semaphoric gesture recognition systems.
Semaphoric approaches may be termed as The power of the HMM lies in its statistical rigor and
communicative in that gestures serve as a universe of ability to learn semaphore vocabularies from exam-
symbols to be communicated to the machine. A prag- ples. A HMM may be applied in any situation in
matic distinction between semaphoric gestures and which one has a stream of input observations for-
manipulative ones is that the former do not require mulated as a sequence of feature vectors and a finite
the feedback control (e.g., handeye, force feedback, set of known classifications for the observed se-
or haptic) necessitated for manipulation. Semaphoric quences. HMM models comprise state sequences.
gestures may be further categorized as being static or The transitions between states are probabilistically
dynamic. Static semaphore gesture systems interpret determined by the observation sequence. HMMs are
the pose of a static hand to communicate the intended hidden in that one does not know which state
symbol. Examples of such systems include color-based the system is in at any time. Recognition is achieved
recognition of the stretched-open palm where flex- by determining the likelihood that any particular
ing specific fingers indicates menu selection, the HMM model may account for the sequence of in-
application of orientation histograms (histograms of put observations. Typically, HMM models for dif-
directional edges) for hand shape recognition, graph- ferent gestures within a semaphoric library are rank
labeling approaches where labeled edge segments are ordered by likelihood, and the one with the great-
matched against a predefined graph of hand poses est likelihood is selected.
that simulate finger spelling, a flexible-modeling In a typical HMM application, Rigoll, Kosmala,
system in which the feature average of a set of hand and Eickeler (1997) were able to train a system to
poses is computed and each individual hand pose is achieve 92.9 percent accuracy in recognizing twenty-
recognized as a deviation from this mean, the ap- four dynamic semaphores using manually segmented
plication of global features of the extracted hand (us- isolated semaphores. This study illustrates the weak-
ing color processing) such as moments and aspect ness of such approaches, in that some form of pre-
ratio to determine a set of hand shapes, model-based segmentation or other constraint is needed.
recognition using three-dimensional model predic- Semaphores represent a miniscule portion of the
tion, and neural net approaches. use of the hands in natural human communication.
In dynamic semaphore gesture systems, some or A major reason for their dominance in the literature
all of the symbols represented in the semaphore li- is that they are the low-hanging fruit.
brary involve predefined motion of the hands or
arms. Such systems typically require that gestures be Conversational Gestures
performed from a predefined viewpoint to deter- Conversational gestures are those gestures performed
mine which semaphore is being performed. naturally in the course of human multimodal com-
Approaches include finite state machines for munication. This has been variously termed ges-
recognition of a set of editing gestures for an aug- ticulation or coverbal gestures. Such gestures are part
mented whiteboard, trajectory-based recognition of the language and proceed somewhat unwit-
of gestures for spatial structuring, recognition of tingly (humans are aware of their gestures in that
gestures as a sequence of state measurements, recog- they are available to subjective description after they
nition of oscillatory gestures for robot control, and are performed, but they are often not consciously
GESTURE RECOGNITION 291
constructed) from the mental processes of language biguation of the multimodal channels and the is-
production itself. The forms of these gestures are de- suing of spatial commands to a map interface. Others
termined by personal style, culture, social makeup have developed systems that resolve speech with
of the interlocutors, discourse context, and other fac- deixes in regular video data.
tors. There is a large body of literature in psychol- In Kendons (1980) parlance, a class of conven-
ogy, psycholinguistics, neurosciences, linguistics, tionalized gestures that may or may not accompany
semiotics, and anthropology in gesture studies that speech are termed emblems. The North American
lies beyond the scope of this article. We will list OK hand gesture is a typical emblem. While the
just two important aspects of gestures here. First, temporal speechemblem relationship is different
hand and arm gestures are made up of up to five from that of free-flowing gesticulation, emblem-
phases: preparation, prestroke hold, stroke, post- atic gestures in conjunction with speech have been
stroke hold, and retraction. Of these, only the stroke proposed for such applications as map interaction.
that bears the key semiotic content is obligatory. Another approach to coverbal gesticulation is to
Depending on timing there may or may not be the parse hand movements into gesture phases. Wilson,
pre- and poststroke holds. Preparations and re- Bobick, and Cassell (1996), for example, developed
tractions may be elided depending on the starting a triphasic gesture segmenter that expects all gestures
and termination points of strokes (a preparation may to be a rest-transitionstroke-transition-rest sequence
merge with the retraction of the previous gesture (ignoring pre- and poststroke holds). They required
phrase). Second, there is a temporal synchrony be- that the hand return to rest after every gesture. In
tween gesture and speech such that the gestural stroke another work Kettebekov, Yeasin, and Sharma (2003)
and the peak of the tonal phrase are synchronized. fused speech prosody and gesticular motion of a tel-
There is a class of gestures that sits between pure evision weather reporter (in front of a green screen)
manipulation and natural gesticulation. This class to segment the phases and recognize two classes of
of gestures, broadly termed deictics (or pointing ges- gestures (deictics and contours). All gestures are con-
tures), has some of the flavor of manipulation in strained to have separate preparations and retrac-
its capacity of immediate spatial reference. Deictics tions. They employed a HMM formalization.
also facilitate the concretization of abstract or dis- Sowa and Wachsmuth (2000) describe a study
tant entities in discourse and so are the subject of based on a system for using coverbal iconic gestures
much study in psychology and linguistics. Following for describing objects in the performance of an
Bolt, work done in the area of integrating direct ma- assembly task in a virtual environment. In this work
nipulation with natural language and speech has subjects wearing electromagnetically tracked gloves
shown some promise in such combination. Earlier describe contents of a set of five virtual parts (e.g.,
work involved the combination of the use of a point- screws and bars) that are presented to them in wall-
ing device and typed natural language to resolve size display. The authors found that such gestures
anaphoric references. By constraining the space of convey geometric attributes by abstraction from the
possible referents by menu enumeration, the deictic complete shape. Spatial extensions in different di-
component of direct manipulation was used to aug- mensions and roundness constitute the dominant
ment the natural language interpretation. Such basic attributes in [their] corpus geometrical
systems have, for example, been employed for query- attributes can be expressed in several ways using
ing geographic databases. A natural extension of this combinations of movement trajectories, hand dis-
concept is the combination of speech and natural tances, hand apertures, palm orientations, hand-
language processing with pen-based gestures. The s h a p e s , a n d i n d e x f i n g e r d i re c t i o n ( w w w
effectiveness of such interfaces is that pen-based ges- .techfak.unibielefeld.de/~tsowa/download/Porto.pdf)
tures retain some of the temporal coherence with . In essence, even with the limited scope of their
speech as with natural gesticulation, and this cotem- experiment in which the imagery of the subjects was
porality was employed to support mutual disam- guided by a wall-size visual display, a panoply of
iconics relating to some (hard-to-predict) attrib- Quek, F. (in press). The Catchment Feature Model: A device for multi-
utes of each of the five target objects were produced modal fusion and a bridge between signal and sense. EURASIP
Journal of Applied Signal Processing.
by the subjects. Quek, F., McNeill, D., et al. (2002). Multimodal human discourse:
This author and colleagues approach conversa- Gesture and speech. ACM Transactions on Computer-Human
tional gestures from the perspective of the involve- Interaction, 9(3), 171193.
ment of mental imagery in language production. The Rigoll, G., Kosmala, A., & Eickeler, S. (1997). High performance
real-time gesture recognition using hidden Markov Models. In
idea is that if gesticulation is the embodiment of the Proceedings of the International Gesture Workshop. Bielefeld,
mental imagery that, in turn, reflects the pulses Germany, September 1997.
of language production, then one might be able to Sowa, T., & Wachsmuth, I. (2000). Coverbal iconic gestures for object
descriptions in virtual environments:
access discourse at the semantic level by gesture- An empirical study. Post-Proceedings of the Conference of
speech analysis. They approach this using the psy- Gestures: Meaning and Use. Porto, Portugal, April 2000.
cholinguistic device of the catchment by which Wilson, A. D., Bobick, A. F., & Cassell, J. (1996). Recovering tempo-
related discourse pieces are linked by recurrent ral structure of natural gesture. Proceedings of the International
Conference on Face and Gesture Recognition. Killington, VT
gesture features (e.g., index to a physical space and Yamato, J., Ohya, J., & Ishii, K. (1992). Recognizing human action in
a specific hand shape). The question becomes time-sequential images using hidden Markov Model. Proceedings
what computable features have the semantic range of the IEEE Conference on Computer Vision and Pattern
to carry the imagistic load. They demonstrate dis- Recognition, 379385.
course segmentation by analyzing hand use, kinds

of motion symmetries of two-handed gestures, ges-
tural oscillations, and space-use distribution.
GOVERNMENT AND HCI

Possibilities of Gesture Use
Gesture use in human-computer interaction is a tan- See Digital Government; Law Enforcement; Online
talizing proposition because of the human capacity Voting; Political Science and HCI
for gesture and because such interfaces permit direct
access to large and three-dimensional spaces. The user
does not even need to manipulate an input device
other than the appendages with which they come. GRAPHICAL USER
We have laid out a purpose taxonomy by which we
can group gesture interaction systems and by which INTERFACE
design may be better understood.
A graphical user interface, or GUI, is a style of human-
Francis Quek computer interaction that presents computer objects
and actions as images on the screen. Graphical
user interfaces are mostly operated with a pointing
FURTHER READING device such as a mouse. It could be argued that with-
out graphical user interfaces personal computers
Bolt, R. A. (1980). Put-that there. Computer Graphics, 14, 262270. would not have become as successful and popular as
Kendon, A. (1980). Gesticulation and speech: Two aspects of the process they are today. The history of GUIs is usually traced
of utterance. Relationship Between Verbal and Nonverbal back to the Xerox Star user interface introduced in
Communication, 207227.
Kettebekov, A., M. Yeasin, & Sharma, R. (2003). Improving continu- 1981, but the underlying technology goes back fur-
ous gesture recognition with spoken prosody. Proceedings of the ther. The first graphics program was Sketchpad, de-
IEEE Conference on CVPR, 1, 565570. veloped by Ivan Sutherland in 1963. The first
McNeill, D. (1992). Hand and mind: What gestures reveal about
thought. Chicago: University of Chicago Press.
mouse-driven interface was developed by Douglas
McNeill, D., Quek, F., et al. (2001). Catchments, prosody and dis- Engelbart in 1968. Another key innovator was
course. Gesture 1, 933. Alan Kay, who developed the concept of the
GRAPHICAL USER INTERFACE 293
Dynabookan icon-driven notebook computer Windowscollections of icons that represent the

in the 1970s. Kay also developed the Smalltalk-80 contents of a folder, directory, or an applica-
language and system (made available for public tion.
use in 19791980), which was the first wholly graph- A trash or wastebasket icon for the removal of
ical user interface environment. At the time that objects.
Smalltalk-80 was developed, the main form of user Scrollbars, which make it possible to view large
interface was the command line, which required numbers of icons in a window by moving hor-
skilled users to type in complex command sequences. izontally or vertically through the window.
The users needed to know the names of the com- Graphical display of text in different fonts and
mands and the syntax of the commands for what- styles, text which the user can edit directly.
ever computer operating system they were using. In addition to the graphical elements on the
Operating systems such as Microsoft Windows screen, the Macintosh Finder interface intro-
and Linux still make a command line interface avail- duced the following supporting concepts:
able, but most people seldom use it. The Edit menu, which supported a common cut,
copy, and paste functions for all applications.
This was supported in turn by the scrapbook,
The Macintosh Finder which made possible the exchange of informa-
The early GUIs ran on specialist hardware that was tion between applications via the Copy buffer.
not available to the general user. It took develop- Dragging as a means of copying or moving items
ments in cheap microelectronics in the early 1980s from one folder to another, or from a folder to
to allow the production of affordable personal com- the wastebasket.
puters. The first of these was the Apple Lisa. Prior to The ability to undo operations via the Undo
producing the Lisa, the newly formed Apple menu item. This made it possible to correct mis-
Computer had produced the Apple II, which had a takes. Similarly, items in the trash could be re-
command-line-driven interface. The Lisa adopted trieved, as they were not immediately deleted.
some of the interface concepts from the Xerox Star Multiple selection of items either by dragging an
interface to become the first personal computer with area around them or using keyboard modifiers
a graphical interface. However, the GUIs memory with mouse selection.
requirements, demands on the central processing Resizable and moveable windows. The screen was
unit, and required hard-disk speed kept the Apple treated as a virtual desktop and the windows as
Lisa from becoming a commercial success. Apple fol- virtual sheets of paper that could be placed on
lowed the Lisa in 1984 with the more successful Apple top of one another or moved around the virtual
Macintosh. The first Macintosh was an all-in-one desktop.
unit with an integrated 9-inch (22.5-centimeter),
black-and-white screen. The Macintosh Finder in- Taken together, the above features provided an
terface had all the GUI components we are famil- interface that supported users in carrying out their
iar with today: tasks quite effectively. Unlike with a command line
interface, users did not have to learn and remember
A mouse for selecting graphical objects. The command names or command syntax. Nor did users
Apple mouse had one button for all actions. have to learn and remember complex hierarchies in
A toolbar of menus for each application, with order to find files and folders. Just as important, soft-
common menus such as File, Edit, and Help ap- ware developers had access to the components of the
pearing in each application. Finder user interface so that they could build graph-
Menuslists of items that represent commands ical user interfaces for the Macintosh. Today there are
to be carried out on computer objects. many other GUI toolkits, including Windows API,
Iconspictorial representations of computer Java Swing, which help programmers build interfaces.
objects, such as files and disks. A GUI toolkit gives the programmer access to buttons,
check boxes, windows, text boxes, scrollbars, menus, between tasks without having to stop and restart their
and icons for building the graphical interface for an work. X Windows uses a three-button mouse, with
application. Apple also provided a set of user inter- the user customizing the action of each button.
face guidelines so that developers could produce
interfaces that were consistent with the Finder and
with other Macintosh applications. The Design of
Graphical User Interfaces
Microsoft Windows All the applications that we use todayword pro-
Microsoft introduced its first version of a desktop cessing, spreadsheets, desktop publishing, e-mail
graphical user interface in 1985. It was not as full- tools, Web browsers, and so forthare built on the
featured as the Macintosh Finder interface, but it had same common WIMP elements: windows, icons,
the advantage of running on the cheaper hardware menus, and pointers. However, although GUIs pro-
of IBM-compatible personal computers and was thus vide the potential for improved interaction, it is pos-
available to more people. There has been an ongo- sible to produce poor ones. The keys to successful
ing debate concerning the relative merits of the Apple graphical user interface design are attention to the
and Microsoft Desktop user interfaces. Usability pro- sizing and layout of graphical items, particularly as
fessionals generally believe that Microsoft caught up these relate to Fittss Law, the appropriate use and
with the quality of the Macintosh interface with the combination of colors, the correct sequencing of ac-
release of Windows 95. Microsoft Windows uses a tions and mouse events, and the involvement of end-
two-button mouse with the left button used for selec- users in the design process.
tion and the right button used for pop-up menus, Fittss Law states that the speed and accuracy with
depending on the context. which a user can select an on-screen object depends
on the size of the object and how far the user has
to move the pointer. The implications for the us-
The X Windows System ability of graphical interfaces are that graphical
At the same time as the Macintosh appeared, re- features should be as large as practically possible
searchers at MIT were developing a graphical user given display constraints and that frequently used
interface for the UNIX operating system, called the features should be grouped near one another to min-
X Windows system. This was another important de- imize the distance that the user has to move the cur-
velopment in the evolution of GUIs, and the de- sor to activate them.
scendants of the X Windows system can be seen today Appropriate use and combination of colors is
on Linux. X Windows had some of the features of important for graphical user interfaces. Overuse
the Macintosh but also exploited the power of UNIX. of bright colors can be distracting; similarly, using
UNIX is a multitasking operating system, which the wrong combination of colors can result in color
means that it can run several applications at the same combinations that lack contrast and make text hard
time. An X Windows user could thus run several ap- to read.
plications at the same time, each in their own win- Graphical user interfaces give users freedom to
dow. This feature was later included in Apple and choose the order in which they carry out tasks.
Microsoft user interfaces; it enables users to switch Certain interactions require a mode or state in the
interface that alerts the user when there is a problem
or confirms an action such as a permanent deletion.
However, the incorrect use of modes can lock
UNIX is simple. It just takes a genius to understand its users in an unwanted state when they should be per-
simplicity. mitted to make their own choices. Such mode errors
Dennis Ritchie can be avoided by constructing dialogue diagrams
of the state of the GUI.
GRID COMPUTING 295
The same graphical interfaces that help the are carefully designed, GUIs can make all manner of
majority of users also help the software developers devices easier to use. With careful attention to the
who build them. The new techniques of visual needs of the intended users, GUIs can greatly assist
programming, using such tools as Visual Basic or us all in our lives, whether at work or play.
Macromedia Flash, enable developers to develop the
graphical user interface rapidly and to add actions David England
to the graphical elements as required. This capabil-
ity supports rapid prototyping of the user interface See also Alto; Mouse
and enables end users to get an early look at the
developing application. At this point end users can
experiment and discuss the interface with the de- FURTHER READING
veloper so that it more closely matches their needs.
Dix, A. (1997). Human-Computer interaction. New York: Prentice-
Hall.
Neilsen, J. (1994). Usability engineering. Boston: Academic Press.
Accessible Graphical Interfaces Preece, J., Rogers, Y., & Sharp, H. (2002). Interaction design. New York:
Popular and useful though they are, GUIs have the John Wiley.
potential to cause problems for users with visual and Shneiderman, B. (2003). Designing the user interface (4th ed.). Reading,
MA: Addison-Wesley.
motor impairment. Such impairment may range
from mild visual problems to complete blindness
and includes color blindness and inability to use stan-
dard pointing devices. To avoid problems for such
users, graphical user interfaces need to be augmented GRID COMPUTING
with accessibility features. For example, screen read-
ers are a computer common tool for the visually im- As the Internet ushered humanity into the Infor-
paired. Graphical user interfaces needed to be built mation Age, communication and access to comput-
with appropriate text labels and audio cues to sup- ing resources and data have become an integral part
port such users. Keyboard shortcuts are also neces- of life in the developed world. Scientists are at-
sary to support those who have difficulty using the tempting to harness the considerable resources made
mouse and similar pointing devices. For people with available through the Internet to offer computing,
color blindness, certain color combinations should communication and data solutions for those who
be avoided, as people with the condition cannot dis- require massive amounts of computer processing
tinguish the colors if they are next to each other. The power. One such solution is grid computing, also
most common form of color blindness is red- known as Internet computing, adaptive computing,
green (inability to distinguish red from green), fol- meta-computing, global computing, and even plan-
lowed by blue-green (inability to distinguish blue etary computing, referring to the much-acclaimed
from green). SETI@home Project, which depends on Internet-
Along with advances in microelectronics and connected computers to Search for Extra-Terrestrial
telecommunications, graphical user interfaces are Intelligence (SETI).
one of the cornerstones of the current digital revo-
lution. Graphical user interfaces remove the bar-
rier of complexity from computer use: People can History and Denition
work through the graphical interface on the task at Ian Foster, a computer scientist at the University of
hand rather than on the task of operating a com- Chicago, and Carl Kesselman, of the Information
puter. Graphical user interfaces have evolved from Sciences Institute at the University of Southern
the 1960s from specialized workstations to every- California, earned world recognition by proposing
ones desktop; now they are spreading to personal a new paradigm in distributed computing in the mid
devices and everyday household appliances. If they 1990s, which they referred to as grid computing.
Grid computing made it possible to use the vast nate over 10 petabytes of the data they expect to gen-
array of new networks, including the Internet, to erate from the new particle accelerator due to begin
bring globally dispersed computing resources to- operations in 2006. SETI@home uses the principles
gether. Grid computing provides computing of grid computing to harness the idle computing re-
power in much the same way that a power grid sources donated by almost 5 million personal com-
creates a single, reliable, pervasive source of energy puter users throughout the world. Commercial
by utilizing electricity generated from many suppli- enterprises find grid computing a viable option for
ers, dispersed through many geographical regions. addressing their fluctuating computing needs.
Fran Berman, Geoffrey Fox, and Tony Hey, editors Organizations have found that subscribing to a
of Grid ComputingMaking the Global Infrastruc- grid network and sharing resources is more eco-
ture a Reality, defined grid computing as follows: [It] nomical than investing in new resources. Many hard-
integrates networking, communication, compu- ware vendors, including IBM, Hewlett-Packard,
tation and information to provide a virtual platform and Dell offer solutions to commercial clientele
for computation and data management, in the same that include such services as computing-on-demand,
way that the Internet integrates resources to form a storage-on-demand, and networking-on-demand.
virtual platform for information (9). In essence, grid Theses services, coupled with specialized applications
computing refers to a set of common standards, pro- and value-added services, make grid solutions very
tocols, mechanisms, and tools that could be imple- desirable to the commercial sector. Grid computing
mented to harness idle computing resources, data also has the capacity to offer its commercial customers
resources, specialized scientific instruments, and appli- end-to-end systems integration, management and
cations in order to create a coordinated and collabo- automation, end-to-end security solutions, disaster
rative virtual supercomputer that would offer almost recovery, higher performance levels, and reduced up-
infinite processing power and storage space. front investments.
Grid Computing Applications Fundamentals of Grid Computing

Although grid computing traces its inception to wide- A grid computing network consists of three core
area distributed supercomputing, today it is used to functionality areas that create a seamless computing
support the needs of myriad disciplines that need environment for the user: grid fabric, grid middle-
dispersed data resources and high computational ware, and grid portals and applications. The grid fab-
power, such as high-energy physics, biophysics, mo- ric comprises all hardware components connected
lecular biology, risk analysis and modeling, financial to the grid. These could range from personal com-
modeling, scenario development, natural disaster puters to supercomputers that run on diverse soft-
modeling, geophysics and astrophysics, weather fore- ware platforms like Windows or UNIX. The grid
casting, computer simulation, and first-response co- fabric could also include storage devices, data banks,
ordination of emergency services. and even specialized scientific instruments like ra-
One major U.S. grid computing project is the dio telescopes. The grid fabric would also contain
National Science Foundations $53 million Tera-Grid, resource management systems that keep track of the
which connects computing resources and provides availability of resources across the grid.
one of the largest grids available. The Tera-Grid per- The next layer of components on a grid network,
forms calculations at a speed of 13.6 teraflops (13.6 known collectively as the grid middleware, may be
trillion floating-point operations per second), offers further divided into two categories core grid mid-
over 0.6 petabytes (millions of gigabytes) of disk space, dleware and user-level grid middleware. Core grid
and has a dedicated network interconnecting all the middleware programs like Globus and Legion pro-
nodes at 40 gigabits per second. The high-energy vide the basic functionality for the grid network.
physics lab of the European Organization for Nuclear They include resource allocation and process man-
Research (CERN) created a data grid to dissemi- agement tools, resource identification and registra-
GRID COMPUTING 297
tion systems, and basic accounting and time man- processing power. When a complex mathematical
agement systems. Most importantly, core grid model or a simulation, which requires immense com-
middleware also includes the core security compo- puting power, has to be processed, a consumer could
nents of the grid, which include local usage policy use the vast distributed computation power avail-
management and authentication protocols. The user- able on the grid to perform the task.
level grid middleware includes programming tools Data grids, on the other hand, are massive data
and resource management and scheduling tools to repositories that often also integrate discipline-based
efficiently utilize globally dispersed grid resources. applications. The par ticle physics data grid
Programming environments such as GrADS help (www.ppdg.net), for example, provides high-energy
computing users develop and execute programs that and nuclear physicists distributed data and com-
suit their unique requirements. Middleware sched- putational resources. Since moving extremely large
ulers like AppLeS and Condor-G provide task sched- data sets over networks is cumbersome and ineffi-
uling to efficiently manage the available computing cient, many data centers offer high-performance
resources to complete queued tasks. computing resources and specialized applications,
The final functionality areas of the grid are the which make data mining and analysis more effective
portals and applications used to access and utilize thus strengthening the global pervasiveness of data
the grid resources. Web-enabled portals allow grids. Finally, service grids provide organizations
users to interact with distributed grid resources and with the ability to adopt Web service technologies to
choose the resources that are most compatible optimize their business processes. Especially in busi-
with their task requirements while adhering to their ness-to-business environments, these service grids
security and financial constraints. Most grids are crucial in creating interoperable, stable, secure
available today offer a suite of applications that are interfaces for businesses to communicate and stream-
fully integrated into the grid network. These appli- line their operations. Similar to a dynamic Web host,
cations can be used to harness the vast computa- a service-grid-based Web host will be able to trans-
tional power of the grid or to access remote data sets fer resources and accommodate peak demand pe-
dispersed throughout the world in order to conduct riods, ensuring that servers will not crumble under
simulations and data mining projects or to perform high demand.
other complex calculations.
A number of available grids cater to specific user
groups with applications that address niche needs. Grid Computing Issues
For instance, the European Data Grid operated by Interoperability and scalability are two concepts fun-
CERN offers specialized computational and data re- damental to grid computing. Thus a standards-based
sources to the high-energy physics community. open architecture that fosters extensibility has
Monash Universitys Virtual Laboratory project been found to be most successful in implementing
(http://www.gridbus.org/vlab/) offers applications a grid network. A standard set of protocols is fun-
and data resources for research in the area of mo- damental to the way resource owners and grid users
lecular biology. NASAs Information Power Grid access, utilize, and share resources. These protocols
(http://www.ipg.nasa.gov/) provides computational govern all aspects of interaction between various
resources for aerospace research communities, while components and consumers while preserving local
the Earth System Grid (https://www.earthsystem- autonomy and the control of the resource owners.
grid.org/) caters to geoscientists working on eco-sys- Resource owners could exercise control over their
tems and climatic modeling. resources by prescribing various limitations and re-
Most grid computing networks could be cate- strictions on when, how, and by whom the resources
gorized into three main classifications based on could be utilized.
applications and user demands: computational grids, The grid should be able to integrate a range of tech-
data grids, and service grids. Computational grids nologies manufactured by various component man-
cater to consumers who require arbitrary levels of ufacturers and running on diverse software platforms
to create a single seamless entity that can handle the create grid-enabled applications or tools to harness
work demands of multiple users. A grid could in- necessary data and computing power to accomplish
clude a few integrated resources and grow to include the desired task. The user then accesses the grid
millions more. With such a complex network, the through a portal, and upon authentication to the
probability of a resource failing is high. Thus grids grid interacts with a resource broker.
should be dynamic, resilient, and adaptive to de- The resource broker is then capable of identify-
tect failed resources and make necessary changes ing the resources that match the computational
to accomplish the assigned tasks using the avail- and data needs. Upon identification of these resources,
able resources effectively and efficiently. This dy- the data and program are transported, scheduled for
namic nature of the grid creates a challenge for execution, processed, and the final results are then
resource management and the scheduling applica- aggregated and delivered to the consumer. The re-
tions that have to keep track of the ever-changing source broker follows the progress of the application
composition of the grid. and data process, making necessary changes to ac-
Grid computing is a collaborative effort that commodate changing grid dynamics and resource
brings together distributed computing resources to failures. All these activities transpire seamlessly across
meet the high computational and data demands of different technologies and software and hardware
consumers efficiently. It differs from other forms of platforms, and the consumer receives the final ag-
high-performance computing systems in several ways. gregated results, unaware of all the machines and tools
Supercomputers are often a single entity running on that cooperated to deliver the final product.
a single platform under one administrative domain Grid computing makes it possible for a user to con-
that can be dedicated to a single task. While they pos- nect to a grid, access programs, data, and instruments
sess the capacity for high-throughput, they are not dispersed throughout the globe, and interact with them
efficient in assembling dispersed data resources nor seamlessly across diverse software and hardware plat-
can they be easily integrated with other technologies. forms. Grid computing is a viable option to meet the
Even though grid computing can offer virtually growing computer needs of a world that is increasingly
endless computing power, supercomputers are more dependent on information acquisition and processing.
effective for tasks requiring low-latency and high-band-
width communications. Cluster computing often uses Cavinda T. Caldera
homogenous interconnected PCs and workstations
within a single administrative domain for high-
throughput applications. While cluster computing FURTHER READING
works much the same way as grids, they are usually ge-
Berman, F., Fox, G., & Hey, T. (2002). Grid computingMaking the
ographically restricted, smaller in the number of sys- global infrastructure a reality. Indianapolis, IN: Wiley.
tems utilized, and rarely made available for public use. Buyya, R. (2002). Economic based distributed resource manage-
ment and scheduling for grid computing. Retrieved February 2,
2004, from http://www.cs.mu.oz.au/~raj/
How the Grid Works Buyya, R., Abramson, D., & Giddy, J. (2000a). An economy driven re-
source management architecture for global computational power
Dispersed computing resources, or the grid fabric grids. Proceedings of the 2000 International Conference on Parallel
(including computers, data bases, and specialized in- and Distributed Processing Techniques and Applications (PDPTA
struments), are integrated to the grid through the 2000). Las Vegas, NV: CSREA Press.
Buyya, R., Abramson, D., & Giddy, J. (2000b). Nimrod-G: An archi-
deployment of core middleware programs like tecture for a resource management and scheduling system in a
Globus that support the basic access requests and global computational grid. The 4th International Conference on
authentication protocols. The core middleware on High Performance Computing in Asia-Pacific Region (HPC Asia
2000). New York: IEEE Computer Society Press.
these machines is then able to recognize and respond Chetty, M., & Buyya, R. (2002). Weaving computational grids: How
to authorized users on the grid. At the same time, analogous are they with electrical grids? Computing in Science
user-level grid middleware like GrADS can be used to and Engineering, 4, 6171.
GROUPWARE 299
Foster, I., & Kesselman, C. (Eds.). (1999). The grid: Blueprint for a fu-
ture computing infrastructure. Burlington, MA: Morgan Kaufman.
Early Groupware Systems: E-mail,
Foster, I., & Kesselman, C. (2003). The grid (2nd ed.). Burlington, MA:
Morgan Kaufmann.
Chat, and the Web
Foster, I., Kesselman, C., & Tuecke, S. (2001). The anatomy of the grid: The first e-mail system was created by the computer
Enabling scalable virtual organizations. International Journal of engineer Ray Tomlinson in 1971 and became gen-
Supercomputer Applications, 15(3). erally available on the Arpanet (Advanced Research
Hagel, J., & Brown, J. S. (2002). Service grids: The missing link in web
services. Retrieved February 2, 2004, from http://www.john- Projects Agency Network, the precursor to the
hagel.com/paper_servicegrid.pdf Internet) in 1972. It rapidly gained popularity as the
Information power grid: NASAs computing and data grid. (2002). Internet grew during the 1970s and 1980s. Although
Retrieved February 2, 2004, from http://www.ipg.nasa.gov/ip-
gusers/globus/1-globus.html
e-mail generally facilitates one-on-one interaction,
National and international grid projects. Retrieved February, 2, 2004, the development of mailing list tools enabled it to
from http://www.escience-grid.org.uk/docs/briefing/nigridp support widely distributed group projects. The
.htm first newsgroups were developed in the early 1980s;
Waldrop, M. M. (2002). Grid computing could put the planets in-
formation-processing power on tap. Technology Review, May 2002.
they were similar to archived mailing lists except that
users would send their text directly to a newsgroup
rather than a list of users.
The first chat-based groupware was the Arpanet
talk command, released in 1972, three years after
GROUPWARE the establishment of the Arpanet in 1969. This com-
mand allowed one user to connect to another and to
Groupware refers to any software system that is communicate by sending lines of text back and forth.
designed to facilitate group work and interaction. It is still available in most UNIX operating sys-
Groupware has been around since the 1970s. In par- tems. Multiuser chat became popular in 1984 with
ticular, e-mail and chat-based groupware have Compuserve's CM Simulator. This software was
long histories. E-mail, mailing lists, bulletin boards, modeled after citizens band radios and provided a
newsgroups, and wikis (collaboratively created web- collection of chat rooms called channels that users
sites) are examples of asynchronous groupware sys- could join. These early groupware systems led di-
tems (that is, there can be a time delay between a rectly to Internet Relay Chat (IRC), ICQ, America
message being sent, read, and then responded to). Online Instant Messaging (AIM), and other in-
Chat systems, multiuser games, group editors, shared stant messaging systems.
whiteboards, and teleconferencing tools are exam- From its inception, the World Wide Web was
ples of synchronous groupware (comments that one viewed as a tool for supporting group work. Tim
participant sends are instantly visible to other par- Berners-Lee, the World Wide Webs inventor, de-
ticipants, and multiple responses can be made in- scribed it as a distributed heterogeneous collabo-
stantly and simultaneously. rative multimedia information system (Berners-Lee
Most groupware systems are designed with the 1991). The collaborative aspect referred to the idea
different place assumption; that is, they assume that anyone with access to a Web server could create
that the users are distributed across the Internet, in- webpages and thereby help build this web of mul-
teracting with one another only through the group- timedia information.
ware itself. There has been some work in developing This vision of collaboration has been further real-
same place groupware systems for users who are ized with the rise of wikis and related systems.Wikis are
present in the same physical space. Two good ex- websites that visitors to the site can edit simply by click-
amples of such collocated groupware systems are ing on the edit link at the bottom of each page and
the Brown XMX shared editor for use in elec- providing user or password data (if required). The
tronic classrooms and the MIT Intelligent Room Wikipedia, a wiki encyclopedia, is a good example of
project. the use of wiki technology. By 2004, it had more than
The Wide World of Wikis
From Wiklopedia, which bills itself as a multilingual free- produced the first implementation of a wiki engine. Some
content encyclopedia that will always belong to everyone. people maintain that only the original wiki should be called
Wiki (upper case) or the WikiWikiWeb. Ward's Wiki re-
mains one of the most popular Wiki sites.
Key Characteristics In the final years of the 20th century, wikis were in-
A WikiWikiWeb enables documents to be authored collec- creasingly recognized as a promising technology to develop
tively in a simple markup language using a web browser. private and public knowledge bases, and it was this poten-
Because most wikis are web-based, the term wiki is usu- tial that inspired the founders of the Nupedia encyclopedia
ally sufficient. A single page in a wiki is referred to as a wiki project, Jimbo Wales and Larry Sanger, to use wiki tech-
page, while the entire body of pages, which are usually highly nology as a basis for an electronic encyclopedia: Wikipedia
interconnected, is called the wiki. was launched in January 2001. It was originally based on the
Wiki wiki means fast in the Hawaiian language, and UseMod software, but later switched to its own open source
it is the speed of creating and updating pages that is one codebase which has now been adopted by many other wikis.
of the defining aspects of wiki technology. Generally,
there is no prior review before modifications are accepted,
and most wikis are open to the general public or at least
Wiki Bus Tours
to all persons who also have access to the wiki server. In fact, There are virtual guided bus tours taking visitors to var-
even registration of a user account is not often required. ious wiki sites. These consist simply of a page on each
participating wiki called TourBusStop, which gives the link
to the next bus stopbasically, a type of web ring. Each bus
History stop page gives some info about that wiki, and one can
Wiki software originated in the design pattern community choose to explore that particular wiki (thus getting off the
for writing pattern languages. The Portland Pattern Repos- bus), or continue to the next wiki in the tour.
itory was the first wiki, established by Ward Cunningham in Source: Wiklopedia. Retrieved March 10, 2004, from http://en.wikipedia.
org/wiki/WikiWiki
1995. Cunningham invented and named the wiki concept, and
150,000 articles in English and was being translated into Design Issues
a dozen languages.Anyone who visits the Wikipedia can The design of a groupware system has a definite ef-
edit existing pages or create new articles, and each ar- fect on the interaction of the people using it. If the
ticle contains a link to all of its previous versions. users do not like the resultant interaction, they may
Another successful web-based system is the Source- be severely hampered in performing the task that the
Forge site, an open-source software development site. groupware was supposed to help them with. In some
Software developers can come to the site and com- cases they may be unable to perform their task at all,
municate asynchronously with other developers about or they may rebel and refuse to use the system.
their code; they can also check out and store multi- To design successful groupware, one must un-
ple versions of their programs at the site. By 2004, derstand the impact the technology will have on the
SourceForge was hosting over 65,000 projects and task the group is trying to perform. Fundamentally,
700,000 registered developers. All visitors to the site designing groupware requires understanding how
have access to the current code base, but only develop- people behave in groups. It also requires a good grasp
ers are allowed to make changes to it. of networking technology and how aspects of that
GROUPWARE 301
technology (for instance, delays in synchronizing take into account. Designers need to know how
views) can affect the user experience. Shortcomings homogeneous the users are, the roles people are likely
in technology can render an otherwise promising to play, and who key decision makers are and what
tool useless, as minor issues of system responsive- influence they have on the decision-making process.
ness and reliability can become very significant when Groupware designers should investigate the effect
coupled with the dynamics of group interaction. the technology will have on the sense of group iden-
Traditional issues of user interface designfor tity, culture, and environment that emerge in long-
example, striving for a consistent interface and of- term collaboration. Additionally, groupware designers
fering helpful feedbackare still relevant, since must consider how the system will deal with sensi-
the technology still involves individuals. However, tive social issues such as anonymity and accounta-
because the target users are groups, there are addi- bility of actions.
tional considerations. For instance, million-person Observational studies, usage studies, surveys, and
groups behave differently from five-person groups, prototyping are all important tools in designing suc-
and the performance parameters of the technologies cessful groupware. Good development may require
required to support the two types of groups are quite a spiral model of user-centered or participatory de-
different. Likewise, groupware must be easier to use velopment, wherein developers observe users using
than software for single users, because the pace of a prototype version of the groupware system, collect
use of an application is often driven by the pace of and analyze data from that study, and redesign the
other users. Consequently, a difficult interface will software accordingly. Multiple cycles of design and
accentuate disparities in user expertise, which can testing are usually necessary to produce quality group-
lead to frustrating delays and serious reductions in ware. Strong analysis tools can help reduce devel-
group productivity. opment tasks by allowing the designer to understand
difficulties that users encounter while trying to per-
form their tasks using the groupware. Methods such
Coordination and Community as ethnomethodology (where the social interaction
Because groupware necessarily supports a commu- between participants is examined in the context of
nity of users performing some task, it must ad- their work) and various forms of discourse analysis
dress not only the work the users perform that relates have been successfully adapted as methods to study
directly to the task, but also the work that users per- the interaction that emerges. By carefully examining
form to stay coordinated during execution of that the recurring problems of coordination that users en-
task. Users need ways to exchange information about counter, the designers can identify what parts of
the task at hand. They need to establish and follow the system need to be redesigned, and they can
conventions for activity, and they must be aware of then create a system that effectively supports the users.
what other users are doing. Users spend a portion of
their time maintaining common ground. One
method for minimizing disparity between users' A Case Study: Instant Messaging
viewpoints, WYSIWIS (what you see is what I see), Instant messaging (IM) is, as its name implies, a tech-
creates systems that give users a similar viewpoint nology that lets people communicate with one an-
on the task. Another approach is to design groupware other synchronously, in real time, as stated on
features that help the users be more aware of each the home page of America Online's Instant Messen-
other's actions. ger. After installing client software, a user of IM tech-
Groupware designers must also keep in mind the nology connects to a central server. The server then
various social issues that arise in collaboration. When informs those users as to the online availability of those
people are dealing with one another remotely, estab- others included in his or her contact (or buddy)
lishing and maintaining identity can become a diffi- list. Likewise, the server informs others who have in-
cult security issue that the groupware designer must cluded the user in their contact list as to the online
presence of the user. Typically, a contact list may con-

tain as many as two hundred screen names. Away Messages
When connected, a user can click on the name of an-
other user who is also online, opening a window to
A
way messages are perceived to be a necessity for many
that person which permits direct, real-time exchanges users of instant messaging. For those who dont want
of messages between them. (Communication is typ- to think up an original message, websites such as
ically in the form of text messages, although the tech- AIMawaymessages.com offer a variety of messages on top-
nology permits audio and visual exchanges as well.) ics from friendship and food to homework and finals.
When a user disconnects from the server, others who You helped me laugh, you dried my tears,
include him or her in their contacts lists are informed Because of you, I have no fears.
that the person is no longer online. Together we live, together we grow,
Different types of users utilize instant messaging Teaching each other what we must know.
in different ways. In a 2000 study, the scholars You came in my life, and I was blessed.
I love you girl, you are the best.
Bonnie Nardi, Steve Whittaker, and Erin Bradner re- Release my hand, and say good-bye,
ported that in the workplace coworkers used IM tech- Please my friend don't you cry.
nology to quickly ask and answer questions and to I promise you this, it's not the end,
clarify issues about ongoing tasks. It was also used to Cause like I said you're my best friend
keep in touch with family and friends through brief
exchanges. In addition, it was used to inquire about I am not currently available right now.
the availability of others for communication in other However, if you would like to be transferred to
another correspondent, please press the number
media as well as to arrange for offline meetings, al- that best fits your personality:
though in a 2002 study the researcher Ellen Issacs and If you are obsessive compulsive, please press
her associates reported that most of the IM conver- 1 repeatedly.
sations they studied, also in the workplace, remained If you are codependant, please ask some-
in IM. Instant messaging is now rapidly spreading in one to press 2.
the work world, as evidenced by mention in newspa- If you have multiple personalitites, please
press 3, 4, and 5.
per accounts, by conferences organized to promote its If you are paranoid delusional, we know who
adoption, and by its inclusion in corporate culture. you are and what you want. Just stay on the line
so we can trace your call.
If you are schizophrenic, listen carefully and
Technological Subversion: the little voice will tell you which number to press.
If you are manic depressive, it doesn't mat-
Away Messages ter what number you press, no one will answer.
Source: AIMAwayMessages. Retrieved March 10, 2004. from http://aimaway-
In a 2001 study for the Pew Internet and American messages.com/
Life project, researchers Amanda Lenhart, Lee Rainie,
and Oliver Lewis describe contemporary teenagers
as the instant-message generation, noting that 74 IM server, he or she is either not near the computer
percent of U.S. teens with Internet access use instant or wishes to remain undisturbed currently. In such
messaging, often to stay in touch with friends and a case, the IM client acts like an answering machine:
relatives who do not live nearby. Many others have It records messages left by others while providing
also reported on the popularity of IM technology feedback in the form of an automated away mes-
among teenagers. Interestingly, these studies do sage sent to those who attempt to contact the
not discuss away messages, a remarkable feature of user. On some clients, this message is visible when-
IM technology. ever a user is away, allowing users to stay appraised
The away feature can inform others that, of each others' status without needing to contact
while the user remains online and connected to the each other directly.
GROUPWARE 303
In AIM, the default away message is I am away

from my computer right now. However, users go to FURTHER READING
great lengths to expand the kinds of messages they
Alterman, R., Feinman, A., Introne, J., & Landsman, S. (2001, August).
post and the uses to which such messages are put. Coordinating Representations in Computer-Mediated Joint
Users, and especially college students who have ac- Activities. Paper presented at the 23rd Annual Conference of the
cess to broadband connections, employ away mes- Cognitive Science Society, Edinburgh, United Kingdom. Retrieved
sages as a kind of bulletin board to inform, to August 7, 2003, from http://www.hcrc.ed.ac.uk/cogsci2001/pdf-
files/0015.pdf
entertain, and to manage social relationships. Berners-Lee, T. (1991). World Wide Web seminar. Retrieved August
The possibilities of away messaging as well as 7, 2003, from http://www.w3.org/Talks/General.html
its popularity are evident in a number of websites Bijker, W. E., & Law, J. (Eds.). (1992). Shaping technology/building
society: Studies in sociotechnical change. Cambridge, MA: MIT
that collect and categorize away messages, serving as Press.
a resource on which users may draw. There is also Bodker, S., Gronbaek, K., & Kyng, M. (1993) Cooperative design:
software that enables users to keep track of those who Techniques and experiences from the Scandinavian scene. In D.
are reading his or her away messages. In addition to Schuler and A. Noamioka (Eds.), Participatory design: Principles
and practices (pp. 157176). Hillsdale, NJ: Lawrence Erlbaum
having these communicative and social uses, the away Associates.
message feature of IM may also serve a psychologi- Clark, H., & Brennan, S. (1991) Grounding in communication. In
cal function, providing users a sense of presence and L. B. Resnick, R. M. Levine, & S. D. Teasley (Eds.), Perspectives on
of attachment. socially shared cognition (pp. 127149). Washington, DC: American
Psychological Association.
Although synchronous communication is still the Ellis, C. A., Gibbs, S. J., & Rein, G. L. (1991) Groupware: Some issues
primary function of IM, it is unlikely that its devel- and experiences. Communications of the ACM, 34(1), 3858.
opers anticipated the extensive and varied uses of asyn- Erickson, T., & Kellogg, W. (2000) Social translucence: An ap-
proach to designing systems that support social processes. ACM
chronous away messages. Like the telephone, which Transactions on Computer-Human Interaction (TOCHI), 7(1),
was developed as a business tool, and like the Internet 5983.
itself, which was developed by the military and ini- Fischer, C. S. (1940). America calling: A social history of the telephone
tially used as a means of communication by a small to 1940. Berkeley and Los Angeles: University of California Press.
Garfinkel, H. (1967). Studies in ethnomethodology. Upper Saddle River,
handful of academic researchers, IM has developed NJ: Prentice Hall.
in ways not necessarily foreseen or intended by its de- Grudin, J. (1990) Groupware and cooperative work. In B. Laurel (Ed.),
signers. It is another reminder of the difficulties in The art of human-computer interface design (pp. 171185). Reading,
predicting how technology will affect society. MA: Addison-Wesley.
Hauben, M. (2003). History of ARPANET. Retrieved August 7,
While the general public tends to believe that a 2003, from http://www.dei.isep.ipp.pt/docs/arpa.html
given technology dictates its uses, sociologists of tech- Hutchby, I. (2001). Conversation and technology: From the telephone to
nology argue that the uses of technology are socially the Internet. Malden, MA: Blackwell.
Issacs, E., Kamm, C., Schiano, D., Walendowski, A., & Whittaker S.
shaped rather than being determined solely by the (2002, April) Characterizing instant messaging from recorded logs.
materiality of technological artifacts or the inten- Paper presented at the ACM CHI 2002 Conference on Human Fac-
tions of their designers. Technologies are often sub- tors in Computing Systems, Minneapolis, MN. Retrieved August 7,
verted by users and employed in ways quite different 2003, from http://hci.stanford.edu/cs377/nardi-schiano/CHI2002.
Isaacs.pdf
from those for which they were originally intended. Lenhart, A., Rainie, L., & Lewis,O., (2001) Teenage life online: The
The case of IM technologies, and especially of IMs rise of instant-message generation and the Internet's impact
away message capability, illustrates the social shap- on friendships and family relations. Retrieved August 7, 2003,
ing of technological change as well as the difficulty from http://www.pewinternet.org/reports/pdfs/PIP_Teens_Report
.pdf
in predicting societal impacts of technology. Nardi, B. A., Whittaker, S., & Bradner E. (2000). Interaction and out-
eraction: Instant messaging in action. In Campbell, M. (Ed.), Pro-
Timothy J. Hickey and Alexander C. Feinman ceeding of the ACM 2000 conference on computer-supported
cooperative work (pp. 7988). Philadelphia: ACM.
Schiano, D. J., Chen C. P., Ginsberg, J., Gretarsdottir, U., Huddleston,
See also Chatrooms; Computer-Supported Cooper- M., & Issacs, E. (2002, April). Teen use of messaging media. Paper
ative Work presented at the ACM CHI 2002 Conference on Human Factors
in Computing Systems, Minneapolis, MN. Retrieved August 7, (Eds.), Design at work: cooperative design of computer systems
2003, from http://hci.stanford.edu/cs377/nardi-schiano/CHI2002. (pp. 6589). Hillsdale, NJ: Lawrence Erlbaum Associates.
Schiano.pdf Shneiderman, B. (1992). Designing the user interface: Strategies for ef-
Suchman, L., & Trigg, R. (1992) Understanding practice: Video as a fective human-computer interaction. Reading, MA: Addison-Wesley.
medium for reflection and design. In J. Greenbaum & M. Kyng Tyson, J. (2003). How instant messaging works. Retrieved August 7,
2003, from http://www.howstuffworks.com/instant-messaging.htm
HACKERS
HANDWRITING RECOGNITION AND RETRIEVAL
HAPTICS
HISTORY OF HCI
HOLLERITH CARD
H
HUMAN-ROBOT INTERACTION
HYPERTEXT AND HYPERMEDIA
at hand to invent or reinvent technology for novel

HACKERS uses or to solve an immediate problem at hand.
Each of the following generations of computer
A computer hacker was originally defined as some- hackers has developed its own culture, its own sense
one who was expert at programming and solving of style, and, perhaps most importantly, its own sense
problems with a computer; later it came to identify a of ethics. These various cultures have generally been
person who illegally gained access to a computer sys- shaped as oppositional or resistant cultures within
tem, often with the intent to cause mischief or harm. broader institutional contexts. There have been sev-
The idea of computer hacking started out with a simi- eral primary contexts for the development of hacker
larly benign meaning: A computer hack was a solu- culture from the mid-1960s to the present.
tion to a hardware or programming problem. One
could argue that the first digital computer, composed
of two telephone relays switches, two light bulbs, a The Original Hacker Ethic
battery, and a tin can keyboard, was itself a hack. In the United States, the origins of hacker culture
When the mathematician and inventor George Stibitz can be traced back to the universities and colleges of
(19041995) put together that device in 1937, he was the 1960s and 1970s. Fueled by government research
doing what generations of computer enthusiasts would funding, primarily from the U.S. Department of
continue to do for decades to come, using materials Defenses Advanced Research Projects Agency
305
(DARPA), the first generation of computer hackers different end. The Youth International Party was per-
were students working during off-hours in computer haps the most representative of these politically ori-
labs. The group that the engineer Joseph Weizenbaum ented students. Led by Abbie Hoffman, Jerry Rubin,
called compulsive programmers and that the sociolo- and others, the Youth International Party used tech-
gist Sherry Turkle and the journalist Steven Levy doc- nology for personal empowerment and activism.
ument as the first generation of hackers was composed The Youth International Party created a newslet-
almost exclusively of computer science students at ter called Technological Assistance Program (or TAP)
major U.S. universities in the 1960s and 1970s. to provide technological knowledge to people who
For the first generation of hackers, hacking was would otherwise have to pay companies what the
a means to come up with clever or unusual solutions group believed were unfair rates for technological
to seemingly intractable problems. Accordingly, services. TAP set out to teach people how the tech-
the notion of hacking was contrary to what most nology worked and how it could be exploited. Initially
computer science programs were teaching: struc- it provided details on how consumer services and
tured programming. Where the curriculum of com- utilities could be used for free as well as on how tech-
puter science instruction followed the philosophy of nology such as voice conferencing on the phone could
finding the single best answer to each problem and be used for long-distance political organizing.
structuring that into code, the hacker ethic preached As the PCs began to emerge as hobby electronics
just the oppositetrying unusual and innovative in the 1970s, small groups of hackers began having
approaches to discover new ways of handling a prob- meetings to share programming ideas and to learn
lem. The hacker ethic rejected conventional wisdom, how to build and modify hardware. These groups,
favoring a more hands-on, bottom-up approach. best represented by San Franciscos Homebrew Com-
This early generation established the basic ethos puter Club, embraced the hacker ethic and began ap-
for hacker culture, best exemplified in Steven Levys plying it to PC culture, sharing software and code and
characterization of the Hacker Ethic: continually finding new ways to innovate and cre-
ate new technology and software. In 1976 two of them,
1. Access to computers should be unlimited and
Steve Jobs and Steve Wozniak, founded Apple
total. Always yield to the Hands-On Imperative.
Computer and, along with a host of other early com-
2. All information should be free.
puter manufacturers, ushered in the age of the per-
3. Mistrust authorityPromote Decentralization.
sonal computer as consumer electronics.
4. Hackers should be judged by their hacking,
These three dimensionsthe desire to invent,
not bogus criteria such as degrees, age, race,
create and explore (taken from the original hacker
or position.
ethic); an engaged hands-on approach, which fo-
5. You can create art and beauty on a computer.
cused on personal empowerment (borrowed from
6. Computers can change your life for the bet-
the political-activist hackers of the 1960s and 1970s);
ter. (Levy 1984, 3949)
and the idea of community and sharing in the
This ethic underlay the development of many of context of the personal computer (in the tradition
the subsequent major technological advances in com- of Homebrew)came together in the 1980s in the
putation, including the creation of the personal com- second generation of hackers.
puter (PC) and the Internet. The hacker ethic, as
originally practiced, was blind to the surrounding
political climate. Through the late 1960s and 1970s, Computer Network Hacking
the vast majority of money that went into funding The emergence of this new group of hackers was the
computer science and technological development in result of the widespread availability of the personal
the United States was given by the military. computer, as well as the popularization of the figure
During the same period that some students of the hacker in the mainstream media. With the re-
were writing code in university labs, their more polit- lease of the film War Games in 1982, thousands of
ically minded peers were deploying technology to a young, inspired hackers went online looking for those
HACKERS 307
who were similarly inclined. The hackers of the 1980s In the 1990s, when the emergence of the World
found their meeting place online, in the form of Wide Web made online commerce feasible, the sit-
newly emerging computer bulletin board systems uation changed. With the growth of e-commerce,
(BBS). Many of them founded BBS, usually run on there was a significant shift in the ways that hack-
a spare phone line and set up in the basement or a ers behaved and the ways in which they were treated.
teenagers bedroom. As hackers discovered increasingly sophisticated ways
Using a dial-up modem, hackers could access the to exploit security flaws, law enforcement developed
bulletin board, and once there, they could swap files an increased interest in hackers behavior.
and trade gossip as well as read the latest technical in-
formation about hacking. Because access to computer
networks was limited, gaining access to computers of- Criminalization of Hacking
ten required breaking into systems (usually owned by Even in the 1990s, most incidents of computer hack-
universities) and then using those computers to ac- ing remained relatively harmless. Hackers were, how-
cess others. Through BBS, loose confederations of ever, being treated as criminals for the first time and
hackers formed. With names like Legion of Doom and were frequently prosecuted under federal wire fraud
Masters of Deception, they made names for them- statutes, which carried heavy penalties including jail
selves by posting files with the latest network exploits time and fines. As the Internet became more closely
or files documenting their latest hacks as trophies. tied to notions of commerce in the public imagina-
As the hacker scene grew larger, two hackers who tion, hacking ceased to be seen as a benign nuisance
called themselves Knight Lightening (Craig Neidorf) and came to be perceived as a public menace, with
and Taran King (Randy King), respectively began to several hackers suffering the consequences. What had
document the underground hacker culture in an on- previously been viewed as pranks or at most petty van-
line journal called Phrack (an amalgam of phreak, dalism has now gained the attention of U.S. govern-
a term used to describe telephone hacking and hack, ment authorities.
which was more specific to computers). Phrack pub- Where previous generations of hackers had
lished articles of interest to the hacker community roamed the networks of universities freely, the hack-
on such topics as how to hack your local telephone ers of the 1990s were finding their options severely
company control office and how to pick locks. The limited. Exploration was being redefined as crimi-
journal also included information about the hack- nality, and high-profile cases, such as the capture and
ers themselves. In sections such as Phrack Prophiles prosecution of Kevin Mitnick, reinforced tensions
and Phrack World News, hackers wrote about the between law enforcement and the hack commu-
personalities and events that has special significance nity. Mitnick had been the subject of a nationwide
for them. manhunt, spurred on by a series of stories in the New
This generation of hackers embraced the ethic of York Times, which branded him cyberspaces Most
open and free information, but found they had in- Wanted. Mitnick became a cause clbre for the hacker
herited a world in which computers and networks community, having been denied a bail hearing and
were proprietary and expensive. In order to ex- spending three years as a pretrial detainee. Perhaps
plore, hackers found they needed to invade other most important, the damage caused by Mitnicks
peoples systems. There was no alternative. The vast hacking was hotly contested. Because Mitnick
majority of such hacking was merely exploratory and never gained financially from any of his hacking, the
harmless. Indeed, it was customary for system ad- defense argued that the damage caused was minimal.
ministrators of the systems that got hacked to men- The prosecution, however, claimed that Mitnick had
tor young hackers when the former caught the latter not only stolen valuable code but had also rendered
in their systems. The worst the hackers usually suf- other code worthless merely by looking at it. The pros-
fered was a phone call to their parents or a stern warn- ecutors set the figure at $100 million in damage,
ing. By the 1980s, hacking also spread throughout the maximum allowed under the statute. Ultimately,
Europe, particularly in Germany and the Netherlands. Mitnick pled guilty to making three fraudulent phone
calls. He received five years in prison for his crimes Techniques include the hacking of webpages and the
and was forced to pay restitution. replacement of the pages original content with po-
Mitnicks arrest and prosecution, along with a litical messages as well as the crashing of websites that
handful of others, sent signals to the computer under- carry messages to which hackers are opposed.
ground that law enforcement was taking computer The themes of innovative problem solving, a
crime seriously, and that they were prepared to make hands-on approach, and political and community
the capture and prosecution of hackers a priority. At action are threads that have run through hacker cul-
the same time, with the development and distribu- ture from its earliest days. As cultural attitudes to-
tion of open-source operating systems such as Linux, wards technology have shifted, so has the nature of
hackers no longer needed to breach other peoples hacker culture and the computer underground, which
systems to explore network security. They were now continues to express the basic principles of hacker
able to experiment on their own machines and culture, attitudes, and actions in creative new ways.
networks without the risk of being arrested.
As the stakes for computer hackers were raised, Douglas Thomas
many hackers turned their attention to system secur-
ity. In the late 1990s a new breed of hackers, so-called See also Law Enforcement; Movies
white-hat hackers, emerged. The premise of white-
hat hacking was that hackers themselves could help
systems defend themselves against black-hat hack- FURTHER READING
ers (hackers who invade computer systems to cause
disruption or damage). Typically, white-hat hackers Levy, S. (1984). Hackers: Heroes of the computer revolution. New York:
Dell.
would release security software, document security Thomas, D. (2002). Hacker culture. Minneapolis: University of
flaws, and hold seminars to inform industry about Minnesota Press.
vulnerabilities and weaknesses. A number of high- Turkle, S. (1984). The second self: Computers and the human spirit.
profile white-hat hackers have even testified before New York: Simon & Schuster.
Weizenbaum, J. (1976). Computer power and human reason: From
Congress about the state of security on the Internet. judgment to calculation. New York: W. H. Freeman & Co.
As white-hat hacking became more accepted, white-
hat hackers began forming collectives and security
companies through which to offer their services.
HANDWRITING
Hacker Activism
In the twenty-first century hackers have turned their RECOGNITION AND
attention to political affairs once again. A new move-
ment, known as hactivism, is based on a fusion of RETRIEVAL
hacker techniques and political activism. A number of
groups and hackers, most notably Hactivismo and the Written information often needs to be electronically
Cult of the Dead Cow, have released information and accessed or manipulated (as in editing). Although
software to help activists in repressive countries com- people generally learn to write by hand before they
municate effectively. Using expertise in networks, learn to type on a keyboard, it is fairly difficult for
cryptography, and steganography (hiding informa- computers to work with handwritten information.
tion in images), these hackers have made it possible In many situations, the handwritten information
for dissidents in a number of countries to organize polit- must be stored in the form of page images, which are
ically and have provided access to otherwise banned difficult for computers to manage (index, search, or
or censored information. The movement has also organize). Because handwriting is such an easy pro-
spawned more direct efforts at political disruption. cess for people, much research has gone into enabling
HANDWRITING RECOGNITION AND RETRIEVAL 309
computers to recognize and retrieve handwritten recognition had been successful. Although the pos-
information. sible lexicon is large, the different fields in postal ad-
Text printed on paper using standard fonts can dresses (postal code, city names, and street names)
usually be recognized with high accuracy using an op- restrict what a given written word may be. In the
tical character recognition (OCR) engine. Commercial United States, the postal service uses machines to rec-
OCR software can recognize printed text with a char- ognize and route a significant proportion of both
acter error rate of about 1 percent, provided the qual- printed and handwritten addresses. Despite those
ity of the printing is good and standard fonts are used. successes, however, the problem of offline handwrit-
The high accuracy is possible because printed char- ing recognition is still unsolved in situations where
acters are very uniform and are usually separated by large, unconstrained lexicons are used, such as in
spaces; OCR software can also be trained to recognize handwritten manuscript collections.
standard fonts. Handwriting recognition is more chal-
lenging because handwriting varies considerably
between writers, and even for a given writer there are Ofine Handwriting Recognition:
often some variations. In addition, the characters in
a word are not always well formed. Preprocessing and Segmentation
Before words in a handwritten document can be rec-
ognized, the document must be cleaned, artifacts
Online and Ofine Handwriting (marks that are unrelated to the written text, such as
Handwriting can be categorized into online and of- creases where paper has been folded) removed,
fline handwriting respectively. With online hand- and the words segmented out and processed for
writing, the act of writing is captured by the device. recognition. Preprocessing may involve operations
Pen stroke and velocity information are, therefore, to improve the quality of the image, to correct for
available to aid the recognition process. With offline the slant of the writing, and to remove noise (which
handwriting, it is assumed that the writing has already may be caused by many factors such as ink blotches,
occurred (often on paper), and all that is available is ink fading, or the scanning process).
a scanned image of the written document. In this sit- The segmentation process involves separating out
uation, information on the pens movements is there- individual words. Current segmentation techniques
fore not available. generally rely on knowledge of the spacing between
In recent years, significant advances have been text. In English, for example, there is space between
made in online handwriting recognition. One ap- lines of text, and the space between words is usually
proach has emphasized teaching people to write char- greater than the space between characters. The seg-
acters in a more distinctive way. For example, the mentation process is usually a two-stage process: First
Graffiti alphabet used by many portable digital as- lines of text are detected and then the words are de-
sistants (PDAs) changes the way characters are con- tected. In some situations, one can also try to detect
structed so that they are easier to recognize. Other the layout of the page.
approaches, such as those used by the Tablet PC, try A common approach to finding lines of text is
to recognize a persons actual handwriting. to analyze the pixel values along each row. Each image
Since offline handwriting offers less information, may be thought of as a matrix of pixel values organ-
it has had more limited success. Successes have ized into rows and columns where the pixel values
been achieved in situations where the lexicon is lim- represent the intensity of the image at each point. Each
ited and additional constraints are available. Bank row may be replaced by a single number obtained
check recognition takes advantage of the fact that the by adding all the pixel values in that row. This creates
amount written in on a handwritten check makes use a vector with as many numbers as rows.
of only about thirty different words. Postal address The values of this column vector when plotted
recognition is another example where handwriting on a graph show a curve with minima (assuming
black text corresponds to low pixel values) at loca- the letterfor example, the top portion of the
tions corresponding to text and maxima at locations lowercase letter l) and descenders (a descender is the
corresponding to the spaces between lines. The lines portion of a lowercase letter that extends below the
can, therefore, be extracted by noting the position of main body of the letterfor example, the lower por-
the maxima. tion of the lowercase letter p). A classifier is trained
using features computed over a number of train-
ing examples of words. Words can then be recog-
Analytic and Holistic Ofine nized using this classifier. The holistic techniques
main advantage is that the difficult problem of
Handwriting Recognition segmenting a word into characters is avoided. On
There are two approaches to recognizing a handwrit- the other hand, holistic techniques must be trained
ten word: the analytic and the holistic. The analytic on each word, which makes it difficult to use with
method involves segmenting a word into characters a large number of different words. Holistic techniques
and then recognizing each character. The word is work best when the vocabulary is small, as with bank
segmented at a number of potential character bound- check recognition.
aries. For each character segment, a classifier sug-
gests possible character choices along with confidence
valuesthat is, the degree of confidence it has in Handwriting Retrieval
those choices. At this point, at each character posi- Handwriting recognition is used to convert images
tion within the word, there is a list of potential char- of words into an electronic form that computers can
acter candidates, each associated with a confidence interpret as textfor example, the American
value. A graph is created whose nodes (points) are Standard Code for Information Interchange (ASCII)
the segmentation points and whose edges (con- data-transmission code. However, this does not by
nections between the nodes) are possible character itself solve the problem of accessing handwritten
choices. The appropriate confidence value is used to documents. For example, suppose one is interested
weight each edgethat is, to suggest that a choice is in locating a particular page from the collected pa-
more or less likely to be correct. Each path from node pers of George Washington. To do this, one needs to
to node through the graph creates a string of char- search the set of pages. Given ASCII text, one can use
acters, only some of which are genuine words. The a search engine to do this. This approach is in fact
cost of each path is obtained by adding the weights used for online handwritten material, as online hand-
(confidences). The path of minimum cost which gives writing can be converted to ASCII with reasonable
a legal word is chosen as the optimal path. accuracy. However, this approach does not work for
The analytic method requires training only on in- offline handwritten material, because handwriting
dividual characters. In English, this is a small set con- recognition for documents with such large vocab-
sisting of upper and lowercase letters, the digits, and ularies is still not practicable. One possible approach,
punctuation marks. The fact that the total set of char- called word spotting, is to segment pages into words
acters is so small makes it practical to obtain training and to cluster word images (for example, one clus-
samples of characters to create a classifier. The main ter might be all the instances of the word inde-
weakness of the analytic technique is that it is so dif- pendence in George Washingtons manuscripts) using
ficult to segment words into characters. image matching. The clusters will have links to the
The holistic technique does not require words to original pages and may, therefore, be used to find the
be segmented into characters. Instead, features are right page.
computed over the entire word image. Examples of An important distinction between recognition
such features include the length of the word, the and retrieval is that the latter usually uses the con-
number of loops, ascenders (an ascender is the por- text supplied by the other words in the page, and this
tion of the letter that extend above the main body of can improve performance. This constraint has not
HAPTICS 311
been applied to offline handwriting recognition or Setlur, S., Lawson, A., Govindaraju, V., & Srihari, S. N. (2002). Large
retrieval yet, but applying such a constraint should scale address recognition systems: Truthing, testing, tools, and
other evaluation issues. International Journal of Document Analysis
improve performance. and Recognition, 4(3), 154169.
Vinciarelli, A., Bengio, S. & Bunke, H. (2003). Offline recognition of
large vocabulary cursive handwritten text. In Proceedings of the
Seventh International Conference on Document Analysis and
Recognition (pp. 11011107). Los Alamitos, CA: IEEE.
The Future
Handwriting recognition and retrieval is a chal-
lenging area. While there have been some successes,
especially in recognizing postal addresses, much work
remains to be done, especially in the area of recog- HAPTICS
nizing and retrieving large-vocabulary documents.
Solving this problem would allow computers to deal Haptic interaction with the world is manipulation
with handwritten material in the same way that they using our sense of touch. The term haptics arises
deal with typed input. from the Greek root haptikos, meaning able to grasp
or perceive. Haptic interaction with computers im-
R. Manmatha and V. Govindaraju plies the ability to use our natural sense of touch
to feel and manipulate computed quantities. Haptic
See also Natural-Language Processing; Optical Char- computer interaction is a relatively new field that has
acter Recognition generated considerable interest in the 1990s and early
years of the twenty-first century.
FURTHER READING
Kim, G., & Govindaraju, V. (1997). Bank check recognition using cross
A New Way To Interact
validation between legal and courtesy amounts. International Jour-
nal on Pattern Recognition and Artificial Intelligence, 11(4), 657674.
with Computers
Kim, G., & Govindaraju, V. (1997). A lexicon driven approach to hand- Initially, computers could deal only with numbers.
written word recognition for real-time applications. IEEE Trans- It took many years to realize the importance of op-
actions on Pattern Analysis and Machine Intelligence, 19(4), 366379. erating with text. The introduction of cathode ray
Kim, G., Govindaraju, V., & Srihari, S. (1999). Architecture for hand-
written text recognition systems. International Journal of Document tube display technology allowed graphics to be dis-
Analysis and Recognition, 2(1), 3744. played, giving people a new way to interact with com-
Madhvanath, S., & Govindaraju, V. (2001). The role of holistic para- puters. As processing power increased over time,
digms in handwritten word recognition. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 23(2), 149164.
three-dimensional graphics became more common,
Madhvanath, S., Kim, G., & Govindaraju, V. (1999). Chain code and we may now peer into synthetic worlds that seem
processing for handwritten word recognition. IEEE Transactions solid and almost real. Likewise, until recently, the no-
on Pattern Analysis and Machine Intelligence, 21(9), 928932. tion of carrying on a conversation with our com-
Madhvanath, S., Kleinberg, E., & Govindaraju, V. (1999). Holistic ver-
ification of handwritten phrases. IEEE Transactions on Pattern
puter was far-fetched. Now, speech technology has
Analysis and Machine Intelligence, 21(12), 13441356. progressed to the point that many interesting appli-
Manmatha, R., & Croft, W. B. (1997). Word spotting: Indexing hand- cations are being considered. Just over the horizon,
written manuscripts. In M. Maybury (Ed.), Intelligent multi-media computer vision is destined to play a role in face and
information retrieval (pp. 4364). Cambridge, MA: AAAI/MIT Press.
Plamondon, R., & Srihari, S. N. (2000). On-Line and off-line handwrit- gesture recognition. It seems clear that as the art of
ing recognition: A comprehensive survey. IEEE Transactions on Pat- computing progresses, even more of the human sen-
tern Analysis and Machine Intelligence, 22(1), 6384. sory palette will become engaged.
Rath, T. M., & Manmatha, R. (2003). Word image matching using dy-
namic time warping. In Proceedings of the IEEE conference on Com-
It is likely that the sense of touch (haptics) will be
puter Vision and Pattern Recognition (pp. 521527). Los Alamitos, the next sense to play an important role in this evo-
CA: IEEE. lution. We use touch pervasively in our everyday lives
and are accustomed to easy manipulation of objects touch (including touch, pressure, vibration), pro-
in three dimensions. Even our conversation is pep- prioception (including joint angle, muscle length
pered with references to touching. The researcher Blake and length rate of change, and tendon tension), and
Hannaford at the University of Washington has com- pain (including itch and tickle) and temperature.
piled a list of verbal haptic analogies: We frequently The first two modalities are the ones that are most
make the analogy to haptics when we speak of our re- important for haptic perception.
lationship to ideas, people, and information. We of- The discriminative touch modality relies on four
ten use phrases like get a feel,poke (into, around), different kinds of receptors in the glabrous (hairless)
put ones finger (on the problem), when referring skin of the hand. They are Meissners corpuscles,
to exploration. We use phrases like (to stay, keep) Pacinian corpuscles, Merkels disks, and Ruffini end-
in touch,tangible (concepts),(a) touchy (subject), ings. Both the Meissners and Pacinian corpuscles
hands-on learning (often used literally), and at my are considered to be rapidly adapting (RA),
fingertips when referring to contact. And phrases like responding mostly to changing stimuli, whereas the
pressing issues, pushy (behavior), hard-hitting Merkels disks and Ruffini endings are considered to
(presentation), get a grasp (of the situation), and be slowly adapting (SA) and continue to fire in the
so forth are used when referring to impact or ma- presence of constant stimuli. Whereas the anatomical
nipulation. In fact, it is quite surprising, given our nat- characteristics of these receptors are known, their pre-
ural propensity to touch and manipulate things, cise role in psychophysical perception is less well under-
that haptic computer interfaces are not common. stood. The Pascinian corpuscles respond to
To explore and interact with our surroundings, high-frequency vibrations such as those encoun-
we principally use our hands. The hand is unique in tered when running ones finger over a textured sur-
this respect because it is both an input device and an face. The Meissners corpuscles are sensitive to sharp
output device: Sensing and actuation are integrated edges, the Ruffini corpuscles to skin stretch, and the
within the same living mechanism. An important Merkels disks to edges and pressure.
question is how best to transmit haptic information The proprioception modality is also of major im-
between a running computer program and a users portance, although its receptors are less well under-
hand. Providing position input to the computer from stood than those of discriminatory touch. The joint
the hand is easy; providing force or torque output angle receptors incorporate Ruffini endings and
from the computer to the hand has proven to be dif- Pascinian corpuscles located at the joints, which re-
ficult. We have not figured out how to invent good spond to pressure applied to the receptor. Interestingly,
haptic devices that will link our hands in some way subjects can resolve changes in angle between thumb
with a running computer programand for that and forefinger as small as 2.5 degrees. Muscle spin-
matter, we do not understand very well how best dles, located between and among muscle fibers, re-
to write programs that can derive and serve up hap- port muscle length and rate of change of length. Being
tic information for our consumption. modified forms of skeletal muscle fibers, muscle spin-
dles not only can send information to the brain,
but can also receive commands causing them to con-
Our Tactile and Kinesthetic Senses tract, resetting their threshold setting. Golgi tendon
How does the skin detect pressure, friction, and vi- organs, which respond to tension force within the
bration? We know the hand is a complicated system tendons, seem to play a role in muscle control.
that includes articulated structure, nerves, muscles In a haptic interface, both the proprioception and
(for output), and senses (for input). But the hands discriminative touch modalities play important roles
sensory capabilities are at best imperfectly under- in perception. Given these considerations, we must ask
stood. Because of this, fully informed design meth- what to put in (or around, or in contact with) the hand
ods for haptic interfaces do not yet exist. in order for a running computer program to impart a
The sensory suite forms part of the somatosen- realistic sensation of touch to a user. There are many
sory system, which has modalities of discriminative possible approaches to answering this question.
HAPTICS 313
Haptic Interface Devices Critical considerations for haptic interface de-

Haptic interaction between a person and a computer vices include degrees of freedom, bandwidth, reso-
requires a special device that can convert a persons lution, and stiffness range. First, haptic interaction
motions into meaningful quantities that can be input with a realistic three-dimensional virtual environ-
into the computer and at the same time can convert ment has proven difficult with devices having fewer
computed quantities into physical forces and torques than six DOFs, but providing six-DOF capability of-
that the person can feel. Many different kinds of de- ten implies high mechanical complexity and cost.
vices have been invented that enable haptic interac- Second, the response of the device should be as lively
tion with the whole hand, or arm, or even the whole as possible; otherwise interaction with the virtual en-
body. We consider here only the type of device that vironment will feel sluggish and unnatural. The hand
can be engaged by the hand. is capable of feeling vibrations having frequencies of
Haptic devices can have several degrees of free- several hundred hertz (cycles per second), but achiev-
dom (DOF). For example, an ordinary mouse device ing the highest possible bandwidth in an electro-
has two DOFs: It can be moved left-right and for- mechanical device is challenging. Third, for the hand
ward-backward, yielding two independent position to feel small nuances in the virtual environment (for
measurements for input to the computer. Similarly, example, fine texture or stick-slip friction) the device
the familiar joystick gaming device can be tilted left- must be able to resolve extremely small changes in
right or forward-backward, yielding two independ- position and force. Finally, when the user moves
ent angles. Neither the mouse nor the joystick are through free space, nothing should be felt, but when
haptic devices, since they provide input only to the a hard virtual surface is contacted, it should feel per-
computer. Adding motors or other actuators to these fectly rigid. In practical systems, neither perfect
devices would permit haptic interaction, since they lack of resistance nor perfect rigidity is achieved, but
could then serve not only as input devices, but also the range of possible stiffness felt by the hand should
as output devices. be maximized.
Haptic devices with three DOFs allow translation Haptic devices primarily engage the propriocep-
in three dimensions, and those with six DOFs allow tive senses in the hand and arm, but also involve
both translation and rotation in three dimensions. the discriminative touch sensors, chiefly through high-
Haptic devices usually employ either serial or par- frequency effects. Collectively, bandwidth, resolution,
allel linkages whose joint positions are measured with and impedance range determine the fidelity of the
encoders and whose joint torques are affected by mo- haptic device. The ideal haptic interface device has
tors. One may recognize that the preceding de- yet to be invented. One popular commercial device
scription also applies to robot arms; in fact, many is a small cable-driven highly back-drivable three-
early haptic interface devices were based on robot DOF arm with three encoders and three motors. A
arms. In the case of a robot arm, the computer com- new approach uses magnetic levitation to provide
mands the arms joint motors to move the arm tip to six-DOF interactions over limited ranges of mo-
a particular position and orientation in space. In the tion with a single moving part.
case of a haptic device, the user holds the arm tip,
moving it to a particular position and orientation as
the computer monitors the joint angles. When the Haptic Rendering
computed position of the tip reaches a place where To be able to feel computed quantities, there must
motion is no longer wanted (according to the virtual- be algorithms and computer programs capable of de-
environment model), the computer turns on motors riving correct force and torque values for a given sit-
to provide stiff resistance to further motion. For this uation to be output to a haptic interface device.
scheme to work, the arm must be back-drivable (that The term haptic rendering is used to describe these
is, the user is able to move the robot around manu- operations, in analogy with the familiar rendering of
ally), or else the users hand force and torque must graphics on a computer display. Normally, haptic ren-
be measured at the tip and used to control motion. dering is accompanied by simultaneous graphical
rendering in what is more properly referred to as a Collision detection for a virtual tool is much
visual-haptic interface. Unlike graphical rendering, more complicated than that for a point. For either a
which can satisfy the eye at update rates of thirty point or a virtual tool, once the virtual object sur-
frames per second or even less, haptic rendering must face is contacted, realistic contact forces and torques
be done at rates approaching a kilohertz (that is, a must be derived. Researchers have developed a num-
frequency approaching a thousand cycles per sec- ber of different algorithms for this. For example, if
ond) to feel right to the hand. the surfaces are supposed to feel rigid, a very stiff
In many cases, one may desire to interact hapti- spring is modeled. If the object is deformable, its sur-
cally with three-dimensional objects modeled in the face is modeled as a network of springs that can
computer. For example, suppose we have modeled a deform much like a mattress deforms. Surfaces of
cube and a cone by using mathematical formulas that virtual objects need not be smooth. Various texture-
define their surfaces, and we wish to be able to and friction-rendering algorithms have been de-
touch these virtual objects with a point-like probe. veloped. In all these methods, there is a trade-off be-
The haptic system establishes a one-to-one corre- tween rendering accuracy and rendering time that
spondence between the (virtual) probe point and the is severely taxing on computer resources. There are
position of the haptic device handle, called a ma- also issues of control stability. When we interact hap-
nipulandum. This is very much like the relationship tically with real objects, energy is almost always dis-
between a computer mouse on the desktop and the sipated, but if the interaction between a haptic device
cursor on the computer display screen. As the user and a virtual environment is not correctly modeled,
moves the probe point about in three dimensions energy can be generated, leading to vibration and
by moving the manipulandum, the computer checks sudden loss of control. Finding the best haptic ren-
whether the point is outside an object in free space or dering algorithms for a given situation continues to
inside an object, an operation termed collision de- be an active area of research.
tection. This test must be done very rapidly, perhaps
a thousand times a second. As long as the probe point
is in free space, the device motors are turned off and Psychophysics of Haptic Interaction
are able to turn freely. As soon as the point is deter- One may characterize the operation of a haptic inter-
mined to be inside the virtual cube or cone, the mo- face in terms of its engineering parameters, but in
tors are turned on, providing torques to the device the end it is the users perception that really matters.
joints, which generate a stopping force on the manip- Physiologists and psychologists have studied the hu-
ulandum. If the user attempts to push the point far- man sense of touch for many decades. Psychophysics
ther into the virtual object, the motor currents are is the scientific study of the relationship between
increased further to resist the motion. The user thus stimuli (specified in physical terms) and the sensa-
experiences a sensation of contacting the surface of a tions and perceptions evoked by those stimuli.
real object. With a three-DOF haptic device, the Researchers have striven to characterize and dis-
user may freely slide the point along the surface of the tinguish psychophysical responses from various dis-
cone or surfaces of the cube, feeling their shapes. criminative touch sensors by performing experiments.
With a six-DOF haptic device, it is possible to For example, the researchers Roberta Klatzky and
make use of a three-dimensional virtual tool instead Susan Lederman conducted an extensive psychophys-
of just a point. The haptic system associates the ical analysis of haptic performance under conditions
position and orientation of the virtual tool with the in which the fingertip is covered by a rigid sheath,
position and orientation of the manipulandum. In held in place by a thin rubber glove. This eliminates
our example, if the user contacts the virtual cube or the spatial pressure gradient normally provided by
cone with, say, a cube-shaped virtual tool, he or the mechanoreceptors (particularly slowly adapt-
she will feel torques as well as forces, as a surface of ing receptors such as the Merkels disks) and provides
the virtual tool rotates into contact with a surface of only a uniform net force (with, possibly, a gradient
the cube or cone. at the edges of the sheath that is unrelated to the
HAPTICS 315
felt surface), simulating conditions encountered when of an engine can pick up and place parts into a com-
using a haptic device. plicated assembly while feeling the fit and friction
Knowledge about the discriminatory touch modal- characteristics. Haptics can also be used for education
ity can inform haptic interface design. It would and medical training. Students can learn physics in-
seem that haptic sensation of subtle effects relating to volving mechanical or electromagnetic forces while
texture and friction, which tend to be high-frequency actually feeling the forces. Virtual surgery can be per-
phenomena communicated mostly through the users formed, with the student feeling modeled viscoelastic
skin, is of major importance. It is precisely these high- properties of tissues. Multidimensional scientific data
frequency effects that permit a person to assess con- might be more easily understood through a visual-
ditions in the real world rapidly. For example, when haptic interface that allowed the user not only to see
a machinist removes a part from a lathe, the first in- the data, but also to feel it at any point. Haptic de-
stinct is to feel the quality of the parts finish; only later vices can also be used as hand controllers for virtual
is it visually inspected and measured. vehicles. For example, in a flight simulator, aerody-
Recently, psychologists have begun performing namic forces and vibration can be transmitted to the
experiments to evaluate the efficacy of haptic in- users hand to provide a more immersive experience.
teraction systems. In these experiments, subjects per- Many other potential applications are under con-
form tasks using only vision, or only haptics, or a sideration, including use by persons who are blind
combination of vision and haptics, and then task or visually impaired. Finally, haptics can be used to
performance is objectively measured. For example, control remote machinery, such as a robot, with
subjects may be asked to fit a virtual peg into a close- forces and torques reflected back to the operator.
fitting virtual hole, with the pegs (or the holes) posi- There is growing research activity in haptics
tion and orientation and the forces and torques on through the efforts of device designers, algorithm
the peg or hole measured in real time as the task pro- developers, and psychologists. As the field evolves,
gresses. Subjects in a control group are asked to fit a these disparate specialists are beginning to work
corresponding real peg into a real hole while mak- together and to share insights, generating new
ing the same measurements. By contriving to have knowledge in a multidisciplinary endeavor. Mean-
the setup for the virtual and real cases nearly iden- while, many future applications of haptics are under
tical, it is possible to assess the degree of haptic trans- consideration.
parency afforded by the haptic computer interaction
system. It is generally found that task performance Ralph L. Hollis
is enhanced when haptic feedback is included, but
subjects experience more difficulty dealing with vir- See also Augmented Reality; Virtual Reality
tual environments than they do with real environ-
ments. Results point to the need to improve both
haptic devices and haptic rendering algorithms.
FURTHER READING
Baraff, D. (1994, July). Fast contact force computation for non-penetrat-

How Haptic Computer Interaction ing rigid bodies. Computer Graphics, Proceedings of SIGGRAPH,
2334.
Can Be Used Berkelman, P. J., & Hollis, R. L. (2000, July). Lorentz magnetic levi-
Many people think that haptic computer interaction tation for haptic interaction: Device design, performance, and inte-
gration with physical simulations. International Journal of Robotics
will have widespread utility in many fields of activ- Research, 19(7), 644667.
ity. A number of application areas are under explora- Bolanowski, S. J., Gescheider, G. A., Verillo, R. T., & Checkosky, C.
tion, but none have as yet entered the mainstream. M. (1988). Four channels mediate the mechanical aspects of touch.
Journal of the Acoustical Society of America, 84(5), 16801694.
One such field is computer-augmented design (CAD), Brooks, Jr., F., Ouh-Young, M., Batter, J. J., & Kilpatrick, P. (1990).
in which countless products are developed for our Project GROPE: Haptic displays for scientific visualization. Com-
daily use. With a visual-haptic interface, the designer puter Graphics, 24(4), 177185.
Burdea, G. C. (1996). Force and touch feedback for virtual reality. New
York: John Wiley and Sons.
Cotin, S., & Delingette, H. (1998). Real-time surgery simulation with
HISTORY OF HCI
haptic feedback using finite elements. IEEE International Conference
on Robotics and Automation, 4, 37393744. The history of human-computer interaction (HCI)
James, D. L., & Pai, D. K. (1999, August). ArtDefo, accurate real time includes the evolution of widespread practices. It also
deformable objects. Computer Graphics, Proceedings of SIGGRAPH, includes people, concepts, and advances in under-
6672.
Jansson, G., Billberger, K., Petrie, H., Colwell, C., Kornbrot, D., Fnger, standing that inspired new developments. Often dec-
J. F., et al. (1999). Haptic virtual environments for blind people: ades elapse between visions or working demonstrations
Exploratory experiments with two devices. The International Journal of concepts and their widespread realization. The field
of Virtual Reality, 4(1), 1020.
LaMotte, R. H., & Srinivasan, M. A. (1990). Surface microgeometry:
of HCI can be understood in terms of existing prac-
Tactile perception and neural encoding. In D. Franzen & J. Westman tices, new visions, and hardware that became sub-
(Eds.), Information processing in the somatosensory system (pp. stantially more powerful year after year.
4958). New York: Macmillan.
Lederman, S. J., & Klatzky, R. L. (1999). Sensing and displaying spa-
tially distributed fingertip forces in haptic interfaces for teleop-
erator and virtual environment systems. Presence, 8(1), 86103. Human Factors before Computers
Massie, T. H., & Salisbury, J. K. (1994).The PHANToM haptic in- Through the centuries people developed highly spe-
terface: A device for probing virtual objects. Proceedings of ASME cialized tools to support carpenters, blacksmiths, and
Winter Annual Meeting, Dynamic Systems and Control, 55,
295301. other artisans. However, efforts to apply science and
McLaughlin, M. L., Hespanha, J. P., & Sukhatme, G. S. (Eds.). (2002). engineering to improving the efficiency of work prac-
Touch in virtual environments. Prentice Hall IMSC Press Multimedia tice became prominent only a century ago. Time-
Series. Upper Saddle River, NJ: Prentice Hall.
Seow, K. (1988). Physiology of touch, grip, and gait. In J. G. Webster
and-motion studies exploited inventions of that era
(Ed.), Tactile sensors for robotics and medicine (pp. 1340). New such as film and statistical analysis. The principles
York: John Wiley and Sons. of scientific management of the U.S. efficiency en-
Tan, H. Z., Lim, A., & Traylor, R. M. (2000). A psychophysical study gineer Frederick Taylor, published in 1911, had lim-
of sensory saltation with an open response paradigm. Proceedings
of the 9th International Symposium on Haptic Interfaces for Virtual
itations, but such principles were applied to U.S.
Environments and Teleoperator Systems, ASME Dynamic Systems assembly line manufacturing and other work prac-
and Control Division, 69(2), 11091115. tices in subsequent decades. World War I motivated
a similar focus in Europe.
World War II accelerated behavioral engineering
as complex new weaponry tested human capabilities.
HEALTH ISSUES One design flaw could cause thousands of casualties.
A legacy of the war effort was an enduring interest in
AND HCI human factors or ergonomics in design and training.
(Another legacy was the creation of the first digital
See Brain-Computer Interfaces; Cybersex; Keyboard; computers.)
Law and HCI; Privacy; Work Early approaches to improving work and the
man-machine interface focused on the nondis-
cretionary (mandatory) use of technology. The as-
sembly line worker was hired to use a system. The
soldier was given equipment. They had no choice in
HELP SYSTEMS the matter. If training was necessary, they were
trained. The goals of workplace study and technol-
See Adaptive Help Systems; Artificial Intelligence; Cog- ogy improvement included reducing errors in op-
nitive Walkthrough; Errors in Interactive Behavior; eration, increasing the speed of operation, and
Information Filtering; Instruction Manuals; User reducing training time. When use is nondiscretionary,
Support small improvements help.
HISTORY OF HCI 317
A Personal StoryHighlights from My Forty Years of HCI
Prehistory
1955 Receive my first desktop word processor when parents take out of storage an old black-framed Underwood type-
writer to do my school papers for junior high.
1957 Discover K&E slide rules at Bronx High School of Science for use in math and physics class. Think it really cool to
have an 18-inch one swinging from its own holster on my belt: a nerd gunslinger.
1959 As I go off to Lafayette College, parents give me my first laptop: a grey Royal portable typewriter, manual, of course.
1963 Leave for Harvard social relations graduate school with new baby-blue Smith Corona portable. The first affordable
electric typewriter, albeit with a manual carriage return.
1964 Make one of the most important decisions in my life, by taking a computer course. Learn to keypunch and use a
counter-sorter: the thing with twelve pockets that movies used to love because it gave good visuals.
1964 Discover that the command do not fold, bend, spindle or mutilate printed on my utility bills was because IBM
cards treated this way would jam in counter-sorters and accounting machines. Henceforth, mutilate all my utility
cards as a 1960s anti-bureaucratic protest and to create more jobs for workers who had to cope by hand with my
de-automated card.
Mainframe
1964 Learn to program in: FAP [early Assembler] and Fortran II and IV. Submit many jobs (stacks of IBM punch
cards) to the Harvard computer center, and sometimes get meaningful output back 10 hours later.
1966 Much of my dissertation data analysis done at this time on the new DataText do-it-yourself statistics program, which
liberates scholars from dependence on technicians to do analyses.
1967 Just before I leave Harvard, I view the remote teletype access between our building and Project Mac at MIT and
ARPAnet. I am amazed, but didnt fully appreciate that this was a precursor of the Internet revolution.
1973 Start using IBM Selectric typewriterwith correction buttonto write papers.
1976 Meet Murray Turoff and Roxanne Hiltz who are developing EIES, one of the first civilian e-mail-like systems, com-
bining messaging and computerized conferencing. I happily join them and have not been offline since.
Remote Access to Mainframe

1979 New HCI interfaces arrive at the University of Toronto. I can submit computer runs from a typewriter-like termi-
nal near my office instead of having to trudge to the mainframe building.
1985 E-mail becomes prevalent on a less-experimental basis: Bitnet in the U.S./Canada and later Netnorth in Canada.
The two systems interconnect and link less easily with European equivalents. I schmooze with people near and
far, and have never stopped.
Stand-Alone Personal Computing

1987 Buy my first stand-alone computer, the hot new 6-megahertz IBM AT with a 20-megabyte hard disk, one of the
hottest machines of the time. No longer must I negotiate with secretaries to type and retype drafts of papers. This
leads to more editing of drafts by me, but also greater productivity in the number of papers published.
The Internet Matures

1990 Cajole the director of the centre to patch together an Ethernet line running through all offices. It connects to a router
that links to the computer centre mainframe
A Personal StoryHighlights from My Forty Years of HCI (continued)
1992 Buy my first laser printer for personal use: a Hewlett-Packard 4M. No need to trudge down the hall to print final
versions. Now have the ability to print pictures and graphs at my own desk.
1992 The modern Internet era begins, with the expansion of e-mail to the rest of the world, especially through dot.com
Internet addresses and commercial Internet Service Providers.
1994 Buy my first computer for the house. Saves me the trouble of going to the office on nights and weekends to do
statistical analyses, e-mail or write.
1995 Buy my first laptop (Dell) to take with me as a visiting professor at Hebrew University. To cope with busy signals,
I set my alarm for 2 am to rise and sign-on to the Internet.
Powerful Personal Computing

1996 Start Netville study with Keith Hampton of a highly wired suburb near Toronto, with 16MB connection. Discover
that even more than high-speed Internet connections, people value the ability to always keep their Internet con-
nections on so that they can quickly share a thought.
1996 Personal computers now fast enough that it is feasible to do large statistical analyses on them. Their graphical in-
terfaces make it easier to compile commands at the cost of some flexibility.
1997 Buy first printer for home use. Now there is even less reason to go to the office.
1998 Early adopter of Google search engine. Miraculously fast, awesome coverage, and it uses social network analytic
principles to identify important sites. Bookmarking now becomes a minor convenience rather than an absolute ne-
cessity for navigating the web.
1999 Obtain high-speed broadband connection (DSL) for the house, making it easy to access websites rapidlyand our
phone line is no longer tied up by computer use.
The Internet Proliferates

2001 Set up my own website which has 35,000 hits by March 2004. People stop writing to me for my papers; they just
go to the site and download
2002 Teach an undergraduate Internet and Society course in a smart classroom where each student has a PC on her
desk, hardwired into the Internet. Biggest challenge is to stop students from emailing, IMing, and web surfing
during class so that they will pay attention to my lectures.
2002 Student assistants reduce their trips to the library, since most scholarly journals now have electronic versions online.
I then accumulate many articles to read because they are so easy to obtain online. Sometimes I even get to read them.
2003 Phone calls have essentially stopped coming to the office, except from my wife and the news media. Get over 100
e-mail messages a day (half of them spam), and send about 50.
2004 Have more than 2,500 e-addresses in my address book.
2004 After forty years of computing, I suffer a herniated (slipped) cervical disk. I have sat at computers for too many
hours, days, and years.
Barry Wellman
Source: Wellman, B. (2004). HCI: A personal timeline. Retrieved April 1, 2004, from http://www.chass.utoronto.ca/~wellman/publications/hci_timeline/timeline-hci.pdf
HISTORY OF HCI 319
Early Visions and Demonstrations Nondiscretionary or discretionary use? Real life

Until transistor-based computers appeared in 1958, lies somewhere along a continuum from the assem-
visions of their future use existed mainly in the realm bly line nightmare satirized in the English actor Charlie
of science fiction because vacuum tube-based com- Chaplins movie Modern Times to utopian visions of
puters had severe practical limitations. The most in- completely empowered people. Although some de-
fluential early vision was Vannevar Bushs 1945 essay sign opportunities and challenges affect the spectrum
As We May Think. Bush, who played a key role in of computer use, distinct efforts focused on discre-
shaping government funding of research in science tionary and nondiscretionary use have proceeded in
and technology during and after World War II, de- parallel with only modest communication.
scribed an inspiring albeit unrealistic mechanical de-
vice that anticipated many capabilities of computers.
Key writings and prototypes by HCI pioneers ap- The First Forty Years
peared during the next decade. J.C R. Licklider, a A key to understanding the evolution of human-
research manager with a background in psychol- computer interaction is to describe the evolution of
ogy, outlined requirements for interactive systems who interacted with computers, why they did so, and
and accurately predicted which would prove easier how they did it. (See Figure 1.)
and which more difficult to fulfill (such as visual dis- The first computer builders did everything them-
plays and natural language understanding, respec- selves. Following the appearance of commercial sys-
tively). Computer scientists John McCarthy and tems, for three decades most hands-on computer
Christopher Strachey proposed time-sharing systems, users were computer operators. Programmers and
crucial to the spread of interactive computing. In 1963 those people who read printed output interacted
Ivan Sutherlands Sketchpad display system with computers but not directly. Labor was thus di-
demonstrated copying, moving, and deleting of hi- vided into these three categories: (1) Operators inter-
erarchically organized objects, constraints, iconic rep- acted directly with a computer: maintaining it,
resentations, and some concepts of object-oriented loading and running programs, filing printouts, and
programming. Douglas Engelbart formulated a broad so on. (2) Programmers, a step removed from the
vision, created the foundations of word processing, physical device, might leave a job in the form of
invented the mouse and other input devices, and con- punched cards to be run at a computer center, pick-
ducted astonishing demonstrations of distributed ing up the cards and a printout the next day. (3) Users
computing that integrated text, graphics, and specified and used a programs output, a printout,
video. Ted Nelsons vision of a highly interconnected or report. They, too, did not interact directly with
network of digital objects foreshadowed aspects of the computer.
World Wide Web, blog (or Web log), and wiki (multi-
authored Web resource) technologies. Rounding out
this period were Alan Kays visions of personal com- Supporting Nondiscretionary Use
puting based on a versatile digital notebook.
These visions focused on the discretionary use of by Computer Operators
technology, in articles that included:Man-Computer In the beginning the computer was so costly that
Symbiosis, Augmenting Human Intellect, and A it had to be kept gainfully occupied for every sec-
Conceptual Framework for Man-Machine Everything. ond; people were almost slaves to feed it. (Shackel
Technology would empower people to work and inter- 1997, 997)
act more effectively and flexibly. These visions inspired For the first half of the computer era, improving
researchers and programmers to work for the decades the experience of hands-on users meant supporting
needed to realize and refine them. Some of the capa- low-paid operators. An operator handled a computer
bilities that the visions anticipated are now taken for as it was, setting switches, pushing buttons, read-
granted; others remain elusive. ing lights, feeding and bursting (separating) printer
paper, loading and unloading cards, magnetic tapes, methods, selection rules) were used to help quin-
and paper tapes, and so on. tessentially nondiscretionary users such as telephone
Teletypes were the first versatile mode of direct operators, people engaged in repetitive tasks involving
interaction. Operators typed commands, the com- little reflection. GOMS added higher level cognitive
puter printed responses or spontaneous status elements to the perceptual-motor focus of knobs
messages. The paper printout scrolled up, one line at and dials human factors.
a time. Displays (called video display units or video A series of ergonomically justified interface guide-
display terminals or cathode ray tubes) were nick- lines culminated in 1986 with the publication of
named glass ttys (glass teletypes) because they func- human factors experts Sidney Smith and Jane Mosiers
tioned much the same, displaying and scrolling up 944 guidelines. Sections were entitled Data Entry,
typed operator commands, computer-generated Data Display, Data Transmission, Data Protection,
responses, and status messages. Most were mono- Sequence Control, and User Guidance. The em-
chrome and restricted to alphanumeric characters. phasis was on supporting operators. The guidelines
The first display to be marketed commercially cost mentioned graphical user interfaces (GUIs), then a
around $50,000 in todays dollars. They were expen- new development, and the major shift and expansion
sive but cost a small fraction of a business computer. of the design space ushered in by GUIs may have been
Typically one console or terminal accompanied a a factor in discontinuing the guideline effort.
computer for use by an operator. By then change was rippling through the indus-
Improving the design of buttons, switches, and try. Mainframe computers and batch processing still
displays was a natural extension of traditional hu- dominated, but time sharing of computers was al-
man factors. In 1959 Shackel published Ergonomics lowing new uses, minicomputers were spreading,
for a Computer, followed in 1962 by Ergonomics and microcomputers were starting to appear. Hands-
in the Design of a Large Digital Computer Console. on computing was becoming available to people who
Little published research followed for a decade. In were not computer professionals, who would use
1970 Shackels HUSAT (Human Sciences and Ad- technology only if it helped them work better.
vanced Technology) research center formed, focused Improving the life of discretionary users had a his-
on general ergonomics. tory in the visions of Bush and others, of course, but
The first influential book was James Martins 1973 also in the support for the other two categories of com-
Design of Man-Computer Dialogues. After a visionary puter users: programmers and users of the output.
chapter that remains interesting to read, the book sur-
veyed existing approaches to supporting operators.
Yet, it was written for programmers and conveyed a Supporting Discretionary Use by
sense of changes ahead.
In 1980 five major HCI books were published; two Computer Programmers
focused on video display terminal (VDT) design and Early programmers used a computer directly when
one on general ergonomic guidelines. Germany pub- they could because doing so was fun and faster.
lished VDT standards in 1981. By threatening exist- However, the cost of computers largely dictated the
ing products, these standards made designing for division of labor noted previously. Working as a pro-
human capabilities a visible economic issue. grammer during the mid-1970s, even at a computer
Also during 1980 Stuart Card, Thomas Moran, company, typically meant writing programs on pa-
and Allen Newells article Keystroke-Level Model per that were then punched onto cards by keypunch
for User Performance Time with Interactive Systems operators. The jobs were run by computer operators,
was published. They wrote: The central idea behind and the programmer received printed output.
the model is that the time for an expert to do a Improving the programmers interface to a com-
task on an interactive system is determined by the puter meant developing constructs (e.g., subroutines),
time it takes to do the keystrokes (397). This model compilers, and programming languages. Grace Hopper
and successors such as GOMS (goals, operators, was a farsighted pioneer in this effort through the 1950s.
HISTORY OF HCI 321
Programmers also worked to advance computer computer specialiststhat is, those people we now
technology. In 1970 the Xerox companys Palo Alto call discretionary users. The preface of the book
Research Center (PARC) was founded, with a focus echoed early visions: Its not enough just to estab-
on advancing computer technology. In 1971 Allen lish what people can and cannot do; we need to spend
Newell proposed a project that was launched three just as much effort establishing what people can and
years later: Central to the activities of computing want to do (viii).
programming, debugging, etc.are tasks that appear Another effort to bridge the gap between pro-
to be within the scope of this emerging theory of the grammer and other professionals emerged in John
psychology of cognitive behavior (quoted in Card and Goulds group at IBM Watson Labs. Like the PARC
Moran 1986, 183). PARC and HUSAT were launched applied psychology group, the Gould group evolved
in 1970 and engaged in a broad range of research but through the 1970s and 1980s to a cognitive focus
with an interesting contrast. HUSAT research was fo- from one that included perceptual-motor studies
cused on ergonomics, anchored in the tradition of and operator support. In order to expand the mar-
nondiscretionary use, one component of which was ket for computers, IBM realized it would be neces-
the human factors of computing. PARC research was sary to make them usable by people who could not
focused on computing, anchored in visions of dis- be expected to program complex systems.
cretionary use, one component of which was also the Many key participants in early HCI conferences,
human factors of computing. PARC researchers ex- including Ruven Brooks, Bill Curtis, Thomas Green,
tended traditional human factors to higher level cog- and Ben Shneiderman, had studied psychology of
nition; HUSAT and European researchers introduced programming. Papers written on programmers as
organizational considerations. users were initially a substantial part of these con-
Thousands of papers written on the psychology ferences but gradually disappeared as programmers
and performance of programmers were published became a smaller subset of computer users.
during the 1960s and 1970s. Gerald Weinberg pub- Other factors contributed to a sense that HCI
lished the book The Psychology of Computer Program- was a new undertaking. Graphic displays dropped
ming in 1971. In 1980, the year when three books in price and became widely used during the late
on VDT design and ergonomics were published, Ben 1970s, opening a large, challenging design space.
Shneiderman published Software Psychology. In 1981 In the United States, academic hiring of cognitive
B. A. Sheil wrote about studies of programming nota- psychology Ph.D.s fell sharply during the late 1970s,
tion (e.g., conditionals, control flow, data types), pro- just when computer and telecommunication com-
gramming practices (flowcharting, indenting, variable panies were eager to hire psychologists to tackle per-
naming, commenting), and programming tasks ceptual and cognitive design issues.
(learning, coding, debugging) and included a section In 1969 Association for Computing Machinery
on experimental design and statistical analysis. (ACM) had formed a special interest group (SIG) for
With time sharing and minicomputers in the late social and behavioral scientists using computers as re-
1970s and 1980s, many programmers became en- search tools. In 1982 this group of discretionary com-
thusiastic hands-on users. Ongoing studies of pro- puter users decided to change its name and charter to
grammers became studies of hands-on users. When the Special Interest Group on Computer-Human
personal computing was introduced, studies shifted Interaction (SIGCHI), focusing on behavioral stud-
to other discretionary users. ies of computer use or human-computer interaction.
The book Human interaction with computers, ed- SIGCHI drew heavily from cognitive psychology and
ited by Thomas R. G. Green and Harold T. Smith and software psychology and from sympathetic pro-
also published in 1980, foreshadowed the shift. With grammers and computer scientists. Many program-
a glance at the human as a systems component, one mers and scientists were unaware of prior human
third of the survey was devoted to research on pro- factors studies of operators. Some cross-publication
gramming and the rest to designing for non-spe- existed between human factors and human-computer
cialist people, meaning people who were not interaction, but the endeavors remained distinct.
In Europe computer companies exerted less in- research on the organization and layout of informa-
fluence, and research boundaries were less distinct. tion focused on both human factors and the psychol-
In 1977 the Medical Research Council Applied ogy of programming.)
Psychology Unit, renowned for theoretically driven MIS also introduced an organizational focus: ap-
human factors research, initiated an IBM-funded HCI proaches to deploying systems, resistance to systems,
project with a focus on discretionary use. In a 1991 effects of using systems. As HCI came to include more
survey with a European perspective, Liam Bannon group support issues, people found and explored
decried the slow shift to a discretionary focus while commonalities in computer-supported-cooperative
also critiquing those people who adopted that fo- work conferences.
cus for mostly considering initial use by new users. Whereas by 1985 almost all programmers were
Some tensions existed between the human fac- hands-on computer users, until the late 1990s most
tors and community interaction communities. managers avoided hand-on use. Managers still dele-
The former felt that its past work was not fully appre- gate much technology use, but most now use some
ciated by the latter. Although methods and goals over- software directly. Not surprisingly, management inter-
lapped, the agendas of the two camps differed. A 1984 est in HCI is growing, with an interest group and sig-
study contrasting performance and preference of nificant symposia and workshops conducted at major
users found evidence that even for a repetitive conferences since 2001.
task, users might prefer an interaction technique that
was pleasant but slower. Although this evidence was
of interest to people studying discretionary use, a Government Role in System Use
leading GOMS proponent recommended suppress-
ing its publication lest it undermine the mission of and Research Funding
maximizing performance. Governments were the major purchasers of comput-
Businesses acquired the first expensive business ers during the decades when feeding the computer
computers to address major organizational concerns. was the norm. In addition to operators, governments
Sometimes merely the prestige of an air-conditioned, employed vast numbers of data entry and other
glass-walled computer room justified the expense, nondiscretionary users. Supporting these people
but most computers were put to work. Output was meshed naturally with the focus on designing to fit
routed to managers. In the field variously called data human capabilities that arose in the world wars.
processing (DP), management information sys- Competitively bid contracts present challenges
tems (MIS), information systems (IS), and infor government acquisition of systems. The govern-
formation technology (IT), the term users referred ment has to remain at arms length from the devel-
to these managers. Like early programmers, they were oper yet needs to specify requirements in advance.
well paid, discretionary, and not hands on. This situation led the U.S. government to participate
Supporting managerial use of computers meant in establishing ergonomic standards during the
improving the visual display of information, first on late 1970s and 1980s. Compliance with interface de-
paper and eventually on displays as well. Because sign standards and guidelines could be specified in
much computer output was quantitative, research a government contract. For example, Smith and
included the design of tables, charts, and graphs Mosiers guideline development effort mentioned
business graphics were one focus of much MIS us- previously was funded by the U.S. Air Force.
ability research. Interaction with information remains Government agencies were early adopters of
central today: Human Computer Interaction stud- computers, but the work conducted by such agen-
ies in MIS are concerned with the ways humans cies changes only gradually. Such agencies no longer
interact with information, technologies, and tasks employ computer operators in large numbers, but
(Zhang et al 2002, 335). (All computer users viewed huge numbers of data entry and handling person-
information. Operators used manuals and displays, nel remain at agencies concerned with such issues as
programmers used flowcharts and printouts. Thus, census, taxes, and health and welfare. Power plant
HISTORY OF HCI 323
Nondiscretionary Use Discretionary Use
HumanCompute
2005
Ongoing priority for
Interaction
government funding Norman
DUX 2003
IS/HCI organized
1995
Smith & Mosier

1985 HCI journal appears
German VDU standards SIG CHI formed
Keystroke-level model Smith & Green

1980 Shneiderman Psychology
Three ergonomics books
of
Martin 1975 Martin Programming
Human Factors
and
HUSAT founded 1970 Xerox PARC founded
Ergonomics
Kay
Shackel 1965 Nelson Visions
Sutherland and Prototypes
Engelbart
Licklider
Transistor-based commercial computers
1955
Hopper
WWII human factors 1945 Bush
WWI training
1915
Taylorism
FIGURE 1. Timeline with publications and events

operators and air traffic controllers are glued to sys- variety of disciplines is researching, developing, and
tems that evolve very gradually. Ground control op- marketing interactive systems.
erations for space launches require highly trained Few computer and software companies that fo-
computer users. Soldiers require training in equip- cused on nondiscretionary use during the mainframe
ment; weaponry grows ever more complex. The and minicomputer eras still do today. Most major ven-
quantity of text and voice intercepts processed by in- dors that thrived then are gone today, and the few that
telligence agencies is immense. Overall, government remain (IBM comes to mind) reformed themselves
remains a major employer of nondiscretionary com- during the transition to discretionary use of com-
puter users. Improving their work conditions is a puters during the 1980s. Most companies that are now
central concern. Small efficiency gains in individual active in human-computer interaction research and
interactions can provide large benefits over time. innovation came into prominence during the 1980s
Government is also a major funding source for and 1990s, by which time it was a good commercial
information technology research. In Europe national strategy to appeal to users who exercised discretion
and European Union initiatives have been the either individual purchasers or organizations.
principal funding source. In Japan the government HCI practitioners started with methods and per-
has funded major initiatives with HCI components, spectives from human factors and experimental psy-
such as the Fifth-Generation Project. Since World chology, much as interactive software developers
War II the U.S. National Science Foundation, armed inherited waterfall modelsdefining a series of steps
services (led by proponents of basic research in the in designing, developing, testing, and maintaining a
Office of Naval Research), and intelligence agen- new productthat had been crafted for other pur-
cies have been major sources of funding, although poses. The need for new methods was obscured by
research laboratories established by telecommuni- the fact that traditional ergonomic goals of fewer er-
cation, hardware, and software companies have also rors, faster performance, quicker learning, greater
been prominent since the 1970s. memorability, and enjoyment still applied, although
U.S. government funding remains focused on not with the same relative priority.
nondiscretionary computer use, with greatest em- Several factors led to change. The notion that
phasis being on speech recognition and natural lan- friendly interfaces are frills in the workplace was
guage understanding. Little research on these two eroded when people asked, Why shouldnt my ex-
topics appears in HCI conferences, even though some pensive office system be as congenial as my home
people hope that the topics will eventually be more computer? Also, as more software appeared,
useful in discretionary situations. The National training people on each application was not feasible.
Science Foundation has also funded substantial work Ease of learning was critical.
on using brainwaves to guide computer displays. This As software came to support more group activ-
is another technology that may have its uses but prob- ities and detailed work practices, lab studies were
ably not in many homes and offices. Research on supplemented with social and ethnographic (cul-
nondiscretionary use is published in specialized jour- tural studies) methods in research and practice.
nals and at conferences, including those of the Contextual design and personas (fully specified
Human Factors and Ergonomics Society and HCI but fictional users) are recent innovations far re-
International. moved from techniques of early HCI.
Finally, the need to seduce discretionary users
grew as software became more capable and com-
Corporate/Academic Role in petitive. The Web promoted design aesthetics (not
a major issue for data entry operators) and market-
System Use and Research ing (previously considered a distraction) as central
Companies employ the information workers whose to human-computer interaction. Reflecting this ex-
discretion has been growing, and a subset of tech- pansion of focus, SIGCHI co-sponsored the Designing
nology company employees and academics from a for User Experiences (DUX) 2003 conference.
HISTORY OF HCI 325
This evolution is captured by the work of Don cant work without it. Increased digitally mediated
Norman. In the first paper presented at the first CHI collaboration forces us to adopt the same systems and
(computer-human interaction) conference,Design conventions for using these systems. If we have choices,
Principles for Human-Computer Interfaces, he we may have to exercise them collectively. This situ-
focused on tradeoffs among the attributes of speed ation in turn motivates people to develop features for
of use, prior knowledge required, ease of learning, customization and interoperation. For example, in
and errors. Twenty years later, in 2003, he published 1985 each member of a team chose a word proces-
Emotional Design. sor and exchanged printed documents with other
members. In 1995 the team members may have had
to use the same word processor to share documents
Related Theory and Disciplines digitally. Today the team members can again use
The early 1980s were marked by a strong effort to different word processors if sharing documents dig-
provide a theoretical framework drawn especially itally in PDF format suffices.
from cognitive psychology for a field previously dom- One constant in the computer era has been the
inated by an engineering perspective. This effort was keyboard and display as a central interface compo-
paralyzed by the rapid changes in the field. Graphical nent. With relentless miniaturization of components
user interfaces and multimedia swept away interest and increase in power, this era is fading. The expan-
in phenomena that been the focus of theoretical sion of human-computer interaction is clearly only a
analysis, such as command naming (selecting mem- beginning.
orable names for computer commands), and in-
troduced a daunting array of new challenges. The Jonathan Grudin
growing focus on computer-mediated interaction
See also Altair; Desktop Metaphor; ENIAC; Graphical
between humans challenged the centrality of cogni-
User Interface
tion, and awareness of the role of design and mar-
keting fur ther reduced the prospects for an
encompassing theoretical framework.
A recent compilation of theory and modeling FURTHER READING
approaches to HCI includes several chapters with a Baecker, R., Grudin, J., Buxton, W., & Greenberg, S. (1995). Read-
cognitive orientation, a few with social science or ings in human-computer interaction: Toward the year 2000. San
cognitive-social hybrids, and one focused on com- Francisco: Morgan Kaufmann.
Bannon, L. (1991). From human factors to human actors: The role of
puter science. As the academic home of HCI moved psychology and human-computer interaction studies in system de-
from psychology to computer science, HCI be- sign. In J. Greenbaum & M. Kyng (Eds.), Design at work (pp. 2544).
came more entwined with software engineering. Hillsdale, NJ: Erlbaum.
Barnard, P. (1991). Bridging between basic theories and the artifacts
Artificial intelligence has also had several points of of human-computer interaction. In J. M. Carroll (Ed.), Designing
contact with HCI. interaction: Psychology at the human-computer interface (pp. 103127).
Cambridge, UK: Cambridge University Press.
Beyer, H., & Holtzblatt, K. (1998). Contextual design. San Francisco:
The Trajectory Morgan Kaufmann.
Bush, V. (1945). As we may think. The Atlantic Monthly, 176, 101108.
The unanticipated arrival and consequences of the Card, S. K., & Moran, T. P. (1986). User technology: From pointing to
Web demonstrated the difficulty of anticipating the pondering. Proceedings of the Conference on History of Personal Work-
future, but a key goal of organizing a history is to stations, 183198.
Card, S. K., Moran, T. P., & Newell, A. (1980). Keystroke-level model
identify trends that may continue. for user performance time with interactive systems. Communi-
Discretionary computer use continues to spread. cations of the ACM, 23(7), 396410.
Nondiscretionary use remains significant and bene- Carroll, J. M. (Ed.). (2003). HCI models, theories and frameworks.
San Francisco: Morgan Kaufmann.
fits from better understanding and interfaces wher- Carroll, J. M., & Campbell, R. L. (1986). Softening up hard science:
ever they originate. For many people software use that Response to Newell and Card. Human-Computer Interaction, 2(3),
was once discretionary has become mandatorywe 227249.
Dyson, F. (1979). Disturbing the universe. New York: Harper & Row. Sutherland, I. (1963). Sketchpad: A man-machine graphical com-
Engelbart, D. (1963). A conceptual framework for the augmenta- munication system. AFIPS, 23, 329346.
tion of mans intellect. In P. Howerton & D. Weeks (Eds.), Vistas Taylor, F. W. (1911). The principles of scientific management. New
in information handling (pp. 129). Washington, DC: Spartan York: Harper.
Books. Weinberg, G. (1971). The psychology of computer programming. New
Engelbart, D., & English, W. (1968). A research center for augment- York: Van Nostrand Reinhold.
ing human intellect. AFIPS Conference Proceedings, 33, 395410. Zhang, P., Benbadsat, I., Carey, J., Davis, F., Galleta, D., & Strong, D.
Fano, R., & Corbato, F. (1966). Time-sharing on computers. Scientific (2002). Human-computer interaction research in the MIS disci-
American, 214(9), 129140. pline. Communications of the Association for Information Sys-
Greenbaum, J. (1979). In the name of efficiency. Philadelphia: Temple tems, 9, 334355.
University Press.
Grudin, J. (1990). The computer reaches out: The historical continu-
ity of interface design. In Proceedings of the SIGCHI conference
HOLLERITH CARD
on human factors in computing systems 90 (pp. 261268). New
York: ACM Press.
Grudin, J., & MacLean, A. (1984). Adapting a psychophysical method
to measure performance and preference tradeoffs in human-com-
puter interaction. In Proceedings of INTERACT 84 (pp. 338342). The Hollerith card, also known as a punch card or
Amsterdam: North Holland. an IBM card, was the preeminent digital medium
Kay, A., & Goldberg, A. (1977). Personal dynamic media. IEEE Com- of data entry and storage for three-quarters of a cen-
puter, 10(3), 3142.
Licklider, J. (1960). Man-computer symbiosis. IRE Transactions of
tury until its replacement by the magnetic floppy
Human Factors in Electronics, 1(1), 411. disk. Hollerith cards were part of a system that spec-
Licklider, J., & Clark, W. (1962). On-line man-computer communi- ified particular relations between human beings and
cation. AFIPS Conference Proceedings, 21, 113128. data processing machinery that were very different
Long, J. (1989). Cognitive ergonomics and human-computer interac-
tion. In J. Long & A. Whitefield (Eds.), Cognitive ergonomics and from the modern relations between human beings
human-computer interaction (pp. 434). Cambridge, UK: Cam- and real-time, networked systems.
bridge University Press. Herman Hollerith (18601929) attended Columbia
Martin, J. (1973). Design of man-computer dialogues. Englewood Cliffs,
NJ: Prentice Hall.
College School of Mines and then gained valuable ex-
Nelson, T. (1965). A file structure for the complex, the changing, perience working on the 1880 U.S. census.A tabulating
and the indeterminate. In Proceedings of the ACM National Con- machine, invented by Charles W. Seaton, did none of
ference (pp. 84100). New York: ACM Press. the counting itself but merely moved scrolls of paper
Nelson, T. (1973). A conceptual framework for man-machine every-
thing. In AFIPS Conference Proceedings (pp. M21-M26). Montvale, between rollers so that clerks could write information
NJ: AFIPS Press.. conveniently.After working with Seaton,Hollerith taught
Newell, A., & Card, S. K. (1985). The prospects for psychological at the Massachusetts Institute of Technology while ex-
science in human-computer interaction. Human-Computer Inter-
action, 1(3), 209242.
perimenting with his own ideas about data automation.
Norman, D. A. (1983). Design principles for human-computer inter- Holleriths first approach was to punch holes in
faces. In Proceedings of the SIGCHI Conference on Human Factors long paper strips, but he quickly switched to cards
in Computing Systems (pp. 110). New York: ACM Press. because correcting errors on punched paper strips
Norman, D. A. (2003). Emotional design: Why we love (or hate) every-
day things. New York: Basic. was difficult. For a century, data had been encoded
Sammet, J. (1992). Farewell to Grace HopperEnd of an era! Commu- as holes in cards, for example, in music boxes, ex-
nications of the ACM, 35(4), 128131. perimental player pianos, and the Jacquard loom that
Shackel, B. (1959). Ergonomics for a computer. Design, 120, 3639.
Shackel, B. (1962). Ergonomics in the design of a large digital com-
controlled complex patterns in weaving cloth.
puter console. Ergonomics, 5, 229241. Hollerith was inspired by seeing train conductors
Shackel, B. (1997). Human-computer interaction: Whence and whither? punch tickets that recorded information about pas-
JASIS, 48(11), 970986. sengers, but he found that the conductors punch
Shiel, B. A. (1981). The psychological study of programming. ACM
Computing Surveys, 13(1), 101120. tool caused painful cramping of the hand. Hollerith
Shneiderman, B. (1980). Software psychology: Human factors in com- filed a series of patents in the 1880s, then demon-
puter and information systems. Cambridge, MA: Winthrop. strated his system with the 1890 census.
Smith, H. T., & Green, T. R. G. (Eds.). (1980). Human interaction with
Holleriths 1890 system combined manual meth-
computers. New York: Academic.
Smith, S. L., & Mosier, J. N. (1986). Guidelines for designing user inter- ods with both mechanical and electric methods. A
face software. Bedford, MA: MITRE. clerk entered data with a manual pantograph punch
HOLLERITH CARD 327
that helped locate the right points on cards and operator, where they could be clearly seen as they
was more comfortable to use than a conductors jumped sideways when each hole was punched, un-
punch tool. To tabulate the data, a clerk would place til they zipped into the output pile. The operator
the cards into a press one at a time. The top of the might place around a drum a previously punched
press held a number of pins that could move up or control card that programmed the keypunch to skip
down, one for each possible hole. If the hole had been or specially punch certain columns.
punched, the pin would make electrical contact with After they were punched, the cards would be
a drop of mercury in a tiny cup, which would acti- placed into card readers, and processed data could be
vate a counter. If the hole had not been punched, the automatically punched on a new set of cards if de-
pin would ride up and fail to close the circuit. To the sired. Cards locked users into the batch processing
right of the press was a sorter consisting of boxes mode, in which users would carefully prepare a com-
with electrically operated lids. When a clerk closed puter run at the keypunch machine, then wait as long
the press, one of the lids would open so that the clerk as several hours for the fanfold computer printout
could place the card into that particular box. For ex- that was the result of the particular job, but users could
ample, if each card represented data about one not interact directly in real time with the computer.
person, the sorter helped a clerk divide males from A typical card was rectangular, with space for
females for later separate analysis. eighty columns of rectangular holes and room for
During the following years Hollerith developed twelve holes in each column, ten of which were marked
each part of this system: punching, counting, and sort- by printed numerals 09. One hole in a given column
ing. Within a decade the cumbersome mercury cups represented a single digit, and a second hole in row 11
and lidded boxes had been replaced, and cards were could mean a minus sign. Letters of the alphabet and
automatically fed at great speed through the tabulat- other characters were represented by the combination
ing and sorting equipment. Sets of electric relays com- of one hole in the first ten rows plus zone punches
bined data from different variables much as transistors in rows 11 and 12 or by multiple punches in the nu-
would do in computers a century later but were pro- merical rows. Multipunched cards were flimsy, so cau-
grammed manually by plugging wires, as in the pe- tious users made duplicate copies of their card decks.
riods telephone operator equipment. Goaded by a One advantage of Hollerith cards over mag-
competitor, Hollerith added electrically driven key- netic data media is that human beings can read the
punch machines in 1914. His system included inno- cards directly. Already in 1902 Holleriths keypunch
vative business practices, such as leasing rather than could typewrite the data across the top of the columns,
selling the machines and making much of his profit and he himself noted how easy it was to sort the cards
by selling the cards themselves. The Tabulating Machine manually by gently putting a knitting needle through
Company he founded in 1896 was a precursor of IBM. the holes. He cut the upper left corner off the cards
Hollerith cards remained one of the major meth- so a user could immediately see if any had been placed
ods for data input for electronic computers as the wrong way. The great fear of card users was that
these machines were introduced during the 1940s they might spill an entire box of two thousand
and 1950s. As late as 1980, many U.S. universities still punched cards, so many users marked lines diago-
had keypunch machines on which scientists and ad- nally across the top edges to facilitate reordering them
ministrative staff would enter data. In terms of hu- if necessary. Hollerith cards may have been incon-
man-computer interaction, these were dramatic and venient in many respects, but they helped launch
rather loud machines that could actually damage the modern information technology and gave users a
users hearing if many machines were in a small more intimate experience of data than do todays fully
room. A typical form was a desk with a fixed key- electronic media.
board and the apparatus above and at the back. A
stack of cards would be placed into a hopper, William Sims Bainbridge
where the cards would feed one at a time into the
machine. They were punched directly in front of the See also ENIAC
uted computing (which allow for more speedy com-

FURTHER READING putation through the use of multiple processors
and/or computers), and more sophisticated user in-
Austrian, G. D. (1982). Herman Hollerith: Forgotten giant of infor- terface design.
mation processing. New York: Columbia University Press.
By the 1980s, robotics was recognized as funda-
mentally interdisciplinary, with major contributions
from mathematics, biology, computer science, con-
HUMAN FACTORS trol theory, electrical engineering, mechanical engi-
neering, and physics. By the 1990s, robots were
ENGINEERING increasingly involved in automated manufacturing
environments, in deep-sea and space exploration, in
military operations, and in toxic-waste management.
See Anthropometry; Keyboard; Task Analysis Predictions abounded that robots would become im-
portant in home and office environments as well.
At the beginning of the twenty-first century, we are
closer to the day when various robot entities may
HUMAN-ROBOT be integrated into peoples daily lives.
Just as computers began as academic and research-
INTERACTION related computational tools but became personal elec-
tronic accessories for the general public, robots now
The relationship between robots and humans is so dif- have the potential to serve not only as high-tech work-
ferent in character from other human-machine rela- horses in scientific endeavors but also as more per-
tionships that it warrants its own field of study. Robots sonalized appliances and assistants for ordinary people.
differ from simple machines and even from complex However, while the study of human-computer inter-
computers in that they are often designed to be mo- action has a relatively long history, it is only recently
bile and autonomous. They are not as predictable as that sufficient advances have been made in robotic per-
other machines; they can enter a humans personal ception, action, reasoning, and programming to allow
space, forcing a kind of social interaction that does not researchers to begin serious consideration of the cog-
happen in other human-machine relationships. nitive and social issues of human-robot interaction.
Background From Human-Computer Interaction

The term robot first entered literature through the
play R.U.R. (1920), by the Czech playwright and nov- to Human-Robot Interaction
elist Karel Capek (18901938); R.U.R. featured hu- In the past, techniques and methodologies developed
manoid devices as servants for humans. In the under the general umbrella of user- or human-
mid-1950s, the first true robots appeared. A human centered computing began by looking at static
operator working from a distance ran these devices, (unintelligent) software applications and their re-
which had the ability to carry out numerical compu- lated input and output devices. Today these tech-
tations and contained mechanisms to control machine niques are being extended to consider issues such as
movement. The rest of the twentieth century saw robot- mobile wireless technology, wearable augmentation
ics continue to make significant advances in such ar- devices (such as miniature heads-up displays and
eas as more flexible motion, refined manipulators cameras), virtual reality and immersive environments,
(e.g., articulated hands and walking devices), and in- intelligent software agents (both cooperative and au-
creased intelligence. Researchers took advantage of tonomous), and direct brain interface technologies.
progress in computer science and software engineer- In addition, mobile robotic agents are now poised
ing, including developments in parallel and distrib- to become part of our everyday landscapein the
HUMAN-ROBOT INTERACTION 329
workplace, in the home, in the hospital, in remote answer phones, open mail, deliver documents to dif-
and hazardous environments, and on the battlefield. ferent departments of a company, make coffee, tidy
This development means we have to look more up, and run the vacuum. Due to the nature of the in-
closely at the nature of human-robot interaction; telligence needed for robots to perform such tasks,
and define a philosophy that will help shape the there is a tendency to think that robots ought to be-
future directions of this relationship. come more like humans, that they need to interact
Human interface and interaction issues continue with humans (and perhaps with one another) in the
to be important in robotics research, particularly since same way that humans interact with one another, and
the goal of fully autonomous capability has not yet that, ultimately, they may replace humans altogether
been met. People are typically involved in the super- for certain tasks. This approach, sometimes termed
vision and remote operation of robots, and interfaces human-centered robotics, emphasizes the study of
that facilitate these activities have been under de- humans as models for robots, and even the study of
velopment for many years. However, the focus of the robots as models for humans.
robotics community can still be said to be on the ro-
bot, with an emphasis on the technical challenges
of achieving intelligent control and mobility. It is only Current Challenges
in the early years of the twenty-first century that Roboticistsscientists who study roboticsare now
the state of the art has improved to such a degree that considering more carefully the work that has been go-
it is predicted that by 2010 there may be robots that ing on in the sister community of human-computer
Carbo-Powered Robots
TAMPA, Fla. (ANS)When modern technology was in His test gastrobota 3-foot-long, wheeled device
its infancy, scientists held out the hope that one day robots uses bacteria to break down the carbohydrate mole-
would cook our meals, do the housework and chauffeur cules in sugar cubes. The process releases electrons that
the children to school. That hope has yet to become real- are collected and turned into electrical current.
ity, but hold on: Here come the gastrobots. Any food high in carbohydrates could be used, the pro-
Powered by carbohydrates and bacteria, these robots fessor says, including vegetables, fruit, grains and foliage.
with gastric systems are taking the science to new dimen- Meat contains too much fat to be an efficient fuel, he
sions by mimicking not just the anatomy and intelligence pointed outso the family pets are safe. A gastrobot would
of humansbut our digestive processes as well. be far happier in an orange orchard, stabbing the fallen
Stuart Wilkinson, an associate professor of mechan- fruit and sucking the juice to propel itself.
ical engineering at the University of South Florida, is Measuring soil moisture and checking for insect infes-
pioneering the new subspecialty. tations, it could then relay its findings via a cell phone
The main thing Im shooting for is a robot that can connection to the farmers desktop computer.
perform some sort of task outdoors for long periods of In its infancy, the new generation of robots has a few
time without anybody having to mess with it, he said. kinks yet to be worked out. At present, his creation is a bit
Traditionally powered by regular or rechargeable bat- of a couch potato, Wilkinson admitted, and requires 18 hours
teries or solar panels, robots lose their efficiency when worth of carbo-loading to move for just 15 minutes.
placed at any distance from a power source or human Then theres the issue of, well, robot poop. We need
overseer. But when powered by foodsay, fruit fallen to develop some sort of kidney, he explained.
to the ground or grass on a lawnthey have the poten- Source: Carbo-powered robot holds promise of relief from drudgery.
American News Service, September 7, 2000
tial to eat and wander indefinitely.
interaction (HCI), which has been studying tech- uations or other situations in which safety is critical.
nology development and its impact on humans since What if the robot has qualities that make the hu-
the 1960s. However, collaboration between HCI man think that it is smarter than it really is? To take
researchers and robotics researchers is not as straight- another example, if the robot is to be used as an as-
forward as one might think. Until recently, much sistant to a disabled person or a senior citizen,
of the work in robotics has focused on integration would it be desirable to program the robot to act
of increasingly intelligent software on the more slowly like it has emotions, even if it doesnt really have any?
evolving hardware platforms. Individual robots with Would this make the users of the robots feel more
some humanoid qualities have been developed with comfortable and happy about using the technology?
amazing capabilities, but it has taken years of ex-
tensive work to produce them, and they are still
not advanced enough to accomplish real tasks in the Current Applications and Case Studies
real world. Human-robot interaction in these ex- Researchers are attempting to address these questions
amples is studied primarily to find out what can by taking their robots out of controlled laboratory en-
we learn from humans to improve robots. On the vironments and having them tackle real-world prob-
other hand, since the late 1990s, much of the HCI lems in realistic settings with real people as users. The
community has adopted an explicitly strong em- results are bringing us closer to a more human-cen-
phasis on human-centered computingthat is, on tered approach to human-robot interaction.
technology that serves human needs, as opposed
to technology that is developed for its own sake, and Urban Search and Rescue
whose purpose and function may ultimately oppose One application is the use of robots for urban search
or contravene human needs or wishes. and rescue (USAR). These are situations in which peo-
Because humans are still responsible for the out- ple are trapped or lost in man-made structures such
comes in human-machine systemsif something as collapsed buildings. For example, after the collapse
goes wrong, it is not the machine that will suffer of New York Citys Twin Towers as a result of the ter-
the consequences or be punishedit is important rorist attack of September 11, 2001, small teams of
that as robots become more independent, they are robots were fielded to give limited assistance to search
also taught how to become more compliant, com- and rescue operations. Because collapsed buildings and
municative, and cooperative so that they can be team associated rubble pose risks not only to the victims
players, rather than simply goal-oriented mechanisms. but also to the rescue workerssecondary collapses
Another challenge that faces researchers is how and toxic gases are constant dangers while the work-
much like a human to make the robot. Does the robots ers are engaged in the time-consuming and painstak-
physical form and personality affect how people re- ing tasks of shoring up entry points and clearing
spond to it? Does the context of the relationship spacesrobot aid is potentially very desirable.
play a role? Are the needs and desires of those who Small, relatively inexpensive, and possibly ex-
will interact with the robots different in the workplace pendable robots may be useful for gathering data
than they are in the home, for example, or different from otherwise inaccessible areas, for monitoring the
in dangerous situations than they are in safe ones, environment and structure while rescue workers are
or in interactions that occur close at hand as opposed inside, for helping detect victims in the rubble, and
to remotely? Interesting work by the sociologist eventually perhaps even for delivering preliminary
Clifford Nass at Stanford University shows that often medical aid to victims who are awaiting rescue. For
people will respond trustingly to technology and will the robots to work effectively, however, they must be
attribute qualities such as intelligence to technology capable of understanding and adapting to the orga-
based on very superficial cues, such as how friendly nizational and information rescue hierarchy. They
or unfriendly the messages generated by the technol- must be able to adapt to episodes of activity that may
ogy are. This has serious implications for the design be brief and intense or long term; they must be
of robots, especially those to be used in hazardous sit- equipped to help different levels of users who will
HUMAN-ROBOT INTERACTION 331
have differing information needs and time pressures. standing the robots orientation (it was cylindrical in
Most of the robots currently available for these kinds shape, with no clearly defined front), in communi-
of hazardous environments are not autonomous and cating spatial directions, and in understanding what
require constant supervision. the robot was doing due to lack of feedback.
The rescue workers will have to adapt as well. Further iterations improved the physical design and
They will need to have special training in order to the interface, and longer studies were conducted in
handle this technology. Currently robotics special- an actual office environment with physically impaired
ists, or handlers, are being trained in search and res- people, who were given the opportunity to use the
cue to supplement rescue teams. However, even robot during their work days to perform tasks such
the specialists are not entirely familiar with the kind as fetching coffee from the kitchen. One of the inter-
of data that the robots are sending back, and there- esting observations from these studies was the insight
fore understanding and interpreting that data in a that although the robot was the personal assistant of
time-critical situation poses additional challenges. one individual, it also affected other people. For ex-
Teams of researchers led by pioneers in the field, such ample, because the robot was not able to pour the cof-
as Robin Murphy of University of South Florida, are fee itself (it did not have any arms), it had to solicit help
now studying these kinds of problems and work from someone in the kitchen to actually get the cof-
on improving the methodologies so that the human- fee into the cup.Another example was that people pass-
robot interaction can be more smoothly integrated ing by in the hallway would greet the robot, although
into the response teams overall operation. from the robots perspective, they were obstacles if they
were in the way. These findings suggest that even if a
Personal Service Robots robot is designed for individual use, it may need to be
Personal service robots also offer many opportunities programmed to deal with a social context if it is to man-
for exploring human-robot interaction. Researchers age successfully in its working environment.
at the Royal Institute of Technology in Stockholm, Robots are working closely with humans in many
Sweden, have been working on the development of a other areas as well. Robotic technology augments space
robot to assist users with everyday tasks such as fetch- exploration in numerous ways, and in the military arena
ing and delivering objects in an office environment. robotic units are being considered for surveillance, sol-
This effort has been targeted at people with physical dier assistance, and possibly even soldier substitutes in
impairments who have difficulty doing these kinds of the future. Of perhaps greater concern are the areas
tasks themselves, and a goal of the project is to develop in which robots will interact with ordinary people, as
a robot that someone can learn to operate in a rela- it remains to be seen whether the robots will be pro-
tively short period of time. From the early stages of this grammed to adjust to human needs or the humans will
project, this group adopted user-centered techniques have to be trained to work with the robots. The robotic
for their design and development work, and, conse- design decisions that are made today will affect the na-
quently, have produced some very interesting results. ture of human-robot interaction tomorrow.
Since ordinary people have little or no experi-
ence in interacting with a robot, a general survey was Erika Rogers
conducted to determine what people would like
such a robot to do, how it should look, how they would See also Affective Computing; Literary Representa-
prefer to communicate with it, and generally how they tions; Search and Rescue
would respond to it. A large proportion of the re-
spondents were positive about having robotic help with
some kinds of basic household or other mundane tasks; FURTHER READING
the majority preferred the service robot not to act
Billings, C. E. (1997). Issues concerning human-centered intelligent
independently, and speech was the preferred mode systems: Whats human-centered and whats the problem?
of communication. Experiments with an early robot Retrieved July 21, 2003, from http://www.ifp.uiuc.edu/nsfhcs/
prototype showed that people had difficulty under- talks/billings.html
Center for Robot-Assisted Search and Rescue. (n.d.). Center for robot- form of buttons, underlined words and phrases, and
assisted search and rescue: CRASAR. Retrieved July 21, 2003, from other hot (interactive) areas on the screen.
http://www.crasar.org/
IEEE Robotics & Automation Society. (1995). Proceedings of the IEEE/
Hypertext is text that uses hyperlinks (often called
RSJ international conference on intelligent robots and systems: Human simply links) to present text and static graphics. Many
robot interaction and cooperative robots. Piscataway, NJ: IEEE websites are entirely or largely hypertexts. Hyper-
Robotics & Automation Society. media extends that idea to the presentation of video,
Interaction and Presentation Laboratory. (n.d.). Human-robot in-
teraction at IPLab. Retrieved July 21, 2003, from http://www.nada. animation, and audio, which are often referred to
kth.se/iplab/hri/ as dynamic or time-based content, or multimedia.
Lorek, L. (2001, April 30). March of the A.I. robots. Interactive Week, Non-Web forms of hypertext and hypermedia in-
8(17), 46. Retrieved August 29, 2003 from http://cma.zdnet.com/
texis/techinfobase/techinfobase/+bwh_qr+sWvKXX/zdisplay.html
clude CD-ROM and DVD encyclopedias (such as
Norman, D. (2001). How might humans interact with robots? Retrieved Microsofts Encarta), e-books, and the online help
July 21, 2003, from http://www.jnd.org/dn.mss/Humans_and_ systems we find in software products. It is common
Robots.html for people to use hypertext as a general term that in-
Rahimi, M., & Karwowski, W. (Eds.) (1992). Human-robot interac-
tion. London: Taylor & Francis.
cludes hypermedia. For example, when researchers
Ralston, A. & Reilly, E. D. (Eds.) (1993). Encyclopedia of computer talk about hypertext theory, they refer to theoreti-
science (3rd ed.). New York: Van Nostrand Reinhold. cal concepts that pertain to both static and multi-
Reeves, B., & Nass, C. (1996). The media equation: How people treat media content.
computers, television, and new media like real people and places.
Stanford, CA: CSLI Publications. Starting in the 1940s, an important body of
Rogers, E., & Murphy, M. (2001, September). Human-robot inter- theory and research has evolved, and many impor-
action: Final report of the DARPA/NSF Study on Human-Robot tant hypertext and hypermedia systems have been
Interaction. Retrieved July 21, 2003, from http://www.csc.calpoly.
edu/~erogers/HRI/HRI-report-final.html
built. The history of hypertext begins with two vision-
National Aeronautics and Space Administration (NASA). (n.d.). ary thinkers: Vannevar Bush and Ted Nelson. Bush,
Robotics. Retrieved July 21, 2003, from http://spacelink.nasa.gov/ writing in 1945, recognized the value of technologies
Instructional.Materials/Curriculum.Support/Technology/Robotics/ that would enable knowledge workers to link docu-
Fong, T., & Nourbakhsh, I. (2003, March). Socially interactive ro-
bots [Special issue]. Robotics and Autonomous Systems, 42.
ments and share them with others. Starting in the
Shneiderman, B. (1997). A grander goal: A thousand-fold increase in mid-1960s, Nelson spent decades trying to build a
human capabilities. Educom Review, 32(6), 410. Retrieved July very ambitious global hypertext system (Xanadu) and
21, 2003, from http://www.ifp.uiuc.edu/nabhcs/abstracts/shnei- as part of this effort produced a rich (though idio-
derman.html
Simsarian, K. (2000). Towards human-robot collaboration. Unpub- syncratic) body of theory.
lished doctoral dissertation, Swedish Institute of Computer Science,
Kista, Sweden. Retrieved July 21, 2003, from http://www.sics.se/
~kristian/thesis/
Takeda, H., Kobayashi, N., Matsubara, Y., & Nishida, T. (1997). Towards
Linear and Nonlinear Media
ubiquitous human-robot interaction. Retrieved July 21, 2003, from A linear communication medium is one we typically
http://ai-www.aist-nara.ac.jp/papers/takeda/html/ijcai97-ims.html experience straight through from beginning to end.
There is little or no choosing as we go. Cinema is a
linear medium. In the world of print, novels are lin-
ear, but newspapers, magazines, and encyclopedias
HYPERTEXT AND are somewhat nonlinear. They encourage a certain
amount of jumping around. The Web and other hy-
HYPERMEDIA pertextual media are strongly nonlinear. Indeed, the
essence of hypertext and hypermedia is choicethe
The terms hypertext and hypermedia refer to web- freedom to decide what we will experience next. You
pages and other kinds of on-screen content that em- can build a website in which the hyperlinks take
ploy hyperlinks. Hyperlinks give us choices when we the user on a single path from beginning to end, but
look for information, listen to music, purchase prod- this would be a strange website, and one can ques-
ucts, and engage in similar activities. They take the tion whether it is really hypertext.
HYPERTEXT AND HYPERMEDIA 333
Ted Nelson on Hypertext and the Web
I DONT BUY IN forced into hierarchical templates! And the semantic web
The Web isnt hypertext, its DECORATED DIRECTORIES! means that tekkie committees will decide the worlds true
What we have instead is the vacuous victory of typeset- concepts for once and for all. Enforcement is going to be
ters over authors, and the most trivial form of hypertext that another problem :) It is a very strange way of thinking,
could have been imagined. but all too many people are buying in because they think
The original hypertext project, Xanadu, has always been thats how it must be.
about pure document structures where authors and read- There is an alternative.
ers dont have to think about computerish structures of files Markup must not be embedded. Hierarchies and files
and hierarchical directories. The Xanadu project has endeav- must not be part of the mental structure of documents.
ored to implement a pure structure of links and facilitated Links must go both ways. All these fundamental errors of
re-use of content in any amounts and ways, allowing au- the Web must be repaired. But the geeks have tried to lock
thors to concentrate on what mattered. the door behind them to make nothing else possible.
Instead, todays nightmarish new world is controlled by We fight on.
webmasters, tekkies unlikely to understand the niceties of More later.
text issues and preoccupied with the Webs exploding alpha- Source: Ted Nelson Home Page. (n.d.) I dont buy in. Retrieved March 29,
2004, from http://ted.hyperland.com/buyin.txt
bet soup of embedded formats. XML is not an improvement
but a hierarchy hamburger. Everything, everything must be
Nodes, Links, and Navigation environment such as a city. In both hypertext navi-
Web designers and others who are interested in gation and physical navigation, we choose the most
hypertext often use the term node to refer to chunks promising route and keep track of where we go. If we
of content. Much of the time a node is simply a web- get lost, we may backtrack to familiar territory or even
page. But there are times when we want to envision return to our home base and start over. In the best
a cluster of closely related webpages as a single unit. case, we gain a mental picture of the overall structure
Also, there are times when one physical webpage re- of the environment (a birds eye or maplike view).
ally behaves like two or more separate chunks of con- At the same time, the concepts of nodes, links,
tent. Furthermore, the page is not the fundamental and navigation have limitations, and their relevance
unit of content in websites built with Flash (an an- and usefulness are being called into question as Web
imation technology from Macromedia and in many technologies become increasingly sophisticated. If
non-Web hypertext systems. Therefore, we do well clicking a link plays an audio sequence, is the au-
to use the term node as the fundamental unit of hyper- dio sequence then a node? Does it matter whether
text content. Links (or hyperlinks) are the pathways the audio sequence is a single word or a three-minute
between nodes. popular song? If clicking a link on a webpage begins
When we click links and thereby display a suc- a video sequence on a portion of that same page, how
cession of webpages (nodes), we are in a sense navi- do we describe what has happened? Is the video se-
gating the website. Navigation is only a metaphor; no quence a kind of subnode embedded within the node
one, of course, travels anywhere. Navigation, how- that is the page as a whole?
ever, is a very natural and useful metaphor because In early hypertext systems links were just simple
exploring a website (or a non-Web hypertext) is much electronic pathways with a fixed origin and destina-
like finding our way through a complex physical tion. But now if you visit an e-commerce website that
you have visited before, you may find an automati- users jump around more freely. For example, users
cally generated, personalized link inviting you to buy can move laterally along the sibling nodes of a single
a new book by an author whose books you have pur- branch and can jump from one branch to another,
chased in the past. Furthermore, this link may be without having to first move up to a higher-level node.
gone the next time you visit the site. Do we need to There is almost always a link from every node back
distinguish between links that everyone sees and links to the homepage (the top of the hierarchy) and there
that only appear under specific circumstances? are usually other kinds of upward links.
A limitation of the navigation paradigm is that it Especially when designing larger hypertexts, de-
does not correspond to the full range of user be- signers must choose between making the hierarchy
havior. At times users do not think spatially; they just wider (putting more nodes on each level) or deeper
click the most promising links they see. Designers, in (adding more levels). One well-established design prin-
fact, have begun employing a different metaphor for ciple is that users navigate a wide hierarchy (one in
Web usethe metaphor of the information which parent nodes have as many as thirty-two links
scent. The idea is that users, like animals foraging or to child nodes) more easily than a deep hierarchy.
hunting for food, look for strong and distinct A great many print documents are hierarchies in
scents that point them toward their desired goals. one significant respect: They are often divided hier-
Designers, therefore, should strive to create links that archically into parts, chapters, sections, and subsec-
give off these strong and unambiguous scents. tions. These divisions create a logical hierarchy
that the user encounters while reading linearly. Cross
references in print invite the reader to jump from
Information Structures one part of the document to another and so are anal-
Designers of websites and other hypertexts must ogous to links in hypertext.
work hard to decide which nodes will be linked to
which other nodes. Only with thoughtful linking will Weblike Structures
users be able to navigate successfully. Fortunately In a weblike structure, any node can be linked to any
there are well-known arrangements of nodes and other. There are no rulesalthough designers must
linksoften called information structuresthat take great care in deciding which links will be most
guide designers as they work. By far the most im- helpful to users. Relatively few weblike websites and
portant of these structures is the hierarchy. Also im- non-Web hypertexts are built. This is because many
portant are the weblike and the multipath structures. subject areas seem to break naturally into a hierar-
chical structure and because users are apt to have
The Hierarchical Structure trouble navigating unsystematic structures. Many
The hierarchy is by far the most important structure weblike hypertexts are short stories and other works
because it is the basis of almost all websites and most of fiction, in which artistic considerations may over-
other hypertexts as well. This is so because hierarchies ride the desire for efficient navigation. Mark
are orderly (making them easy to understand) and yet Bernstein, who is the founder and chief scientist at
provide ample navigational freedom. On a hierarchi- Eastgate, a hypertext development and publishing
cally organized website, users start at the homepage, company, questions the belief that weblike structures
descend the branch that most interests them from are necessarily hard to navigate. He has been a cham-
among a number of possible branches, and make pion of weblike and other unorthodox hypertext
further choices as the branch they have chosen di- structures for both fiction and nonfiction.
vides. At each level, the information on the nodes
becomes more specific. Branches may also converge. Chains and Multipath Structures
Hierarchical structures are supplemented by sec- As noted earlier, content linked as a linear sequence
ondary links that make them more flexible. The of nodesa simple chain structureprobably does
secondary links function mainly as shortcuts; they let not qualify as hypertext because the users choice is
HYPERTEXT AND HYPERMEDIA 335
highly restricted. Linear sequences, however, are reg- make greater use of voice commands and commands
ularly included within hierarchical websites, often issued by hand gestures.
taking the form of a tutorial, demo, or tour. These and other advancements will surely change
It is possible to build a sequence of nodes that hypertext and hypermedia. For example, websites
is in large part linear but offers various alternative may provide much improved site maps consisting of
pathways. This is the multipath structure. Often we a three-dimensional view of the site structure, per-
find multipath sections within hierarchical websites. haps using the metaphor of galaxies and solar sys-
For example, a corporate website might include a his- tems. The Web may well become more intelligent,
torical section with a page for each decade of the com- more able to generate personalized links that really
panys existence. Each of these pages has optional match our interests. The Web may also become more
digressions that allow the user to explore events socialwe may routinely click links that open up live
and issues of that decade. One may also find a mul- audio or video sessions with another person.
tipath structure in an instructional CD-ROM in which As a communications medium changes, theory
learners are offered different pathways through the must keep pace. Otherwise, it becomes increasingly
subject matter depending on their interests or mas- difficult to understand the medium and design suc-
tery of the material. cessfully for it. We will therefore need to extend the
hypertext concepts of nodes, links, and navigation and
augment them with new concepts as well.
Node-Link Diagrams, Sketches,
David K. Farkas
and the Design Process
Because node-link diagrams show the overall struc- See also Website Design
ture of a website, Web developers often create them
as part of the design process. Some Web authoring
tools create these diagrams automatically. Using both FURTHER READING
node-link diagrams and mock-ups of webpages, de-
signers can effectively plan out how the site as a whole Bernstein, M. (1991). Deeply intertwingled hypertext: The navigation
should be linked and how to design the linking of problem reconsidered. Technical Communication, 38 (1), 4147.
individual pages. Bolter, J. D. (1991). Writing space: The computer, hypertext, and the his-
tory of writing. Hillsdale, NJ: Lawrence Erlbaum Associates.
When webpages are well designed, the placement Bush, V. (1996). As we may think. Interactions, 3(2), 3546.
of the links on the page along with the phrasing of Farkas, D. K., & Farkas J. B. (2002). Principles of Web design. New
the links enables a user to grasp, at least in part, the over- York: Longman.
Hodges M. E., & Sasnett, R. M. (1993). Multimedia computing: Case
all site structure, the users current location, and whether studies from MIT Project Athena. Reading, MA: Addison Wesley.
he or she is moving down, across, or up in the hierar- Landow, G. P. (1997). Hypertext 2.0. Baltimore, MD: Johns Hopkins
chy. Many websites provide site maps for users.Although University Press.
site maps differ greatly in appearance and usefulness, Larson, K., & Czerwinski, M. (1998). Web page design: Implications
of structure, memory, and scent for information retrieval. In Pro-
they resemble node-link diagrams in that they provide ceedings of ACM CHI 98 Human Factors in Computing Systems
the user with a birds eye view of the site structure. (pp. 2532). Los Angeles, CA: ACM Press.
McKnight, C., Dillon A., & Richardson J. (1991). Hypertext in context.
Cambridge, UK: Cambridge University Press.
Nelson, T. H. (1992). Literary machines 93.1. Sausalito, CA: Mindful Press.
Future Developments Nielsen, J. (1994). Multimedia and hypertext: The Internet and beyond.
Computing and the Web will continue to evolve in Boston, MA: Academic Press.
a great many ways. Monitors may give way to near- Nyce, J. M., & Kahn, P. (Eds.). (1991). From Memex to hypertext:
Vannevar Bush and the minds machine. Boston: Academic Press.
eye displays, at least for mobile computing. Virtual Parunak, H. V. D. (1991). Ordering the information graph. In E. Berk
reality may become more widespread and may be & J. Devlin (Eds.), Hypertext/hypermedia handbook (pp. 299325).
routinely incorporated into the Web. We may New York: McGraw-Hill.
Pirolli, P., & Card, S. (1999). Information foraging. Psychological Review, Rosenfeld, L. & Morville, P. (2002). Information architecture for the
106(4), 643675. World Wide Web (2nd ed.). Sebastopol, CA: OReilly.
Powell, T. A. (2000). Web design: The complete reference. Berkeley, CA: Rouet, J., Levonen, J. J., Dillon, A., & Spiro, R. J. (Eds.). (1996). Hypertext
Osborne: McGraw-Hill. and cognition. Mahwah, NJ: Lawrence Erlbaum.
ICONS
IDENTITY AUTHENTICATION
IMPACTS
INFORMATION FILTERING
INFORMATION ORGANIZATION
I
INFORMATION OVERLOAD
INFORMATION RETRIEVAL
INFORMATION SPACES
INFORMATION THEORY
INSTRUCTION MANUALS
INTERNETWORLDWIDE DIFFUSION
INTERNET IN EVERYDAY LIFE
ITERATIVE DESIGN
become popular to the point that standards for them

ICONS have emerged within some operating systems, such
as Microsoft Windows and Apple Macintosh.
Icons are visual symbols of objects, actions, or ideas. One result of such standards is consistency.
In computer software, these visual symbols are used Consistency allows users to easily use the same features
to identify everything from a disk drive or file on a among different programs and lessens the time needed
computer to the print command in a word- to learn different programs. The graphical represen-
processing program. Icons are used in graphical user tation of an icon can range from the abstract to the
interfaces (GUIs), where a user selects them with a concrete. An abstract icon is a symbol that reflects a
pointer that is manipulated via a mouse, trackball, convention of what the represented object, action, or
or related device. idea actually is. Such a symbol is often of simple shape,
Icons are intended to facilitate the use of a GUI line, and color. A concrete icon contains a more de-
by all levels of computer users. For novice users icons tailed, more graphical representation of the object, ac-
represent a program or command that would other- tion, or idea, which makes interpretation easier.
wise need to be remembered and typed. For more The link between the user (and his or her in-
experienced users icons are easier to remember terpretation) and the object, action, or idea that
and are quicker to activate than a typed command. the icon represents is a prime example of human-
The use of icons in graphical user interfaces has computer interaction. Although icons are not the
337
sole answer to the question of how to make tech- ware programs are complex and contain many fea-
nology easier to use, icons provide a means of con- tures for users to access. In order to accommodate
ducting work in a variety of environments for a wide many features, users can sometimes customize the
range of computer users. size of icons displayed on the monitor screen.
Other software programs let users select which icons
are displayed on the screen or move icons around the
History screen (as a toolbar or in their own window). The
The notion of icons originated before the graphi- customization of icons can reach the file system as
cal user interface or even computers were in use. well. Whereas some systems use abstract icons for
Traditionally icons were religious images. How- documents, other systems use other options. Some
ever, the notion of icons has evolved. In 1935 the U.S. systems abandon generic representations of an im-
scientist and philosopher Charles Peirce defined an age file in favor of more detailed representations.
icon as a sign resembling an object. An icon has at- Some recent image editing programs, such as Adobe
tributes that resemble those of the object that it rep- Photoshop, can produce a miniature version of the
resents in reality. In even earlier references during image in the file rather than a generic image file icon
the early 1900s, philosophers likened an icon to a as the files icon. The rationale is to provide an im-
sign that resembles an object and contains its mediate recognition of an image files contents that
properties. An example of this resemblance per- the file name may not provide.
spective of an icon is a painting of a person. The As the World Wide Web has become popular with
painting resembles the person and is thus a repre- a wide range of people, people who otherwise would
sentation of the person. The notion that icons re- not create their own computer software have had the
semble objects in reality was popular for several years. opportunity to create personal webpages containing
When the graphical user interface became icons. The difference between the icons in software
commonplace for computers, the notion of icons programs and the icons on webpages is that rather
as representations was maintained. What has changed than activate a feature, icons on webpages will gen-
are the extent of the representations and their size. erally take the user to a different webpage. The abil-
Representations can range from abstract to photo- ity of people to design their own icons and webpages
realistic and appear even three dimensional. The size and the ability to visit other websites, representing
can vary from software program to software program, a variety of services, information, and diversions, have
but an icon is usually less than 2.54 centimeters square. widened the range of icons in use by a wide range
Icons were used in graphical user interfaces of of people.
early computers, such as the Xerox Star in 1973. In
1984 Apple released the Apple Macintosh, contain-
ing the first commercially successful operating sys- Advantages and Disadvantages
tem with a graphical user interface. As technical One advantage of using icons instead of text labels
innovations progressed, the Macintosh interface is that icons are smaller than the corresponding text
evolved, Microsoft Windows became successful, and description of many objects, actions, or ideas. When
graphical user interfaces became common in other many icons are displayed on a screen in order to
operating systems. Icons have been a mainstay allow a user to access a wide variety of features, pre-
throughout the evolution of the GUI. serving as much screen space as possible to maxi-
The visual aesthetics of icons evolved alongside mize the users workspace is essential. Icons are
the graphical capabilities of computers, and now a space-efficient reminders of the functions they rep-
range of styles of icons is used in software. For ex- resent. For example, the floppy disk icon repre-
ample, some software programs arrange icons in sents the save function in many software programs.
groups to form a toolbar, whereas others arrange them Although many computer users save their work to
on the screen more creatively. The complexity of pro- their hard drive (as opposed to a floppy disk drive),
grams has affected how icons are utilized. Some soft- the icon serves as a reminder that the floppy disk rep-
ICONS 339
resents the ability to save the current document to a Use an appropriate amount of visual detail and
storage device. color in the icon.
To minimize the frustration and time that users Select an appropriate image because the image
need to learn a computer program, many interface should be associated with the object, action, or
designers use a metaphor. Some interfaces have an idea that one wishes to represent.
underlying theme, such as a desktop. A main com-
However, using heuristics is not a guarantee that
ponent of a successful metaphor is that of carefully
an icon will be successful. Users can be uncertain
selected icons that apply to the metaphor. The use
about the use of an icon if a designer does not char-
of such icons helps users and by representing objects
acterize them appropriately.
or actions that are recognized as relevant to the over-
Many software companies, including Microsoft
all interface metaphor.
and Apple, have standards for their developers to use
However, the use of a metaphor can also have
when designing user interfaces for operating systems.
negative connotations for icons. In many metaphors
Such standards are also guides for third-party com-
icons are an approximation that falls apart in terms
panies to use when designing user interfaces for ap-
of its ability to represent the characteristic of an
plication software that will run under the operating
object or action that would exist in the world that the
systems. These standards are intended to provide users
metaphors represent. For example, consider the
a consistent look and feel for the applications and
metaphor of a desktop and the icon of a trashcan that
the operating system. However, third-party compa-
represents the delete function on the screen (the
nies do not always follow the guidelines fully. Instead,
desktop). The metaphor is ineffective because people
some companies, such as Adobe and Macromedia,
do not place their trashcan on top of their desk.
develop their own user interface look and feel for
Many expert computer users like to use keyboard
their software.
shortcuts that enable tasks to be completed without
taking their hands off of the keyboard. When such
keyboard shortcuts are not available, and expert users
must use a mouse to select icons, expert users can Problems and Possibilities
become frustrated with the interface. Productivity A computer user must be able to interpret the
and satisfaction decrease. An example is the use of a symbolic meaning of an icon in order to use it.
word processor by users who are so expert that the Whether abstract or concrete representations are
mouse (and thus icons) is never used. If keyboard used in icons, problems may arise when a user
shortcuts were removed from the word processor, does not have the same cultural knowledge of an
such users productivity would decrease and result icons designer or does not have the visual acuity
in a need to change how they work. Although this to interpret the information represented by an icon.
drawback to icons is not specific to icons themselves, For example, the icon for an U.S. telephone booth
it relates to the use of icons in an interface. will likely not have any meaning to someone from a
remote village in China. When an operating sys-
tem is intended to be released to multiple coun-
tries with different cultural references, developers
Guidelines and Standards must keep special considerations in mind. The con-
The design of icons is based on heuristics (aids in
ventions of one country may not be the same of an-
learning) and GUI standards. Heuristics require
other country in terms of the shapes or colors selected
no mathematical proof or modeling, which makes
for an abstract icon. A red X in the United States
them easy to use by developers of user interfaces.
may not mean stop or warning in other parts
Icon heuristics include:
of the world. Concrete icons have similar issues
Be simple and direct. because the object, color, or context used in one cul-
Be clear in terms of what object, action, or idea ture may not be appropriate in other cultures. For
one wants to represent in the icon. example, an icon showing men and women together
in an office is acceptable in many cultures, whereas sues of interpretation by diverse groups of users
such a graphical representation is not acceptable in remain, research will continue to support the use of
other cultures. icons in user interfaces.
Careful icon design is required for the inter-
national distribution of a software system. Advice Stephanie Ludi
from an expert in the culture of a target environment
can save a software company embarrassment and the See also Graphical User Interface
loss of business. Because many software programs
are released internationally, different icon sets are
developed for different countries.
Currently most icons are staticsymbols that FURTHER READING
do not move. However, with all of the advantages of
icons, some researchers see the potential to extend Apple Computer. Apple developer connectionIcons. (2003).
the information capacity of icons with animation. Retrieved January 4, 2004, from http://developer.apple.com/ue/
aqua/icons.html
Researchers have studied the use of animated icons Apple Computer. Apple human interface guidelines. (2003). Retrieved
in general use and in use by computer users who have January 7, 2004, from http://developer.apple.com/documentation/
impaired vision. In both cases animated icons have UserExperience/Conceptual/OSXHIGuidelines/index.html#//
demonstrated benefits in comparison to traditional, apple_ref/doc/uid/20000957
Baecker, R., & Small, I. (1990). Animation at the interface. In B. Laurel
static icons. For example, some visually impaired (Ed.), The art of human-computer interface design (pp. 251267).
computer users can recognize an animated version Reading, MA: Addison-Wesley.
of an icon at a smaller size than a static icon. Design- Bergman, M., & Paavola, S. (2001). The Commens dictionary of Peirces
terms. Retrieved December 19, 2003, from http://www.helsinki.fi/
ers can use this fact to maximize the amount of work- science/commens/terms/icon.html
space that visually impaired users can utilize when Caplin, S. (2001). Icon design: Graphic icons in computer interface de-
creating a document in a word-processing or e-mail sign. New York: Watson-Guptill Publications.
program. Cornell University Common Front Group. (n.d.). Concepts of user
interface design. Retrieved January 11, 2004, from http://cfg.cit
Traditional icons are visual. However, computer .cornell.edu/cfg/design/concepts.html
users who cannot discern visual symbols can discern Dix, A., Finlay, J., Abowd, G., & Beale, R. (1993). Human-computer
auditory signals. Thus, researchers have developed interaction. New York: Prentice Hall.
audio-based interfaces, including auditory icons that Gajendar, U. (2003). Learning to love the pixel: Exploring the craft
of icon design. Retrieved January 11, 2004, from http://www
use sounds from everyday objects and allow com- .boxesandarrows.com/archives/learning_to_love_the_pixel_
puter users to interact with a computer system with exploring_the_craft_of_icon_design.php
sound. For example, a computer user can drag a doc- Haber, R. (1970). How we remember what we see. Scientific American,
222, 104112.
ument icon across the screen and hear the sound Nielsen, J. (n.d.). Icon usability. Retrieved December 20, 2003, from
of pieces of paper being dragged across a table. http://www.useit.com/papers/sun/icons.html
Different sounds can accommodate different file sizes Sun Microsystems Incorporated. (2001). Java look and feel design guide-
or other attributes. Although the visual aspect of an lines. Retrieved January 4, 2004, from http://java.sun.com/products/
jlf/ed2/book/
icon is not represented in an auditory icon, the no- Bayley, A. (2000). KDE user interface guidelines. Retrieved January
tion of a symbolic (albeit auditory) representation 11, 2004, from http://developer.kde.org/documentation/
is consistent. design/ui/index.html
Icons have earned a prominent place in graph- Ludi, S., & Wagner, M. (2001). Re-inventing icons: Using animation
as cues in icons for the visually impaired. In M. J. Smith, G. Salvendy,
ical user interfaces by representing everything D. Harris, & R. J. Koubeck (Eds.), Proceedings of the Ninth
from warnings to software features to programs and International Conference on Human-Computer Interaction. New
files. Icons represent information in a small package, Orleans, LA: HCI International.
Microsoft Corporation. Microsoft Windows XPGuidelines for
whether that information is an object, action, or idea. applications. (2002). Retrieved December 20, 2003, from http://
Careful design, in terms of heuristics and standards, www.microsoft.com/whdc/hwdev/windowsxp/downloads/
can maximize the usefulness of icons. Although is- default.mspx
IDENTITY AUTHENTICATION 341
Peirce, C., Hartshorne, C., Weiss, P., & Burks, A. (Eds.). (1935). Collected be read easily by the host system. Hashed passwords
papers IVIII. Cambridge, MA: Harvard University Press. or encrypted passwords avoid these problems.
Preece, J., Rogers, Y., Sharp, H., Benyon, D., Holland, S., & Carey, T.
(1994). Human-computer interaction. Reading, MA: Addison-
Wesley. Hashed Passwords
Shneiderman, B. (1998). Designing the user interface (3rd ed.). Reading, Hash functions are used to produce hashed pass-
MA: Addison-Wesley. words. Hash functions take blocks of text as input
and produce output that is different from the input.
A good hash function is irreversible: It is impossible
to reconstruct the original data from the hashed data.
IDENTITY Also, if the hash function is good, then it will be
nearly impossible to build a data block that would
AUTHENTICATION produces the given hashed data. A hash function
must always produce the same output for the same
Authentication is the process of verifying that some- given inputit must not contain any anomaly that
one or something is what it claims to be. One can leads to randomness of output for the same input.
verify a persons identity on the basis of what he or Systems using hashed passwords follow this se-
she knows (passwords, personal information, a per- quence: The first time a person logs on to the sys-
sonal identification number, and so on), what he tem, the system accepts the password, applies a hash
or she possesses (an ID card, key, or smart card, for function to the password, and stores this hashed value
example), or biometrically, based on fingerprints, in the system. Next time, when the user logs on,
DNA, or other unique features. This article exam- the system requests the password and the hash func-
ines four sorts of authentication systems used in tion is applied to the data the user has entered. If the
human-computer interaction: password authen- resultant output matches the stored record, then it
tication, Kerberos, digital signatures, and biometric means user has entered the correct password and
authentication. hence is authenticated.
Encrypted Passwords
Password Authentication Encryption is a process used to scramble data to
Password authentication is the most common au- make it unreadable to all but the recipient. With en-
thentication technique. With password authentica- cryption, passwords are encrypted using some encryp-
tion, the user supplies his or her user name and a tion algorithm, and the encrypted text is stored in
secret word (something only the intended user the system. Even the host system cannot read the en-
knows) to prove his or her identity. The information crypted text. When user logs on by supplying his
submitted by the user is compared with the infor- or her password, it is encrypted using the same algo-
mation stored in the authentication system to vali- rithm, with the resultant output being checked
date the user. Passwords can be plain text, hashed against the stored encrypted password. If both en-
text, or encrypted text. crypted texts match, then the user is authenticated.
Plain-Text Passwords Problems with Password Authentication

Plain-text passwords are the simplest passwords. Using Although there are different ways in which passwords
plain-text passwords reduces administrative over- can be used, all password authentication systems suf-
head because no preprocessing is required. But plain- fer from certain disadvantages. First, passwords
text passwords have many disadvantages. On the can be guessed, since generally users have a tendency
Internet, it is easy for an intruder to sniff out such to pick familiar words as passwords. Password-
passwords and then use them to pass as the rightful breaking programs that use combinations of alpha-
user. Also, passwords that are stored in plain text can bet letters can crack word passwords quickly.
Second, password authentication requires two him- or herself with the KDC by giving unique iden-
types of data: a user name and a password. It can be- tifiers. On the users request the KDC selects a ses-
come difficult for user to remember the pairs for sion key (a password limited to a particular session
multiple systems. A user may forget either his or her of communication), generates something called a
user name or his or her password. When that hap- ticket (a unique data message containing the users
pens, users must create a new user name or a new identity and the time range for which the ticket is
password, or both. valid), combines the ticket with the session key, and
Third, because of the problem of remembering encrypts it with the users secret key. Since the users
multiple passwords, many people use a single pass- secret key is exposed in the ticket, the ticket is used
word for many systems. This means that once the to generate a second ticket from the TGS. This sec-
password is cracked, all the systems using same pass- ond ticket is encrypted by the session key of the TGS,
word are endangered. so there is no risk of users secret key being exposed.
Using this second ticket, the client is authenticated
Policies to Make the Password and a secure connection with the server is established.
Authentication System Stronger
Even though password authentication systems have Benefits and Limitations of Kerberos
inherent disadvantages, they can be used efficiently The Kerberos system has the advantage of prevent-
by following certain strategies. First, a policy can be ing plain-text passwords from being transmitted over
enforced that requires passwords that are harder to the network. User names and passwords are stored
break. Long passwords containing combination of centrally, so it is easier to manage that informa-
alphabet letters and numerals are generally harder to tion. Furthermore, since passwords are not stored
break. Making passwords case-sensitive (that is, hav- locally, even if one machine is compromised, there
ing a recognition system capable of distinguishing are no additional compromises.
between uppercase and lowercase letters) also However, Kerberos has limitations. It is not ef-
helps. Additionally, a single sign-on system, which fective against password-guessing attacks, and it
gives a user access to multiple systems with a single requires a trusted path over which to send the pass-
password, can eliminate the need to remember mul- words. If a hacker can sniff communication between
tiple passwords. Finally, users can be reminded that the user and the initial authentication program,
they must never reveal their passwords to others. the hacker may be able to impersonate the user.
Kerberos Digital Signature

Kerberos is an authentication protocol that was In any online transaction, the most important and
developed by MIT for authenticating a request for a critical consideration is who you are dealing with.
service in a computer network. It relies on secret-key Digital signatureselectronic signaturesare one
cryptography to securely identify users. There are method of confirming someones identity.
five basic entities involved in the Kerberos authen- Digital signatures are based on public-key
tication system: the user, the client (computer), which cryptography, a form of cryptography that uses two
acts on behalf of the user, the key distribution cen- keys, a public key and a private key, to ensure au-
ter (KDC), the ticket-granting service (TGS), and thentication. The private key is used to create the dig-
the server, which provides the requested service. ital signature, and the public key is used to verify it.
In the Kerberos system, users have secret keys The private key is known only by signer, who uses
(passwords) that are stored at the KDC. A user ini- it to sign the document; the public key, in contrast,
tiates the authentication procedure by requesting is freely available and many people can share it. Thus,
credentials (permission for access) from the KDC. a document signed by one person can be verified
To acquire these credentials the user first identifies by multiple receivers. Although the public key and
IDENTITY AUTHENTICATION 343
the private key are related to each other, it is com- Biometrics Authentication System
putationally impossible to derive the signers private Biometric authentication verifies the user by mea-
key from the public key. Use of digital signature in- suring certain physiological or behavioral character-
volves two processes: digital signature creation, which istics such as fingerprints or retinas. The measurable
the signer performs, and digital signature verifica- characteristics used in any biometric system are
tion, which the receiver performs. unique. Using biometrics for identity authentication
typically involves the following processes.
Digital Signature Creation
For a message, document, or any other information Registering the User in the System
that needs to be signed digitally, an extract of the This process is also called enrollment. In this step,
message or document is generated using a hash func- the users biometric characteristic is measured using
tion. The signer further encrypts this hashed mes- an input device. This step must be performed very
sage by using his or her private key, and the doubly carefully, since future authentications of the user de-
encrypted text becomes the digital signature. This pends on this sample.
digital signature is then attached to the message and
stored or transmitted along with message. Processing the Biometric Characteristic
The sample recorded by the input device is then
Digital Signature Verification processed and its features are extracted. Before ex-
To verify the signers digital signature, the recipient tracting the features, biometric samples can be
creates a new hash value of the original message is checked in order to ensure their quality. The num-
using the same hash function that was used to create ber of samples needed for processing depends on the
first hash value. Then, using the freely available pub- biometric characteristic that the system is using
lic key, this newly generated hash value is compared for authentication.
with the hash value attached to the message. By com-
paring these two hash values, the identity of the signer Storage
can be verified. Apart from identity authentica- The processed sample, called the master template, is
tion, digital signatures can also be used for mes- then stored in the database for the future use.
sage authentication. By comparing the hash values Biometric systems can be used both for identifi-
generated by the signer and receiver, the integrity of cation and for verification. When they are being used
the message can be checked. If the message is altered for identification, the processed biometric charac-
by an intruder or damaged while in transit, the hash teristic is compared with the entire set of master tem-
value generated by the receiver will not match with plates stored in the database. By this means, the sys-
the original hash value attached by the signer. tem ensures that same person is not trying to enroll
under two different names. When being used for
Advantages and Disadvantages of verification, the system compares the processed
Digital Signature biometric characteristic with the master template
Digital signatures have both advantages and disad- stored during enrollment. If it matches the master
vantages. On the plus side, digital signatures cannot template, then user is authenticated.
be copied. They ensure that data has not been tam-
pered with after it has been signed. And, since digi-
tal signatures are created and verified electronically, Types of Biometrics
they are safe from unauthorized influence. On the Among the physiological and behavioral charac-
minus side, digital signatures are costly. Users must teristics that are used by biometric systems are fin-
pay to obtain a digital signature, and recipients of gerprints (unique even for identical twins), face
digital signatures need special software to verify recognition, voice recognition (which has limita-
the signature. tions, as a persons voice can change with age and
different input devices can give different results), iris basic use is to prevent identity theft. Digital signa-
recognition, signature verification (which measures tures are tamper proof, but they are only useful for
such characteristics as the writers pressure and speed authenticating the sender (or content) of a docu-
when signing his or her name), and hand and finger ment. They cannot be used in situations in which
geometry. authentication is required to give access to confi-
dential data, for example. As for biometrics, because
Advantages and Disadvantages of they rely on peoples physical features, they are dif-
Biometric Authentication System ficult to use for authenticating participants in online
One advantage of biometric systems is that because transactions. Thus, every authentication system has
they recognize the authenticated user him- or herself some features but because of limitations cannot be
and not information that he or she has, they avoid used everywhere. Currently, then, people rely on a
the problem of lost, stolen, or forgotten passwords or combination of different authentication systems for
identification numbers or cards. They are also fast maximum security.
and easy to use, and generally do not cost much.
However, some biometric characteristics are sub- Ashutosh Deshpande and Parag Sewalkar
ject to change as a person ages and hence must be
updated periodically. Additionally, some biometric See also Privacy; Security
systems, such as signature verification or voice recog-
nition, must operate within a tolerance range since
it is very difficult to produce exactly the same sig-
nature or to speak in exactly the same tone of voice FURTHER READING
each time. Establishing the correct tolerance range
Bellovin, S. M., & Merritt, M. (1991). AT&T Bell Labs limitations of
can be difficult. When relying on a biometric system, the Kerberos authentication system. Retrieved February 17, 2004,
one must be sure that the system is not producing from http://swig.stanford.edu/pub/summaries/glomop/kerb_
too many false rejections. (False rejections should be limit.html
below 1 percent.) Depending on the biometric char- Brennen, V. A. A basic introduction to Kerberos. Retrieved February 17,
2004, from http://www.cryptnet.net/fdp/crypto/basic_intros/
acteristic being measured, some people may be ex- kerberos/
cluded. For example, mute people cannot be put How Stuff Works. (19982004). How do digital signatures work?
through voice recognition. Finally, although bio- Retrieved March 18, 2004, from http://www.howstuffworks.com/
metric systems are generally low cost, some input question571.htm
Jaspan, B. (1995). Kerberos users frequently asked questions 1.14.
devices may need regular maintenance, which in- Retrieved February 17, 2004, from http://www.faqs.org/faqs/
creases the cost. kerberos-faq/user/.
Kohl, J. T. (1991). The evolution of the Kerberos authentication ser-
vice. Proceedings of the Spring 1991 EurOpen Conference.
Retrieved February 17, 2004, from http://www.cmf.nrl.navy.mil/
Implications of CCS/people/kenh/kerberos-faq.html#whatis
MIT Kerberos. (n.d.). Kerberos: The network authentication protocol.
Using Authentication Systems Retrieved March 18, 2004, from http://web.mit.edu/kerberos/www/
Podio, F. L., & Dunn, J. S. Biometric authentication technology: From
Clearly, all authentication systems have advantages the movies to your desktop. Retrieved March 18, 2004, from http://
and disadvantages; there is no one authentication www.itl.nist.gov/div895/biometrics/Biometricsfromthemovies.pdf
system that is suitable for all situations. Password Smeraldi, F., & Bigun, J. (2002). Retinal vision applied to facial fea-
authentication is the cheapest authentication sys- tures detection and face authentication. Pattern Recognition Letters,
23(4), 463475.
tem, so if you want to authenticate identity but your Syverson, P., Y Cevesato, I. (2001). The logic of authentication pro-
budget is limited, password authentication is a good tocols. In R. Focardi & R. Gorrieri (Eds.), Foundations of security
choice. With Kerberos, it is assumed that you are us- analysis and design (pp. 63136). Heidelberg, Germany: Springer
Verlag
ing trusted hosts on non-trusted networks. The Treasury Board of Canada. (2001, 2003). PKI questions and answers
basic goal of Kerberos is to prevent plain-text pass- for beginners. Retrieved March 18, 2004, from http://www.cio-dpi
words from being sent across the network, so its .gc.ca/pki-icp/beginners/faq/faq_e.asp
IMPACTS 345
database matching more efficient). Table 1 shows

IMPACTS how basic research and system development, re-
mote and direct effects, and intentional and unin-
Often the effects of computerization on society are tentional effects interrelate. It is important to
remote and outside the control of the designer of remember, however, that while the table presents
any specific system. But design decisions in par- discrete categories, each pair actually represents
ticular systems may also have more direct social ef- a continuum, with the names of each category rep-
fects, and these decisions are ones over which resenting endpoints on the continuums. Effects
computer professionals may have influence. This will fall somewhere on the continuum.
article will first review some of the large-scale so- Research is basic to the extent that it is solving
cial effects of computing as well as some of the intellectual rather than practical problems, though
more direct social effects that computer systems clearly many projects have both theoretical- and
can have, along with some of the methodologies applied-research facets. An effect is direct when its
that are being proposed to help predict and ame- outcome depends little upon the presence or ab-
liorate these direct effects. sence of other factors (such as, for example, the
In addition to distinguishing between remote presence of a union, or regulations, or reuse of the
and direct impacts, we can also distinguish between system for a different purpose). An effect is inten-
intentional effects (efficiency, unemployment) and tional to the extent that the developers and client
unintentional effects (deskilling, component reuse foresee and desire that outcome and design the sys-
in other applications) of computing systemsboth tem to produce it.
of which may be either remote or direct. Even This categorization shows clearly that computer
basic research in computing, such as the devel- research produces some social effects that system de-
opment of systems of fuzzy logic, can have both signers are aware of and over which they have some
intentional effects (making medical diagnosis more control; it also produces effects that designers may
reliable) and unintentional ones (making computer not foresee and may not be able to control (in some
A U.S. Post Office first-day cover from 1973 notes the impact of microelectronics on the development
of advanced technology. Photo courtesy of Marcy Ross.
TABLE 1. Examples of Different Kinds of Social Impacts of Computing
Basic Research (example: development System Development (example: data

of fuzzy logic systems) monitoring in banking call center)
remote direct remote direct
Intentional Increased Utilization of Improved Increased
reliability of categorization integration of productivity in
medical diagnosis algorithms in customer data the call center
difficult areas across sales
and support
Unintentional Increased privacy Revisions of Increased job
violations by conceptions of turnover among
marketers human rationality workers. Decreased
making use of job satisfaction
data-matching among call
techniques center workers
cases they may share control with many other ac- across state lines in the United States to catch people
tors). Again, this is a continuum rather than a di- filing for welfare benefits in more than one juris-
chotomy. Creative designers can often find technical diction. The potential for privacy violation is clear,
solutions to address even remote effects of system but from the corporate point of view, data mining is
design, once they are aware of them. an effective way to discover important markets, or
to reveal fraud. Computer systems are also used to
monitor the workplace behavior of individuals, usu-
Remote Social Impacts of ally with an eye to reducing unwanted behavior or
to increasing productivity. The successand in some
Specic Types of Systems cases the legalityof such schemes depends crucially
Much work on the social impacts of computing con- on how the employees are involved in the planning
centrates on the specific effects of specific types of and implementation of the systems. Approaches to
systems, such as computer games, monitoring soft- legislation of these privacy issues differ radically
ware, safety-critical systems, or medical applications. around the world, with Europe being very system-
atic and the United States taking a more fragmented
Computer Games and Violence approach.
Extensive playing of computer games has been doc-
umented to lead to thoughts, feelings, and reactions Safety-Critical Systems
that are clear precursors of violent behavior. The link Much computer software is directly integrated into
between use of the games themselves and violent systems that run sensitive machinery in real time.
behavior on the part of players is reasonably well This might be medical machinery, for example, or it
substantiated, but not surprisingly is a matter of might be machinery in missile launchers. In every
controversy. case, the complexities of process, the range of pos-
sible input states, the close coupling of action
Data Mining, Work Monitoring, and Privacy links, and time-based interdependencies make it dif-
Privacy threats are posed by systems designed to ficult to verify the safety of these systems. Nancy
match data from databases without the knowledge Levesons 1995 Safeware: System Safety and Computers
of the users. These systems match data, for example, and Neil Storeys 1996 Safety-Critical Computer
IMPACTS 347
Therac-25: Safety Is a System Property
Normally, when a patient is scheduled to have radiation

therapy for cancer, he or she is scheduled for several ses- simple programming errors
sions over a few weeks and told to expect some minor inadequate safety engineering
skin discomfort from the treatment. The discomfort is poor human computer interaction design
described as being on the order of a mild sunburn over a lax culture of safety in the manufacturing or-
the treated area. However, a very abnormal thing hap- ganization

pened to one group of patients: They received severe inadequate reporting structure at the company level
radiation burns resulting in disability and, in three and as required by the U.S. government.
cases, death. As noted by Nancy Leveson and Clark S. Turner
The Therac-25 was a device that targeted electron or (1993, para. 2), the authors of the study that investigat-
X-ray beams on cancerous tissue to destroy it. Electron ed the effects of Therac-25: Our goal is to help others
beams were used to treat shallow tissue, while photon learn from this experience, not to criticize the equip-
beams could penetrate with minimal damage to treat ments manufacturer or anyone else. The mistakes that
deep tissue. Even though operators were told that there were made are not unique to this manufacturer but are,
were so many safety mechanisms that it was virtually unfortunately, fairly common in other safety-critical
impossible to overdose a patient, this is exactly what systems.
did occur in six documented cases (Leveson and Clark Chuck Huff
1993). Source: Leveson, N., & Turner, C. S. (1993). An inves-
tigation of the Therac-25 accidents. IEEE-Computer
These massive radiation overdoses were the result of 26(7), 1841. Retrieved March 19, 2004, from http://
courses.cs.vt.edu/~cs3604/lib/Therac_25/Therac_1
a convergence of many factors including: .html
Systems provide good overviews of the issues in but then released to the entire world. Websites can be
designing safe systems in this area. subject to denial-of-service attacks by actors across
the globe. The legal issues of how actors will be held
responsible for their actions in a global network are
Social Impacts Associated with the still being resolved and are embedded in larger cul-
tural issues, such as the importance attached to pri-
Networked Nature of Computing vacy or property, and attitudes toward censorship.
Networked computing has increased the potential
scale of effect of human action. Messages or programs
written on a single machine by a lone actor can now More General Social Issues
propagate to millions of machines across the globe in There are several more general social issues associ-
a few minutes. This increase in the scale of action is ated with computing that are not directly associated
available to anyone who can buy or rent access to with the reach of the global network.
the global network. One effect of this globalization of
action over computer networks is that action can rou- The Digital Divide
tinely cross legal jurisdictional boundaries. Pictures The simple issue of access to computing technol-
can be created and uploaded in a jurisdiction where ogy and to the benefits of the increasingly wired world
their content is legal, but then viewed from another has generated concern over what is called the digi-
jurisdiction, in another country, where they are ille- tal divide, or the lack of access to current technology,
gal. Computer viruses can be designed in jurisdictions connectivity, and training that people may face based
where it is more difficult to prosecute the designers, on gender, race, class, or national wealth.
tem and out of the immediate control of the design-

Lo! Men have become the tools of their tools. ers or implementers of the systems. But all systems
Henry David Thoreau are implemented within some social context. If
they are successful, they will at least have the direct,
intentional effects desired by their designers and im-
Intellectual Property plementers. However, they are also likely to have
Computing technology is based on movement and indirect and unintentional effects. The software en-
transformation of information from one form to an- gineer Donald Gotterbarn and computer ethicist
other. The ease with which information can be trans- Simon Rogerson make the case in The Ethics of
ferred and software copied has made intellectual Software Project Management (2001) that using sys-
property theft a major issue. The chapter on intel- tematic methods to become aware of these effects will
lectual property in Deborah Johnsons Computer improve the quality of the delivered software, help
Ethics (third edition published in 2001) provides a avert expensive software project failures, and in the
good overview of the issues in this area. end save both developers and clients money. Since
the methods they and others advocate are relatively
Employment new and untried, there is little empirical evidence for
Finally, one of the direct, intentional goals of much these claims, but the logic of the case studies they
software development is to make businesses and other present is compelling. Other scholars also provide
organizations more efficient. But along with increases case study evidence that these approaches help to pro-
in efficiency come job redundancies. There seems duce better-designed products and may save clients
little evidence that technological advancement money, but the methods are not widely enough used
produces unemployment over the long run, but yet to make systematic evaluation possible. Still, it
individual implementations often produce unem- is worth reviewing these proposed methods.
ployment as a direct intentional outcome. The soci- What all the methods have in common is making
etal problem is to help resolve the structural issues sure that early design decisions take into account larger
that make it difficult for these workers to be reem- contextual issues rather than merely narrowly focused
ployed in comparable and meaningful work. technical and economic ones. The scientist Ben
Similarly, deskilling of workers (simplifying their Shneidermans original proposal (published in 1990)
jobs so management has more control or can pay lower was based on an analogy with the government re-
rates) is often the direct, intentional outcome of soft- quirement in the United States that construction proj-
ware development in business. Computer scien- ects complete an environmental impact statement
tists in Scandinavia have worked in cooperation with to determine the projects effect on the surrounding
labor unions and employers to develop software that environment. He proposed a social impact statement
adds to the skills and autonomy of workers while be incorporated into the software development process
also increasing the quality of the producta process of projects that were large enough to warrant it.
called participatory design. It is unclear whether this This statement would begin by listing the stake-
model will be more widely adopted in countries holders (people and groups likely to be effected by the
where unions have less power and legitimacy. software), would attempt to determine the systems ef-
fect on those stakeholders, and then propose modifi-
cations in the requirements or design of the system to
reduce negative and enhance positive effects. Ben
Addressing the Social Impacts of Shneiderman and Anne Rose reported on an attempt
Particular Implementations of to implement this approach in 1996. The psycholo-
gist Chuck Huff has proposed the integration of par-
Computing ticular social-science methods into the social
Many of the social issues reviewed so far are often re- impact statement as a way to increase its sensitivity to
mote from the implementation of any particular sys- the social context in which a system is designed.
IMPACTS 349
The social scientist Enid Mumford developed a called value-sensitive design that stresses the im-
method that was inspired by her early association portance of iterations of conceptual, technical, and
with the Tavistock Institute in London, a center empirical tasks throughout the software design proj-
founded in 1946 for the study of the interplay be- ect in an attempt to account for important human
tween technical systems and human welfare. Her values that are affected by the system. One starts with
method has been applied reasonably successfully philosophically informed analysis of relevant values.
to other technical systems and has seen a few suc- One then identifies how existing and potential tech-
cessful implementations in software systems as well, nical designs might enhance those values, and then
making it the best substantiated approach currently uses social-science methods to investigate how those
available. She gives seven steps for a sociotechnical values affect various stakeholders related to the
design process: system. Like most other approaches, this methodol-
ogy is a new development, and though it has had
1. Diagnose user needs and problems, focusing
some interesting successes, it still awaits more care-
on both short- and long-term efficiency and
ful validation.
job satisfaction;
Two other approaches deserve mention as meth-
2. Set efficiency and job satisfaction objectives;
ods to address social issues in software design:
3. Develop a number of alternative designs and
computer-supported cooperative work (CSCW) and
match them against the objectives;
participatory design. Participatory design was men-
4. Choose the strategy that best achieves both sets
tioned earlier in the context of the employment ef-
of objectives;
fects of system implementation. It is an approach
5. Choose hardware and software and design
that attempts to integrate democratic values into sys-
the system in detail;
tem design by involving potential users of the system
6. Implement the system; and
in intensive consultation during an iterative design
7. Evaluating the system once it is operational.
process. Computer-supported cooperative work is
Gotterbarn provides a method and supporting an area in which systems are designed to support the
software for investigating the social issues associated work of groups. The work in this area focuses on
with the implementation of a system. His approach those values that are most relevant in CSCW sys-
is as simple as asking what the effects of each of the tems, such as privacy and trust.
systems tasks will be on each of the stakeholders who The importance of social issues in computing
are relevant to that task. But this simplicity is illusory, was recognized by the early innovators in the field
because of the explosion of combinations that occurs of computing. The 1946 founding of the Tavistock
when you cross all tasks with all stakeholders and ask Institute has already been mentioned, and as early
a series of questions about each of these combina- as 1950, the mathematician Norbert Wiener ad-
tions. The SoDIS (software development impact state- dressed many of the issues considered in this arti-
ment) software Gotterbarn outlined in a 2002 article cle in his The Human Use of Human Beings. Social-
helps to control this complexity. To complete a impact issues in computing then lay fallow for many
software development impact statement, one must: years until revived in 1968 by Don Parker, who was

followed in 1976 by Joseph Wiezenbaum. There was
Identify immediate and extended stakeholders;
Identify requirements, tasks, or work breakdown
an explosion of work in the 1980s and 1990s con-
comitant with the proliferation of personal com-
packages;
Record potential issues for every stakeholder re-
puters and the emergence of the Internet as tool
for widespread use. At present numerous rigorous
lated to every task; and
Record the details and solutions to help mod-
and systematic methods are emerging to take account
of social-impact issues, ethics, and values in the
ify the development plan.
design of software systems.
The researchers Batya Friedman, Daniel Howe,
and Edward Felton are proponents of a design process Chuck Huff
See also Computer-Supported Cooperative Work; Friedman, B., & Nissenbaum, H. (1996). Bias in computer systems.
Digital Divide; Value Sensitive Design ACM Transactions on Information Systems, 14(3), 330347.
Gotterbarn, D. (2002). Reducing software failures: Addressing the eth-
ical risks of the software development lifecycle. Australian Journal
of Information Systems, 9(2). Retrieved December 10, 2003, from
FURTHER READING http://www.inter-it.com/articles.asp?id=195
Gotterbarn, D. & Rogerson, S. (2001). The ethics of software project
management. In G. Colllste (Ed.), Ethics in the age of information
Adam, A. (2001). Gender and computer ethics. In R. A. Spinello & technology (pp. 278300). Linkping Sweden: Linkping University
H. T. Tavani (Eds.), Readings in cyberethics, (pp. 6376). Press.
Sudbury, MA: Jones and Bartlett. Greenbaum, J., & Kyng, M. (Eds.) (1991). Design at work: Cooperative
Anderson, C. A., & Bushman, B. (2001). Effects of violent video games design of computer systems. Hillsdale, NJ: Lawrence Erlbaum.
on aggressive behavior, aggressive cognition, aggressive affect, phys- Huff, C. W. (1996). Practical guidance for teaching the social impact
iological arousal, and pro-social behavior: A meta-analytic review statement. In C. W. Huff (Ed.), Proceedings of the 1996 Symposium
of the scientific literature. Psychological Science, 12, 353359. on Computers and the Quality of Life (pp. 8690). New York: ACM
Anderson, C. A., & Dill, K. E. (2000). Video games and aggressive Press.
thoughts, feelings, and behavior in the laboratory and in life. Journal Huff, C. W., & Martin, C. D. (1995, December). Computing conse-
of Personality & Social Psychology, 78, 772790. quences: A framework for teaching ethical computing. Commu-
Baase, S. (1997). A gift of fire: Social, legal and ethical issues in com- nications of the Association for Computing Machinery, 38(12), 7584.
puting. Upper Saddle River, NJ: Prentice Hall. Johnson, D. (2001). Computer ethics (3rd. ed.). New York: Prentice-
Betts, M. (1994, April 18). Computer matching nabs double dip- Hall.
pers. Computerworld (p. 90). Johnson, R. C. (1997). Science, technology, and black community de-
Bjerknes, G., Ehn, P., & Kyng, M. (Eds.). (1987). Computers and democ- velopment. In A. H. Teich (Ed.), Technology and the future (7th ed.;
racy: A Scandinavian challenge. Aldershot, UK: Avebury. pp. 270283). New York: St. Martins Press.
Bdker, S., Ehn, P., Sjgren, D., & Sundblad, Y. (2000). Co-operative Kling, R. (Ed.). (1996). Computerization and controversy: Value con-
design: Perspectives on 20 years with the Scandinavian IT design flicts and social choices (2nd ed.). San Diego, CA: Academic Press.
mode (Report No. CID-104). Retrieved June 16, 2003, from Kraut, R., Kiesler, S., Boneva, B., Cummings, J., Helgeson, V., &
http://cid.nada.kth.se/pdf/cid_104.pdf Crawford, A. (2002). Internet Paradox Revisited. Journal of
Braman, S. (2004). Change of state: An introduction to information Social Issues, 58(1), 4974.
policy. Cambridge, MA: MIT Press. Kretchmer, S., & Carveth, R. (2001). The color of the Net: African
Brown R. (1967). Review of research and consultancy in industrial Americans, race, and cyberspace. Computers and Society, 31(3),
enterprises: A review of the contribution of the Tavistock Institute 914.
of Human Relations to the development of industrial sociology. Kusserow, R. P. (1984). The government needs computer matching to
Sociology, 1, 3360. root out waste and fraud. Communications of the ACM, 27(6),
Bynum, T. (2000). Ethics and the information revolution. In G. Colllste 542545.
(Ed.) Ethics in the age of information technology (pp. 3255). Kyng, M., & Mathiassen, L. (Eds.) (1997). Computers and design in
Linkping Sweden: Linkping University Press. context. Cambridge, MA: MIT Press.
Camp, T. (1997). The incredible shrinking pipeline. Communications Leveson, N. (1995). Safeware: System safety and computers. Reading,
of the ACM, 40(2), 103110. MA: Addison-Wesley.
Colin, B. J. (1992). Regulating privacy: Data protection and public policy Leveson, N., & Turner, C. S. (1993). An investigation of the Therac-
in Europe and the United States. Ithaca, NY: Cornell University Press. 25 accidents. IEEE-Computer 26(7), 1841.
Collins, W. R., Miller, K., Spielman, B., & Wherry, P. (1994). How good Lewis, S. G., & Samoff, J. (Eds.). (1992). Microcomputers in African de-
is good enough? An ethical analysis of software construction velopment: Critical perspectives. Boulder, CO: Westview Press.
and use. Communications of the ACM, 37(1), 8191. Miller, M. (1992, August 9). Patients records are a treasure trove for
Compaine, B. (2001). The digital divide: Facing a crisis or creating a a budding industry. Wall Street Journal (p. A21).
myth. Cambridge, MA: MIT Press. Mumford, E. (1996). Systems design: Ethical tools for ethical change.
Douthitt, E. A., & Aiello, J. R. (2001). The role of participation and London: MacMillan.
control in effects of computer monitoring on fairness perceptions, Mumford, E., & MacDonald, B. (1989). XSELs progress. New York:
task satisfaction, and performance. Journal of Applied Psychology, Wiley.
86(5), 867874. Parker, D. (1968). Rules of ethics in information processing. Commu-
Friedman, B. (Ed.). (1997). Human values and the design of com- nications of the ACM, 11(3), 198201.
puter technology. London: Cambridge University Press. Rocco, E. (1998). Trust breaks down in electronic contexts but can be
Friedman, B., Howe, D. C., & Felten, E. (2002). Informed consent in repaired by some initial face-to-face contact. In Proceedings of CHI
the Mozilla browser: Implementing value-sensitive design. In 1998 (pp. 496502). New York: ACM Press.
Proceedings of the 35th Annual Hawaii International Conference Rosenberg, R. S. (1997). The social impact of computers (2nd ed.). San
on System Sciences [CD-ROM OSPE101]. Los Alamitos, CA: IEEE Diego, CA: Academic Press.
Computer Society: Shade, L. (1996). Is there free speech on the Net? Censorship in the
Friedman, B., & Kahn, P. H. (in press). A value sensitive design ap- global information infrastructure. In R. Shields (Ed.), Cultures of
proach to augmented reality. In W. Mackay (Ed.), Design of aug- the Internet: Virtual spaces real histories, living bodies. Thousand
mented reality environments, Cambridge MA: MIT Press. Oaks, CA: Sage.
INFORMATION FILTERING 351
Shattuck, J. (1984). Computer matching is a serious threat to indi-

vidual rights. Communications of the ACM, 27(6), 538541). HELP SYSTEMS Computer-based systems that answer
Shneiderman, B. (1990). Human values and the future of technology:
A declaration of empowerment. Computers & Society, 20(3),
user questions on a programs functions and help users
16. lear n how to accomplish g iven tasks using the
Shneiderman, B., & Rose, A. (1996). Social impact statements: Engaging program.
public participation in information technology design. In C. Huff
(Ed.), Computers and the quality of life: The proceedings of the
Symposium on Computers and the Quality of Life (pp. 9096). New
York: ACM Press. tering suggests leaving certain information out,
Sproull, L., & Faraj, S. (1995). Atheism, sex, and databases: The Net as but the term has been used in different contexts. One
a social technology. In B. Kahin & J. Keller (Eds.), Public access to
the Internet (pp. 6281). Cambridge, MA: MIT Press.
context emphasizes blocking, which refers to auto-
Storey, N. (1996). Safety-critical computer systems. New York: Addison- matic censorship of the material available to a par-
Wesley. ticular user or kind of user. Another definition
Swinyard, W. R., Rinne, H., & Keng Kau, A. (1990). The morality of emphasizes selecting things from a larger set of pos-
software piracy: A cross cultural analysis. Journal of Business Ethics,
9, 655664.
sibilities and presenting them in a prioritized or-
Tavani, H. (2003). Ethics and technology: Ethical issues in an age of der based on a users interests.
information and communication technology. Hoboken, NJ: John A conceptual framework for the design of IF sys-
Wiley & Sons. tems comes from two established lines of research:
Trist E. (1981). The evolution of socio-technical systems. Ontario, Canada:
Ontario Quality of Working Life Center. information retrieval and user modeling. Information
U.S. Congress Office of Technology Assessment. (1986). Technology retrieval attempts to retrieve as many relevant items
and structural unemployment: Reemploying displaced adults (OTA- as possible while minimizing the amount of irrele-
ITE-250). Washington, DC: U.S. Government Printing Office.
Wiener, N. (1950). The human use of human beings. New York: Da
vant information. A relevant item is one that helps
Capo Press. satisfy an information need. Because the most dis-
Weizenbaum, J. (1976). Computer power and human reason: From judg- tinctive characteristic of the field of IF is the exis-
ment to calculation. New York: Freeman. tence of a profile to model the information need,
Wresch, W. (1998). Disconnected: Haves and have-nots in the infor-
mation age. In K. Schellenberg (Ed.), Computers in society (7th ed.;
most of the work on IF is being conducted within
pp. 207212). Guilford, CT: Dushkin/McGraw Hill. the framework of user modeling. A user model con-
Yakel, E. (in press). The social construction of accountability: tains pertinent information about the user; the IF
Radiologists and their recordkeeping practices. Information Society. system uses such models to guide its personalization
of information for the user.
Filtering mechanisms depend on accurate rep-
resentations of their users and theirs users needs
INFORMATION in order to perform their task successfully. Charac-
teristics of user models can vary according to the sys-
FILTERING tem, the user, and the task being performed. The
information scientist P. J. Danielss profiles include
The amount of information we create and exchange user background, habits, topics of interests, and user
is far more than a person can easily manage. One way status. Robert B. Allen suggests that the model should
to overcome information overload is to personalize also include situation, task, and environmental in-
information delivery, that is, to present information formation. For user profiles of potential customers
that matches users preferences. Selective dissemi- of electronic shops, the scholars Liliana Ardissono
nation of information, alerting services, collabora- and Anna Goy suggest domain expertise, lifestyle,
tive filtering, help systems, social filtering, social and intended use of the items purchased as perti-
data-mining systems, and user-adaptive systems can nent. Based on these elements, an IF system can
collectively be called information-filtering (IF) determine which product to recommend, how to
systems. Personalized information systems are based present product descriptions, how much technical
on the information preferences, profiles, or user mod- information to include, and which linguistic form
els that these IF systems produce. Information fil- to employ.
Filtering systems frequently present material to be gleaned an initial model of their background
users based on how well it corresponds with a list of knowledge. The customized hypertext of a user who
topics or subjects that the user has selected as be- claims to have a solid understanding of a concept
ing of interest. In 1997 Raya Fidel and Michael will have the concept relegated to a glossary, while
Crandall, scholars in the field of library and infor- the users who dont know the concept will find it
mation science, published results of an empirical as a live link providing a definition in the main body
study evaluating users perception of filtering per- of the hypertext. In addition, a combination of cog-
formance. In the study, they examined criteria users nitive and biological criteria such as eye movements,
employed in judging whether or not a document was pulse, temperature, and galvanic skin response,
relevant. Although the profiles in the users filter- can be used for individual optimization. Dara Lee
ing system were based on topics, the researchers Howard and Martha Crosby performed experiments
found that users employed many criteria in addition to understand individual viewing strategies for bib-
to topics or subjects (for example, their current be- liographic citations. Analysis of the users eye move-
liefs, knowledge, and working situation) to deter- ments illuminated their actual behavior. The
mine a documents relevance. researchers found that relevant material was read se-
quentially while non-relevant material was viewed
non-sequentially.
User Models for Filtering
Although researchers differ in how they categorize Explicit Modeling
models, most agree that a model can be either canon- Explicit models are built directly and explicitly by
ical or individual in its modeling and that its ac- users. Users are presented with direct questions, and
quisition of information can be either explicit or their answers comprise their profile. Advantages of
implicit. the explicit model are that users are more in control
and have a good understanding of what the system
Canonical Modeling is doing. Thomas Malone and his colleagues found
Canonical, or stereotypical, modeling models a typ- that the rules that constitute the profile of InfoLens,
ical user, while individual modeling models a a filtering system that helps the user to filter incom-
particular user. Because stereotypes are collection ing email messages. Users write the rules instructing
of attributes that co-occur in users, inferences about the email system on the actions to take depending
a user that rely on stereotypical modeling can be on the sender, message type, and date. Daniela Petrelli
made on a smaller sample of user behavior. The in- and her colleagues categorized visitors to muse-
heritance properties of some representation meth- ums according to classical dimensions, provided
ods allow the system to infer data and predict by each visitor, such as age, profession, education,
actions. When the user is new, the system relies and specific knowledge or background. Visitors also
on the stereotype; as the system learns, less data is provided situational dimensions such as available
taken from the stereotype and more from the up- time for the visit, motivation for the visit. Language
dated profile of the actual user. Stereotype mod- style (expert versus naive) and verbosity used in the
els have their limitations: It is difficult to achieve individualized guide was constructed from the
a sufficiently personal expression with a stereotype user profile.
model.
Implicit Modeling
Individual User Modeling Implicit models are inferred by the responses of
Individual models are exemplified by the Personalized the user assigning relevance to the information they
Text (PT) system, which aims to customize hyper- are presented or by monitoring users actions and
text. In PT, readers are asked to do a concept in- behavior. The usefulness of feedback for adjusting a
ventory, which is a short dialogue from which can profile has been long recognized. Luz Quiroga and
INFORMATION FILTERING 353
Javed Mostafa conducted experiments to compare Interactive Information Retrieval

the performance of explicit and implicit modes of The information scientist Nicholas Belkin developed
profile acquisition. Results of the study suggested the theory of the anomalous state of knowledge
that context plays an important part in assess- (ASK). The theory suggests that it is better to em-
ments of relevance. Implicit feedback mechanisms, ploy interactions than to model explicit knowledge.
as well as the explicit ones, based on surveys and Belkin points out that people engage in information
questionnaires, are criticized for imposing an extra seeking to achieve some goal for which their current
burden on the user. Monitoring user behavior, on status of knowledge is inadequate. Thus, the users
the other hand, appears to be an efficient way to cap- may not know, let alone be able to specify, what their
ture the users profile transparently. Other implicit information needs are.
profiling systems examine actions such as whether
an electronic message is read or ignored, saved or Intelligent Help Systems
deleted, and replied to or not. User interest in vari- The computer scientist Johan Aberg designed a sys-
ous topics can also be ascertained based on the time tem that combines automatic and human help in a
users spent reading NetNews articles. personalized and contextualized way. He describes
his system as a live help system that integrates hu-
man experts in the processes of advice-giving by
allowing users to communicate with dedicated ex-
Information Filtering and pert assistants through the help system (Aberg 2002,
4). Human experts thus complement computer-based
User Modeling Techniques and Tools help but do not replace it. A disadvantage of this sys-
Among the techniques and tools that are useful for tem is that it is difficult to provide human aug-
information filtering are virtual agents, collective mentation to the automatic help due to cost and
profiles, interactive information retrieval, intelligent availability of the human expert.
help systems, collaborative filtering, mining social
data, and bookmarking. Collaborative Filtering
With collaborative filtering, the system gives sug-
Virtual Agents gestions on various topics based on information
Virtual agents are software programs that act on be- gleaned from members of a community or peer
half of the user, delivering selected, prioritized in- group. The system identifies users with similar pref-
formation to an IF system. Pattie Maes (1994), of erences and makes recommendations based on what
MITs Media Lab, considers IF to be one of the many those others have liked. MovieLens is a home video
applications facilitated by agents, which act as per- recommendation system that chooses movies from
sonal assistants to their clients, alleviating poten- a large and relatively stable database of movie titles,
tial information overload. based on peers expressed taste in films. This approach
differs from approaches that choose items based only
Collective Profiles on what a user has chosen before.
Collective profiles allow individual users to organ-
ize information that relates to their families and Social Data Mining
friends. SETA, a shopping recommender, maintains The process of mining social data reduces the cogni-
models of all the beneficiaries for whom the user tive burden that collaborative filtering places on the
selects goods while shopping. Research on person- user. Vannevar Bush (1945) suggested that the rich
alized guides to museums suggested that such sys- trails followed by scholars through information repos-
tems should support family profiles and family itories could guide others with similar interests.
discussions and interaction in addition to main- Statistics on how much time people spend reading var-
taining individualized museum guide capabilities. ious parts of a document, counts of spreadsheet cell
recalculations, and menu selections are sources for so- but more empirical evaluations are needed if we are
cial data mining. Another approach exploits the struc- to ever understand how humans find, organize, re-
ture of the World Wide Web itself. This approach has member, and use information.
its roots in bibliometrics, which studies patterns of co-
citationsthe citation of pairs of documents in a third Luz M. Quiroga and Martha E. Crosby
document. A link from one website to another may in-
dicate that the two sites are similar. Bell Labs Brian See also Information Overload; Information Retrieval;
Amento and his colleagues used their system TopicShop User Modeling
to investigate how well link-based metrics correlate
with human judgments. Features collected by
TopicShop could predict which websites were of high- FURTHER READING
est quality, which made it possible for users to select
better sites more quickly and with less effort. A vari- Aberg, J. (2002). Live help systems: An approach to intelligent help for
ant approach, called social navigation, has focused on Web information systems (Linkopings Studies in Science and
extracting information from web usage logs, record- Technology, Dissertation No. 745). Linkopings, Sweden:
Department of Computer and Information Science, Linkopings
ing browsing history, finding commonly traversed links. University.
Peter Pirolli, a researcher at Xeroxs Palo Alto Research Abrams, D., Baecker, R., & Chignell, M. (1998). Information archiv-
Center, and his colleagues researched data mining in ing with bookmarks: Personal Web space construction and or-
conjunction with organization theories. Their infor- ganization. Proceedings of the SIGCHI Conference on Human Factors
in Computing Systems, 4148.
mation foraging theory attempts to understand how Allen, R. B. (1990). User models: Theory, method, and practice.
people allocate their resources to find information. International Journal of Man-Machine Studies, 32, 511543.
It has been applied to information-seeking tasks, pre- Amento, B., Terveen, L., & Hill, W. (2003) Experiments in social
data mining: The TopicShop system. ACM Transactions on
senting users with a clustered navigable overview of Computer-Human Interaction, 10(1), 5485.
the content of a document collection. Ardissono, L., & Goy, A. (1999). Tailoring the interaction with users
in electronic shops. In J. Kay (Ed.), Proceedings of the 7th
Bookmarking International Conference on User Modeling (pp. 3544). New York:
Springer-Verlag.
David Abrams and his colleagues envision book- Belkin, N. (2000). Helping people find what they dont know.
marking as a common strategy for dealing with in- Communications of the ACM, 43(8), 5861.
formation overload. They surveyed 322 web users Belkin, N., & Croft, B. (1992). Information filtering and informa-
and analyzed the bookmark archives of 50 users to tion retrieval: Two sides of the same coin? Communications of
the ACM, 35(12), 2938.
learn why people make a bookmark and how they Bush, V. (1945, July). As we may think. Atlantic Monthly, 176(1),
organized and use them. One of their recommenda- 101108.
tions is to use automated filters to improve visuali- Daniels, P. (1986). Cognitive models in information retrieval: An eval-
uative review. Journal of Documentation, 42(4), 272304.
zation and reuse. Rushed Kanawati and Maria Malek Ellis, D. (1990). New horizons in information retrieval. London: The
proposed CoWing (Collaborative Web Indexing), a Library Association.
system in which an agent interacts with other agents Fidel, R., & Crandall, M. (1997). Users perception of the performance
to fetch new bookmarks that match its clients needs. of a filtering system. In Proceedings of the 20th annual ACM SIGIR
Conference on Research and Development in Information
All the solutions to information overload de- Retrieval (pp. 198205). New York: ACM Press.
scribed above emphasize the need for specialization Hill, W., & Hollan, J. (1994). History-enriched digital objects:
in information retrieval services using IF techniques. Prototypes and policy issues. Information Society, 10(2), 139145.
User modeling, which explores what affects users Howard, D. L., & Crosby, M. (1993). Snapshots from the eye: Towards
strategies for viewing bibliographic citations. In G. Salvendy &
cognitive loads, is critical for reducing information M. Smith (Eds.), Advances in human factors/ergonomics: Human-
overload. Currently, user modeling and information computer interaction: Software and hardware interfaces (Vol. 19B,
filtering techniques are employed in fields such as pp. 488493). Amsterdam: Elsevier Science.
Kanawati, R., & Malek, M. (2002). A multi-agent system for collab-
e-business, information retrieval, alerting systems, orative bookmarking. In P. Georgini, Y. LEsprance, G. Wagner, &
finance, banking, and communications. Research E. S. K. Yu (Eds.), Proceedings of the Fourth International Bi-
evaluating IF techniques and systems is on the rise, Conference Workshop on Agent-Oriented Information Systems
INFORMATION ORGANIZATION 355
(pp. 11371138). Retrieved December 16, 2003, from http://sunsite templates for the internal organization of
.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-59/ documents;
8Kanawati.pdf
cataloging and metadata;
Kay, J., & Kummerfeld, B. (1994). User models for customized hy-
pertext in J. Mayfield and E. C. Nicholas (Eds.). Advances in hy- knowledge organization systems (KOS): (faceted)
pertext for the World Wide Web (4769). New York: Springer-Verlag. classification schemes, taxonomies, ontologies,
Maes, P. (1994). Agents that reduce work and information overload. and thesauri; knowledge representation.
Communication of the ACM, 37(7), 3040.
Malone, T., Grant, K., Turbak, F., Brobst S., & Cohen, M. (1987).
Intelligent information-sharing systems. Communications of the
ACM, 30(5), 390402.
Morita, M., & Shinoda, Y. (1994). Information filtering based on user
The Entity-Relationship
behavior analysis and best match text retrieval. In Proceedings of
the Seventh Annual ACM-SIGIR Conference on Research and
Approach
Development in IR. ( pp. 272281). New York: Springer-Verlag. Information organization depends on object char-
Oard, D. (1997). The state of the art in text filtering. User Modeling acteristics (or properties), often expressed as state-
and User-Adapted Interaction, 7, 141178.
Petrelli, D., De Angeli, A, & Convertino, G. (1999). A user-centered
ments: entities (nouns) are connected through
approach to user modeling. Proceedings of the 7th International relationships (verbs), for example:
Conference on User Modeling, 255264.
pecan pie has ingredient (shelled pecans, 2 cups,
Pirolli, P., James, P., & Rao, R. (1996). Silk from a sows ear: Extracting
usable structures from the Web. Proceedings of the SIGCHI
for taste)
Conference on Human Factors in Computing Systems: Common
ground (pp. 118125). Retrieved December 16, 2003, from http:// Figure 1 shows an E-R conceptual schema
www.acm.org/sigchi/chi96/proceedings/papers/Pirolli_2/pp2.html for foodsa list of statement patterns, each defin-
Quiroga, L., & Mostafa J. (2002). An experiment in building profiles
in information filtering: The role of context of user relevance feed-
ing a type
back. Information Processing and Management, 38, 671694.
Rich, E. (1979). User modeling via stereotypes. Cognitive Science, 3,
335366. Food product hasName Text
Riedl, J., & Konstan, J. (2002). Word of mouse: The marketing power of Food product hasDescription Text
collaborative filtering. New York: Warner Books.
Food product hasHomePrepTime Time duration
Food product isa Food
Food product comesFromSource Food source[plant
or animal]
INFORMATION Food product

Food product
comesFromPart
hasIngredient
Anatomical
(Food product,
ORGANIZATION Amount [number
and unit], Purpose)
[(Chocolate, 50g, for
We organize informationin our minds and in taste), (BHT, 0.1g
information systemsin order to collect and record preservation)]
it, retrieve it, evaluate and select it, understand it, Food product underwentProcess (Process, Intensity,
process and analyze it, apply it, and rearrange and Purpose)[(broil, low
reuse it. We also organize things, such as parts, mer- heat, to brown)]
chandise in a store, or clothes in a closet, using Food product containsSubstance (Substance,
similar principles for similar purposes. amount)[(fat,13 g)
Using data on foods as an example, this article vitamin A, 4000
IU)]
introduces the following concepts:
Food product intendedFor Type of diet[low-
the entity-relationship (E-R) approach as the ba- fat, low-salt]
sis for all information organization;
database organization: relational databases, FIGURE 1. Entity-relationship (E-R) schema for a food
object-oriented databases, and frames; product database
foodName ingredient no unit purpose foodName: shelled pecans

pecan pie flaky 1 count body fromSource: pecan tree
piecrust fromPart: seed
pecan pie shelled 2 cup taste
process: shelling
pecans
foodName: eggs
pecan pie eggs 5 count body
fromSource: chicken
pecan pie white 1 cup taste
sugar fromPart: egg (part of animal)
Diet Coke carbon- 355 ml body
ated water FIGURE 3: Sample frames with slots and slot fillers
Diet Coke aspartame 200 mg taste
food product diet (in databases or in the mind) use the mechanism of
pecan pie normal hierarchical inheritance for efficient data input
Diet Coke low cal
and storage; for example, a frame for chocolate pecan
pie simply refers to the pecan pie frame and lists only
split pea soup normal
additional slots, such as
unsalted butter low salt
ingredient: (chocolate, 50 g, for taste).
ice cream normal
frozen yogurt low cal
FIGURE 2. Tables (relations) in a relational database The Internal Organization of

Documents: Templates
A recipe is a simple document describing a food prod-
uct, structured into a standard outline or document
This E-R schema is the basis for several ap- template (a frame applied to documents) with slots
proaches to storing and presenting data and for or- based on relationships (Figure 4). A template can be
ganizing a database for access and processing. encoded using XML (eXtensible Markup Language)
tags. Each tag is defined in an XML schema (not shown)
and identifies a type of information. (The ability to de-
Database Organization fine tailor-made tags for each application gives XML
In a relational database, data corresponding to one its power.) Each piece of information has a beginning
relationship type are expressed in a table (also called tag and a corresponding end tag. Once the informa-
relation, Figure 2); data about one Aobject@, such tion is encoded using XML, it can be used for many
as a food product, are distributed over many tables. purposes: to display a recipe in print or on the World
Tables are very simple data structures that can be Wide Web, produce a cookbook with table of contents
processed simply and efficiently. and an index, find all recipes that use certain ingredi-
Object-oriented databases store all the informa- ents, compose the ingredient label for a food (ingre-
tion about an object in one frame that has a slot for dients in order of predominance), compute the nutrient
every object characteristic as expressed in a relation- values for a serving (using a nutrient value table for ba-
ship (Figure 3). A frame can also call procedures op- sic foods).
erating on its data, such as computing fat content by As this example shows, organization of data in data-
adding fat from all ingredients. Frames are complex bases and structuring text in documents are alike. In
data structures that require complex software. Frames Figure 4, ingredients are given in a database-oriented
<foodProduct>
<foodName>pecan pie</foodName>
<unitsMade><number>8</number><unit>serving</unit></unitsMade>
<timeToMake><number>1.5</number><unit>hour</unit></timeToMake>
<description>A custard pie, loaded with pecans.</description>
<ingredients>
<foodProduct>flaky pie crust</foodProduct><number>1</number>
<unit>count</unit>
<foodProduct>shelled pecans</foodProduct><number>2</number><unit>cup</unit>
<foodProduct>eggs</foodProduct><number>5</number><unit>count</unit>
...
</ingredients>
<processingSteps>
<step>1</step><text>Prebake crust. Place pecans on baking sheet and bake</text>
<step>2</step><text>Start the filling</text><step>3</step>
<text>Beat the eggs. Beat in the sugar, salt, and butter</text>
...
</processingSteps>
</foodProduct>
FIGURE 4. Recipe following a standard outline (template), encoded with XML
mode (each element tagged separately), processingSteps perature and duration tagged separately.) These data
in a text-oriented mode. (Just the <text> tag; for can then be formatted for text output.
database-oriented tagging, steps would be broken down
into separately tagged processes, with data, such as tem-
Cataloging and Metadata
The recipe/food database or the catalog of a Web
store organizes the actual data from which users
title format questions can be answered. A library catalog or-
creator identifier ganizes data about books, which in turn contain the
subject source data to answer questions; the library catalog stores
description language data about data or metadata, as do Web search
publisher relation engines and catalogs of educational materials.
contributor coverage Metadata are stored and processed just like any other
date rights kind of data; whether a data item should be called
type metadata or just data is often a matter of perspec-
tive. The Resource Description Framework (RDF)
has been designed to encode metadata but can be
FIGURE 5.The Dublin Core (dc) for the description of used to encode any data represented in the E-R
document-like objects approach.
Food type Food source Plant/animal part Process Substance
side dishes plant food source plant part mechanical process food substance
. appetizers Juglandaceae . below ground . shelling . bulk nutrient
. soups Juglans (walnut) root . peeling carbohydrate
. salads Carya (Hickory) tuber . slicing sugar
vegetable C. illinoensis . above ground . grating starch
grain/starch dishes (pecan) stem . crushing fiber
. pasta . compositae leaves cooking process soluble f.
. grains Cichorium fruit . c. with dry heat protein
. breads C. intybus (anat. part) baking fat
. pizza C. endivia seed broiling . trace nutrient
fish, poultry, meat animal food source animal part . c. w. microwave vitamin
fish vertebrates . skeletal meat . c. w. moist heat mineral
poultry fish . organ meat boiling non-food substance
meat bird liver steaming . preservative
sweet baked dishes mammal . egg fruit . c. with fat or oil BHT
pies, tarts, pastries Bovidae (anat. part) freezing . package glue
cookies, brownies, Bos (cattle)
and cakes
FIGURE 6. Excerpts of a faceted classification for the food domain
There are many standards defining metadata brary arranges books on one subject together and
elements for different kinds of objects, for exam- like subjects close to each other. Such arrangement
ple the Dublin Core (Figure 5). These are often en- requires a classification (or taxonomy), such as
coded in XML, for example: Figure 6, column 1, for foods, or the Dewey
Decimal Classification for all subjects. To describe
<dc:title> How to cook everything </dc:title> foods by their characteristics, we need, for each
<dc:creator> Mark Bittman </dc:creator> characteristic or facet, a classification of the pos-
<dc:subject> cookbook </dc:subject> sible values (the possible fillers for a given frame
<dc:publisher> Macmillan </dc:publisher> slot); examples of facets, each with a partial clas-
sification of values, as shown in Figure 6.
(Not all records use all dc elements.) A classification is a structure that organizes
(The pecan pie example is based on a recipe in concepts into a meaningful hierarchy, possibly
this cookbook, which also inspired the food type in a scheme of facets. The classification of living
classification) things is a taxonomy. (The term taxonomy is in-
creasingly used for any type of classification.) A
classification is now often called an ontology, par-
Knowledge Organization Systems ticularly if it gives richer concept relationships.
A classification deals with concepts, but we need
(KOS) terms (words or phrases) to talk about concepts.
For the benefit of the user, a cookbook or a gro- However, the relationships between language and
cery store arranges like foods together, just as a li- concepts are complex. A concept can be expressed
Belgian endive Symbols used

DFVegetable consisting of the DF Definition SEMANTIC WEB A common framework that allows
leaves of Chicorium inty-bus, UF Used For USE data to be shared and reused across application, enter-
growing in a small, cylindrical BT Broader Term prise, and community boundaries.
head. NT Narrower Term
COmbination: vegetable : Cicho- RT Related Term
rium intybus : leaves
UFchicon computer can traverse along links from one con-
chiccory (vegetable) [spelling
cept to the next (a process called spreading acti-
variant]
chicory (vegetable) vation). Conceptual and terminological re-
French endive lationships can be encoded for computer storage
witloof using the Topic Map standard or RDF, both im-
BThead vegetable plemented in XML.
salad vegetable
RT chicory (coffee)
Outlook
FIGURE 7. A typical thesaurus entry Information organization is important for people
to find and understand information. It is also
important for computer programs to process in-
by several terms, such as Belgian endive, French en- formation to make decisions or give recommen-
dive, witloof, chicory, and chicon, which all refer to the dations, for example in medical expert systems and
same vegetable; these terms are in a synonym electronic commerce (e-commerce) and semantic
relationship with each other. Conversely, a term Web applications (where information organization
may refer to several concepts, such as chicory, which is called knowledge representation). These appli-
refers (1) to a vegetable and (2) to a coffee substi- cations require well-thought-out conceptual struc-
tute made from the root of the same plant; such a tures, which must be developed by beginning
term has the property of being a homonym (in in- from scratch or by refining existing knowledge or-
formation retrieval, a character string with multi- ganization systems (KOS). The most serious chal-
ple meanings). A thesaurus is a structure that (1) lenge is ensuring the interoperability of KOS and
manages the complexity of terminology by group- metadata schemes worldwide so that different sys-
ing terms that are synonymous to each other and tems can talk to each other.
disambiguating homonyms by creating a unique
term for each meaning and (2) provides concep- Dagobert Soergel
tual relationships, ideally through an embedded
classification/ontology. A thesaurus often selects See also Expert Systems; Information Retrieval;
from a group of synonyms the term, such as Belgian Markup Language; Ontology
endive, to be used as descriptor for indexing and
searching in a given information system; having one
descriptor for each concept saves the searcher from
having to enter several terms for searching. The de- FURTHER READING
scriptors so selected form a controlled vocabulary
Bailey, K. D. (1994). Typologies and taxonomies: An introduction to clas-
(authority list, index language). Figure 7 shows a sification techniques. Thousand Oaks, CA: Sage Publications.
typical thesaurus entry. Dodds, D. (2001). Professional XML metadata. Hoboken, NJ: Wrox.
Rich conceptual relationships can be shown Jonassen, D. H., Beissner, K., & Yacci, M. (1993). Structural knowledge:
Techniques for representing, conveying and acquiring structural
graphically in concept maps, which are used par- knowledge. Hillsdale, NJ: Lawrence Erlbaum.
ticularly in education to aid understanding; they Lancaster, F. W. (1972). Vocabulary control for information retrieval.
represent semantic networks, which a user or a Washington, DC: Information Resources Press.
Lynch, P., & Horton, S. (2002). Web style guide: Basic design princi- In the Beginning
ples for creating Web sites (2nd ed.). New Haven, CT: Yale University During the 1960s mainframe computers were pro-
Press.
Milstead, J., & Feldman, S. (1999). Metadata: Cataloging by any
grammed with punch cards and magnetic tapes that
other name . . . Metadata projects and standards. Online, 23(1), made such innovations as automated accounting sys-
2440. Retrieved January 22, 2004, from www.infotoday.com/on- tems and space travel possible. In a short period of
line/OL1999/milstead1.html time, the development of personal computers and
Mondeca topic organizer. Retrieved January 22, 2004, from http://www
.mondeca.com/ computer networking made it possible for people to
Ray, E. (2003). Learning XML (2nd ed.). Sebastopol, CA: OReilly have the power of a mainframe on their desktop.
Rob, P., & Coronel, C. (2004). Database systems: Design, implemen- People could write their own documents and memos
tation, and management (6th ed.). Boston: Course Technology.
Rosenfeld, L., & Morville, P. (2002). Information architecture for the
and use spreadsheet software to perform numeri-
World Wide Web: Designing large-scale web sites (2nd ed.). Sebas- cal analysis with relative ease. Today the presence
topol, CA: OReilly. of the Internet puts the world at ones fingertips. Any
Skemp, R. R. (1987). The psychology of learning mathematics. Hillsdale, product of service that a person wants to buy is read-
NJ: Lawrence Erlbaum.
Soergel, D. (1974). Indexing languages and thesauri: Construction and
ily available, not from two companies, but from over
maintenance. New York: Wiley. two hundred.
Soergel, D. (2000). ASIST SIG/CR Classification Workshop 2000: The terms information overload, infoglut, scholo-
Classification for user support and learning: Report. Knowledge glut and data smog all refer to the same topictoo
Organization, 27(3), 165172.
Soergel, D. (2003). Thesauri and ontologies in digital libraries. Retrieved much information to comprehend. The following
January 22, 2004, from http://www.clis.umd.edu/faculty/soergel/ examples demonstrate a variety of contexts for in-
SoergelDLThesTut.html formation overload:
Sowa, J. F. (2000). Knowledge representation: Logical, philosophical and
computational foundations. Pacific Grove, CA: Brooks/Cole. A person wants a book on gardening. A search
Staab, S., & Studer, R. (Eds.). (2004). Handbook on ontologies in in- of gardening books on the Amazon.com website
formation systems. Heidelberg, Germany: Springer.
Taylor, A. G. (2003). The organization of information (2nd ed.). yields 30,707 possible books.
Westport, CT: Libraries Unlimited. An executive begins the day by logging into her
Vickery, B. C. (1960). Faceted classification: A guide to construction and e-mail. She opens Microsoft Outlook only to find
use of special schemes. London: Aslib.
Vickery, B. C. (2000). Classification and indexing in science. Burlington,
she starts the day with sixty e-mail messages; ten
MA: Butterworth-Heinemann. of them are advertisements.
Zapthink. (2002). Key XML specifications and standards. Retrieved A person is watching television. He has Direct TV
January 22, 2004, from http://www.oasis-open.org/commit- with subscriptions to movie and sports channels
tees/download.php/173/xml%20standards.pdf XXX
and can choose from over 500 different stations.
A national security advisor receives 200 messages
a day about potential threats to national security.
Limits to Human Cognition

INFORMATION The first study related to information overload was
done in 1956 by George A. Miller. In his work on the
OVERLOAD human ability to retain pieces of information, Miller
discovered that people generally can retain seven plus
Information overload has a personal, social con- or minus two pieces of information at any point in
text too. Information overload is the confusion that time. When the number of items of information is
results from the availability of too much informa- much greater than that, cognitive overload occurs
tion. When a person is trying to find information or and retention is hindered. Miller also described
solve a problem, and there is an overabundance of chunking of information. By that he meant that
information available, the result is confusion and an a person can retain more than seven pieces of in-
inability to filter and refine the onslaught of infor- formation by grouping, or chunking, like pieces of
mation so that it is possible to make a decision. information together. This seminal study in psy-
INFORMATION OVERLOAD 361
chology went on to form the practical basis for graph- people, when faced with massive amounts of in-
ical user interface design, providing guidance on how formation to make a decision, simply shut down and
many application windows should be open at one make their decisions based up instinct. However,
time and how to nest menus. quitting or ignoring information is not a viable way
A model for information overload that defines to make many decisions. Instead, the problems of in-
characteristics and symptoms of information formation overload created by technology can also
overload was developed by Schneider (1987). Factors be alleviated by technology in the following ways:
that influence and exacerbate information overload
include uncertainty, ambiguity, novelty, complexity, Filtering tools: A good example of informa-
intensity, and amount and rate of input. These faction filtering is using a web browser. When you
tors and organizational or environmental conditions search for Saturn (the planet), you may end up
work together to create information overload. with links to cars, planets, and nightclubs. You
Primary symptoms of overload listed by Schneider can further refine your search to include the word
are loss of integration, loss of differentiation, and planet and exclude the words car and club. The
confusion. narrower you define your search, the more likely
The amount of information we have and the you are to get the information you are looking
massive access we have to it are two of the major con- for. E-mail applications also come with filters. If
tributors to information overload. The age of ubiqui- you get a great number of junke-mail messages
tous computing being able to access, use, and asking you to refinance your mortgage, you
communicate by computer, anytime, any place, can set a filter to block mail from a specific ad-
anywhereis here. Current estimates are that over dress or mail that contains specific keywords. The
2 billion webpages exist on the Internet. There are messages you filter out can be deleted from your
over ten thousand scholarly journals and databases. computer without your viewing them. If you set
The amount of information conveyed is gigantic. the filter narrowly enough, there will be a re-
Equally amazing is the ease with which we can ac- duction in the amount of information you need
cess Internet, databases, phone, radio, and television to process. The danger is that you may exclude
from anywhere we travel. Professionals often take information that could be of potential value.
laptops and check e-mail while on vacation. Even Intelligent agents: An intelligent agent is a soft-
children carry cellular phones. Technology exists ware program that can actively seek information
in every aspect of our daily lives and there is little for you based on parameters you set. It differs
escape from it. from a search engine or information filter in that
Increasingly, people identify the social ills cre- it actively seeks specific information while you
ated by information overload. In a speech by Tim are doing other things. If you are an osteopath,
Sanders, a Yahoo executive, information overload for example, your agent can actively, continually
was blamed for the rise in work stress (Soto 2003). seek the latest research on bone fractures. When
Co-workers are continually interrupted by tech- it finds something relevant, it will bring it to your
nologies such as instant messaging and e-mail. attention.
Theyve become so confined by technology that they Web agents work using cookies (pieces of infor-
send a message to someone 5 feet away rather than mation from websites recorded on your hard
talk to that person face-to-face. drive) to track user preferences and provide ad-
ditional information of value to users. For exam-
ple, if you are shopping for a digital camera, a web
Technical Solutions to agent can supply you with special deals on cam-
eras, competing products, and accessories avail-
Information Overload able for the camera in which you are interested.
Some estimates indicate that 80 percent of people Prioritizing schemes: One of the weakest aspects
abandon electronic purchases at the checkout. Other of information-seeking technology is that it does
not often rate the quality of the information it chy in which like items are grouped. If the item
finds. A search engine provides information based group is not relevant to the user, the user can eas-
on algorithms for indexing pages and the con- ily ignore the information that is not relevant.
tent (meta tags, header information, page key- For example, if I go to the Newsweek site to
words) of the pages that are indexed. Different read about business, I can quickly ignore featured
search engines yield different results because they stories, opinion polls, and multimedia offerings
use different indexing techniques. Pages, some- because they are all located and distinguished
times unrelated to what you are seeking, are somewhere else on the site from the magazines
found because the authors designed them for regular business columns. Good design can make
success in online searches. To overcome this, some finding information easy. However, sites are also
searching tools, such a periodical indexes and site designed to get a person to do something. While
searches, do provide an estimate of the quality the users goal might be to buy the cheapest cam-
of the found information. Often, the rating of era, the purpose of the site may be to sell over-
quality is given as a percent or as an icon show- stocked items. In this case the design could draw
ing a fill bar indicating relevance. the user to special offers instead of facilitating
the search for a quality camera.
Many applications and websites offer quick pri-
oritization schemes to allow the user to see infor-
mation based on what is important to them. For
example, in purchasing a digital camera, I am allowed
Making Optimal Decisions
Information is important because it assists us in mak-
to order my information according to price, popu-
ing decisions. With perfect information, we can make
larity, brand, or relevance.
the optimal decision. When faced with problems,
Prioritizing gives users a quick way to focus on
people seek as much information as they can find to
what they find relevant or important. However, the
support a rational decision that will best help them
prioritization provided by the website or database is
solve their problems. They optimize their deci-
usually a generic scheme to help all users. A more
sions based on sought-out information. Given too
personalized approach to handling information can
little, too much, or contradictory information, people
be found through personal web portals.
are forced to find solutions that are satisfactory. This
Personalized portals: A personalized portal is is called satisficing, or choosing a solution that will
an access point for information based on a users work, even if it is not the optimal solution.
personal preferences. For example, a farmer could In his work on information filtering (1987),
use a portal that monitored weather, costs asso- Thomas Malone concedes that the value of tech-
ciated with harvesting and delivery of food, and nology is not so much in eliminating unwanted in-
competitors prices. The rich information that formation as it is in seeking information that is
the farmer needs to make decisions is contained relevant. To this end we see an explosion of tech-
in one portal interface, showing no distracting nological devices and sources filling the market. As
or confounding additional information. the number of products increases, so do the tech-
Design: In addition to filtering and seeking technological innovations for managing those products.
nologies, design and visualization can assist with Consider, as an example, the invasiveness of tele-
information overload too. In web development, marketers in the United States. Their access to
simple design concepts can assist the user in find- individual phone numbers and other private in-
ing the information they seek. The concept of formation led to the development of new telecom-
proximity putting like items together helps munications blocking tools: telephone technologies
the user chunk large amounts of information now that display who is calling and provide call-
into fewer, more manageable pieces. Information blocking and call-waiting services, all for a fee.
intensive websites provide a clear visual hierar- Ultimately, a national Do Not Call list was im-
INFORMATION RETRIEVAL 363
plemented so people could elect to have their educational materials for a learning objective;
numbers protected from telemarketer harassment. digital cameras for taking family photos;
Now technology is used to check the list before the recipes that use ingredients on hand;
call is made, rather than users having to purchase facts needed for deciding on a company merger.
their own blocking devices.
The main trick is to retrieve what is useful while
The growth of information, access, and products
leaving behind what is not.
ensures that overload is a consequence of the in-
formation age. As a result, innovations and tools to
cope with information overload will continue to The Scope of IR
be developed, along with social policies and norms IR systems are part of a family that shares many
to reduce overload. principles (Figure 1).
Two distinctions are of particular importance:
Ruth A. Guthrie
1. A system for unstructured information deals
with such questions as: The economic impact
of the Reformation, The pros and cons of
FURTHER READING
school uniforms, or Find a nice picture of
Malone, T. W., Grant, K. R., Turbak, F. A., Brobst, S. A., & Cohen, M. D. my niece. It finds documents that are more or
(1987). Intelligent information sharing systems. Communications less useful; the user must then extract the data
of the ACM, 30(5), 390402. needed. In contrast, a system for well-struc-
Miller, G. A. (1956). The magical number seven, plus or minus two:
Some limits on our capacity for processing information. Psycho-
tured information deals with precise questions
logical Review, 63, 8197. and returns precise answers, exactly the small
Schneider, S. C. (1987). Information overload: Causes and conse- pieces of data needed: the salary of Mrs. Smith;
quences. Human Systems Management, 7, 143153. the population of China; the winner of the
Soto, M. (2003). The toll of information overload: Too much tech-
nology diminishes work relationships. (2003, August 8). Seattle
1997 World Series.
Times, (p. C1). 2. Finding versus creating answers. IR and database
Tushman, M. L., & Nadler, D. A. (1978). Information processing as an systems merely find what is already there: for ex-
integrated concept on organizational design. Academy of ample, from a patient database, a patients symp-
Management Review, 3, 613624.
toms; from a disease database, the diseases these
symptoms point to (or a medical textbook from
which to extract this information); and from a
drug database, the drugs that treat a disease. A
INFORMATION physician must then absorb all this information,
derive a diagnosis, and prescribe a drug. A med-
RETRIEVAL ical expert system goes beyond just finding the
factsit creates new information by inference:
Information retrieval systems are everywhere: Web It identifies a disease that explains the patients
search engines, library catalogs, store catalogs, cook- symptoms and then finds a drug for the disease.
book indexes, and so on. Information retrieval (IR),
also called information storage and retrieval (ISR or
ISAR) or information organization and retrieval, is the
The Objects of IR
Traditionally, IR has concentrated on finding whole
art and science of retrieving from a collection of items
documents consisting of written text; much IR re-
a subset that serves the users purpose; for example:
search focuses more specifically on text retrieval
webpages useful in preparing for a trip to Europe; the computerized retrieval of machine-readable text
magazine articles for an assignment or good read- without human indexing. But there are many other
ing for that trip to Europe; interesting areas:
finding answers and information

that already exist in a system
creating answers and
Search by navigation new information by
(following links, as in a analysis and
subject directory and the Search by query inference
Web generally) (as in Google) based on query
Unstructured Hypermedia systems IR systems
information (Many small units, such (Often dealing with
(text, images, as paragraphs and whole documents, such
sound) single images, tied as books and journal
together by links) articles)
Structured Database management Data analysis systems
information systems (DBMS) Expert systems
FIGURE 1. The IR system family
Speech retrieval, which deals with speech, often people or hardware items, and this article deals with
transcribed manually or (with errors) by auto- IR broadly, using document as stand-in for any
mated speech recognition (ASR). type of object. Note the difference between retriev-
Cross-language retrieval, which uses a query in ing information about objects (as in a Web store cat-
one language (say English) and finds documents alog) and retrieving the actual objects from the
in other languages (say Chinese and Russian). warehouse.
Question-answering IR systems, which retrieve
answers from a body of text. For example, the
question Who won the 1997 World Series? Utility, Relevance, and
finds a 1997 headline World Series: Marlins are
champions. IR System Performance
Image retrieval, which finds images on a theme Utility and relevance underlie all IR operations. A
or images that contain a given shape or color. documents utility depends on three things, topical
Music retrieval, which finds a piece when the user relevance, pertinence, and novelty. A document is
hums a melody or enters the notes of a musical topically relevant for a topic, question, or task if it
theme. contains information that either directly answers the
IR dealing with any kind of other entity or ob- question or can be used, possibly in combination
ject: works of art, software, courses offered at a with other information, to derive an answer or
university, people (as experts, to hire, for a date), perform the task. It is pertinent with respect to a user
products of any kind. with a given purpose if, in addition, it gives just the
information needed; is compatible with the users
Text, speech, and images, printed or digital, carry background and cognitive style so he can apply the
information, hence information retrieval. Not so for information gained; and is authoritative. It is novel
other kinds of objects, such as hardware items in a if it adds to the users knowledge. Analogously, a
store. Yet IR methods apply to retrieving books or soccer player is topically relevant for a team if her
abilities and playing style fit the team strategy, per- software? In the literature, the term relevance is
tinent if she is compatible with the coach, and novel used imprecisely; it can mean utility or topical rel-
if the team is missing a player in her position. evance or pertinence. Many IR systems focus on find-
Utility might be measured in monetary terms: ing topically relevant documents, leaving further
How much is is it worth to the user to have found selection to the user.
this document? How much is this player worth Relevance is a matter of degree; some docu-
to us? How much did we save by finding this ments are highly relevant and indispensable for
Query description Document titles Relevant
Production and uses of plastic pipes 1 The production of copper pipes

As the examples show, simple word 2 Cost of plastic pipe manufacture /
match is often not enough; retriev-
3 Polyethylene water pipes /
ing documents and assessing rele-
vance require knowledge: The system 4 Steel rod manufacture
needs to know that polyethylene and
5 Spiral PVC tubes as cooling elements /
PVC are plastics, that tube is another
word for pipe, that artery in the con- 6 Innovative plastic surface for new city artery
text of 6 means a major street and in
7 Artificial arteries help heart bypass patients /
7 a pipe in the body, usually made of
plastic. 8 Plastic mouthpieces in making smoking pipes
Bioinformatics 1 Bioinformatics /
Bioinformatics is the application of 2 Computer applications in the life sciences /
sophisticated computer methods to
studying biology. This is another 3 Biomedical informatics /
illustration of the variability of lan- 4 Modeling life processes /
guage IR systems must deal with.
5 Modeling traffic flow
6 Modeling chemical reactions in the cell /
Jewish-Gentile relations 1 We played with our non-Jewish friends. /
This could be a question to the 2 We were taunted in school. /
Shoah Foundations collection of
3 Aryan people had many advantages.
transcribed testimonies from
Holocaust survivors. None of the 4 My mother talked often to the neighbors. /
stories that shed light on this ques-
tion has the query phrase in it. 5 Jews were deported to concentration camps.
Relevance must be inferred from the
6 Jews were forbidden to attend concerts. /
entire context.
FIGURE 2. Query descriptions compared with document or story titles

the users tasks; others contribute just a little bit and (1) actively find out what users need, (2) acquire
could be missed without much harm (see ranked re- documents (or computer programs, or products,
trieval in the section on Matching). or data items, and so on), resulting in a collec-
From relevance assessments we can compute tion, and (3) match documents w ith needs.
measures of retrieval performance such as Determining user needs involves (1.1) studying user
needs in general as a basis for designing responsive
recall = How good is the system at find-
systems (such as determining what information stu-
ing relevant documents?
dents typically need for assignments), and (1.2) ac-
discrimination = How good is the system at re-
tively soliciting the needs of specific users, expressed
jecting irrelevant documents?
as query descriptions, so that the system can pro-
precision = Depends on discrimination,
vide the information (Figure2). Figuring out
recall, and the # of relevant
what information the user really needs to solve a
documents
problem is essential for successful retrieval.
Evaluation studies commonly use recall and Matching involves taking a query description and
precision or a combination; whether these are the finding relevant documents in the collection; this
best measures is debatable. With low precision, the is the task of the IR system (Figure 2).
user must look at several irrelevant documents for The simplest text retrieval systems merely com-
every relevant document found. More sophisti- pare words in the query description with words in
cated measures consider the gain from a relevant the documents (title, abstract, or full text) and rank
document and the expense incurred by having documents by the number of matches, but results
to examine an irrelevant document. For ranked are often poor (Figure 2). A good IR system pro-
retrieval, performance measures are more com- vides the access points required to respond to user
plex. All of these measures are based on assess- needs in retrieval and selection. This means prepar-
ing each document on its ow n, r ather than ing user-oriented document representations
considering the usefulness of the retrieved set as (Figure 3) that describe a document by several state-
a whole; for example, many relevant documents ments using <relationships> as verbs and Entities as
that merely duplicate the same information just subjects and objects. The allowable Entity Types and
waste the users time, so retrieving fewer rele- <relationship types> define what kinds of infor-
vant documents would be better. mation the system can store; they make up the con-
ceptual schema.
Fo r s o m e e n t i t y t y p e s ( i n t h e e x a m p l e
How Information Retrieval Person, Text, Phrase, and URL), values can be freely
chosen; for others (Subject and Function), values
Systems Work come from a controlled vocabulary that fixes the
IR is a component of an information system. An term used for a concept. For example, pipe is used
information system must make sure that everybody for the concept also known as tube, so the user
it is meant to serve has the information needed to needs to enter only one term. If the user enters tube,
accomplish tasks, solve problems, and make deci- the system (or the user) follows the thesaurus cross-
sions, no matter where that information is avail- reference
able. To this end, an information system must tube USE ST pipe (ST = Synonymous Term)
The thesaurus also includes conceptual cross-

EXPERT SYSTEM A computer system that captures and references:
stores human problem-solving knowledge or expertise pipe BT hollow object (BT = Broader Term)
so that it can be used by other, typically less-knowl-
edgeable people. and
pipe NT capillary (NT = Narrower Term)
Statement Data field
Document <written by> Person John Smith Author
Document <has title> Te xt Artificial arteries help heart ... Title
Document <has abstract> Te xt A clinical study ... showed that ... Abstract
Document <contains word or phrase> Phrase artificial arteries Free text
Document <relevant for> Subject Blood Vessel Prosthesis Descriptor
Document <describes tool for> Function Coronary Artery Bypass Function
Document <has URL> URL www.healtheduc.com/heart/... URL
FIGURE 3. Document representation as a group of statements
(For the structure of thesauri, see the article automated process of making statements about a doc-
on Information Organization.) The conceptual ument, lesson, person, and so on, in accordance with
schema and the thesaurus must of course reflect the conceptual schema (see Figure 3). We focus
user needs. here on subject indexingmaking statements about
If an entity (such as a document or a data file) a documents subjects. Indexing can be document-
is sought as a source of data/information, the data oriented the indexer captures what the document
about the entity are used as metadata (data de- is about, or request-oriented the indexer assesses the
scribing data); thus, the data in Googles catalog documents relevance to subjects and other features
of Web pages are used primarily as metadata. of interest to users; for example, indexing the testi-
monies in Figure 2 with Jewish-Gentile relations,
marking a document as interesting for a course, or
Steps in the IR Process marking a photograph as publication quality. Related
An IR system prepares for retrieval by indexing doc- to indexing is abstracting creating a shorter text
uments (unless the system works directly on the doc- that describes what the full document is about (in-
ument text) and formulating queries, resulting in dicative abstract) or even includes important results
document representations and query representa- (informative abstract, summary). Automatic summa-
tions, respectively; the system then matches the rep- rization has attracted much research interest.
resentations and displays the documents found Automatic indexing begins with raw feature
and the user selects the relevant items. These processes extraction, such as extracting all the words from a
are closely intertwined and dependent on each other. text, followed by refinements, such as eliminating
The search process often goes through several itera- stop words (and, it, of), stemming (pipes Y pipe),
tions: Knowledge of the features that distinguish rel- counting (using only the most frequent words), and
evant from irrelevant documents is used to improve mapping to concepts using a thesaurus (tube and
the query or the indexing (relevance feedback). pipe map to the same concept). A program can an-
alyze sentence structures to extract phrases, such
Indexing: Creating Document Representations as labor camp (a Nazi camp where Jews were forced
Indexing (also called cataloging, metadata assign- to work, often for a company; phrases can carry much
ment, or metadata extraction) is the manual or meaning). For images, extractable features include
color distribution or shapes. For music, extractable document was published. A query can simply give
features include frequency of occurrence of notes or features in an unstructured list (for example, a
chords, rhythm, and melodies; refinements include bag of words) or combine features using
transposition to a different key. Boolean operators (structured query). Examples:
Raw or refined features can be used directly for Bag of words: (pipe tube capillary plastic polyeth-
retrieval. Alternatively, they can be processed fur- ylene production manufacture)
ther: The system can use a classifier that combines Boolean query: ( p i p e O R t u b e O R c a p i l l a r y )
the evidence from raw or refined features to assign AND(plastic OR polyethylene)
descriptors from a pre-established index language. AND(production OR manufacture)
To give an example from Figure 2, the classifier uses
the words life and model as evidence to assign bioin- The Boolean query specifies three ANDed con-
formatics (a descriptor in Googles directory). A clas- ditions, all of which are necessary (contribute to the
sifier can be built by hand by treating each descriptor document score); each condition can be filled by any
as a query description and building a query for- of the words joined by OR; one of the words is as
mulation for it as described in the next section. Or good as two or three. If some relevant documents
a classifier can be built automatically by using a train- are known, the system can use them as a training set
ing set, such as the list of documents for bioinfor- to build a classifier with two classes: relevant and not
matics in Figure 3, for machine learning of what relevant.
features predict what descriptors. Many different Stating the information need and formulating
words and word combinations can predict the same the query often go hand-in-hand. An intermediary
descriptor, making it easier for users to find all doc- conducting a reference interview helps the user think
uments on a topic. Assigning documents to (mu- about the information need and find search terms
tually exclusive) classes of a classification is also that are good predictors of usefulness. An IR system
known as text categorization. Absent a suitable can show a subject hierarchy for browsing and find-
classification, the system can produce one by clus- ing good descriptors, or it can ask the user a series
tering grouping documents that are close to each of questions and from the answers construct a query.
other (that is, documents that share many features). For buying a digital camera, the system might ask
the following three questions:
Query Formulation: What kind of pictures do you take (snapshots,
Creating Query Representations
stills, ...)?
Retrieval means using the available evidence to pre- What size prints do you want to make (57,
dict the degree to which a document is relevant or
810, . . .)?
useful for a given user need as described in a free- What computer do you want to transfer im-
form query description, also called topic description
ages to?
or query statement. The query description is trans-
formed, manually or automatically, into a formal Without help, users may not think of all the features
query representation (also called query formulation to consider. The system should also suggest syn-
or query for short) that combines features that onyms and narrower and broader terms from its
predict a documents usefulness. The query expresses thesaurus. Throughout the search process, users fur-
the information need in terms of the systems con- ther clarify their information needs as they read
ceptual schema, ready to be matched with document titles and abstracts.
representations. A query can specify text words or
phrases the system should look for (free-text search) Matching the Query Representation with
or any other entity feature, such as descriptors as- Entity Representations
signed from a controlled vocabulary, an authors or- The match uses the features specified in the query to
ganization, or the title of the journal where a predict document relevance. In exact match the sys-
Query term
housing (weight 2) conditions (1) Siemens (2) labor camps (3) Score
(weight in query)
idf, log(idf) 10,000, log=4 100, log=2 100,000, log=5 10,000, log=4
term(tf ) Doc. 1 barracks (5 times) conditions (3) Siemens (2) labor camps (4)
(tf = 40 + 6 + 20 + 48 = 114
frequency
Doc. 2 housing (3 times) conditions (2) Siemens (2) labor camps (4) 96
of the
term in Doc. 3 housing (3 times) conditions (4) Siemens (1) labor camps (4) 90
each Doc. 4. housing (3 times) conditions (3) Siemens (2) labor camps (3) 86
document) Doc. 5 housing (2 times) conditions (10) labor camps (1) 48
FIGURE 4. Computing relevance scores
tem finds the documents that fill all the conditions example), scores are normalized to a value be-
of a Boolean query (it predicts relevance as 1 or 0). tween 0 and 1.
To enhance recall, the system can use synonym ex-
pansion (if the query asks for pipe, it finds tubes as Selection
well) and hierarchic expansion or inclusive search- The user examines the results and selects relevant
ing (it finds capillary as well). Since relevance or use- items. Results can be arranged in rank order (ex-
fulness is a matter of degree, many IR systems amination can stop when enough information is
(including most Web search engines) rank the re- found); in subject groupings, perhaps created by
sults by a score of expected relevance (ranked re- automatic classification or clustering (similar items
trieval). Consider the query Housing conditions in can be examined side by side); or by date. Displaying
Siemens labor camps. Figure 4 illustrates a simple title + abstract with search terms highlighted is most
way to compute relevance scores: Each terms con- useful (title alone is too short, the full text too long).
tribution is a product of three weights: The query Users may need assistance with making the con-
term weight (the importance of the term to the user), nection between an item found and the task at hand.
the term frequency (tf) (the number of occurrences
of the term in the document, synonyms count also), Relevance Feedback and Interactive Retrieval
and the rarity of the term or inverse document fre- Once the user has assessed the relevance of a few
quency (idf) on a logarithmic scale. items found, the query can be improved: The sys-
tem can assist the user in improving the query
If document frequency = .01
by showing a list of features (assigned descriptors;
(1 % or 1/100 of all documents include the term),
text words and phrases, and so on) found in many
then idf = 100 or 102 and log(idf) = 2.
relevant items and another list from irrelevant
For example, in Figure 4 the contribution of items. Or the system can improve the query au-
housing to relevance score of Document 1 is tomatically by learning which features separate
relevant from irrelevant items and thus are good
query weight 2 * log(idf) 4 * tf
predictors of relevance. A simple version of auto-
(term frequency in document) 5 = 40
matic query adjustment is this: increase the weights
(Google considers, in addition, the number of of features from relevant items and decrease the
links to a webpage.) Usually (but not in the simple weights of features from irrelevant items.
IR System Evaluation Outlook: Beyond Retrieval

IR systems are evaluated with a view to improvement Powerful statistical and formal-syntax-based meth-
(formative evaluation) or with view to selecting the ods of natural language processing (NLP) extract
best IR system for a given task (summative evalua- meaning from text, speech, and images and create
tion). IR systems can be evaluated on system char- detailed metadata for support of more focused
acteristics and on retrieval performance. System searching. Data mining and machine learning dis-
characteristics include the following: cover patterns in large masses of data. Sophisticated

database and expert systems search and correlate huge
the quality of the conceptual schema (Does it
amounts of different types of data (often extracted
include all information needed for search and
from text) and answer questions by inference. New
selection?);
the quality of the subject access vocabulary (in-
visualization techniques using high-resolution dis-
plays allow users to see patterns and large networks
dex language and thesaurus) (Does it include
of linked information. Sophisticated user models al-
the necessary concepts? Is it well structured?
low intelligent customization. IR can be integrated
Does it include all the synonyms for each
into day-to-day work: A medical IR system can
concept?);
the quality of human or automated indexing
process a patients chart, find several relevant arti-
cles, and prepare a tailor-made multi-document sum-
(Does it cover all aspects for which an entity is
mary, or it can deduce the drugs to be prescribed.
relevant at a high level of specificity, while avoid-
A legal IR system can take an attorneys outline of
ing features that do not belong?);
the nature of the search algorithm;
the legal issues in a case, find relevant cases or sec-
the assistance the system provides for informa-
tions of cases, and arrange them according to the
outline to give the attorney a running start on
tion needs clarification and query formulation;
writing a brief. All these advances contribute to an
and
the quality of the display (Does it support
unprecedented level of support for problem solving,
decision making, and intellectual work.
selection?).
Measures for retrieval performance (recall, dis- Dagobert Soergel
crimination, precision, novelty) were discussed in
the section Relevance and IR system performance. See also Information Filtering; Information
Requirements for recall and precision vary from Organization; Ontology; Search Engines
query to query, and retrieval performance varies
widely from search to search, making meaningful
evaluation difficult. Standard practice evaluates sys- FURTHER READING
tems through a number of test searches, computing Baeza-Yates, R., & Rubiero-Neto, B. (1999). Modern information re-
for each a single measure of goodness that combines trieval. Reading, MA: Addison Wesley.
recall and precision, and then averaging over all Berners-Lee, T., Hendler, J., Lassila, O. (2001). The semantic web.
the queries. This does not address a very impor- Scientific American, 284(5), 3443, Retrieved January 22, 2004,
from http://www.sciam.com
tant system ability: the ability to adapt to the specific Blair, D. C. (1990). Language and representation in information re-
recall and precision requirements of each individual trieval. Amsterdam: Elsevier Science.
query. The biggest problem in IR evaluation is to Brin, S., & Page, L. (1998). The anatomy of a large-scale hyper-
identify beforehand all relevant documents (the textual web search engine. 7th International World Wide Web
Conference (WWW7). Computer Networks and ISDN Systems,
recall base); small test collections have been con- 30(XXX). Retrieved January 22, 2004, from www-db.stanford.edu/
structed for this purpose, but there is a question of ~backrub/google.html, www7.scu.edu.au/programme/fullpapers/
how well the results apply to large-scale real-life col- 1921/com1921.htm, decweb.ethz.ch/WWW7/00/
Boiko, B. (2002). Content management bible. New York: Hungry Minds.
lections. The most important evaluation efforts of Retrieved January 22, 2004, from http://metatorial.com/index.asp
this type today are TREC and TDT (see Further Chu, H. (2003). Information representation and retrieval in the digi-
Reading). tal age. Medford, NJ: Information Today.
INFORMATION SPACES 371
Feldman, S. (1999). NLP Meets the Jabberwocky: Natural language

processing in information retrieval. ONLINE, May 1999. Retrieved DATA MINING The process of information extraction
January 22, 2004, from www.onlinemag.net/OL1999/feldman5.html
Feldman, S. (2000). The answer machine. Searcher, 8(1), 121, 5878.
with the goal of discovering hidden facts or patterns
Retrieved January 22, 2004, from http://www.infotoday.com/ within databases.
searcher/jan00/feldman.htm
Frakes, W. B., & Baeza-Yates, R. (Eds.). (1992). Information retrieval:
Data structures and algorithms. Englewood Cliffs, NJ: Prentice Hall.
Hert, C. A. (1997). Understanding information retrieval interactions: Information spaces can facilitate finding and us-
Theoretical and practical applications. Stamford, CT: Ablex. ing information. They involve representation, most
Jackson, P., & Moulinier, I. (2002). Natural language processing for on- often spatial or similar (e.g. locations and inter-
line applications: Text retrieval, extraction, and categorization.
Amsterdam: John Benjamins.
connections in graph form).
Negnevitsky, M. (2001). Artificial intelligence: A guide to intelligent sys-
tems. Reading, MA: Addison Wesley.
Soergel, D. (1985). Organizing information: Principles of database and
retrieval systems. Orlando, FL: Academic Press.
The Geometric Metaphor
Soergel, D. (1994). Indexing and retrieval performance: The logical
People often consider information spaces in geo-
evidence. Journal of the American Society for Information Science, metric terms, and geometry is more often than not
4(8), 589599. the basis for sophisticated graphical display of
Sparck Jones, K., & Willett, P. (1997). Readings in information retrieval. data. Consider the following setting: We have three
San Francisco, CA: Morgan Kaufmann.
Text REtrieval Conference (TREC), cosponsored by the National books characterized by four themes or three indi-
Institute of Standards and Technology (NIST) and the Defense viduals characterized by four behavioral character-
Advanced Research Projects Agency (DARPA). Retrieved Jan- istics. This semantic aspect is not immediately
uary 22, 2004, from http://trec.nist.gov/
Wilson, P. (1973). Situational relevance. Information Storage and
relevant to us. A data table or cross-tabulation of our
Retrieval, 9(8), 457471. books and themes can be represented in an infor-
Witten, J., & Bainbridge, D. (2002). How to build a digital library. mation space as three points (rows), each possess-
San Francisco, CA: Morgan Kaufmann. ing four coordinates or dimensions. We could equally
well consider a space of four points (columns), each
possessing three coordinates or dimensions. Various
mathematically well-defined algorithmic techniques
are available to us as software to display such an
INFORMATION SPACES information space or to process it by giving priority
to important aspects of it. One such technique,
In all walks of life the flood of data is ever increas- known as principal components analysis (PCA),
ing. Data are raw measurementsmeteorological, redefines the coordinates or dimensions in order
stock market, or traffic flow data, for example. In- to determine a better expression of the points con-
formation is data that have more structure. Such data sidered. Interestingly, a close mathematical rela-
have been subjected to preprocessing and linked with tionship exists between the analysis of the four
other data, leading to greater meaningfulness to three-dimensional points and the three four-
people as they make decisions in their work or leisure. dimensional points. Simultaneous display is possi-
In addition to data and information, a third term that ble in a kindred technique called correspondence
characterizes the life cycle of human decision mak- analysis.
ing is knowledge. Thus, we have a simple path that The geometric metaphor is a powerful one. It al-
starts with raw quantitative or qualitative data (plu- lows any quantitatively characterized data set to be
ral of datum, Latin for given). This path leads on displayed and processed. It also allows qualitative
to information, which we could call potentially data sets to be processed through the simple device
meaningful data or the set of components that is of taking the qualitative attributes as quantitative
usable in making decisions. Finally the path leads to ones (e.g., presence versus absence scored as 1 or 0).
knowledgeinformation that has been deployed in Origins or sources of data can be varied. In
decision making and is relevant to people. databases data are structured in defined fields and
in specified relationships. An online data feed in fi- and as the user interacts. The progression is nat-
nance or meteorology, where data are continually ural enough that the visualization becomes the
captured or generated, may be partially structured. user interface.
On the other hand, data from the World Wide Web
or from the news mediathink of national finan- Information retrieval was characterized in terms
cial and business confidence indicators, for exam- of semantic road maps by Doyle in 1961. The spa-
pleare partially structured at best. Raw text or tables tial metaphor is a powerful one in human infor-
drawn from published articles are usually taken as mation processing and lends itself well to modern
unstructured. A first task in viewing such data in in- distributed computing environments such as the
formation space terms is to impose structure on the Web. The Kohonen self-organizing feature map
data. People often use the term coding for assign- (SOFM, originally developed by T. Kohonen who
ing numerical values to keywords or reformulating works in Helsinki, Finland) method is an effective
originally numerical data. A second task is nor- means toward this end of a visual information re-
malization, that is, ensuring that the data are con- trieval user interface.
sistent and inherently comparable. Some of our Our emphasis is on both classical approaches to
measurements cannot shout louder than others. data analysis and recent approaches that have proved
Subsequent tasks are defined by our objectives in their worth in practical and operational settings.
seeking information. A large range of statistical, neu- Such algorithms start with data taken from real
ral network, machine learning, or data mining ap- life, which is multi-faceted or multidimensional.
proaches to processing data could be relevant. Then an algorithm such as principal components
analysis projects multidimensional input data (ex-
pressed as a set of vectors or points) into a more prac-
Visualization-Based User Interfaces tical and observable low-dimensional space, which
Visualizing information and data stored in databases
in practice is usually the best-fitting plane. PCA is
or in unstructured or semistructured repositories is
implemented using linear algebra, where the
important for the following reasons:
eigenvectors (i.e. mathematically most important
1. It allows the user to have some idea before sub- underlying facets of the multidimensional cloud of
mitting a query as to what type of outcome is points under investigation) of the covariances or cor-
possible. Hence, visualization is used to sum- relations (expressing a large set of pairwise rela-
marize the contents of the database or data col- tionships between the multidimensional data points)
lection (i.e., information space). serve to define a new coordinate system. Corres-
2. The users information requirements are of- pondence analysis is similar to PCA. It is particularly
ten fuzzily defined at the outset of the infor- suitable for data in the form of frequency counts
mation search. Hence, visualization is used to or category memberships (e.g., frequencies of oc-
help the user in information navigation by sig- currence in a set of discrete categories); on the other
naling related items, by showing relative den- hand PCA is suitable for continuously changing
sity of information, and by inducing a (possibly measurement values in our input data. Like PCA and
fuzzy) categorization on the information space. correspondence analysis, multidimensional scaling
3. Visualization can therefore help the user before also targets a best-fitting low-dimensional space (e.g.,
the user interacts with the information space best planar or cartographic fit, rather like a street
map). Multidimensional scaling takes all possible
ranks as input and owes its origins to application
INFORMATION SPACES Representations, most often domains where ranks are easier to define compared
spatial or similar (e.g. locations and interconnections in to more precise measurement. Examples of where
graph form), that facilitate can facilitate finding and us- ranks are more easily obtained include perceptual
ing information. studies in psychology and aptitude studies in
education.
HTML page. The image file itself is created and links

made active in it using a batch script on the
server side (i.e., the users communicating partner,
which is serving content to the user) and pro-
vided to the user as an inlined image (i.e., in-
serted in the HTML page served to the user). The
second input data framework is based on linkage
data. A multigraph (i.e. a graph with possibly more
than one link between any pair of node or object)
of links connecting three or more types of objects
is first created. Such objects include author name,
publication title, and content-related keyword (e.g.,
astronomical object name). In this case a Java
(the programming language often used for web soft-
ware applications) application is used to construct
the interactive cartographic representation in real
time. (See Figure 1.)
Figure 1 shows a visual and interactive user in-
FIGURE 1. Visual interactive user interface to the jour- terface map using a Kohonen self-organizing feature
nal Astronomy and Astrophysics based on eight thou- map. The original of this map is in color, and it is
sand published articles. Relative intensity represents show n here in monochrome. Relative color
relative document density. intensitybrightnessis related to density of doc-
ument clusters located at regularly spaced nodes of
the map, and some of these nodes/clusters are an-
notated. The map is deployed as a clickable image
The Kohonen map achieves a similar result map, allowing activation on user command of CGI
through iterative optimization, a different algorithm (Common Gateway Interface: protected web server
that has implications for practical deployment in area for the support of executable programs) pro-
that it is usually slower. Importantly, the Kohonen grams accessing lists of documents andthrough
map output is highly constrained: Rather than a con- further linksin many cases the full documents.
tinuous plane, it is instead (nearly always) a regu- Such maps are maintained for thirteen thousand ar-
lar grid. A regular grid output representation space ticles from the Astrophysical Journal, eight thousand
offers an important advantage in that it easily pro- from Astronomy and Astrophysics, and more than two
vides a visual user interface. In a Web context it thousand astronomical catalogues.
can be made interactive and responsive. In Figure 1 strongly represented concepts, span-
ning wide areas of observational astronomy, are shown
in bold: ISM = interstellar matter, COSMO = cos-
Kohonen Self-Organizing mology, MHD = magneto-hydrodynamics, SUN,
STARS, GALAXIES. MHD processes exist in the sun,
Feature Maps which in turn is a star. Cosmology research is usu-
Two visual user interface frameworks are based ally galaxy oriented. This divide between solar system
on two types of input data. The first input data research and cosmology research outside the solar sys-
framework is based on document/index term de- tem is fairly well represented in this map. Near
pendencies. From such data a cartographic repre- COSMO we see terms such as gravitational lensing,
sentation is periodically (i.e., in batch mode as redshift, dark mattergamma ray bursts, and so on. This
opposed to real-time) updated and made available annotation of the map was informed by the keywords
as a clickable image map contained in an used to construct the map and was carried out
manually. Consideration was given to automated an- Summarizing the types of information space dis-
notation but left for future work. play that we have covered, we can distinguish be-
Portable computing platforms allow for a new tween the following two types of input for maps of
area of application of visual user interfaces, given the information spaces. Both are potentially of relevance
normal mode of interaction using a stylus. Not all for data with varying degrees of structure, including
Web browsers on PDA (personal digital assistant) data originating in databases and in HTML Web files.
platforms support image maps. One that does is the Keyword based: The bibliographic maps exem-
FireViewer application running on a Palm operat-
plified in Figure 1 are of this type. The keywords
ing system (OS) platform.
or index terms provide the dimensions of a geo-
metric space in which our objects are located.
Sparse graph: This is likely to be the case
Related Kohonen-Type Maps whenever XML richer link functionality, com-
Hyperlink-rich data present an interesting case for
pared to the relatively limited forward links sup-
taking a visualization tool further. The extensible
ported by HTML, is used as the basis for
markup language (XML) format is more appropri-
associations between our objects.
ate than the more mundane hypertext markup
language (HTML). The latter is limited in document If an interdependency graph containing a great
linkage and supports little description detail. HTML amount of data is available, a convenient way to
most notably lacks any special support for document process such data is to project the objects, using these
structure. interdependencies, into a geometric space. We can
Essentially such a visualization tool is a Web do this using principal coordinates analysis, which
browser with specialized functionality. The proto- is also referred to as classical multidimensional scal-
type of one such tool was developed for XML data. ing and metric scaling.
The data related to astronomers, astronomical ob-
ject names, and article titles. They were open to the
possibility of handling other objects (images, sum- Ontologies: Support for Querying
mary tabulations, etc.). Through weighting, the var- We have noted that the mathematical notions of geo-
ious types of links could be given priorities. An metric spaces can be found behind many aspects
algorithm was developed to map the nodes (objects) of how we think about information spaces and be-
to a regular grid of cells, which were clickable and hind many approaches to displaying and visualizing
provided access to the data represented by the clus- information spaces. However, having information
ter. Given the increasingly central role of XML in ac- spaces cooperate and collaborate such that a user can
cess to Web information and data, the importance draw benefit from more than one information space
of such clustering for data organization and for at one time leads us to requirements of a different
knowledge discovery can be underscored. sort. For interoperability of information systems
Such an interactive visual user interface works of any type, we need to consider a common language
in the following way. We consider a set of documents. to support formal querying or more informal search-
The units clustered are authors, titles, and astro- ing for information resources. A term that has come
nomical objects. The map is arranged to give a to be much in vogue of late is ontologythe termi-
central position to a selected unit (e.g., a person nology underpinning a common language. An on-
an astronomer). The annotations of paper titles or tology lists the terminology used in a particular area
of astronomical objects shown in the regular grid are (e.g. a particular field of business, or engineering)
representative or important ones. Clicking on a lo- and some of the relationships between these terms.
cation provides considerably greater detail in an ad- Its aim is to help user searching, since the salient in-
ditional panel relative to what is presented in a formation aspects of an area are essentially summa-
global visual view in the clickable visual interface. rized by such an ontology.
Ontology describes a terminology hierarchya This Eurostat database supports three concept hier-
helpful basis for supporting querying. Such a con- archies: branches, themes, and countries. The total
cept hierarchy defines a sequence of mappings from number of branch concepts is 423, the total num-
a set of low-level concepts to higher-level, more gen- ber of theme concepts is 30, and the total number
eral concepts. These concepts may be defined within of country concepts is 23. An example extract of
two structures. A hierarchical structure is such that the branch concept hierarchy is shown as follows.
a so-called child node cannot have two parents,
but a parent node can have more than one child node. Total Industry
A lattice structure is such that a child node can have Total industry (excluding construction)
two parents. A concept hierarchy can be explicitly Mining, quarrying and manufacturing
generated by expert users before the data are queried Intermediate goods industry
and will be static, or it can be generated automati- Energy
cally, and the user may reform the hierarchy when Intermediate goods industry, excluding
needed. industry
The concept hierarchy can be based on hierar- Capital goods industry
chical structure and generated by, for example, econ- Consumer goods industry
omists. This will be illustrated by a case study using Durable consumer goods industry
Eurostat (the Statistical Office of the European Union, Non-durable consumer goods industry
the prime business and economic statistical data Mining and quarrying
agency at European level) databases and in particu- Mining and quarrying of energy producing
lar documents from the Eurostat Economic Bulletins. materials
Key National
Accounts Labor Costs Documents Related to Climate
and and
Germany International Germany
and
R&D
F I G U R E 2 . The distribu-
Investment tion of topics found as a
result of a query discussed
in the article and based on
an ontology for this do-
Editorial
Labor Cost main of economic infor-
and mation.
East and
Employment
West Germany Note: GB = Great Britain
and
Great Britain
Economy and
German and France
Economy Asia
Micro-Macro
Economy
and
East Germany
Mining and quarrying except energy produc- compressing of compressed medical or other imagery
ing materials and zooming to locations of interest; Web browser
Manufacturing scrolling; accessing information content such as one
Electricity, gas and water supply or a number of video streams based on user inter-
Construction est as manifested by his or her eye-gaze dwell time;
and manipulating objects in two-dimensional and
A query is formed from the cross-product of the
three-dimensional spaces. Such interaction with com-
entries of the concept hierarchy categories. For
puter displays is certainly feasible, as we have shown.
example:
However, issues related to user acceptability and er-
Branch: Total Industry, Energy gonomics have yet to be fully investigated.
Theme: Production
Country: Germany, United Kingdom Fionn Murtagh
In this case the result query will be:
Total Industry and Production and Germany See also Data Visualization; Information Retrieval;
Total Industry and Production and United Ontology
Kingdom
Energy and Production and Germany
Energy and Production and United Kingdom FURTHER READING
We can seek documents having at least one of Benzcri, J.-P. (1992). Correspondence analysis handbook. Basel,
the preceding combinations. (See Figure 2.) Figure Switzerland: Marcel Dekker.
2 shows in schematic form the type of result that we Doyle, L. B. (1961). Semantic road maps for literature searchers. Journal
of the ACM, 8, 553578.
may find. A possibly important consideration in such Farid, M., Murtagh, F., & Starck, J. L. (2002), Computer display con-
work is that the information resulting from a user trol and interaction using eye-gaze. Journal of the Society for
query be built on the fly into an interactive graphi- Information Display, 10, 289293.
cal user interface (GUI). Guillaume, D., & Murtagh, F. (2000). Clustering of XML documents.
Computer Physics Communications, 127, 215227.
Hoffman, P. E., & Grinstein, G. G. (2002). A survey of visualizations
Outlook for high-dimensional data mining. In U. Fayyad, G. G. Grin-
stein, & A. Wierse (Eds.), Information visualization in data mining
Summarizing information is necessary. A visual sum- and knowledge discovery (pp. 4782). San Francisco: Morgan
mary is often a natural way to summarize. When we Kaufmann.
add the possibilities for human-computer interac- Kohonen, T. (2001). Self-organizing maps (3rd ed.). New York: Springer-
Verlag.
tion, visual user interfaces become a toolset of Murtagh, F., & Heck, A. (1987). Multivariate data analysis. Dordrecht,
significance. Netherlands: Kluwer.
The ergonomics of interaction based on visual Murtagh, F., Taskaya, T., Contreras, P., Mothe, J., & Englmeier, K.
user interfaces is still under investigation. Although (2003). Interactive visual user interfaces: A survey. Artificial
Intelligence Review, 19, 263283.
human understanding is greatly aided by maps Oja, E., & Kaski, S. (1999). Kohonen maps. Amsterdam, Elsevier.
and drawings of all sorts, we have yet to find the most Poinot, P., Murtagh, F., & Lesteven, S. (2000). Maps of information
appropriate visual displays for use in visual user spaces: Assessments from astronomy. Journal of the American Society
for Information Science, 51, 10811089.
interfaces. Shneiderman, B. (2002). Leonardos laptop: Human needs and the new
Beyond the uses of information space visualiza- computing technologies. Cambridge, MA: MIT Press.
tion described here, we can consider other modes Torgerson, W. S. (1958). Theory and methods of scaling. New York:
of interacting with information spacesthrough ges- Wiley.
Wise, J. A. (1999). The ecological approach to text visualization. Journal
ture or voice, for example. Operations that can be of the American Society for Information Science, 50, 12241233.
controlled by eye-gaze dwell time (manifesting
continuing subject interest through the approximate
point of gaze remaining fairly steady) include de-
INFORMATION THEORY 377
into a signal, which is sent along a channel affected

INFORMATION THEORY by a noise source. The received signal (transmitter
signal plus noise) enters the receiver, which decodes
Information theory generally refers, especially in the it and converts it into a message for the destina-
United States, to a theory of communication origi- tion. In a local area Ethernet, for example, a user
nated by mathematician Claude E. Shannon at Bell (information source) types in an e-mail message
Telephone Laboratories in 1948 and developed to send to a second user (destination). The first
further by him and others. The highly abstract, math- users computer system (transmitter) encodes
ematical theory has been influential, and sometimes the message and converts it into an electrical sig-
controversial, in digital communications, mathe- nal that is sent over the computer network (chan-
matics, physics, molecular biology, and the social and nel), where electrical noise (noise source) can affect
behavioral sciences. More broadly, information the- it. The second users computer system (receiver) picks
ory refers to both Shannons approach and other up the signal, decodes it, and displays a message
probabilistic methods of analyzing communication hopefully what was sent.
systems. Other theories of information are common The hallmark of Shannons theory is a quan-
in statistics and information science. titative measure of information, which he adapted
from the work of Bell Labs researcher Ralph
Hartley. In 1928 Hartley proposed a simple loga-
Shannons Theory and rithmic measure of information transmitted in a
digital system that did not consider noise or the
Its Development probability of selecting symbols. Drawing on his
Shannons theory, comprehensively established own work in cryptography during World War II,
in A Mathematical Theory of Communication Shannon went beyond Hartley to treat the statis-
(Shannon 1993) uses probability theory and the tical aspects of messages, noise, and coding prob-
concept of entropy to determine how best to en- lems comprehensively. He related the amount of
code messages in order to transmit information ef- information generated by a source, assumed to
ficiently and reliably in the presence of noise. be statistically regular, to the uncertainty and choice
The theorys general model (Figure 1) applies involved in selecting messages from an ensemble
to any communication system, whether it involves of messages. The greater the uncertainty and choice,
machines, humans, other living things, or any com- the more information the source produces. If there
bination of these. An information source selects is only one possible message, there is no infor-
a message from a set of possible messages. The mation; maximum information exists when the
transmitter encodes that message and converts it choice of messages (or symbols) is random.
IN F O R M ATI O N
SOURCE TRANSMITTER RECEIVER DESTINAT I ON
SIGNAL RECEIVED
SIGNAL FIGURE 1. Schematic di-
MESSAGE MESSAGE agram of a general com-
munication system
NO ISE
SOURCE
Information in this sense does not measure data or bilities of when these symbols occur to calculate
knowledge; the meaning of messages is irrelevant to the entropy of the source as bits per symbol, and
the theory. then devise a code to match that entropy: 0, 10,
Throughout the theory, uncertainty and choice 110, 111. The new code requires only bits per
are expressed by equations of the form symbol to encode messages in this language, ver-
sus 2 bits per symbol, a compression ratio of .
H = p log2 p [for the digital case] No method uses fewer bits per symbol to encode
H = p log p [for the analog case] this source.
Despite many claims to the contrary, Shannon
where p is the probability of occurrence of an did not avoid the term information theory. He
eventfor example, a symbol. The expressions are used it, for example, in the titles of early talks on
mathematically analogous to physical entropy in the subject, in the original paper and other early
statistical mechanics. In the digital case, the en- papers, in an encyclopedia article, and in an edi-
tropy of a source gives the average number of torial (Shannon 1993).
bits per symbol or bits per second required to Much research has strengthened and extended
encode the information produced by the source. Shannons theory. In the 1950s, mathematicians
Channel capacity, defined as the maximum value reset the theorys foundations by rigorizing proofs
of source entropy minus the uncertainty of what of the main theorems and extending them. The
was sent, gives the maximum rate of transmis- theory has been further developed in such areas as
sion of information in bits per second. The paper error-correcting codes, rate distortion theory (lossy
seems to mark the first appearance of the term bit data compression), multiuser channels (network
(a contraction of binary digit) in print. information theory), and zero-error channel ca-
A prominent feature of the theory is the cod- pacity (zero-error information theory).
ing theorem for a noisy digital channel. This states
the surprising result that if the entropy of a source
is less than channel capacity, a code can be devised Other Theories of Information
to transmit information over such a channel When Shannon published his path-breaking paper
with an arbitrarily small error. The tradeoffs are in 1948, it vied with three other mathematical the-
complex codes and long delays in the transmitter ories of information. The most established was the
and receiver. The upper bound of transmission theory of estimation of British statistician and ge-
in an important analog case is neticist Ronald Fisher. In the 1920s and 1930s, Fisher
defined the amount of information to be expected,
C = W log (1 + P/N)
with regard to an unknown statistical parameter,
where C is channel capacity, W is bandwidth, P is from a given number of observations in an experi-
average transmitter power, and N is the average ment. The measure was mathematically similar to,
power of white thermal noise. These and other the- but not equivalent to, that for entropy. In 1946 British
orems establish fundamental limits on data com- physicist Denis Gabor defined a quantum of infor-
pression (encoding) and transmitting information mation, expressed in logons, in terms of the prod-
in communication systems. uct of uncertainties of time and frequency of an
A simple coding example for a noiseless digi- electrical signal. Gabor used the concept to analyze
tal channel, drawn from the paper, illustrates waveforms in communication systems. The third
Shannons approach. Consider a language with just theory was proposed by American mathematician
four symbols: A, B, C, D. Symbol A occurs of the Norbert Wiener in his well-known book Cybernetics
time, B of the time, C and D of the time each. published in 1948. Wiener independently derived a
A direct method to encode these symbols uses 2 measure of information similar to Shannons, except
bits per symbol, for example, as 00, 01, 10, and 11, that he defined it as negative rather than positive en-
respectively. Alternatively, one can use the proba- tropy. It thus measured order rather than disorder
INFORMATION THEORY 379
(uncertainty). Contemporaries often referred to the form a subfield of research in the IEEEs Information
entropy concept of information as the Shannon- Theory Society. Shannon Theory, as it has been
Wiener measure or the Wiener-Shannon formula. called since the early 1970s, has remained at the cen-
At the first London Symposium on Information ter of the discipline. The IEEE honors Shannon as
Theory, held in 1950, British physicist Donald the founder of Information Theory, and many text-
MacKay brought all three measures into a unified books view his approach as nearly synonymous with
Information Theory. MacKay included Fishers the topic.
and Gabors work under the new category of Following a suggestion made in 1949 by Warren
Scientific Information Theory, the realm of the physi- Weaver, an American mathematician who directed
cist, and Shannons and Wieners work under Com- the natural sciences division of the Rockefeller Foun-
munication Theory, the realm of the engineer. dation, numerous researchers have tried to make
However, MacKays efforts did not resolve the dif- Shannons theory the basis for a semantic theory
ferent meanings of information or information of information. These range from the highly math-
theory. Shannons followers, especially those in ematical theory of logicians Rudolf Carnap and
the United States, employed information theory Yehoshua Bar-Hillel in the early 1950s to numerous
exclusively to describe his approach. Yet the Pro- quantitative and non-quantitative attempts by work-
fessional Group on Information Theory in the ers in the interdisciplinary field of Information
Institute of Radio Engineers, founded in 1951 and a Science. None have gained the scientific status of the
forerunner of the present-day Information Theory non-semantic theories.
Society of the Institute of Electrical and Electronics
Engineers (IEEE), considered both Shannons the-
ory and Wieners theory of prediction and filtering Inuence of Information Theory
to be in its purview. Information theory became something of a fad in
Shannon himself included the latter field in scientific circles in the 1950s when numerous re-
Information Theory, an article he wrote for the searchers enthusiastically applied the new sci-
Encyclopedia Britannica in the mid-1950s. British ence to a variety of fields. These included physics,
electrical engineer Colin Cherry observed in 1957 artificial intelligence, behavioral and molecular bi-
that the research of physicists such as MacKay, Gabor, ology, physiology, experimental and cognitive psy-
and Leon Brillouin on scientific method,is referred chology, linguistics, economics, organizational
to, at least in Britain, as information theory, a term sociology, and library and information science.
which is unfortunately used elsewhere [that is, in the Ironically, communication engineers were skeptical
United States] synonymously with communica- until the 1960s, when Shannons theory was used
tion theory. Again, the French sometimes refer to to encode messages in deep space communications.
communication theory as cybernetics. It is all very Although most applications, adaptations, and
confusing (Cherry 1957, 216). modifications outside of mathematics and engi-
Subsequently the mathematical and electrical en- neering proved to be unfruitful, the language of
gineering communities in the United States viewed information theory became ingrained in such fields
these interpretations of information as comple- as molecular biology (gene as carrier of informa-
mentary concepts, not as competitors, a position that tion), economics (markets as information proces-
holds today. Gabors measure is now prominent in sors), and artificial intelligence (semantic information
electrical circuit theory, Fishers in classical statistics. processing). The biological and behavioral sciences
Wieners work on prediction and filtering defines the describe the operation of all forms of life, from the
area of statistical communication theory. Although DNA molecule to society, in terms of information
it was a core discipline of information theory for transfer, storage, and processing. Technical appli-
many years, Wieners theory moved outside the main cations have proven themselves in physiology (for
stream of information theory in the 1960s (Viterbi instance, the informational capacity of sense organs)
1973, 257). Yet signal detection and estimation still and experimental psychology ( for instance, the
relation between the amount of information in a Fisher, R. A. (1935). The design of experiments. London: Oliver & Boyd.
stimulus to the response time to the stimulus). A re- Gabor, D. (1946). Theory of communication. Journal of the Institution
of Electrical Engineers, Pt. III, 93, 429459.
cent textbook notes that information theory inter- Hartley, R. V. L. (1928). Transmission of information. Bell System
sects physics (statistical mechanics), mathematics Technical Journal, 7, 535563.
(probability theory), electrical engineering (com- Kay, L. (2000). Who wrote the book of life? A history of the genetic code.
munication theory) and computer science (algo- Palo Alto: Stanford University Press.
Machlup, F., & Mansfield, U. (Eds.). (1983). The study of information:
rithmic complexity) (Cover and Thomas 1991, 1). Interdisciplinary messages. New York: Wiley.
Applications of Shannons theory to information MacKay, D. M. (1969). Information, mechanism, and meaning.
technology increased dramatically following the in- Cambridge: MIT Press.
Shannon, C. E. (1993). A mathematical theory of communication. In
vention of the microprocessor in the 1970s and in- N. J. A. Sloane & A. D. Wyner, (Eds.), Claude Elwood Shannon, col-
creasing levels of semiconductor integration. lected papers (pp. 583). New York: IEEE Press. (Original work
Complex error-correcting codes and data compres- published 1948)
sion schemes are pervasive in digital communica- Slepian, D. (1973). Information theory in the fifties. IEEE Transactions
on Information Theory, 19(2), 145148.
tions. They help make possible such technologies as Slepian, D. (Ed.). (1973). Key papers in the development of Information
hard-disk drives, high-speed memories, cell phones, Theory. New York: IEEE Press.
compact discs, DVDs, digital television, audio and Verd, S. (Ed.). (1998). Information Theory: 19481998. IEEE
video compression, and video conferencing on the Transactions on Information Theory, 44(6), 20422272.
Viterbi, A. J. (1973). Information theory in the Sixties. IEEE Transactions
Internet. on Information Theory, 19(3), 257262.
Perhaps the most pervasive influence of infor- Weaver, W. (1949). Recent contributions to the mathematical the-
mation theory has been indirect. Social theorists from ory of communication. In C. E. Shannon & W. Weaver, The math-
ematical theory of communication (pp. 93117). Urbana: University
Marshall McLuhan in the 1960s to Manuel Castells of Illinois Press.
in the 1990s, drawing on the popularization and wide Webster, F. (1995). Theories of the information society. London:
application of information theory, have helped Routledge.
create a public discourse of information that pro- Wiener, N. (1948). Cybernetics: Or control and communication in the
animal and the machine. Cambridge and New York: Technology
claims the dawning of an information age, economy, Press and Wiley.
and society.
Ronald Kline
See also Information Overload; Theory INSTANT MESSAGING

See Chatrooms; Collaboratories; Cybercommunities;
FURTHER READING E-mail; Groupware; Internet in Everyday Life; Social
Psychology and HCI
Aspray, W. (1985). The scientific conceptualization of information:
A survey. IEEE Annals of the History of Computing, 7, 117140.
Attneave, F. (1959). Applications of information theory to psychology:
A summary of basic concepts, methods, and results. New York: Henry
Holt.
Capurro, R., & Hjorland, B. (2003). The concept of information. Annual
INSTRUCTION MANUALS
Review of Information Science and Technology, 37, 343411.
Cherry, C. (1957). On human communication: A review, a survey, In the world of computing, an instruction manual
and a criticism. Cambridge: MIT Press. is a book that explains how to use a software pro-
Cover, T. M., & Thomas, J. A. (1991). Elements of information theory.
New York: Wiley. gram or a hardware device. If we accept a relatively
Dahling, R. L. (1962). Shannons Information Theory: The spread broad definition of manual, we can say that manu-
of an idea. In Stanford University, Institute for Communication als dealing with such technologies as architecture,
Research, Studies of Innovation and of Communication to the Public
(pp. 117139). Palo Alto: Stanford University Press.
mining, and agriculture have been written for many
Edwards, P. N (1996). Closed world: Computers and the politics of centuries. Manuals that approximate contemporary
discourse in Cold War America. Cambridge: MIT Press. manuals in their design generally date from the nine-
INSTRUCTION MANUALS 381
teenth century. The first author of computer man- to computer users. Little by little, however, help
uals was J. D. Chapline, who worked on the Binac systems began to compete successfully with print
and Univac I computers from 1947 to 1955. Chapline documentation.
borrowed many design ideas from military manuals
and, later, from automotive manuals.
Interest in computer documentation increased Kinds of Computer Manuals
with the growth of the computer industry. This in- It is important to understand the various kinds of
terest intensified considerably when personal com- computer instruction manuals and how each kind
puters (microcomputers) burst onto the scene in is designed. The largest class of manuals are those
the early and mid 1980s. Prior to personal comput- that document software applications. These manu-
ers, most computer users were either computer pro- als are usually divided into three categories: tutori-
grammers and other computer specialists or scientists als, users guides, and reference manuals. These three
and engineers who were prepared to master a new and genres of instruction manuals are meant to work to-
daunting technology. Now, however, a much larger gether to support the users complete cycle of prod-
and more diverse audience, including business people, uct use from installation and initial learning, through
graphic designers, students, and hobbyists were using ongoing use, to returning to the product or some
computers and reading documentation. product feature after time has passed. Other signif-
Computer companies, heeding widespread com- icant classes of manuals are programmer language
plaints about confusing and tedious manuals, began manuals and hardware manuals.
to regard clear and engaging documentation as an
important aspect of their products and a competi- Tutorials
tive advantage in the marketplace. Industry leaders Tutorials are intended to introduce users to the prod-
such as IBM, Apple, and Microsoftalong with uni- uct. They are slow paced and provide detailed ex-
versities and other organizationsdeveloped us- planations and instructions along with numerous
ability testing programs and undertook research screen captures (images of how the screen will look)
studies. These and other computer companies issued that show how the system will respond to the
insightfully designed, expensively produced manu- users actions.
als with refined page layout, ample illustrations, and Tutorials are organized as a series of lessons in
color printing. Similarly, corporations in a wide range which the designer carefully chooses the most im-
of industries improved their internal documenta- portant features of the product and explains the fea-
tion to achieve greater efficiency, although budgets tures in the sequence that will result in the most
for internal documentation have traditionally been effective learning. Rarely does a tutorial cover every-
lower than for product documentation. thing the product can do. The lessons are generally
Commercial publishers discovered that many built around extended examples (often called sce-
people would pay for a better manual than the one narios). For example, the tutorial for a database man-
shipped with the product or for a manual that met agement system might guide the user through
specific needs. A new publishing business in third- building a database to keep track of the wine in a wine
party computer books emerged. Software compa- cellar. Tutorials are written in a friendly, conversa-
nies assisted third-party publishers, recognizing that tional manner. In many tutorials, the user is encour-
customers value software products with strong third- aged to tackle a new lesson and is congratulated at
party book support. the end of the lesson. Having completed the tutorial,
During this period most software products in- the user will presumably graduate to the users guide
cluded some kind of online (on-screen) help system. and will be able to learn whatever additional features
Online help, however, was often crudely designed. are required to accomplish the users own tasks.
Furthermore, the appearance and operation of help Print tutorials have drawbacks for both computer
systems varied greatly from one software product to companies and users. They are expensive to write,
another, and the whole idea of online help was alien and because they have a high page count, they are
expensive to print and ship. While some users rely group of related procedures may be introduced by
on tutorials, others do not have the patience for slow- overview paragraphs, often located at the beginning
paced documentation. Another significant problem of a chapter.
(which can be avoided in online tutorials) is it takes Writing users guides poses many challenges. First,
just a single mistake to put the user can out of sync while almost everyone in the world of computer doc-
with the tutorial. Still another problem is the rele- umentation embraces the idea of carefully study-
vance of the scenario to the users work (transfer ing the products users and writing user-centered,
of learning). A tutorial dealing with a wine cellar task-oriented documentation, these are not easy
database may not be very useful to someone in- things to accomplish. There are major challenges
tending to build a database to log laboratory data. in learning how the users of a product understand
Finally, while a well-written tutorial can lead the user their work, what background knowledge and men-
successfully through tasks, it is not clear how much tal models they bring to bear, and the cognitive
of this knowledge is retained. processes they employ when using the product, par-
ticularly when they encounter difficulties.
Users Guides Among the many thorny design issues is how
When most people think about a computer manual, much detail to include in procedures. Too much in-
they are thinking of a users guide. This is the cen- formation is tedious; insufficient information leaves
tral piece of the print documentation set, the piece users puzzled or unable to carry out tasks. Hitting
that will be used most of the time. If only one man- the proper balance is difficult, especially for a
ual is provided, it is very likely a users guide. product that is used by a wide range of individu-
The users guide consists of procedures (inals. These kinds of problems apply to online as
structions) for carrying out all the tasks (or at least well as print documentation.
all the mainstream tasks) that can be performed with
the product. Technical communicators carefully or- Reference Manuals
ganize users guides for maximum usefulness. Broadly Whereas users guides are organized by tasks, refer-
speaking, the sequence of the chapters corresponds ence manuals are organizedoften alphabetically
to the sequence in which users are likely to carry out by the names of commands. The manual describes
tasks. So, for instance, the chapter on creating a new each commands purpose and options along with
document will precede the chapter on printing. the location of the command in the applications
Highly specialized chapters come toward the end. menu structure and the keyboard shortcut for exe-
Within a chapter, basic tasks precede specialized tasks. cuting the command.
As much as possible, procedures are written as Reference documentation assumes a sophisti-
independent modules that the user can consult in cated user who understands the task he or she wishes
any order. The writing style is straightforward, with- to carry out and who can identify the commands
out scenarios or motivational comments. Examples that are necessary. Often people consult reference
are brief and appear within an individual procedure. manuals for a review of commands they have used
Procedures generally consist of a title, a concep- in the past.
tual element, and numbered steps. The title identi-
fies the procedure. The conceptual element (usually
a paragraph or two) makes clear the purpose of Programming Language
the procedure and, if necessary, provides such in-
formation as the prerequisites that must be met Documentation
before the procedure can be carried out. The steps Manuals for documenting computer languages take
are the actions that users will take and, at times, the form of tutorials and references. Tutorials ex-
descriptions of how the system will respond to these plain the basic concepts of the programming lan-
actions. If the purpose is clear from the title, there guage and guide the programmer through the
may be no reason for the conceptual element. A creation of simple programs. Code samples and
INSTRUCTION MANUALS 383
explanations of likely errors are included in these tu- It is also true that online help has greatly ma-
torials. References are the heart of programmer doc- tured, and a strong argument can be made that the
umentation. They explain the statements, variables, needs of users are better served by online help. While
operators, and other constructs of the programming print manuals are more legible than on-screen doc-
language. Syntax diagrams and code examples are umentation, online help provides faster access to in-
generally included. formation. Clicking links is faster than turning pages,
and when help content is integrated with the ap-
plication (context sensitivity), users can instantly
Hardware Manuals display information pertinent to the portion of the
Hardware manuals vary greatly because computer interface they are working with. Other advantages
hardware encompasses hand-held devices, standard of online help include the ability to deliver anima-
desktop computers, mainframes, computer compo- tion, audio, and video, and to allow the user to di-
nents, and more. When hardware devices include rectly access more detailed content stored on the
built-in software, a display, and a keypad, the hard- vendors support website. Even so, many users re-
ware documentation may resemble a software man- main loyal to print manuals.
ual. Hardware manuals also explain, with illustrations
as well as text, the procedures for setting up, main-
taining, and repairing the device. Users may well Acrobat as a Compromise Solution
be unscrewing plates, connecting plugs, installing Adobe Acrobat enables publishers of paper docu-
circuit boards, and measuring voltages. Hardware ments to distribute these documents as computer
documentation is much less likely to be delivered files with confidence that the formatting will be pre-
online, if only because online presentation is gener- served regardless of the recipients computer system
ally unavailable when the user is assembling or start- and printer. Some software vendors, therefore, pre-
ing the device or when the device is malfunctioning. pare handsomely formatted manuals and distribute
them free of charge (often from a tech support web-
site) in Acrobat (PDF) format. Users can have printed
The Rise of Online Help manuals if they accept the trouble and cost of the
Starting in the early 1990s, computer companies be- printing.
gan reducing the size of the print documentation set Furthermore, with such features as a clickable
and placing greater emphasis on the online help sys- table of contents, text search, thumbnail images of
tem and other forms of on-screen documentation. pages, and the ability to add notes and highlights,
Now products are likely to ship with only a users Acrobat makes it possible to create manuals that can
guide or a brief guide to getting startedor they may be used effectively on the computer screen. On-screen
ship with no print at all. If there is a tutorial, it is PDF manuals cannot be integrated with the appli-
probably an online tutorial, very possibly delivered cation, but users do get many of the features of a help
over the Web. system, along with book-like page layout.
This change was driven in large part by the need Computer companies now have various ways to
to control costs. Another reason is the desire to provide product documentation. They will make
streamline the product development cycle. Once a choices based on the nature of their products, the
manual is finished, there is often a six-week lag while preferences of their customers, and cost factors.
the manual is at the printer. Not only cant the prod- Instruction manuals, both comprehensive and scaled
uct be shipped while the manual is being printed, down, will share the stage with online help sys-
but if last-minute corrections to the product code tems, and these instruction manuals will be shipped
change the products appearance and behavior (for with products, printed by users from PDF files,
instance, if a buggy feature is modified or removed), utilized as on-screen PDF manuals, and purchased
the documentation cannot be updated to reflect these in the form of third-party books. If we continue to
changes. learn more about human-computer interaction and,
in particular, how people use computer documen-

tation, and if computer companies and third-party INTERNETWORLDWIDE
publishers strive to create the best documentation
they can, computer use will become a more pro- DIFFUSION
ductive and enjoyable experience.
The Internet grew rapidly in the 1990s. Although we
David K. Farkas are still far from universal Internet connectivity, a
large percentage in the developed world had access
See also Adaptive Help Systems to the Internet by the end of 2003, as did signifi-
cant percentages of people in developing coun-
tries. Among the factors affecting the rate of diffusion
FURTHER READING of Internet around the globe are economics, gov-
ernment policies, and the original dominance of the
Agricola, G. (1950). De re metallica (H. C. Hoover & L. H. Hoover, English language on the Web.
Trans.). New York: Dover.
Barker, T. T. (2003). Writing software documentation: A task-oriented
approach (2nd ed.). New York: Longman.
Brockmann, R. J. (1998). From millwrights to shipwrights to the twenty- The Growth of the Internet
first century: Explorations in the history of technical communication Although there are no reliable data on the size of the
in the United States. Creskill, NJ: Hampton Press.
Carroll, J. M. (1990). The Nurnberg funnel: Designing minimalist in- worlds online population, estimates show that use
struction for practical computer skills. Cambridge: MIT Press. of the Internet has diffused rapidly. The number of
Carroll, J M. (Ed.). (1998). Minimalism beyond the Nurnberg funnel. Internet users around the globe has surged from
Cambridge: MIT Press.
Farkas, D. K. (1999). The logical and rhetorical construction of pro-
an estimated 4.4 million in 1991 to 10 million in
cedural discourse. Technical Communication, 46(1), 4254. 1993, 40 million in 1995, 117 million in 1997, 277
Hackos, J. T., (1994). Managing your documentation projects. New York: million in 1999, 502 million in 2001, to more than
Wiley. 600 million in 2002. Thus, the global penetration
Haramundanis, K. (1997). The art of technical documentation (2nd
ed.). Woburn, MA: Butterworth-Heinemann.
rate of the Internet has increased from less than a
Haydon, L. M. (1995). The complete guide to writing and produc- tenth of a percent in 1991 to 2 percent in 1997, 7 per-
ing technical manuals. New York: Wiley. cent in 2000, to over 10 percent of the total world
Horton, W. (1993). Lets do away with manuals before they do away population in 2002. Projections for 2004 place the
with us. Technical Communication, 40(1), 2634.
Horton, W. (1994). Designing and writing online documentation: number of global Internet users between 700 mil-
Hypermedia for self-supporting products (2nd ed.). New York: Wiley. lion and 945 million. Estimates for the Internets
Jordan, S. (Ed.). (1971). Handbook of technical writing practices. global penetration rate in 2004 are between 11 per-
New York: Wiley-Interscience.
Microsoft Corporation. (2003). Manual of style for technical publica-
cent and 15 percent.
tions. Redmond, WA: Microsoft Press. Despite rapid worldwide diffusion of the Internet,
Price, J., & Korman, H. (1993). How to communicate technical infor- a disproportionate number of users are concentrated
mation: A handbook of software and hardware documentation. in more developed countries, especially the United
Redwood City, CA: Benjamin/Cummings.
Schriver, K. A. (1997). Dynamics of document design. Creating texts for
States. In 2002, 169 million Americans were on-
readers. New York: Wiley. line, accounting for about 60 percent of the coun-
Simson, H., & Casey, S. A. (1988). Developing effective user docu- trys total population and 29 percent of the worlds
mentation: A human-factors approach. New York: McGraw-Hill. Internet population. There were 172 million users
Steehouder, M., Jansen, C., van der Poort, P., & Verheijen, R. (Eds.).
(1994). Quality of technical documentation. Amsterdam: Rodopi. in Europe (28 percent of the worlds Internet popu-
Ummelen, N. (1996). Procedural and declarative information in soft- lation, 182 million in Southeast and East Asia,
ware manuals: Effects on information use, task performance, and including 145 million in China, Japan, and Korea
knowledge. Amsterdam: Rodopi.
Vitruvius, P. (1950). The ten books on architecture (M. H. Morgan,
(23 percent). South America was home to 29 million
Trans.). New York: Dover Publications. users (5 percent), while there were 11 million in
Oceania (1.9 percent), and 10 million in Africa
(1.6 percent).
INTERNET-WORLDWIDE DIFFUSION 385
Countries with the largest population of on- men. Yet the digital divide is not a binary yes-no
line users are the United States with an estimated question of whether basic physical access to the
159 million, China with 59 million, Japan with Internet is available. Access does not necessarily equal
57 million, Germany with 34 million, South Korea use. What matters is the extent to which people reg-
with 26 million, and the United Kingdom with ularly use a computer and the Internet for mean-
25 million Internet users. By proportion of popu- ingful purposes.
lation, Nordic countries top the list, with Iceland, The digital divide is shaped by social factors as
Sweden, Denmark, Finland, and Norway in the top much as technological factors, with systematic vari-
ten. Over half to three-quarters of these countries ations in the kinds of people who are on and off
populations are online. Other countries leading in the Internet. Moreover, the digital divide cuts across
the rate of Internet penetration include the United nations and changes over time. Not only are there
States, Canada, the Netherlands, South Korea, and socially patterned differences within countries in
Singapore. who uses the Internet, there are major differences
To be sure, these data are rough approximations. between countries: the global digital divide. Thus,
Getting a perspective on the Internet is like tracking there are multiple digital divides, varying within
a perpetually moving and changing target. Mean- countries as well as between countries, both devel-
ingful comparisons and knowledge accumulation oped and developing.
are hindered by the lack of comparability of data
from different countries. This has led researchers
studying the global diffusion of the Internet to rely Seven National Examples
on statistics gathered country-by-country that often To show the varied diffusion of Internet use, sum-
employ different measurements. The definition of maries of the development of Internet use in seven
the online population often differs between studies: quite different countries are provided here and com-
Some focus on adult users, while others include chil- pared in Table 1.
dren and teenagers. There is no standard definition
of who is an Internet user: Some studies embrace United States
everyone who has ever accessed the Internet as a user, The United States has always had the largest num-
while others count only those who use the Internet ber of Internet users. In 1997, 57 million Americans
at least once a week. Some studies use households, were using the Internet, representing 22 percent of
not individuals, as the unit of analysis. This also the total U.S. population. The number of users
masks how individuals within a household use the climbed to 85 million in 1998 (33 percent), 116 mil-
Internet. To increase reliability and comparability, lion in 2000 (44 percent), 143 million in 2001 (54 per-
this article relies on data from national representa- cent), and reached 169 million in 2002. Between 1997
tive surveys conducted by government agencies, and 2001, while the number of Americans using com-
scholarly researchers, and policy reports issued by puters increased by 27 percent from 137 million to
international organizations. 174 million, the online population rapidly increased
by 152 percent. By 2001, 66 percent of American
computer users were also online. Likely because
Digital Divides Internet use is so widespread, all forms of digital di-
Internet users are not a random sample of a coun- vides are shrinking in the United States (and Canada).
trys population: They differ from nonusers in so-
cioeconomic status (education, occupation, income, Germany
wealth), age, gender, race or ethnicity, stage in the The Internet penetration rate has risen in Germany
life-course, urban-rural location, and language. This since the mid 1990s. Among the German popula-
digital divide has been present since the onset of tion aged 14 and older, 7 percent used the Internet
computerization and the Internet, when most users in 1997, 10 percent in 1998, and 18 percent in 1999.
were North American, young, well-educated, white Unlike in North America, there was a substantial
gap in Germany between computer ownership and 69.4 million in 2002. The Japanese are the worlds
Internet use as late as 1999, when 45 percent of house- heaviest users of mobile phone-based Internet ser-
holds in Germany owned a computer but only about vices, comprising about one-third of the worlds
one-quarter of those households (11 percent of all users, with the less-educated and women accessing
households) were connected to the Internet. Internet the Internet from Web-enabled mobile phones at a
diffusion has accelerated since then. Twenty-nine per- higher rate than from PCs.
cent of the German population was wired in 2000, 39
percent in 2001, and 44 percent in 2002. Despite the Korea
diffusion of the Internet, the socioeconomic and gen- The number of Internet users in Korea increased
der digital divides are increasing in Germany, as new- threefold between 1998 and 1999, from 3 million to
comers to the Internet are disproportionately men of 11 million. By 2001, 24 million Koreans over 7 years
high socioeconomic status. old (57 percent) were online. The number of Internet
users grew to 26 million by June 2002, a figure nearly
Italy nine times large than the figure from five years ear-
Only five percent of Italian households had Internet lier. There were also 27 million mobile Internet sub-
access in 1998. Italys low Internet penetration rate scribers in Korea in June 2002, although many of
has been associated with low computer ownership. them presumably also had PC-based access. With
But the situation is changing, as Italians have been 14 broadband subscribers per 100 inhabitants in June
rapidly adopting personal computers (PCs) and the 2001, Korea has become the world leader in broad-
Internet since the late 1990s. In 2000, 31 percent of band Internet access. Although only 14,000 Korean
Italian households owned a PC, of which 60 percent households had a broadband connection in 1998,
were connected to the Internet. In one year, the nearly 9 million of them were using broadband con-
Internet penetration rate increased by one-third, from nections by 2002. Koreans are also heavy users of
14 percent in 1999 to 21 percent in 2000. It more mobile-phone-based Internet services.
than doubled in two years, reaching 33 percent
(19 million) in 2001. However, Italy still has relatively China
low rates of PC and Internet penetration com- China is a relatively late starter in Internet use but has
pared with other western European nations. The gen- been catching up quickly. Because Chinas popula-
der gap has remained significant, with women using tion is so large, the low penetration rate of less than
the Internet far less than men. 5 percent provides both a great many users and much
room for growth. There has been a dramatic increase
Japan in Internet users, from 620,000 in 1997 to 22 million
The percentage of Japanese households owning PCs in 2001, and about 60 million in 2003. The number
more than doubled from 1996 to 2002. About 22 per- of Internet-connected computers has increased from
cent of Japanese households owned a PC in 1996, about 0.3 million in 1997 to 12 million in 2002.
29 percent in 1997, 33 percent in 1998, 38 percent in Chinas Internet population probably ranks second
1999, 51 percent in 2000, and 58 percent in 2001. in the world and is growing rapidly. Currently, use is
However, there has been a gap between PC access concentrated in major urban areas near the east coast.
and Internet access. The diffusion of the Internet, Public access points, such as Internet cafes, accom-
especially the PC-based Internet, started relatively modate many. The majority of new users are
late in Japan. For instance, while 40 percent of Am- young, especially university students, creating a prob-
erican households were online in 1999, only 12 per- able demand for continued Internet use in later life.
cent of Japanese households were online that year.
The number of Internet users (6 years and older) was Mexico
12 million in 1997, 17 million in 1998, 27 million in In Mexico a much higher percentage of the popu-
1999, 47 million in 2000, 56 million in 2001, and lation has a computer than has Internet access. For
INTERNET-WORLDWIDE DIFFUSION 387
TABLE 1. Summary of Internet Access in Seven Countries
Socioeconomic Geographic
Country Gender Life Stage
Status Location
Declining yet Half of Internet users Declining yet Declining yet

U.S.
persistent are female. persistent persistent
Declining yet
Germany Increasing Increasing Declining
persistent
Younger
Italians cur-
rently more
Northern Italy is
Deep digital divide likely to access
Italy Increasing leading the south in
based on education. and use the
Internet diffusion.
Internet. Trend
is not avail-
able.
Major cities have

Declining, and
Declining yet persist- Younger are higher Internet dif-
Japan reversed for mobile
ent more involved fusion than smaller
(webphone) Internet
cities.
Declining. Seoul is
South Korea Increasing Persistent Increasing the most wired area
in the country.
Huge, yet declining Slightly Huge, yet declining

China Declining yet persistent
slightly declining slightly
Younger Very uneven. Users

Less than half of Mexicans are concentrated in
Mexico Huge Internet users are make up the the central districts,
women. majority of Guadalajara, and
Internet users. Monterrey.
Wenhong
Wenhong ChenChen and
and Barry Barry
Wellman 2003 Wellman 2003
example, at the start of the Internet era in 1994, al- all countries, developed as well as developing. In some
though nearly 5 percent of the Mexican popula- countries, the digital divide is widening even as the
tion owned a PC, only 0.04 percent of the population number and percentage of Internet users increases.
accessed the Internet. Since then, Mexicans have This happens when the newcomers to the Internet
adopted the Internet rapidly. The penetration rate are demographically similar to those already online.
surpassed the 1 percent mark in 1998 and increased The diffusion of the Internet is not merely a mat-
to 2.6 percent in 1999. It continued to grow to ter of computer technology. The digital divide has
2.8 percent (2.7 million) in 2000 and 3.7 percent profound impacts on the continuation of social
(3.6 million) in 2001, and reached 4.7 million in 2002. inequality. People, social groups, and nations on the
Unreliable telephone service hinders home-based wrong side of the digital divide may be increas-
Internet use, especially in rural and impoverished ingly excluded from knowledge-based societies
areas. Instead, public Internet terminals provide a and economies.
significant amount of connectivity.
Wenhong Chen, Phuoc Tran, and Barry Wellman
See also Digital Divide

The Diffusion of the
Internet in Context
FURTHER READING
The diffusion of the Internet (and accompanying
digital divides) has occurred at the intersection of Chen, W., & Wellman, B. (2004). Charting Digital Divides: Within
international and within-country socioeconomic, and Between Countries. In W. Dutton, B. Kahin, R. OCallaghan,
& A. Wyckoff (Eds.), Transforming Enterprise. Cambridge, MA:
technological, and linguistic differences. Telecom- MIT Press.
munications policies, infrastructures, and educa- CNNIC (China Internet Network Information Center). (2003,
tion are prerequisites for marginalized communities July). 12th statistical survey report on the Internet development in
to participate in the information age. High costs, China. Retrieved January 23, 2004, from http://www.cnnic.org.cn/
download/manual/en-reports/12.pdf
English language dominance, lack of relevant con- Fortunati, L., & Manganelli, A. M. (2002). A review of the literature
tent, and lack of technological support are obstacles on gender and ICTs in Italy. In K. H. Srensen & J. Steward (Eds.),
to disadvantaged communities using computers Digital divides and inclusion measures: A review of literature and
and the Internet. For instance, while about one-half statistical trends on gender and ICT (STS Report 59-02). Trondheim,
Norway: Centre for Technology and Society.
of the worlds Internet users are native English speak- Light, J. (2001). Rethinking the Digital Divide. Harvard Educational
ers, about three-quarters of all websites are in Review, 71(4), 709733.
English. Ministry of Public Management, Home Affairs, Posts, and Telecom-
munications (Japan). (2003). Building a new, Japan-inspired IT
The diffusion of Internet use in developed coun- society. Tokyo: Author.
tries may be slowing and even stalling. Currently, National Telecommunications and Information Administration (2002).
Internet penetration rates are not climbing in several A nation online: How Americans are expanding their use of the
of the developed countries with the most penetra- Internet. Washington, DC: U.S. Department of Commerce.
Norris, P. (2001). Digital divide? Civic engagement, information poverty
tion. This is a new phenomenon, but is too soon to and the Internet in democratic societies. Cambridge, UK: Cambridge
tell if it represents a true leveling-off of the pene- University Press.
tration rate or a short-term fluctuation as Internet Nua Internet (2004). How many online? Retrieved January 23, 2004,
use continues its climb to triumphant ubiquity. from http://www.nua.com/surveys/how_many_online/index.html.
Reddick, A., & Boucher, C. (2002). Tracking the dual digital divide.
With the proliferation of the Internet in devel- Ottowa, Canada: Ekos Research Associates. Retrieved January
oped countries, the digital divide between North 23, 2004, from http://olt-bta.hrdc-drhc.gc.ca/resources/digital-
American and developed countries elsewhere is nar- divide_e.pdf
Servicios de Telecomunicaciones. (2003). Usuarios estimados de Internet
rowing. However, the digital divide remains sub- en Mxico [Estimated Internet users in Mexico]. Retrieved January
stantial between developed and developing countries. 23, 2004, from http://www.cft.gob.mx/html/5_est/Graf_
The divide also remains substantial within almost internet/estimi1.pdf
INTERNET IN EVERYDAY LIFE 389
Soe, Y. (2002). The digital divide: An analysis of Koreas Internet dif-

fusion. Unpublished masters thesis, Georgetown University,
First AgeThe Internet as
Washington, DC.
Thomasson, J., Foster, W., & Press, L. (2002). The diffusion of the Internet
Dazzling Wonder
in Mexico. Austin: University of Texas at Austin, Latin America When the Internet moved from the arcane world
Network Information Center. of library computer terminals to homes and offices
TNS Interactive. (2001). Asia-Pacific M-commerce report. Retrieved during the 1990s, it was heralded as a technologi-
January 23, 2004, from http://www.tnsofres.com/apmcommerce/
UCLA Center for Communication Policy. (2003). The UCLA Internet cal marvel. Early adopters congratulated themselves
report: Surveying the digital future year three. Retrieved January on being progressive elites, and techno-nerds rejoiced
8, 2004, from http://ccp.ucla.edu/pdf/UCLA-Internet-Report-Year- in newfound respect and fame. Bespectacled, nerdy
Three.pdf
van Eimeren, B., Gerhard, H., & Frees, B. (2002). ARD/ZDF-Online-
Microsoft founder Bill Gates was as much a super-
Studie 2002. Entwicklung der Online-Nutzung in Deutschland: Mehr star as rock singers and professional athletes. All
Routine, Wengier Entdeckerfreude [The development of Internet use things seemed possible. The cover of the Decem-
in Germany: More routine, less fun of discovery]. Media Perspektiven, ber 1999 issue of Wired magazine, the Internets great-
8, 346362.
World Internet Project Japan. (2002). Internet usage trends in Japan:
est cultural champion, depicted a cyberangel leaping
Survey report. Tokyo: Institute of Socio-Information and Commu- from a cliff to reach for the ethereal sun. The angels
nication Studies, Tokyo University. graceful posture pointed upward, placing seemingly
boundless faith in an unfettered cyberfuture.
The first clear pictures from the frontier of cyber-
space came from early studies of online culture.
INTERNET IN Investigators, peering into online communities
such as the Whole Earth Lectric Link (WELL) and
EVERYDAY LIFE LambdaMOO, provided insight into how early
adopters multitasked and negotiated identities, given
The increasing presence of the Internet in every- a paucity of social cues. Early adopters activities pro-
day life has created important issues about what the vided material for stories in the mass media and
Internet means to people concerning access to re- ethnographies (cultural studies). Instead of travel-
sources, social interaction, and commitment to ing to remote places, ethnographers only had to
groups, organizations, and communities. log on to the Internet and tune in to online virtual
What started in 1969 as a network between four communities. Fascinating stories abounded of color-
computers in southern California has metamor- ful characters and dangerous situations such as
phosed into a global system of rapid communica- virtual transvestites and cyberstalkers.
tion and information retrieval. The Internet was Some pundits went too far, extrapolating from
designed to be decentralized and scalable from the such esoteric online settings to the generalized
beginning. These design features have given the Internet experience. However, as the Internet became
Internet room to expand to immense proportions broadly adopted, people saw that communication
and to keep growing. By the end of 2003, the Internet would not primarily be with far-flung mysterious
had an estimated 600 million regular users, and others in virtual worlds, but rather with the people
the Google search engine could reach more than about whom users already cared most: family, friends,
3.4 trillion unique webpages. The Internet was a novel and workmates. Nevertheless, ideologies of the
curiosity in academia in the 1980s, but it went main- unique, transformative nature of the Internet per-
stream in the early 1990s when e-mail was joined by sisted as enthusiasts failed to view it in historical
the World Wide Web. Since then the Internet has be- perspective (presentism). For example, long-distance
come so pervasive that in many parts of the world community ties had been flourishing for generations,
its presence is taken for granted. Such widespread using automobiles, telephones, and airplanes.
nonchalance about such a powerful set of techno- Other pundits assumed that only online phe-
logical tools illustrates how deeply the Internet has nomena are relevant to understanding the Internet
embedded itself into everyday life. (parochialism). They committed the fundamental
A Personal StoryInformation Technology and

Competitive Academic Debate
They say a transformation occurs when something new becomes so much a part of an activity that you cannot think of do-
ing it otherwise. I have witnessed something on the order of a transformation in doing research for competitive aca-
demic debate as both a participant and a coach of at both the high school and college levels. Research is essential to
constructing competitive arguments and strategies in the activity. In high school in the early 1990s, most of this was
done in libraries as preparation before going off to a tournament. By the time I reached college in the mid 1990s, we also
had access to Lexis-Nexis. This was particularly significant because we could dial into the service to adapt and update
our arguments while away at tournaments. The ability to adapt arguments over the course of a tournament is crucial to
maintain a competitive edge.
The next change came as the university I attended began to wire all the buildings with Internet and the Internet
gradually became more central to debate research. With the latest journals and reports from various think tanks and avail-
able online, we could do the bulk of our research without leaving the debate officewhich came in handy during those
cold Iowa winters. Some became so dependent on the Internet that they never went to the library. At this point, however,
the Internet was still not portable, meaning we, like most teams, did not have Internet access when we were out of town
at tournaments.
Over time, however, the ability to access the vast diversity of resources on the Internet became essential to stay com-
petitive with other teams. Given the literally hundreds of dollars we would incur in hotel phone charges, we began to
network computers together so that multiple people could be online or using Lexis-Nexis while using a single phone
line. Now that most major hotels have high speed Internet access, the costs of connectivity have dramatically declined.
Some tournaments even offer onsite wireless access to all participants so that their coaching staffs can continually update
and adapt argument strategies throughout the day. Internet-based research, once not even possible, became a competi-
tive edge for those who had it, and is now a competitive necessity for debate programs at home and on the road.
Michael J. Jensen
sin of particularism, thinking of the Internet as a so widely used in developed countries that it was be-
lived experience distinct from the rest of life. This coming routinized. Familiarity breeds cognitive neg-
sin often shaded into elitism because only the small lect, and as with the telephone and the automobile
percentage of the technologically adept had the before it, exotic stories diminished just as the
equipment, knowledge, time, and desire to plunge widespread diffusion of the Internet increased its
so fully into cyberspace. true social importance.
The social exuberance for all things technolog-
ical departed quickly in 2000. For one reason, that
years dot.com stock market bust curbed enthusiasm Second AgeThe Internet Embedded
and media attention. Special newspaper Internet
sections shrank in the wake of instantly vanishing in Everyday Life
dot.com vanity ads, and the pages of Wired maga- The story of the Internet after the hype is more in-
zine shrank 25 percent from 240 pages in September teresting, if less fashionable. The Internet plugs in to
1996 to 180 pages in September 2001 and another existing social structures: reproducing class, race,
22 percent to 140 pages in September 2003. When and gender inequalities; bringing new cultural forms;
the rapidly contracting dot.com economy was and intersecting with everyday life in both un-
brought down to earth, it took Internet euphoria conventional and conventional ways. Attention now
with it. At the same time, the Internet had become focuses on the broader questions of the Internet in
society rather than on Internet societies. The thrust Most people in most developed countries use the
of research is moving from using culture-oriented Internet to find information or to contact friends,
small sample studies and abstract theorizing to using but many people are not online. Surveys and ethno-
surveys to study the more diffuse impact of this new graphies have shown how racial minorities, the eco-
communication and information distribution me- nomically disadvantaged, and those people who
dium in the broad population of Internet users (and do not read English use the Internet less than others.
nonusers). Whereas the first age of the Internet was This situation has serious social consequences as
a period of exploration, hope, and uncertainty, the companies and government agencies place more ser-
second age of the Internet has been one of routin- vices exclusively online. Thus,digital divides mean
ization, diffusion, and development. that the lack of Internet access and use can in-
Research shows that computer networks actively crease social inequality. Digital divides exist within
support interpersonal and interorganizational social countries and among countries. Moreover, different
networks. Far from the Internet pulling people apart, countries have different sorts of digital divides. For
it often brings them closer together. Internet users example, Italian women access the Internet much
are more likely to read newspapers, discuss impor- less often than do Italian men or northern European
tant matters with their spouses and close friends, women. Overall, however, the gaps in income, lo-
form neighborhood associations, vote, and partici- cation, culture, and language are shrinking between
pate in sociable offline activities. The more they meet those people who are comfortable with computeri-
in person or by telephone, the more they use the zation and those who are not. The gender gap has
Internet to communicate. This media multiplexity already disappeared in some places.
means that the more people communicate by one Digital divides constitute more than an access/no
medium, the more they communicate overall. For access dichotomy. People have concerns about the
example, people might telephone to arrange a social quantity of information flowing through the Internet
or work meeting, alter arrangements over the and the quality of the experience. The quality of the
Internet, and then get together in person. Rather than Internet experience is a key concern for reducing
being conducted only online, in person, or on the social inequality. First, the ability to perform a com-
telephone, many relationships are complex dances plex and efficient search online is not a skill learned
of serendipitous face-to-face encounters, scheduled by osmosis, but rather by experience and openness
meetings, telephone chats, e-mail exchanges with to the potential of the technology. Second, bloated
one person or several others, and broader online dis- software that inundates the user with ambiguous op-
cussions among those people sharing interests. tions and icons can intimidate novice users instead
Extroverts are especially likely to embrace the ways of providing the best framework for learning. Third,
in which the Internet gives them an extra and effi- content providers must consider the time lag be-
cient means of community. However, introverts can tween experienced early adopters, late adopters, and
feel overloaded and alienated. newbies (new users). These populations can have
Internet-based communications have always fos- significantly different expectations about what to do
tered social networks serendipitously. Even eBay, the online and how to do it. Fourth, many websites are
Internet auction enterprise, helps create commu- available only in English. Fifth, one can have diffi-
nication between hitherto-disconnected specialized culty using the Internet to communicate if ones con-
producers and collectors. Many software developers tacts are not online.
have focused on identifying, using, and analyzing At one time analysts expected all societies to use
these social networks. Participants in online net- the Internet in similar ways. However, comparative
working websites, such as Friendster, not only de- research shows different national patterns. The extent
scribe themselves online (e.g., single, male truck to which such media as e-mail or instant messaging
driver aged thirty-five), but also list their friends. (IM) is used depends on a complex interplay between
They hope that friends of friends will be able to con- peoples tastes, financial resources, culture, location
tact each other. geographically, location in the social structure, and
Finding Work Online
In an interview for my dissertation on work in the Internet technologies such as programs to maintain contact in-
industry, a freelance writer described how she found one formation, e-mail lists, and new networking services. Jobs
of her jobs. Members of the online communities she in a different city can be easily researched, and looking
belonged toincluding e-mail lists that connected mem- for jobs that require specialized skills across the country is
bers of New Yorks Internet industryfreely shared job easier than before the use of the Internet for job hunting.
leads with one another and often announced when they The same changes that make it easier for a job seekers
were looking for a new job. One position, however, came to get information about job openings, however, make it
when an employer met her through the list. She used more difficult for them to differentiate themselves from
the story to describe the importance of maintaining ones other, now more numerous jobseekers. Through a kind of
reputation online: He said, I saw your post on the list information overload, online job databases may, in fact,
and saw that you know how to use a semicolon. You seem be strengthening the role that personal connections play
super competent. She got the job without ever having in ultimately helping workers find jobs. One group I stud-
met him. ied had a novel approach to the online/offline searching
My interview data show that her experience is not un- techniquesthey stood on street corners in New York City
usual. Online social connections can feel as real as other passing out fliers for a website where they posted their re-
ways of expressing knowing someone, especially in realms sumes. Rather than relying solely on the ways in which
like job hunting. Through e-mailed job announcements, they were linked to prospective employers online, they in-
people can demonstrate what anthropologists call gift ex- creased the number of their social connections.
change through their social networks. They use informa- Still, getting a job, as the sociologist Mark Granovetter
tion as a resource to share with people they are in once wrote, comes down to the friends of friends, even in
contact. The rise of these announcements and online job a digital era, even if some of those people are known only
databases make the process of job hunting easier, as does online.
the expansion of acquaintance networks through computer Gina Neff
national infrastructure. At times people face not a mat- With physical co-presence continuing to be
ter of personal choice but rather a matter of social important, the Internet supports glocalization
constraint: It is foolish for a person to send e-mails or rather than the Canadian philosopher Marshall
instant messages if few people are reading them. For McLuhans imagined global village. In the com-
example, Catalans in Spain mostly use the Internet munity and at work the Internet facilitates physically
for acquiring information and shoppingtrain sched- close local ties as well as physically distant ties. People
ules, theater ticketsand less for communicating often use the Internet to communicate quickly with
by e-mail. Catalonia is a local society in a healthful cli- nearby others without the disturbance of a phone
mate where people gather in cafes to chat face to face. call or in-person visit. For example, one study of
To take another example, teenagers in developed coun- Netville, the pseudonym for a Toronto suburb, found
tries communicate more by mobile phone and instant that active Internet users knew the names of more
messages than by e-mail. In Japan the proliferation of neighbors, had visited more of them, and used the
Web-enabled phones means that two hitherto-sepa- Internet effectively to mobilize against their real
rate communication media are becoming linked: estate developer. Another Toronto study found
Japanese teenagers and young adults frequently ex- that co-workers were more likely to use the Internet
change e-mails on their mobile phones or use their when they worked in the same building, in part
personal computers to send short text messages to because they had more tasks and concerns in
mobile friends. common. Even many long-distance ties have a lo-
cal component, as when former neighbors or office- The Social Possibilities

mates use the Internet to remain in touch. E-diasporas
abound as migrants use the Internet to stay linked of the Internet
with their old country: communicating with friends The Internetor any other technologydoes not
and relatives, reading newspapers online, and pro- simply cause anything, just as a light switch placed
viding uncensored information. high on a wall does not make access impossible for
Peoples increased use of the Internet is corre- children. Understanding the implications of the
lated with their decreased time doing housework, Internet calls for understanding some of the possi-
watching television, and being with family members ble social activities that can be accomplished by
in person. Experts have suggested two models to ex- using it. For example, people can take advantage of
plain this situation. The hydraulic model treats time real-time communication with partners via instant
as a fixed amount. Hence, an increase in the use of messaging, access daily journals (blogs), and send
the Internet directly corresponds with a decrease instant announcements to one or one hundred spe-
in other activities. According to the efficiency model, cific people through e-mail.
people on the Internet can use time more effectively Yet, the Internet does not determine the nature
and may get more communication and information of communication. Indeed, a line of media rich-
out of their day. The efficiency model appears to ness research during the early 1990s failed to
be more accurate as people combine e-mail, IM, and show much fit between the nature of a communi-
mobile phone use with other activities. Indeed, with cation medium and what it was used for: People used
multitasking, we can say that some people live a what their friends and their co-workers used. The
thirty-six-hour day, performing tasks concurrently Internet lends itself to particular styles of commu-
online and offline. nication that are parts of a persons overall ensem-
The Internet also has affected social networks at ble of everyday communication via the telephone
work. Early efforts at software to support people and in-person encounters. Thus, the Internets tech-
working groups were fitful, in part because most nical characteristics provide a means of organizing
knowledge workers do not work in one group. Many relationships with other people, not a blueprint of
now are engaged in distributed work, operating how the relationships will or should take place.
through geographically dispersed and loosely knit
social networks. Rather than being parts of tradi-
tional bureaucratic hierarchies (with organizational Asynchronous Communication
structures looking like inverted trees) in which each Most communication takes place in real time.
worker fits into a single group, many knowledge On the telephone and in person, people assume that
workers are partial members of multiple teams communication is reciprocal and that the delay be-
and report to multiple superiors. Many teams are ge- tween utterances is brief. By contrast, e-mails, like
ographically dispersed so that much communica- letters, allow people to communicate on their own
tion is done by the Internet. Moreover, those workers time. Yet, unlike letters, e-mails reach their desti-
who spend the day working on personal comput- nation within minutes. As long as systems are not
ers often turn to the Internet to acquire information overloaded, the only significant delay in e-mail com-
rather than ask a co-worker in a nearby cubicle. They munication is the time lag set by the users attention.
form communities of practice with fellow work- E-mail, as a form of asynchronous communication,
ers whom they may never have met in person, ex- gives greater individual autonomy. People can choose
changing know-how and empathy online. However, when to turn on their computers and their e-mail
proximity still has its advantages because it provides programs, choose to whom they wish to respond,
a broad bandwidth of multisensory communication and choose who else in their network they want to
people learn more when they see, hear, smell, and include in their e-mail communication. The cost
touch each otherand enables the exchange of phys- of this autonomy is uncertainty regarding when and
ical objects. if the receiver will read the e-mail and reply.
Bandwidth worthy of its name. Ubiquity means the widespread

The number of bits (units of computer information) availability of usable computing and computer-
that can be pushed through a computer network con- mediated communication. Travelers in the devel-
nection has risen from 110 bits per second (bps) in oped world are coming to expect to be able to
the mid-1970s to 30,000 bps in the early 1990s and co n n e c t to t h e In te r n e t w h e re ve r t h e y a re
upward of 1 million bits for a high-speed connec- through cybercafes, high-speed links in hotels, wire-
tion today. High capacity bandwidth is important less hotspots, and the local offices of their or-
for speed so that text messages and webpages become ganizations. Workers can now be reached on
readable without distracting delays. Greater band- vacation. Cybercafes proliferate in the less-devel-
width affords richer content. Greater bandwidth can oped world, catering to people who cannot afford
mean the difference between sending terse and ugly to own a computer or who do not have reliable
text messages and sharing photos or music and phone and electricity connections. Other solutions
seeing one another via Internet-connected cam- are developing, such as wirelessly connected com-
eras (webcams). puters carried by Indian postal workers on their
Greater bandwidth usually allows computers to rounds. The continuing development of global
be connected continuously. The study of the Netville availability means that even more people and or-
wired suburb found this always-on feature of its net- ganizations will be reachable online. All users could
work to be more valued than sheer speed. Such a con- be connected to all, through either direct ties or
nection affords people the ability to send e-mail or short chains of indirect ties.
check the Web whenever the inclination strikes them.
Employers now complain about workers use of the
Internet for personal matters, and family members Wireless Portability
complain that their loved ones are tied to their com- People are shifting from wired computing
puters during supposed leisure hours. connected to the Internet through cablesto
High-speed, always-on connections allow a portable, wireless computing. Portability means that
different relationship to the Internet than that pos- people can take computing with them: One does not
sible with micromanaged dial-up connections have to depend on others equipment to connect.
with slow file access and concerns about usurping Already more wireless mobile phones than wired
telephone lines. Whereas dial-up connectivity facili- phones are in use worldwide. Although wires still
tates a discrete Internet session, always-on Internet carry the most bandwidth, mobile phones are be-
encourages spontaneous use. coming integrated with the multifunctional capac-
ity of computers. This integration lets people be less
rooted to place and gives them the ability to connect
Globalized, from anywhere. Portability means that much work
is being carried home from the office on laptop com-
Ubiquitous Connectivity puters or high-capacity storage devices.
Computerization has oscillated between centralized
control (computer centers) and personal control
(standalone computers). The current situation Relational Data
networked computingmeans that information Vannevar Bush, the great-grandfather of the Internet,
(and control) flows up and down between central once suggested that information could be organized
servers and somewhat autonomous personal com- into trails of association rather than grouped into
puters. Yet, despite organizational control, many discrete categories (as in an encyclopedia). This sug-
people in organizations also use their computers for gestion has been translated into Web surfing. People
social and personal matters. move through Web networks, from one link to an-
Computer networks are expanding as the World other, rather than exhaust a particular category. In
Wide Web is becoming more comprehensive and some instances these links are dynamic, such as the
recommendations on an Amazon.com book web- ular open source operating system (GNU/Linux)
page, whereas others are more static, such as a per- contains many millions of lines of computer code.
sonal list of favorite links. Another open source product, Apache, runs most of
Most computer operating systems now allow the worlds Web servers. Without the Internet to con-
users to have their own settings, e-mail accounts, and nect specialized developers and distribute their code,
desktop aesthetics. Instant messaging accounts, ac- open source work would have remained a slow-
cessible from any Internet terminal, are tailored to moving, poorly communicated, and badly coordi-
the person, not to the house or the particular com- nated activity for hobbyists.
puter. Ubiquitous computing could soon mean that Yet, even open source work does not exist ex-
whenever people log on to a communications de- clusively on the Internet. Linux user groups popu-
vice, the device knows who they are, where they are, late most North American cities, popular face-to-face
and their preferences. Such personalization, even conferences generate revenue, and developers like to
at its early stages, is fostering societal shifts from talk to each other in person. The Internet has facil-
place-to-place connectivity (a particular computer) itated, not monopolized, this type of production.
to person-to-person connectivity (a particular users The most notorious exchange of complex in-
account). formation on the Internet is the sharing of music,
The Internet has partially democratized the pro- computer programs, and movies. Because only com-
duction and dissemination of ideas. More people are puter bits, not material goods, are exchanged, many
providing information to the public than ever be- downloaders feel that they have the right to obtain
fore. E-mail discussion groups, Web-based chat- such material for free. The complex interplay be-
rooms, and Usenet newsgroups foster conversations tween immaterial media, their physical containers
among (more or less) like-minded people. All of these such as CDs, their copyright licenses, and the costs
communication media are based on many-to-many of distribution has challenged producers, consumers,
communication, in contrast to e-mail and instant and the legal system.
messaging, which usually are based on one-to-one
communication. Although some people have feared
that the like-minded will talk only to each other, in The Turn toward
practice much diversity exists in these interactions,
and interesting ideas can be copied and forwarded Networked Individualism
to others. Rather than result in inbred sterilization, The Internet lets people who can afford recent tech-
such media are concerned with flaming (making nological services communicate when, where, and
offensively rude comments) and spamming (send- with whom they want and have their experiences per-
ing off-topic comments). sonalized. Indeed, the Internet (along with mobile
For those people who want more complex means phones and automobiles) is fostering a societal turn
of communication, easy-to-use software now facil- away from groups and toward networked individ-
itates do-it-yourself webpages. Software has trans- ualism: People connect to each other as individuals
formed the creation of simple webpages from an arcane rather than as members of households, communities,
art to a straightforward creation of non-specialists. At kinship groups, workgroups, and organizations.
one time most webpages were relatively static, but Especially in the developed world, this flourishing
now blogs make frequent updating simple. Many of person-to-person connectivity has also been fos-
blogs combine personal reflections, social commu- tered by social changes such as liberalized divorce
nication, and links to other websites. laws and by technological changes such as the pro-
This democratization of computing is not just liferation of expressways, mobile phones, and air
recreational. The Internet also supports open source travel. The turn to a ubiquitous, personalized, wire-
development of computer code: a peer-production less world fosters personal social networks that sup-
system where members of a team contribute and dis- ply sociability, support, information, and a sense
tribute computer code freely and openly. One pop- of belonging. The individual user is becoming a
switchboard between his or her unique set of ties and Castells, M. (2000). The rise of the network society (2nd ed.). Oxford,
networks; people separately operate their specialized UK: Blackwell.
Castells, M. (2001). The Internet galaxy: Reflections on Internet, busi-
ties to obtain needed resources. Although people re- ness, and society. Oxford, UK: Oxford University Press.
main connected and supportive, individuals in Castells, M., Tubella, I., Sancho, T., & Wellman, B. (2003). The network
unique networks have supplanted the traditional society in Catalonia: An empirical analysis. Barcelona, Spain:
Universitat Oberta Catalunya.
organizing units of the household, neighborhood,
Chen, W., & Wellman, B. (2004). Charting digital divides: Within and
kin group, and workgroup. between countries. In W. Dutton, B. Kahin, R. OCallaghan, &
Networked individualism is having profound ef- A. Wyckoff (Eds.), Transforming enterprise. Cambridge, MA: MIT
fects on social cohesion. Rather than being part of a Press.
Cross, R., & Parker, A. (2004). The hidden power of social networks:
hierarchy of encompassing groups like nesting Understanding how work really gets done in organizations.
Russian dolls, people belong to multiple, partial com- Boston: Harvard Business School Press.
munities. They move from person to person, not place Deibert, R. J. (2002). Dark guests and great firewalls: The Internet and
to place. Increasing specialization of tastes and com- Chinese security policy. Journal of Social Issues, 58(1), 143159.
Gibson, W. (1984). Neuromancer. New York: Ace Books.
bination of roles are not products of the Internet. Hampton, K., & Wellman, B. (2003). Neighboring in Netville: How
Yet, the design of the Internet, culturally rooted in a the Internet supports community and social capital in a wired sub-
specific brand of individualism, considers the per- urb. City and Community, 2(3), 277311.
Haythornthwaite, C., & Wellman, B. (1998). Work, friendship and me-
son regardless of place and regardless of a socially dia use for information exchange in a networked organization.
imposed structure such as a kinship network. Social Journal of the American Society for Information Science, 49(12),
coordination may be usurped as the role of main- 11011114.
taining ties with kin becomes either less important Hiltz, S. R., & Turoff, M. (1993). The network nation (2nd ed.).
or more overloaded. Teenagers parents do not get Hinds, P., & Kiesler, S. (Eds.). (2002). Distributed work. Cambridge,
to approve of buddy lists or websites visited. MA: MIT Press.
The development of computer networks, and the Howard, P. N., & Jones, S. (Eds.). (2004). Society online: The Internet
in context. Thousand Oaks, CA: Sage.
flourishing of social networks are building upon each Ito, M. (Ed.). (2004). Portable, personal, intimate: Mobile phones in
other to support the rise of networked individual- Japanese life. Cambridge, MA: MIT Press.
ism. Just as the flexibility of less-bounded, spatially Katz, J., & Aakhus, M. (2002). Perpetual contact: Mobile communica-
dispersed, social networks creates demand for col- tions, private talk, public performance. Cambridge, UK: Cambridge
University Press.
laborative communication and information sharing, Katz, J. E., & Rice, R. E. (2002). Social consequences of Internet use:
the rapid development of computer-communications Access, involvement, and interaction. Cambridge, MA: MIT Press.
networks nourishes societal transitions from group- Kendall, L. (2002). Hanging out in the virtual pub: Masculinities and
relationships online. Berkeley, CA: University of California Press.
oriented societies to a society of networks. Kiesler, S. (Ed.). (1997). Culture of the Internet. Mahwah, NJ: Lawrence
Erlbaum.
Barry Wellman and Bernie Hogan Kim, A. J. (2000). Community building on the Web. Berkeley, CA:
Peachpit Press.
Kraut, R., Kiesler, S., Boneva, B., Cummings, J., Helgeson, V., &
See also Computer-Supported Cooperative Work; Crawford, A. (2002). Internet paradox revisited. Journal of Social
Cybercommunities; Digital Divide; Internet Issues, 58(1), 4974.
Worldwide Diffusion Madden, M., & Rainie, L. (2003). Americas online pursuits. Washington,
DC: Pew Internet and American Life Project.
Manovich, L. (2002). The language of new media. Cambridge, MA:
MIT Press.
National Telecommunications and Information Agency. (1999). Falling
FURTHER READING through the Net: Defining the digital divide. Washington, DC:
U.S. Department of Commerce.
Bradner, E., Kellogg, W. A., & Erickson, T. (1999). The adoption and Preece, J. (2000). Online communities: Designing usability, support-
use of Babble: A field study of chat in the workplace. In S. Bodker, ing sociability. New York: John Wiley & Sons.
M. Kyng, & K. Schmidt (Eds.), ECSCW 99: Proceedings of the Sixth Raymond, E. (1999). The cathedral and the bazaar: Musings on
European Conference on Computer Supported Cooperative Work Linux and Open Source by an accidental revolutionary. Sebastopol,
(pp. 139158. Dordrecht, The Netherlands: Kluwer. CA: OReilly.
Bush, V. (1945). As we may think. The Atlantic Monthly, 176(1), Rheingold, H. (2000). The virtual community (Rev. ed.). Cambridge,
101108. MA: MIT Press.
ITERATIVE DESIGN 397
Rheingold, H. (2002). Smart mob: The next social revolution. New York: When software developers develop software
Perseus. for their own use, they rely on iterative design. An
Smith, M. A., & Kollock, P. (Eds.). (1999). Communities in cyberspace.
London: Routledge.
early description of iterative design used as a for-
Sproull, L., & Kiesler, S. (1991). Connections. Cambridge, MA: MIT malized software development method is given in
Press. a 1975 article by the computer scientists Victor Basili
Turkle, S. (1995). Life on the screen: Identity in the age of the Internet. and Albert Turner, who called it iterative enhance-
UCLA Center for Communication Policy. (2003). The UCLA Internet ment. In a 1985 article, John Gould and Clayton Lewis
report: Surveying the digital future year three. Retrieved January broadened the idea and impact considerably by sug-
8, 2004, from http://ccp.ucla.edu/pdf/UCLA-Internet-Report-Year- gesting that iterative design should be one of the fun-
Three.pdf
Watts, D. J. (2003). Six degrees: The science of a connected age. New
damental principles for collaboration between system
York: W. W. Norton. developers and the prospective users of the system
Wellman, B. (2001). Physical place and cyberspace: The rise of per- being developed. Reinhard Budde and his colleagues
sonalized networks. International Urban and Regional Research, clarified the concept in a 1992 article by making a
25(2), 227252.
Wellman, B., & Haythornthwaite, C. (Eds.). (2002). The Internet in
distinction between the purpose of prototyping and
everyday life. Oxford, UK: Blackwell. the relation between the prototype and the final sys-
tem. They drew the connection between iterative
design and evolutionary prototyping, which is de-
scribed as a continuous process for adapting an
application system to rapidly changing organiza-
ITERATIVE DESIGN tional constraints. The relation between the proto-
types and the application system depends on the
Iterative design is a product development process form of iterative design that is employed, as discussed
based on repeated cycles. In each cycle, designers below.
elaborate, refine, and experiment with the design. Iterative design processes in the field of human-
The work done in each cycle feeds into the next cycle. computer interaction typically involve a substantial
Although iterative design can be used with many element of initial analysis, which forms the basis for
aims, the most common goal is usability. Where us- developing a first prototype. This prototype is grad-
ability is the goal, each iteration experiments with ually redeveloped, and through a number of itera-
some aspect of users experiences with the prod- tions it becomes the system. Often no aspect of the
uct, and user evaluations form a large part of the initial design remains untouched through the many
feedback. In this way, iterative design is a form of iterations, and it is argued that the iterations con-
user-centered design. tribute to a significantly better product.
Iterative design dovetails well with prototyping
when it comes to software development. Prototyping
is described in general in a separate article in this vol- Forms of Iterative Design
ume. Iterative design is sometimes distinguished In practice, there are three distinct forms of iterative
from iterative development, in that iterative design design: abstract iterative design, experiential itera-
aims to produce a design solution. This is also called tive design, and embedded iterative design.
design prototyping, as the prototype is a key element
in describing the design. Iterative development, by Abstract Iterative Design
contrast, aims to produce a software system. This is This simple form of iterative design does not involve
also called production prototyping, with the proto- any operational prototypes. There may be only a sim-
type becoming the system. While iterative develop- ple two-stage process in which the designer creates
ment and iterative design are thus conceptually the design and then meets with users. The designer
different, they are usually conflated both in the lit- explains the design and solicits feedback from the
erature on iterative design and in system develop- users. This user feedback forms the basis for design
ment practice. revisions and further cycles of user review. The
designers may use simple aids such as drawings or struction cycles until the design stabilizes and the
replicas of system screen images, components, and production prototype passes a final user acceptance
so on. test. The design and the system itself synchronously
The simple aids used in iterative design are of- evolve into the final development product.
ten referred to as paper prototypes or mock-ups. Embedded iterative design is usually the basis for
They are a form of design artifact that functions as short-cycle systems development, which is also called
a concrete embodiment of how the designer plans agile development or Internet-speed development.
to meet users requirements. This artifact performs This form of systems development is similar to
an important social role in making user-designer production prototyping, except that the design is
communication accurate; it provides registration never really considered to be stable. The software ar-
and alignment of the abstract ideas of the designer tifact evolves through a continuous series of releases,
with the abstract ideas of the user by providing a growing and adapting to changes in its environment.
physical object on which to anchor terminology. Embedded iterative design is suitable in highly com-
petitive environments and in emergent organizations.
Experiential Iterative Design
More commonly, user review in iterative design will
include some form of experience with a working pro- Managing Iterative Design
totype of system components. The users will have Iterative design is an ideal response to adaptive sit-
a simulated but dynamic experience with various uations. It is carried out in a sequence of iterations
system functions and will examine input format and or cycles, each involving a number of activities. These
system output. The experiential process is similar to activities include analysis of evaluations of previ-
the abstract process, but the design artifacts are con- ous versions of the system and requirements for
siderably richer, more sophisticated, and dynamic. additional functions, design of software elements that
The user has the opportunity to experience how the integrate the new functions into the existing version,
new system will look and feel. This form of iterative implementation of the design through the building
design is sometimes called design prototyping, and of new artifacts, and evaluation of those new artifacts
the design artifacts are sometimes called throwaway by prospective users. After the evaluation, the cycle
prototypes. (They are called throwaway prototypes begins again. These cycles continue until the de-
because they are not actually part of the final system; sign, and usually the artifact, is stable and acceptable
they serve only as models for the design.) to all stakeholders involved. (Stakeholders can include
the systems designers, users, testers, and owners.)
Embedded Iterative Design Iterative design may seem simple and appealing,
Embedded iterative design embeds the design ac- but it is definitely not without its challenges. Iterative
tivity in the construction activity in cycles that pro- design depends fundamentally on repeated activi-
duce working system components that will be part ties, which makes management particularly prob-
of the final production system. In embedded iterative lematic, because it becomes hard to exercise the two
design, the design artifact is the emerging produc- basic management functions of planning and con-
tion software system. Users experiment with the var- trol. Plans are unstable because usability goals are
ious system components as these are produced in hard to specify in advance and are supposed to
rough form and then refined. Based on users expe- change with each cycle. Control is also difficult to
rience and evaluation, designers refine and develop maintain, as it is difficult to measure progress in us-
their designs. An iterative cycle will include design, ability, especially if an iteration leads to a shift in de-
implementation, and evaluation. Usability testing sign goals. In addition, progress depends on user
with prospective users is a key part of evaluation. cooperation that at times may not be forthcoming.
Embedded iterative design is the basis of pro- Management of iterative design must include
duction prototyping. This form of prototyping getting managers, analysts, programmers, and users
evolves a software system through design and con- to agree to the exact objectives of the process. Man-
ITERATIVE DESIGN 399
Risk Management Cycle

1
Prototyping Cycle
2 Define
Specify Risks F I G U R E 1 . Iterative
Consequences
design as an inter-
play between a pro-
3
Assign
totyping cycle and a
4
Priorities Select Resolution risk management
Strategies cycle.
agers must keep users from dominating the designer- Looking Forward
user interaction, otherwise users may inflate the scope Iterative design is a fundamental element in most
of development. Managers must also keep designers high-speed systems development approaches, in-
from becoming domineering, because designers may cluding agile development methods. The growing
deflate the scope in order to reduce the program- use of such approaches and methods, especially in
ming work. Designers must often work with in- highly competitive global settings, will drive increasing
complete materials and limited time for user reviews. use of iterative design for many years to come.
The iterative design process must carefully define the
contents of the next prototype in each cycle, because Richard Baskerville and Jan Stage
user reviews may uncover a vast array of potential
revisions and improvement directions. Designers See also Prototyping, User-Centered Design
should not run in too many different and possibly
unproductive directions. The process must accu-
rately gauge progress and nearness to completion so FURTHER READING
that design artifacts are not prematurely accepted
before user reviews are stable. Alavi, M. (1984). An assessment of the prototyping approach to in-
A risk management approach works well for con- formation systems development. Communications of the ACM,
27(6), 556563.
trolling iterative design processes. This approach en- Baecker, R. M., Nastos, D., Posner, I. R., & Mawby, K. L. (1993). The
ables appropriate risk resolution strategies to be user-centered iterative design of collaborative writing software. In
placed in effect before the design process breaks Proceedings of InterCHI93 (pp. 399405). New York: ACM Press.
Basili, V., & Turner, A. (1975). Iterative enhancement: A practical tech-
down. As illustrated in Figure 1, a four-stage risk nique for software development. IEEE Transactions on Software
management cycle is introduced into the design Engineering, SE-1(4), 390396.
cycle to evaluate the current stage of the project. First Baskerville, R. L., & Stage, J. (1996). Controlling prototype develop-
risks are defined. Next, the consequences of those ment through risk management. Management Information Systems
Quarterly, 20(4), 481504.
risks specified by considering what undesirable sit- Baskerville, R., & Pries-Heje, J. (in press). Short cycle time systems de-
uation will result from the risk and ranking the prob- velopment. Information Systems Journal, 14(2).
ability and potential severity for each risk. Then Boehm, B., Gray, T., & Seewaldt, T. (1984). Prototyping versus spec-
the risks are assigned priority, with high-probabil- ifying: A multiproject experiment. IEEE Transactions on Software
Engineering, SE-10(3), 290303.
ity or high-consequence risks receiving high prior- Budde, R., Kautz, K., Kuhlenkamp, K., & Zllighoven, H. (1992). What
ity. Finally, resolution strategies are developed for is prototyping? Information Technology & People, 6(23, 8995).
urgent risk factors. The process must designate Connell, J. L., & Schafer, L.B. Structured Rapid Prototyping. Englewood
Cliffs, NJ: Yourdon Press.
resolution strategies for the two to four risks that Dieli, M. (1989). The usability process: Working with iterative design
have the highest ranks, and these strategies form the principles. IEEE Transactions on Professional Communication, 32(4),
basis for managing the next design cycle. 272279.
Ehn, P. (1989). The art and science of designing computer artifacts. Norman, D. (1988). The psychology of everyday things. New York: Basic
Scandinavian Journal of Information Systems, 1, 2142. Books.
Gould, J. D., Boies, S. J., & Lewis, C. (1991). Making usable, useful, Plaisant, C., Marchionini, G., Bruns, T., Komlodi, A., & Campbell, L.
productivity-enhancing computer applications. Communications (1997). Bringing treasures to the surface: Iterative design for the
of the ACM, 34(1), 7486. Library of Congress National Digital Library Program. In
Gould, J. D., & Lewis, C. (1985). Designing for usability: Key princi- Proceedings of Human Factors in Computing, CHI 97 (pp. 518525).
ples and what designers think. Communications of the ACM, 28(3), New York: ACM Press.
300311. Sullivan, K. (1996). The Windows 95 user interface: A case study in
Khanna, N., Fortes, J. A. B., & Nof, S. Y. (1998). A formalism to usability engineering. Proceedings of Human Factors in Computing
structure and parallelize the integration of cooperative engi- Systems, CHI 96, 473480. New York: ACM.
neering design tasks. IIE Transactions, 30(1), 116. Vonk, R. (1990). Prototyping: The effective use of CASE technology. New
Nielsen, J. (1993). Iterative user-interface design. IEEE Computer, York: Prentice Hall.
26(11), 3241.
THE KEYBOARD
keyboard. The Type-Writer had alphabetically

THE KEYBOARD arranged keys that used an understroke mechanism;
depressing a key caused a lever to swing the type-
The keyboard is the main input device for the per- head upward against carbon paper over a piece of
sonal computer and its design retains features first stationery that was held against the underside of a
developed for typewriters over a hundred years ago. platen. The typist could not see what was being typed
and the device was prone to frequent key jams, which
slowed typing.
Keyboard History From 1874 to 1878 the Remington Company
In England in 1714, Queen Anne granted Henry Mills in New York manufactured a new design, the Sholes
the first keyboard patent. In Italy in 1808, Pellegrino & Glidden Type-Writer, which had a new keyboard
Turri built the first typewriter. In 1870 in Denmark, layout with keys organized into three rowstop,
a pastor named Malling Hansen produced the home, and bottom. The new layout, developed by
writing ball, commonly regarded as the first com- Amos Densmore, brother of Sholess chief finan-
mercial typewriter. In 1868 in the United States, cial backer, James Densmore, was and designed to
Christopher Latham Sholes, C. Glidden, and make the overall task of typing faster by slowing typ-
S.W. Soule, three Milwaukee inventors, patented the ing to minimize key jams. Based on letter-pair fre-
Type-Writer, the forerunner of todays computer quency and named after the leftmost six keys on the
401
top row, QWERTY was patented in 1878. The same timate keyboard layout; he called it the Dvorak
year a shift key was added to allow either upper- Simplified Keyboard. Both keyboard layouts have
or lower-case type, The Typewriter Magazine was been officially recognized by the American National
published for the first time, and Scientific American Standards Institute (ANSI X4.22-1983). A com-
used the term typewriting in an article. parison of the two keyboard layouts and the claimed
Most early typists learned a four-finger technique. performance differences for them when typing in
In 1876, a law clerk named Frank E. McGurrin de- English are shown in Figure 1.
veloped and taught himself all-finger touch typ- However, in spite of the claimed advantages,
ing, and six years later a Mrs. M. V. Longley began research in the mid-1990s failed to demonstrate either
a Shorthand and Typewriting Institute to teach the performance or postural benefits for the Dvorak lay-
all-finger method. In Cincinnati in 1888 a typing out, and QWERTY remains the de facto alphabetic
competition was held between McGurrin and Louis keyboard layout in the United States. But many mod-
Traub (a skilled four-finger typist), which McGurrin ern computer keyboards use a Dvorak layout for
easily won. After that, QWERTY became the stan- number and punctuation keys, and, in addition, in-
dard keyboard layout and touch typing became the corporate numerous other key groupingsfor ex-
standard technique. ample, a horizontal row of function keys is above the
Typewriter technology continued to improve. In top row of the keyboard. This arrangement maps
1896 Franz Wagner of the Underwood Company most closely to the horizontal menu arrangements
developed the up strike mechanism. The Blickens- of most software, and experiments show that the hor-
derfer Company developed the first electric type- izontal arrangement gives the fastest response times.
writer in 1902 and the first portable typewriter in Internet keyboards may also provide an addi-
1913. As technology improved there was less need tional row of dedicated buttons above the function
for the QWERTY layout to slow typing. In 1932 keys. A numeric keypad is usually positioned to
August Dvorak, a professor at Washington State the right of the keyboard, and keys are arranged in
University, developed what he claimed was the ul- calculator layout (789, 456, 123, 0) rather than phone
QWERTY keyboard layout

! # $ % _ & ( ) *
1 2 3 4 5 6 7 8 9 0 =
Top-row words 52 %
1 /4 Home-row words 32 %
Q W E R T Y U I O P 1
/2
Bottom-row words 16 %
: @ # words typed on home row 120
A S D F G H J K L ; c Typing speed 100 %
Typing accuracy 100 %
. ? Finger travel (8 hours/day) 16 miles
Z X C V B N M , . / Learning time 60 hours
Dvorak keyboard layout

! @ # $ % C & * ( ) [ *
1 2 3 4 5 6 7 8 9 0 ] =
Top-row words 22 %
. ?
P Y F G C R L Home-row words 70 %
, . / Bottom-row words 8%
_ # words typed on home row 3000
A O E U I D H T N S . Typing speed 115120 %
Typing accuracy 115120 %
: Q J K X B M W V Z Finger travel (8 hours/day) 1 mile
; Learning time < 20 hours
FIGURE 1. Comparison of the QWERTY and Dvorak keyboard layouts specified in ANSI X4.22-1983
THE KEYBOARD 403
layout (123, 456, 789, 0), even though studies find specifically for people who have been injured or who
that the calculator layout is slightly slower and less have a physical disability that affects their typing abil-
accurate. Keyboards also have keys for cursor con- ity, such as a missing hand. Others are designed for
trol (up, down, left, right) and these are arranged a more general user population and offer significant
either as a cross or an inverted T. Recommended health and performance benefits to a wide range of
modern keyboard design requirements are specified people. The key to good ergonomics is always to
in the BSR/HFES 100 standard (2002), which are match the design of the technology to the needs and
summarized in Table 1. characteristics of the user.
Ergonomic keyboard designs focus on improv-
ing the posture or the position of the hands when
Ergonomic Keyboards typing, on decreasing the keying forces, and on re-
Early in the twentieth century typists began reducing the amount of repetition. They tend to be
porting a variety of upper body musculoskeletal in- more expensive than conventional keyboards, but
juries which were attributed to the design of the most of them offer health benefits by reducing in-
keyboard. In 1915 Fritz Heidner was granted a U.S. jury risk factors and some of the designs also im-
patent for a series of split-keyboard designs. In prove productivity. The following seven alternative
Germany in 1926, E.A. Klockenberg also recom- ergonomic keyboard designs are currently available:
mended a split-keyboard design to improve the bent Fixed-angle split keyboards (for example, Microsoft
position of the hands. However, little changed in the naturalwww.microsoft.com). These keyboard de-
design of keyboards until the 1960s and 1970s when signs split the alphanumeric keys at a fixed angle and
Karl H. E. Kroemer began experimental investiga- they slightly tent the keyboard. There is some re-
tions of split designs for computer keyboards. Since search evidence of reduced discomfort because of
then there have been numerous redesigns of the com- reduced ulnar deviation (lateral bending of the
puter keyboard. hands). These designs work better for people with
Ergonomic keyboards use different physical broader or larger frames and for pregnant women
designs to try to improve typing safety and perfor- because they put the arms in a better position to reach
mance. Some of the keyboards have been designed around the front of the body. However, the designs
TABLE 1. Mandatory Keyboard Design Requirements Specified in BSR/HFES100 (2002)
Feature Requirement
Keyboard Layout Numeric keypads shall be provided when users primary task involves
data entry. These keys shall be grouped together.
Cursor control If cursor keys are provided, they shall be arranged in a two-dimen-
sional layout (as a cross or inverted-T).
Keyboard Height and The slope of conventional tabletop-mounted keyboards shall be
Slope between 0 and 15 degrees.
Key Spacing Centerline distances between the adjacent keys within a functional
group shall be between 18 and 19 mm horizontally and between 18
and 21mm vertically.
Key Force The force to activate the main alphabetic keys shall be between 0.25
and 1.5 N.
Key Displacement Vertical displacements of the alphabetic keys shall be between 1.5 and
6.0 mm.
Keying Feedback Actuation of any key shall be accompanied by tactile or auditory feed-
back, or both.
usually address the issue of wrist extension, whereas typing speed. Completely split keyboards are also
the upward bending of the hands turns out to be a hard for hunt-and-peck typists to use, and some of
more important musculoskeletal injury risk factor them are very expensive.
than ulnar deviation. Hunt-and-peck typists will find Vertically split keyboards (for example, Safetype
split keyboards more difficult to use, and the key- www.safetype.com). The keyboard resembles an ac-
boards are generally more expensive than conven- cordion and users type with their hands facing each
tional keyboards; they also tend to be larger and other. This design works well to reduce ulnar devi-
wider, which in some situations can put the mouse ation and wrist extension, but if the keyboard is too
too far out to the side of the keyboard. The multi- high the chest and shoulders can become fatigued.
touch fixed-angle split keyboards (Touchstream The design is nearly impossible for hunt-and- peck
www.fingerworks.com) do not use conventional keys typists to use because the keys cannot be seen easily,
but have a touch-sensitive surface that allows users and because it is a specialist keyboard it is expensive.
to key and mouse on the same physical area. This de- Chordic keyboards (for instance, BATwww
sign also allows users to control many computer com- .aboutonehandtyping.com). Chord keyboards have
mands with simple finger gestures performed on the a smaller number of keys, and letters and digits are
same physical area. It takes some time to learn but generated by combinations of keys in chords. One-
as users become proficient, the overall speed of their handed and two-handed designs are available.
computer work performance can increase by over 80 Research shows that it takes about eighty hours to
percent. gain moderate proficiency using the chords that cor-
Adjustable-angle split keyboards (for example, respond to characters. And although the keyboards
Goldtouch, www.goldtouch.com). These keyboard are more expensive than regular keyboards, they can
designs allow users to change the split angle to suit be useful to some users, especially those with special
their own needs. Often the split angle is linked to the needs, such as impaired vision or severely arthritic
degree of tenting of the keyboard as well. There is hands.
some research evidence of reduced discomfort with Specialist keyboards (for instance, Datahand
this kind of design because of reduced ulnar devia- www.datahand.com or Orbitouchwww.keybowl
tion, but these designs do not usually address wrist- .com). Several different keyboard designs have been
extension issues. The fact that users have to decide developed to assist users who have some physical
on the split angle means that they may need some limitation or who wish to type in a different way. The
training, which suggests that some users might end Datahand allows users to rest their hands on a series
up with a split angle that is inappropriate for of switches that detect different directions of fin-
them. There is also a multitouch adjustable-angle ger movements, and these generate the characters.
split-keyboard (Touchstream LPwww.fingerworks The Orbitouch lets users rest their hands on two
.com). Split keyboards are always difficult for domed surfaces and then move these surfaces to gen-
hunt-and-peck typists to use, and these designs are erate the characters. Specialist keyboards often re-
often fairly expensive. sult in slower typing and learning to use them can
Completely split keyboards (for instance, Kinesis take time, so they arent a good choice for most
www.kinesis.com). In these designs the left hand and people. And like other alternative keyboard designs,
right hand portions of the keyboard are com- they are also expensive.
pletely split apart. In some designs the keys are One-handed keyboards. Sometimes users have a
presented in a scooped design that allows the hands physical limitation, such as a missing hand, or they
to rest in a more neutral posture for typing. There is perform work where one hand needs to key while
some research evidence of reduced discomfort be- the other does something else. Several alternative de-
cause of reduced ulnar deviation and also reduced signs for one-handed keyboards are available. The
wrist extension. However, it takes time to learn to Half-QWERTY (www.aboutonehandtyping.com)
use a split keyboard, and research shows that ini- uses the same keys found on a regular keyboard, but
tial performance can suffer a 50 percent slowing of each key functions in two modes, allowing the user
THE KEYBOARD 405
to generate all the characters of a regular keyboard QWERTY layout and some have an alphabetic lay-
in a smaller area. The Frogpad (www.frogpad.com) out. Although adequate for short text messaging and
works in a similar way. One-handed chordic key- email, thumb typing is too slow for large documents
boards (for instance, Twiddlerwww.handykey.com) and overuse injuries of the thumb (for instance,
and one-handed multitouch keyboards (like Mini DeQuervains tenosynovitis) can occur with inten-
www.fingerworks.com) are also available. sive thumb keyboard use.
Conventional keyboards have also changed their
design over the past twenty years: Keyboards are flat- Alan Hedge
ter, function keys are on a top row, key mecha-
nisms have become lighter, requiring less force, and
keyboards have a cursor key pad and a numeric FURTHER READING
key pad. These features were not always available on BSR/HFES 100. (2002). Human factors engineering of computer work-
older keyboards. For the average user (average size stations. Santa Monica, CA: Human Factors and Ergonomics
and average skill), the modern conventional com- Society.
Heidner, F. (1915). Type-writing machine. Letters Patent 1,
puter keyboard is a familiar and cost-effective de- 138474. United States Patent Office.
sign, and to date, no ergonomic design has gained Klockenberg, E. A. (1926). Rationalisierung der Schreibmaschine und
widespread acceptance. ihrer Bedienung (Rationalization of typewriters and their operation).
Berlin: Springer.
Kroemer, K. H. E. (1972). Human engineering the keyboard.
Human Factors, 14, 5163.
Thumb Keyboards Kroemer, K. H. E. (2001). Keyboards and keying: An annotated bib-
The development of PDAs (personal digital assis- liography of literature from 1878 to 1999. UAIS, 1, 99160.
Office machines and suppliesalphanumeric machinesalternate key-
tants) and wireless email products, such as the board arrangement (revision and redesignation of ANSI X4.22-
Blackberry, have resulted in the development of small 1983) (formerly ANSI X3.207-1991 (R1997)). Washington, DC:
thumb-operated keyboards. Some of these have a American National Standards Institute.
LANGUAGE GENERATION
LASER PRINTER
LAW AND HCI
LAW ENFORCEMENT
LEXICON BUILDING
L
LIQUID CRYSTAL DISPLAYS
LITERARY REPRESENTATIONS
Since the 1960s, NLG technology has matured

LANGUAGE GENERATION from experimental systems for story generation to
well-established methodologies that are used in a
The success of human-computer interaction de- wide range of current applications. For example,
pends on the ability of computers to produce NLG systems are used to summarize the medical con-
natural language. Internally, computers represent ditions of patients in emergency care units based on
information in formats they can easily manipulate, the patients records, describe objects displayed in
such as databases and transaction logs. In many a museum gallery based on facts from the museums
cases, however, interpreting data in such formats knowledge base and user interests, answer questions
requires a considerable amount of effort and ex- about flight reservations presented in a database,
pertise from the user. The methodology for trans- generate explanatory captions for graphical presen-
lation of nonlinguistic data into human language tations, and produce stock market reports from a set
is the focus of natural-language generation (NLG). of stock quotes.
This research topic draws on methodologies from
linguistics, natural-language processing, and arti-
ficial intelligence, including user modeling, dis- Methodology
course theory, planning, lexical semantics, and A generation system creates a text from a semantic
syntax. input that is given in some nonlinguistic form.
407
Examples of such input are stock transactions, nu- tence chunks and selects an appropriate syntactic
merical weather data, and scores of sport events. construction for each sentence. For example, when
To transform a given nonlinguistic representation describing the performance of fallen stocks, we may
into a text, the system needs to address two issues: either generate a separate sentence for each stock
what to say (content planning) and how to say it (lin- in that category, or aggregate this information into
guistic realization). The first issue operates at the one complex sentence. At this point, the genera-
macro level; here we want to specify what our text tion system is still operating with a nonlinguistic rep-
will be about, given a rich nonlinguistic represen- resentation of information. It is the job of the lexical
tation of information. The second issue, on the mi- chooser to translate this representation into natural
cro level, deals with how to verbalize the selected language. The translation is performed using a sys-
information in natural language. Both issues raise tem lexicon that specifies how each semantic con-
multiple questions. At the macro level, the questions cept can be verbalized and what the constraints on
include: In which order should the selected topics be the different alternatives are. In the stock domain,
presented? What are the connections between dif- the concept drop (for stocks that have fallen) can
ferent topics? How should these connections be be verbalized as plummeted when the drop is more
described in the text? At the micro level, the ques- than ten points and as dipped when the drop is
tions include: Which verbalization is the most ap- around five points. Once all the information units
propriate, given the unit of semantic information are translated into words, the system still has to se-
and the context in which this unit of information lect the appropriate word ordering and insert auxil-
appears? What word ordering will produce a iary words, such as determiners and prepositions.
grammatical sentence? This last task is performed by the surface realizer,
A typical generation system has two levels. The based on the rules of natural-language grammar.
macro level consists of content selection and con-
tent organization modules. The micro level consists
of a sentence planner, a lexical chooser, and a sur- Implementation:
face realizer. To illustrate the functionality of each
component, consider the task of automatically gen- Traditional Approaches
erating daily reports of stock market activity. Since Traditionally, there have been two approaches to the
typical financial reports include not only informa- implementation of generation systems: template
tion about current transactions but also historic stock generation and multilayed linguistic generation.
information, the input to the system consists of tables Templates delineate output strings containing var-
specifying stock performance in the past and all iables that can be instantiated with particular values.
the transactions recorded for a given day. Given They are relatively easy to implement and are there-
the large amount of information, the task of the con- fore commonly employed in applications in which
tent selection component is to identify which trans- only a few different types of sentences are being gen-
actions are to be included in the generated text. erated. However, template-based approaches are not
For example, we may decide to describe a general effective in complex domains since they are not ro-
trend in the stock market on a given day and name bust to new types of input and are difficult to main-
the stocks that fluctuated the most. Once the tain as systems expand. Linguistic-based generation
information is selected, the content organization is a preferred alternative for applications in which
component groups together relevant pieces of in- variability of the output and scalability is an issue.
formation and establishes a linear order among them. The following discussion applies to linguistic-based
In our example, a target text may start with a dis- generation.
cussion of the general trend, followed by a list of
stocks that have risen or fallen the most. Content Planning
The next component of the generation system, As mentioned earlier, the task of the content plan-
the sentence planner, divides information into sen- ner is to select relevant material and to order it into
LANGUAGE GENERATION 409
a coherently flowing sequence. Content selection and tionary that lists possible words corresponding to el-
content organization are performed in a single step ementary semantic concepts. Sample entries might
by most generation systems. These tasks can be be [Parent [sex:female]], with the mappng mother,
done at many levels of sophistication. One of the most mom; or [love x, y], with the possibilities x loves
simple approaches is to write a hard-coded text plan- y, x is in love with y. Entries of the mapping dic-
ner, which produces text with a standardized content tionary can be augmented with information that en-
and structure. This approach is particularly suitable codes grammatical features of the word as well as
for domains in which text variability is not an issue, constraints on its usage, including stylistic, contex-
such as many technical domains. Other systems em- tual, and idiosyncratic constraints.
ploy artificial-intelligence techniques for planning, Finally, the linguistic realizer generates sentences
considering content selection and organization to in a grammatical manner, taking care of agreement,
be a multistep process aimed at reaching a specific morphology, word order, and other phenomena. The
communication goal. While this approach can yield linguistic realizer is the most extensively studied com-
flexible and powerful systems, in practice it is difficult ponent of generation. Different grammar theories
to implement because of the amount of knowledge have led to very different approaches to realiza-
that must be encoded. tion. Some of the grammars that have been used suc-
The most common approach today makes use of cessfully in various NLG systems include systemic
a schema. A schema is a text-planning language grammars, meaning-text grammars, and tree-ad-
that captures style-specific principles of text organ- joining grammars. Typically, NLG systems rely on
ization. It operates at the level of semantic mes- one of several general-purpose realization engines,
sages and the discourse relations that hold among such as FUF/Surge, KPML, and RealPro.
them. Typically, a schema is augmented with domain
communication knowledge that instantiates it with
semantic predicates specific to a domain. For instance, Implementation:
a schema of an encyclopedia entry may include the
following information: (1) identification of an item Corpus-Based Approaches
as a member of some generic class, (2) description of Most of the modules in existing generation sys-
an objects function, attributes, and constituency, tems are domain and application specific. For
(3) analogies made to familiar objects, and (4) ex- instance, the guiding principle for selecting infor-
amples (McKeown 1985). mation in the stock market domain is unlikely to
carry over to the weather domain. Content planners
Linguistic Realization are, therefore, typically developed anew for each ap-
The sentence planner must decide how to group se- plication; the same holds for the content organizer,
mantic units into sentences and what syntactic mech- sentence planner, and lexical chooser. Typically,
anism should be used to implement the desired human experts construct complex rule-based mod-
combinations. Although there are several obvious ules by analyzing large amounts of domain text. As
constraints on the aggregation process, such as the a result, the development of a generation system takes
length of the resultant sentences, the number of po- significant time and human effort.
tential aggregations is still vast. In most cases, hu- In recent years, the focus of research in the gen-
man experts analyze types of aggregation that occur eration community has shifted to data-driven ap-
in a corpus and then encode corpus-specific rules proaches, in which generation systems learn necessary
based on their findings. data from samples of texts. Data-driven approaches
At the next stage, lexical choice (choosing which are particularly effective for tasks in which the selec-
words to use) is commonly implemented as a rewrit- tion between alternative outputs involves a variety of
ing mechanism that translates domain concepts and constraints and therefore is hard to specify manually.
their semantic relations into words and syntactic re- Surface realization is a case in point. While some
lations. The lexical chooser relies on a mapping dic- choices in the surface realization component are
uniquely determined by grammar rules (for instance, domain-independent character of these applications
in English, the subject is always placed before the preclude the use of techniques developed for con-
verb), the realization of other grammatical constructs cept-to-text applications. Consequently, text-to-text
depends on semantic and discourse constraints, and generation systems must rely solely on input texts
in some cases this selection is idiosyncratic. Instead and knowledge that can be automatically derived
of considering the complex interaction between these from those texts.
features, we can rule out implausible candidates by Natural-language generation systems are typi-
considering corpus statistics. For instance, it is un- cally oriented towards production of written lan-
likely to find the noun phrase a money in any guage. Spoken responses, however, have different
corpus of well-formed English sentences. Based on characteristics from written ones. For example, long,
this intuition, Kevin Knight and Vasileios Hatzivas- complex sentences are usually inappropriate in speech.
siloglou developed the first statistical realizer. Their Further research is needed to incorporate the results
realizer uses a few syntactic rules to generate a lattice of work in linguistics on spoken language constraints.
of possible verbalizations of a given input and then
selects the optimal path in this lattice based on lan- Regina Barzilay
guage model scores. Today, statistical methods are ap-
plied to other modules in the generation process. See also Dialog Systems; Machine Translation; Natural-
A commonly used resource for such learning is a col- Language Processing
lection of texts annotated with semantic information.
For instance, Pablo Duboue and Kathleen McKeown
(2001) learn domain ordering constraints by ana- FURTHER READING
lyzing patterns in the distribution of semantic con-
Barzilay, R. & Lee, L. (2002). Bootstrapping lexical choice via multi-
cepts in the corresponding text. Other researchers ple-sequence alignment. Proceedings of Empirical Methods in Natural
automatically induce lexicons for generation systems Language Processing, 164171.
by aligning semantic concepts with matching phrases Biber, D. (1988). Variation across speech and writing. Cambridge, UK:
(Barzilay and Lee; Reiter, Sripada and Robertson). Cambridge University Press.
Dubuoe, P., & McKeown, K. R. (2001) Empirically estimating order
constraints for content planning in generation In Proceedings of
the ACL-EACL 2001, July 611, Toulouse, France.
Future Directions Hovy, E. H. (1988). Generating natural language under pragmatic con-
It is safe to say that at the present time one can build straints. Hillsdale, NJ: Lawrence Erlbaum.
ILEX. (n.d.). Intelligent labeling explorer: A project at the Univer-
a natural-language generation system for a specific sity of Edinburgh into dynamic hypertext generation. Retrieved
application. Future research incorporating machine- March 23, 2004, from http://www.hcrc.ed.ac.uk/ilex/
learning techniques may help speed up the devel- Joshi, A. K. (1987). The relevance of tree adjoining grammar to gen-
eration. In G. Kempen (Ed.), Natural language generation: Recent
opment and increase the coverage of NLG systems. advances in artificial intelligence, psychology, and linguistics
Most of the current methods of NLG require a man- (pp. 233252). Dordrecht, Netherlands: Kluwer Academic Publishers.
ually annotated corpus, which is not available in Kittredge, R., Korelsky T., & Rambow, O. (1991). On the need for do-
many domains. Further research in the direction main communication language. Computational Intelligence,
7(4), 305314.
of weakly supervised and unsupervised machine- Klein, S. (1965). Automatic paraphrasing in essay format. Mechanical
learning methods is required. Translation 8 (3), 6883.
Most current research in NLG focuses on the task Knight, K., & Hatzivassiloglou, V. (1995). Two-level, many paths gen-
of text production from semantic input. However, eration. In Proceedings of the 33rd annual meeting of the Association
for Computational Linguistics (pp. 252260). San Francisco: Morgan
in many applications there is a need for text-to- Kaufmann.
text generation, that is, for the rewriting of input that Kukich, K. (1983). Knowledge-based report generations: A technique
is already in a textual form. Example of such ap- for automatically generating natural language reports from
databases. In Sixth ACM SIGIR Conference (pp. 246250). New
plications include summarization, text simplifica- York: ACM Press.
tion, and information fusion from multiple texts. Mann, W. C., & Matthiessen, C. M. I. M. ( 1985). Demonstration of
Lack of semantic representation in the input and the the Nigel text generation computer program. In J. D. Benson &
LASER PRINTER 411
W. S. Greaves (Eds.), Systemic Perspectives on Discourse, 1, 5083. materials with high fidelity. Today copiers can re-
Norwood, NJ: Ablex. produce color or black and white materials with high
McKeown, K. R. (1985). Text generation: Using discourse strategies and
focus constraints to generate natural language text. Cambridge, UK:
speed and high quality. This copying process, how-
Cambridge University Press. ever, was a reproduction process only and not one
McKeown, K. Jordan, D., Feiner, S., Shaw, J., Chen, E. Ahmad, S., et al. that could readily create the original material. Much
(2002). A study of communication in the cardiac surgery intensive of the creative process was done by conventional page
care unit and its implications for automated briefing. Retrieved March
23, 2004, from http://www1.cs.columbia.edu/~shaw/papers/ assembly methods, as had been done for years.
amia00.pdf. In 19671968 Gary Starkweather, an optical en-
Melcuk, I. A., & Polguere, A. (1987). A formal lexicon in the mean- gineer at Xerox, became interested in this creative
ing-text theory (or how to do lexica with words). Computational
Linguistics, 13(34), 261275.
process and thought that a combination of optics,
Mittal, V., Moore, J., Carenini, G., & Roth, S. F. (1998). Describing electronics, and the xerographic process might solve
complex charts in natural language: A caption generation system. the creative problem. One can make up pages us-
Computational Linguistics, 24(3), 431467. ing rectangular or circular points, but this often re-
Reiter, E., Sripada, S., & Robertson, R. (2003). Acquiring correct knowl-
edge for natural language generation. Journal of Artificial Intelligence
quires using multiple-sized shapes, which is often
Research, 18, 491516. a slow process. Starkweather believed that the ideal
Seneff, S., & Polifroni, J. (2000). Dialogue management in the Mercury approach would be to create page images using points
flight reservation system. Paper presented at the Satellite Workshop, or zero dimensional objects. Additionally, if one
ANLP-NAACL 2000, Seattle, WA.
could create the pages fast enough, such a process
would permit not only creation of the original or
master image but also the copies. Books and other
types of documents, for example, could be printed
LASER PRINTER as needed and with the type sizes required at the time
of need.
Ever since the German inventor Johannes Gutenberg Two critical problems had to be solved, however.
invented movable-type printing in 1436, printing The first was how to make an imaging system that
has become an increasingly valuable technology. The generates the points precisely and at the right posi-
challenge in printing has always been how to format tions and how to determine when the spot should
and assemble the elements of the page to be printed and should not be generated; the second was de-
in a form that allows rapid replication of large num- signing a digital system that could generate the data
bers of books, papers, documents, and so forth. Three stream with which to drive the imaging system.
basic page elements exist: text, graphics, and pictures. For example, for white areas of the page, no points
Many page creation devices can produce one or per- should be printed, and for black areas points should
haps two of these elements but not all three. For a be printed. Some process had to provide the correct
new technology to be truly useful, it had to be able signals to the spot generator and at sufficient speeds.
to produce all three elements in high quality and Three critical technologies were becoming available
in a rapid manner. during this time frame. The first was the digital com-
puter. This technology was not as yet a personal tech-
nology, but the start of the personal technology was
Invention of the Laser Printer there. The second technology was the integrated cir-
In October 1938 the U.S. inventor Chester Carlson cuit, which started with the U.S. scientist Jack Kilby
invented the process known as xerography. This in 1968. The third technology was the laser, invented
process was the driving force behind the creation and in 1961 by Arthur Schalow and Charles Townes. The
growth of the copier industry by Xerox Corporation. laser was critical because without the high bright-
The word xerography was coined from two Greek ness capability of the laser one could not expose the
words, xeros and graphein, which mean dry writ- required points of light fast enough. The computer
ing. The xerographic process, also known as electro- and integrated circuit would eventually combine
photography, was capable of reproducing printed to permit computing the image at a fast enough rate.
nology resulted in the worlds first commercial

laser printer in 1977, known as the Xerox 9700. This
printer generated ninety thousand image points
per 6.4 square centimeters and produced two pages
per second.
Technical Principles
A laser printer is a rather eclectic assemblage of tech-
nologies. First, one needs a light source with which
to illuminate the photoconductor of an electro-
photographic machine such that sufficient energy is
deposited in the available time. This time can be as
brief as a few billionths of a second. Additionally, an
A Hewlett Packard LaserJet 4050T. optical system is needed to focus the laser beam to a
spot size of a couple of thousandths of an inch across
the entire image, which might be 28 by 43 centimeters
or even larger. In order to image the beam across this
Why was such speed needed? The vision for size at sufficient speed, one must have a beam de-
the laser printer required producing pages at a rate flector. Since the invention of the laser printer, the
of about one page per second. The number of points principal deflector technology has been the poly-
required to produce a reasonable page was about gonal scanner. This can be thought of as a disc
ninety thousand per 6.4 square centimeters mini- with flat mirrors on its periphery. As the disc spins,
mum. If the standard page to be printed had a for- each mirror in its turn intercepts the light beam and
mat of 20 by 25 centimeters, then at least 80 by 90,000 scans it across the region to be imaged. To under-
or 7.2 million points of light would have to be com- stand the extent of the task, if one is to image one
puted and printed in one second. Would this be pos- page 28 centimeters long per second at six hun-
sible? At the time when Starkweather was working dred scans per inch, the scanner must deliver
on the printer technology, other scientists and engi- about 6,600 scans per second. As the scans occur, a
neers were assessing computer technology. Xerox in small optical detector synchronizes the data with the
1970 established a research center in Palo Alto, beam position. For a scan of 28 centimeters to oc-
California, that became famous as Xerox PARC. cur at a page rate of one page per second, the opti-
Starkweather transferred to Xerox PARC in 1971 and cal beam is moving at about 1,300 meters per second
combined his efforts with those of Charles Thacker, or about Mach 4. The laser, of course, must be turned
Butler Lampson, and others who were working on a on and off to correspond to the data to be printed,
personal computer that became known as the Alto. and some lasers, such as gas lasers, require a beam
In order for the laser printer to generate points modulator. Smaller laser printers such as those used
at a rate of at least 7 million per second, the laser beam in personal applications utilize solid state lasers that
had to be focused and scanned across the surface of can be directly modulated and do not require ex-
a sensitive photoconductor material, as used in Xerox ternal modulators. Although such systems sound
copiers. A polygonal mirror was used and combined complex, research and engineering work has ren-
with a novel optical system to generate the large num- dered such scan subsystems reliable today.
ber of points precisely and reliably. The first laser Subsequent refinement of laser technology
printer used at Xerox PARC was combined with the pioneered at Xerox PARC resulted in personal laser
Alto computer and some special interface electron- printers at lower speeds but at lower costs as well.
ics to produce a printer that generated 15 million The first personal laser printer was known as the
points per second. Gradual refinement of this tech- Hewlett-Packard LaserJet and used the basic de-
LAW AND HCI 413
sign pioneered in Starkweathers earlier work. Later, on-demand, real-time, electronic transmission of
Hewlett-Packard, Xerox, Canon, Apple computer, information and knowledge became the standard
and others developed laser printers with higher page mode of correspondence and its major currency, with
quality. Today one can purchase laser printers that time a premium and speed a commodity. Real estate
image in color and print at page rates ranging became invisible and the value of property assets
from a minimum of four pages per minute at low in the global marketplace determined by whether
cost to as high as 180 pages per minute. domain name ownership was in a dot.com, dot.net,
Recently Hewlett-Packard announced that it had or dot.org. Privacy was stripped, security breached,
shipped its 30 millionth laser printer. Today the great crime pervasive, theft untraceable, identity trans-
bulk of electronic printing is done with laser print- parent, and piracy commonplace in the concealed
ers. The newest applications of laser printers involve world of cyberspace. Success was ascertained by
what is known as demand or sort-run printing, the speed of deliverables, and power became pro-
by which several thousand documents can be gen- grammable. Robust communication now could be
erated as needed. Even books are now beginning to conducted with and through computers, robots, in-
be demand printed, thus fulfilling Starkweather and formation systems, and the Internet, not just with
Xerox PARCs vision of what the personal computer people. As David Johnson and David Post assert,The
combined with laser printer could become. rise of an electronic medium that disregards geo-
graphical boundaries throws the law into disarray by
Gary K. Starkweather creating entirely new phenomena that need to be-
come the subject of clear legal rules but that can-
See also Alto; Fonts not be governed, satisfactorily, by any current
territorially based sovereign (Johnson and Post 1996,
1367, 1375).
FURTHER READING How could virtual reality, called cyberspace
(Gibson 1984, 51), be legally harnessed and this world
Elzinga, C. D., Hallmark, T. M., Mattern Jr., R. H., & Woodward, J. M. of downloads, networks, interfaces, and avatars un-
(1981). Laser electrophotographic printing technology. IBM Journal
of Research and Development, 25(5), 767773.
derstood by the courts? Over which space would
Fleischer, J. M., Latta, M. R., & Rabedeau, M. E. (1977). Laser- jurisdiction attachcyberspace or real spaceor
optical system of the IBM 3800 printer. IBM Journal of Research multiple non-coordinating jurisdictions? In partic-
and Development, 21, 479. ular, how could the law keep pace with the rapid pro-
Laser printing. (1979). SPIE Proceedings, 169, 1128.
Starkweather, G. K. (1980). High speed laser printing systems. Laser liferation of ubiquitous, high-bandwidth, embedded,
Applications, 4, 125189. miniaturized, portable, and invisible dissolution of
Starkweather, G. K. (1985). A high resolution laser printer. Journal high-functionality systems in the environment and
of Imaging Technology, 11(6), 300305.
Urbach, J. C., Fisli, T. S., & Starkweather, G. (1982). Laser scanning for
its accompanying vulnerabilities? Machines had be-
electronic printing. Proceedings of the IEEE, 70(6). come complex interacting systems, sometimes hav-
ing a mind of their own and at times failing as a result
of bugs. Now computers were interacting to form
networks. Just as human-computer interaction (HCI)
was spawned from the Information Age, so, too,
LAW AND HCI did new legal practice areas evolveInternet law as
well as unprecedented issues such as the copyright of
Knowledge is power. However, who owns knowl- a computer program and the patent of a click.
edge? Knowledge is something that can be sold while
simultaneously kept and whose value can either ap-
preciate or vanish through time. With the Internet A Brief History of Law
geographic constraints on and distinctions between In the United States rules are established and en-
communication and computation blurred. Online, forced through one of three legal systems found at
the federal, state, and local levels. The Constitution (described below) were recognized as new forms
together with the laws of Congress, decisions by of agreements. Pornography was freely downloaded.
the federal courts, executive orders of the president, Music was easily swapped. Proprietary writings were
and regulations adopted by the executive agencies infringed. Distance learning would enter the educa-
constitute the federal level. In addition, each of the tional fray, with the extensible enterprise becom-
fifty states has its own constitution. On that level laws ing the business norm. Snail mail would be the last
can be enacted by state legislatures, decisions held resort used to reach out and touch someone. Change
by courts, and regulations promulgated by their would be the only constant. The letter e would en-
respective agencies. Laws created by either federal or ter the mainstreame-commerce, e-mail, e-sign,
state court are usually based on precedenceor the e-governance, e-discovery, e-courts. The legal sys-
past, which serves as an example of prior decisions tem would be challenged to create new practice ar-
made on the same or similar issues. That fact is why eas while trying to understand and interpret evolving
the United States is characterized as a common law communication standards and protocols.
country, with the theory that through time laws will
adapt to new circumstances.
New issues without precedence to guide the Privacy
courts create what is commonly known as a case of One contentious issue surrounding the Internet is
first impression. privacy. Although advocacy groups allege that a right
The Information Age raised new concerns that to privacy exists, legally it does not. The legal con-
continue to perpetuate at all three levels of the le- cept of the right to privacy can first be found in an
gal system. HCI will up the ante even more as the in- 1890 Harvard Law Review article entitled The Right
tegration of human with machine becomes ever more to Privacy, written by Samuel Warren and Louis
pervasive and invisible and our virtual realities grad- Brandeis when they were law firm partners. Warren
ually augmented as thinking, feeling, and intelligent and Brandeis claimed that the right to privacy al-
systems and environments become commonplace. ready existed in the common law and gave each per-
son the choice to share or not to share information
about his or her private life. Their intent was merely
Personal Computing to establish the right to privacy as a legal protec-
Considered a revolutionary device at the time, the tion in their day. Neither did either man coin the
personal computer (PC) began to gain widespread phrase the right of the individual to be let alone, as
acceptance by consumers and businesses alike in found in U.S. Supreme Court Justice Brandeiss dis-
1981. One click was about to change our lives for- sent in Olmstead v. United States (1928), which is of-
ever. The word mouse would take on an entirely new ten quoted by privacy champions and is the first case
meaning. The word Google would enter our lexi- in which the U.S. Supreme Court considered the con-
con to signify a new research methodthat of a stitutionality of electronic surveillance. Warren and
search engine. Internet service providers (ISPs) would Brandeis in their 1890 article interpreted the Fifth
become commonplace as they granted us entrance Amendment to the United States Constitution: No
beyond the physical world and permission into a person shall . . . be deprived of life, liberty, or prop-
global cyberspace. erty, without due process of law . . . to read that a
However, the PCs magnitude, scope, social im- person has an inherent right to be let alone and to
pact, and legal implications were yet unrealized. Our privacy. Their interpretation was their legal theory
world was about to become electronically wired, net- and their view of a more general right to enjoy life.
worked, sensored, interfaced, and imaged. Signatures Even so, with the onset of the Internet several
could be obtained electronically. Auctions, gambling, well-recognized organizations were formed to assert
sweepstakes, promotions, and games could be played a persons rights within a network of global commu-
online. Contracts, negotiations, and disputes could nication: The Center for Democracy and Technology
be altered in an instant. Clickwraps and shrinkwraps (CDT) was established to promote democratic values
LAW AND HCI 415
and constitutional liberties in the digital age; the sharing the customers personal information with
Electronic Frontier Foundation (EFF) was estab- third parties.
lished to defend the right to think, speak, and The privacy requirements of the GLBA are di-
share ideas, thoughts, and needs using new tech- vided into three principal parts: the Financial Privacy
nologies, such as the Internet and the World Wide Rule, the Safeguards Rule, and pretexting provisions.
Web; and the Electronic Privacy Information Center Eight federal agencies, together with the states, have
(EPIC), a public interest research center, was es- authority to administer and enforce the Financial
tablished to focus on emerging civil liberties issues Privacy Rule and the Safeguards Rule, which apply
and to protect privacy, the First Amendment, and to all financial institutions.
constitutional values. Although the Financial Privacy Rule governs per-
Just as no right to privacy exists, no privacy pol- sonal financial information collected and disclosed
icy is required to be posted on a websiteconsidered by financial institutions, it also applies to non-financial
to be a form of online advertising. However, should companies that may receive such information. The
a privacy policy be displayed on an personal website Safeguards Rule requires financial institutions that
or a business website, it then may be subject to collect customer information, as well as those that re-
civil liability or criminal sanctions should the owner ceive it from other financial institutions, to design,
not abide by its own policies. implement, and maintain safeguards to protect that
The U.S. Federal Trade Commission (FTC) is information. The pretexting provisions protect con-
charged with guarding against unfairness and decep- sumers against companies that have obtained per-
tion (Section 5 of the Federal Trade Commission Act, sonal information under false pretenses such as calling
15 U.S.C. 41-58, as amended) by enforcing privacy a bank pretending to be a customer, also known as
policies about how personal information is collected, pretexting.
used, shared, and secured. In its 1998 report, Privacy The Childrens Online Privacy Protection Act
Online: A Report to Congress, the FTC described the is designed to offer parents control over information
fair information practice principles of notice, choice, gathered online and provided by their children
access, and security in addition to enforcementto and the subsequent use of that information. COPPA
provide sanctions for noncomplianceas critical com- applies to commercial websites that collect personal
ponents for online privacy protection. information from children under the age of thirteen,
Today the FTC plays a central role in imple- requiring that the websites follow several rules to
menting rules and safeguarding personal informa- safeguard a childs privacy while obtaining the par-
tion under the Gramm-Leach-Bliley Act (GLBA), the ents consent before collecting such personally iden-
Childrens Online Privacy Protection Act (COPPA), tifiable information. Any website directed at children
and the Fair and Accurate Credit Transaction Act under the age of thirteen must comply with COPPA.
(FACTA). The Fair and Accurate Credit Transaction Act,
The Gramm-Leach-Bliley Act, also known as the signed into law by President George W. Bush in
Financial Modernization Act of 1999, was enacted December 2003, amends the Fair Credit Reporting
to protect personal information held by a financial Act (FCRA) by requiring the nationwide consumer
institution. The act applies to banks; securities firms; reporting agencies (CRAs) to provide a yearly credit
insurance companies; consumer loan lenders, bro- report at no cost to consumers. FACTA prohibits a
kers, and servicing entities; companies preparing in- CRA from circumventing such a requirement by
dividual tax returns or providing financial advice, clearly illustrating what constitutes circumvention.
credit counseling, or residential real estate settlement
services; debt collectors; and enterprises transferring
or safeguarding money. It requires these institutions Spam
to provide privacy notices with an opt-out provision In addition to enforcing privacy policies, the FTC en-
to their customers. If the opt-out provision is cho- forces the Controlling the Assault of Non-Solicited
sen by a customer, the institution is prohibited from Pornography and Marketing Act of 2003 (CAN-SPAM
Act) to combat unsolicited commercial e-mail access to long term care services and coverage, to
advertising, fraudulent and deceptive chain letters, simplify the administration of health insurance, and
and pyramid and other get-rich-quick schemes. The for other purposes. The Office of Civil Rights is
CAN-SPAM Act additionally includes a protection responsible for enforcement of HIPAA. It establishes
against unmarked sexually oriented or pornographic a framework for the standardization of electronic
material. data interchange (EDI) in health care, in particu-
Signed into law on 16 December 2003 and effec- lar, protections for the privacy and security of in-
tive 1 January 2004, the CAN-SPAM Act, an opt- dividually identifiable health information (IIHI).
out law, impacts all U.S. online businesses marketing Compliance with the HIPAA Security Rule is required
their services or products through e-mail transmis- to be met by health care plans, providers, and
sions, defining a commercial electronic mail message clearinghouses in 2005.
as one whose primary purpose is the commercial ad- With HIPAA the federal government introduced
vertisement or promotion of a commercial product a complex regulatory scheme with broad implica-
or service, including content on an Internet website tions for health plan administration and health plan
operated for a commercial purpose. sponsors. It subjected not only health care entities,
The CAN-SPAM Act governs nearly any business but also health care providers that conducted cer-
e-mail, including electronically submitted newslet- tain electronic transactions, along with health care
ters and stand-alone promotional e-mails. It pro- clearinghouses, group health plans, group plans, and
hibits fraudulent or deceptive subject lines, headers, plan sponsorsin other words, all employers who
or returned address, requires that e-mail advertisers provide group health plans for their employeesto
identify their messages as advertisements or solici- the Privacy Rule. It forbade an employer from ob-
tations in a clear and conspicuous manner, and retaining IIHI from its health plan unit and using it to
quires that a postal mailing address be included in decide work assignments, promotions, firings or lay-
the e-mail message. These requirements apply not offs, employee discipline, or any other employment-
only to spammers, but also to those people or busi- related issue.
nesses that may procure spammers services. Non- To be HIPAA compliant, an employer has to pro-
compliance with the CAN-SPAM Act could result in vide notice of his or her privacy policies and prac-
civil enforcement by the FTC or state attorneys gen- tices, together with confirmation of receipt of such
eral, resulting in both criminal sanctions and civil notice; designate a privacy officer; train personnel
penalties. ISPs may bring civil lawsuits against vio- handling IIHI in privacy and security compliance;
lators who adversely affect those providers. The FTC have a documented policy and procedure for privacy
now is considering the establishment of a national and security violations; have in place mechanisms
Do not spam list similar to the Do not call reg- for sanctioning employees violating the privacy and
istry restricting telemarketing phone calls. security policies; allow individuals the right to ac-
cess, amend, and receive accountings of their IIHI;
establish procedures for mitigating harmful effects
Health Information Privacy of improper uses or disclosures of IIHI; and have
On 14 April 2003, the U.S. Department of Health whistleblower provisions in place to not retaliate
and Human Services (HHS) mandated compliance against those people who may exercise their rights
with the Privacy Rule of the Health Insurance under the Privacy Rule. In addition, an employer
Portability and Accountability Act of 1996 (HIPAA), now was obligated to establish firewalls to ensure
Public Law 104-91. HIPAA amends the Internal that IIHI handled by his or her group health plan or
Revenue Code of 1986 to improve portability and plan sponsor was segregated from the rest of the em-
continuity of health insurance coverage in the group ployers operations. Pursuant to the Privacy Rule, a
and individual markets, to combat waste, fraud and person, entity, or third-party administrator involved
abuse in health insurance and healthcare delivery, to in any activity involving the use or disclosure of IIHI
promote the use of medical accounts, to improve now was required to sign a business associate agree-
LAW AND HCI 417
ment, thereby adding another layer to the already ex- telephone calls. One legal concern is how VoIP should
tensive requirements of the Privacy Rule. be regulated and global laws harmonized.
The U.S. Federal Communications Commission
(FCC) is investigating the regulatory status of VoIP
Radio Frequency Identication to identify the migration of communication services
Radio frequency identification (RFID) tags have rev- to Internet-based platforms. The result of the in-
olutionized the concept of HCI by changing how vestigation will have a direct impact on taxation.
computing works, affecting data collection and in- However, because the data bits are encrypted,
ventory management. RFID devices, using a varia- and no standardized method exists for distin-
tion of a bar code with smart chips and wireless guishing voice calls from the terabits (one trillion
capabilities, can track, monitor, search, and scan bits) of other data on the Internet, the technical
people continuously without their knowledge, as if limitations of VoIP preclude law enforcements abil-
people were wearing a sign on their back that flashes: ity to wiretap conversations and accurately locate
I was here, thought about this, and purchased that. 911 calls. The Federal Communications Commis-
Anonymity disappears. In addition to privacy con- sion is investigating whether Internet telephone
cerns, RFID opens a new target area for hackers at- providers will need to rewire their networks to gov-
tempting to penetrate through security. ernment specifications in order to provide law
As of 2005, the U.S. Department of Defense (DoD) enforcement with guaranteed access for wiretaps.
is requiring its suppliers to begin using RFID devices, The result will call into question whether VoIP is
and the Food and Drug Administration (FDA) is a phone service. The 1994 Communications Assist-
encouraging their adoption among drug wholesalers, ance for Law Enforcement Act (CALEA) will need
manufacturers, and retailers. The challenge to imple- to be revisited in order to address the law en-
menting the use of RFID tags will be coordinating dis- forcement and national security issues raised by
parate databases into a synchronized infrastructure these applications.
to manage the data. Handheld readers containing
RFID capabilities to collect data already generated are
becoming a reality. Although the technology dates E-commerce Jurisdiction
as far back as World War II, todays RFID applications The global information highway has no rules of
are expected to change how retail business is con- the road that apply to all participants. International
ducted, resolving problems such as shoplifting, in- boundaries are nonexistent to those people who con-
ventory shortages, and logistical errors while reducing duct business from and through a website on the
manual labor, inventory checks, and the scanning of Internet. Local regulations have international ram-
bar codes. ifications. In such an ethereal realm, the effects of
online conduct may be felt at the other end of the
world. By establishing an online presence, a business
Internet Telephony may be subjected to the laws and courts of juris-
The Internet enables the transmission of telephone dictions outside the location of its operations. The
calls using the Internet protocol (IP), the same pro- basis for such jurisdictions may be either subject mat-
tocol that sends data from one computer to another. ter or personal. Whereas subject matter jurisdic-
Placing a telephone call over voice over IP (VoIP) tion refers to the competence of a particular court
requires a network connection and a PC with a to hear a particular issue, personal jurisdiction de-
speaker and a microphone. In some cases, software termines whether a defendant can be brought into
may also be required. VoIP, through fiber-optic net- the court that claims to have actual subject matter
works and broadband connections, will have the jurisdiction.
capability to tie together voice with e-mail, instant One area of subject matter jurisdiction involv-
messaging, videoconferencing, and caller ID as well ing e-commerce is the illegal distribution of copy-
as reduce the cost of long-distance and international right materials, giving the U.S. federal courts potential
rights to hear infringement cases involving viola- As a result of corporate ethical failures high-
tions of the U.S. Copyright Act. Cyberspace trans- lighted by the scandals involving the energy com-
actions often involve matters of personal jurisdiction. pany Enron and the accounting company Arthur
However, for website owners, the issue of most con- Andersen, President George W. Bush signed into law
cern is the scope of expanded special jurisdiction the Sarbanes-Oxley Act of 2002 (SOX), affecting cor-
whether and to what extent the courts of a state other porate governance, internal controls, and disclosure
than the state of the website owners incorporation obligations for publicly traded companies. SOX cre-
or principal office assert jurisdiction simply because ated new crimes, with severe penalties for the de-
the residents of that state or country can access struction, alteration, or tampering of records to
that owners website. One such issue rose to a influence, obstruct, or impede a federal investiga-
criminal prosecution case by a court in Paris, France, tion, bankruptcy case, or official proceeding with
and was closely watched around the world. In 2000, intent to alter, destroy, or conceal such records or
Frances Union of Jewish Students and the Inter- documentspunishable by imprisonment of up to
national Anti-Racism and Anti-Semitism League twenty years.
sued Yahoo for selling Nazi paraphernalia on its auc- As a result of SOX and HCI, computer forensics
tion pages. French criminal statutes prohibit the pub- developed as a field to analyze a companys computer
lic display of Nazi-related uniforms, insignia, or servers to discover evidence of wrongdoing. The
emblems, and so the French court issued an order Arthur Andersen company was convicted because a
directing Yahoo to deny access to the Nazi artifacts jury found that it had destroyed documents after it
by Internet users in France, demanding it re-engineer had become aware of a Securities and Exchange
its United States content servers for recognition of Commission investigation of its client Enron. Docu-
French Internet Protocol addresses or it would other- ment retention and records management now
wise face severe penalties. Although the French crim- were of utmost importance. Courts were requiring
inal court dismissed all charges, the case would have companies to produce electronic documents in lit-
opened up Internet providers to possible prosecu- igation discovery. The party who would bear the cost
tion anywhere in the world even if their activities of retrieving electronic documents, and the risks as-
were legal in their home base. sociated with spoliation of evidence and failure to
Because laws are not harmonized, civil, crimi- preserve documents, was determined by the
nal, and regulatory jurisdictions over cyberspace courts, as evidenced in Zubulake v. UBS Warburg LLC
overlap. Laws do not default from one country to the (2003), a suit over gender discrimination and illegal
other, causing jurisdictional conflicts and potential retaliation.
global risks, in particular, transborder cybercrimes
such as unauthorized system intrusions, online fraud,
intellectual property and identity theft, cyberter- Computer Contracts
rorism, stalking, manipulation of data, and economic The anticipatory aspect of the law is best seen through
espionage. the drafting of contracts. In negotiating a contract,
lawyers attempt to foresee the future by predicting
what may happen between the parties and provide
Sarbanes-Oxley Act of 2002 for contingencies by stipulating a remedy to protect
E-mail can be deleted, but it never really and truly their client.
ceases to exist. Once posted, electronic content is im- Unlike most contracts by which services are pro-
mortal. It is forever and permanently in cyberspace. vided or title to a product is sold, title to software re-
It can be retrieved by employers or prosecutors at mains with the vendor who grants a license for its
any time in the future. Your mouse footprints are use. A license allows someone other than the owner
electronically dropped with every website visit and the right to use the software in limited ways, through
purchase. the conveyance of a software license agreement.
LAW AND HCI 419
If the software is prepackaged for mass mar- the DMCA have appeared and continue to appear
keting and bought off the shelf, then an agreement before the courts.
is included with a consumers purchase. Merely open-
ing the box containing the software, or using it, con-
stitutes assent to the terms and conditions of the The Future
agreement. No signature is necessary between the The invention of computers took engineering com-
parties. These types of agreements are called plexity into a whole new realm. The complexity is
shrinkwrap or self-executing licensesthe soft- being driven by the technology but also, even more
ware and the agreement are in the same box, and the importantly, by the new ways people want to use tech-
terms and conditions are nonnegotiable. nology. Computer science will borrow from biology.
The other type of software license agreement is Within the next fifteen years microprocessors may
a clickwrap or point-and-click agreement. When a become obsolete, making room for molecular elec-
person visits a website, before downloading a docu- tronics (nanocomputing), redefining what is meant
ment or a particular software, often the website re- by a computer and reducing it to the size of a
quires that the person agree or not agree to the terms blood cell. Equipped with nanotube transistors
and conditions of the agreement. In order to have ten thousand times thinner than a human hair
a valid agreement, the person must give conspicu- computers may one day be able to mimic the cells
ous assent to the terms; otherwise no rights will be ability to self-replicate and outperform the most ad-
licensed. vanced models of silicon transistors, increasing pro-
cessing capabilities multifold.
Intellectual Property Sonia E. Miller

In the area of HCI, and clearly in all science and tech-
nology innovation, intellectual property plays a See also Political Science and HCI; Privacy
key role, defining not only what is owned, but also
what can be owned. A legal practice area, intellectual
property covers patent law, copyright and trademark
law, and trade secrets. Patent law is a specialized field FURTHER READING
within intellectual property. In order to be admitted
Agre, P. E., & Rotenberg, M. (1997). Technology and privacy: The
to practice patent law before the U.S. Patent and new landscape. Cambridge, MA: MIT Press.
Trademark Office, an attorney must have a science Auletta, K. (1995). The highwaymen: Warriors of the information su-
or engineering background and fulfill additional perhighway. New York: Harcourt Brace.
Ballon, I. C. (2001). E-commerce and Internet law: Treatise with forms,
requirements. 1, 2, & 3. Little Falls, NJ: Glasser Legal Works.
The development of information technology Battersby, G. J., & Grimes, C. W. (2001). Drafting Internet agreements.
continues to have an enormous impact on intellec- New York: Aspen Law & Business.
tual property law and rights. One key result of the Brin, D. (1998). The transparent society: Will technology force us to
choose between privacy and freedom? Reading, MA: Addison-Wesley.
impact of HCI on intellectual property was the Brooks, F. P. (1975). The mythical man-month. Reading, MA: Addison-
enactment of the Federal Digital Millennium Wesley.
Copyright Act (DMCA) in 1998. The DMCA was Burnham, S. J. (1987). Drafting contracts. Charlottesville, VA: Mitchie
enacted as an attempt to begin updating national Company.
Bush, V. (1945, July). As we may think. The Atlantic Monthly
laws for the digital age. Its goals were to protect (pp. 101108).
intellectual property rights and promote growth and Davies, S. (1997). Re-engineering the right to privacy: How privacy
development of electronic commerce. The DMCA has been transformed from a right to a commodity. In P. E. Agre
& M. Rotenberg (Eds.), Technology and privacy: The new landscape
is used to stop circumvention technology, with hefty (p. 143). Cambridge, MA: MIT Press.
civil and criminal liability for bypassing circum- Dertouzos, M. L. (1997). What will be: How the new world of infor-
vention technology. Many contentious cases citing mation will change our lives. New York: Harper Edge.
Dertouzos, M. L. (2001). The unfinished revolution: Human-centered Warren, S., & Brandeis, L. (1890). The right to privacy. Harvard Law
computers and what they can do for us. New York: HarperCollins. Review, 4(5), 193.
Fernandez, R., & Picard, R. W. (2003). Modeling drivers speech un- Court Cases
der stress. Speech Communication, 40, 145159. Olmstead v. United States, 277 U.S. 438, 473 (1928).
Garfinkel, S. (2000). Database nation: The death of privacy in the Zubulake v. UBS Warburg LLC, 02 Civ. 1243 (SAS) (S.D.N.Y. May 13,
21st century. Sebastopol, CA: OReilly & Associates. 2003).
Gelernter, D. (1992). Mirror worlds. New York: Oxford University Press.
Gershenfeld, N. (1999). When things start to think. New York: Henry
Holt.
LAW ENFORCEMENT
Gibson, W. (1984). Neuromancer. New York: Ace Books.
Glancy, D. J. (1979). The invention of the right to privacy. Arizona Law
Review, 21(1), 139.
Harris, M. S. (2002). Update on e-commerceJurisdiction. NY Business
Law Journal, 6(1), 2128. Information technology (IT) has the potential to rev-
Johnson, D., & Post, D. (1996). Law and bordersThe rise of law in olutionize the work done by law enforcement.
cyberspace. Stanford Law Review, 48, 1367. Although information has always been a cornerstone
Kiesler, S., & Hinds, P. (2004). Human-robot interaction. Human-
Computer Interaction, 19(12), 18.
for police and other law enforcement agencies, un-
Lessig, L. (1999). Code and other laws of cyberspace. New York: Basic til recently such agencies have not typically viewed
Books. information systems as a valuable asset. As IT has
Martin, J. (2000). After the Internet: Alien intelligence. Washington, proliferated in the private sector, IT in most police
DC: Capital Press.
Meyer, C., & Davis, S. (2003). Its alive: The coming convergence of in- organizations is still in its infancy.
formation, biology, and business. New York: Crown Business. However, as the world changes and as govern-
Miller, S. E. (2003). A new renaissance: Tech, science, engineering and ments review national and local security processes,
medicine are becoming one. New York Law Journal, 230(70),
57.
the need for police to capitalize on cutting-edge tech-
Monassebian, J. (1996). A survival guide to computer contracts: How to nologies has never been greater. The challenges for
select and negotiate for business computer systems. Great Neck, NY: system and interface designers developing these
Application Publishing. technologies stem from the distinctive context of po-
Moore, G. E. (1965). Cramming more components onto integrated
circuits. Electronics, 38(8).
lice work and how this context affects how infor-
Moran, T. P., & Dourish, P. (2001). Context-aware computing. Human- mation technology is used by police.
Computer Interaction, 16(24), 18.
Myers, B. A. (1998). A brief history of human computer interaction
technology. ACM Interactions, 5(2), 4454.
Picard, R. W. (2003). Affective computing: Challenges. International The Use of Information
Journal of Human-Computer Studies, 59(12), 5564.
Picard, R., & Healey, J. (1997). Affective wearables. Personal Technologies: in the Police Context
M.I.T. Media Laboratory Perceptual Computing Section Technical
Report, 467(1), 231240.
Information is a central feature of modern societies,
Picard, R. W., Vyzas, E., & Healey, J. (2001). Toward machine emo- and it is the central feature of policing. Police agen-
tional intelligence: Analysis of affective physiological state. IEEE cies use information to determine resource alloca-
Transactions on Pattern Analysis and Machine Intelligence, tions to police divisions, to determine when and
23(10), 11751191.
Rosen, J. (2000). The unwanted gaze: The destruction of privacy in
where to patrol, and to determine who might be in-
America. New York: Random House. volved in a crime and which crimes might be solved.
Rotenberg, M. (2000). The privacy law sourcebook 2000: United States Aggregation of information in the form of statistics
law, international law, and recent developments. Washington, DC: also helps police and the community better un-
Electronic Privacy Information Center.
Schneier, B. (2000). Secrets and lies: Digital security in a networked derstand important trends in crime in specific ar-
world. New York: Wiley Computer Publishing. eas. The knowledge of crime trends in their cities
Shneiderman, B. (2002). Leonardos laptop: Human needs and the new help police understand the underlying problems in
computing technologies. Cambridge, MA: MIT Press.
Siebel, T. M., & House, P. (1999). Cyber rules: Strategies for excelling at
neighborhoods. Being able to plan and allocate re-
e-business. New York: Doubleday. sources to areas with higher needs leads to more
Waldrop, M. M. (2003). Autonomic computing: An overview of the con- proactive crime prevention. Decision making at
cept and exploration of the public policy implications (Woodrow all levels of police agencies is based upon police
Wilson International Center for Scholars Foresight and Governance
Project, 2003-7). Washington, DC: Woodrow Wilson.
information.
LAW ENFORCEMENT 421
Information is also used in police investigations Controlling Systems (1990), Diana Gordon reported
not only to identify and apprehend a criminal sus- that in the United States in 1989, the federal National
pect, but also to provide evidence in a court of law Crime Information Center (NCIC) database housed
to convict those guilty of a crime. Police put together more than 20 million records and processed more
pieces of fragmented information to try to under- than 1 million transactions per day.
stand the events that have occurred in order to iden- In addition to text-based data, police use tech-
tify and apprehend an offender. In order to convict nologies that capture multimedia data, such as tele-
an offender, police might testify in a court of law, phone conversations, surveillance camera video, and
presenting information that provides evidence crime scene pictures. Police also use geographical in-
that the offender is indeed guilty of a crime. The formation systems that allow them to map time and
more complete and irrefutable the information, the space information for crime analysis. The develop-
more likely an offender is to be convicted. The pieces ment of information technology that manages multi-
of information build a case and provide the sub- media information becomes more important as these
stantiation that enables police to catch an offender. types of data become more prevalent.
The ability of police to share information both Information technology facilitates the sharing
within and outside of their agency is another way by of information. Because criminal activity is not bound
which information can be accessed and utilized. by geographical jurisdictions, police realize that the
Although police must acquire information, meth- ability to share information across agencies is im-
ods of handling and presenting it are of utmost portant. As more police organizations store data in
importance. In the private sector companies protect electronic systems, sharing information through net-
their competitive secrets. A breach of such secrets works becomes more important and more feasible.
can result in a decline in a companys profitability. The federal government has developed several
In the law enforcement domain, the violation of po- national initiatives to help U.S. law enforcement agen-
lice information may have more serious conse- cies deal with issues in the development and use of
quences, such as the inability to convict a dangerous information technology. For example, the Office of
felon or even the death of an officer or innocent by- Justice Programs Integrated Justice Information
stander. Consequences of errors in handling infor- Technology Initiative (part of the U.S. Department
mation can lead to violation of individual rights and of Justice) was developed in 1997 to coordinate fund-
physical harm. Legal and moral issues include the ing and technical assistance to support the design
incarceration of a wrongly accused person and the and implementation of information technology for
inability to incarcerate the real perpetrator of a crime. information sharing. The National Institute of
Justices Office of Science and Technology has also
developed programs to address issues in systems in-
Importance of IT teroperability to facilitate information sharing
through the use of IT.
in Police Organizations
Given the value of information in all aspects of
police organizations and the importance of how in- Challenges of Designing IT for Police
formation is handled in this context, police are rely- Given its importance, IT is becoming less of an en-
ing on IT to manage their information needs. Although hancement and more of a necessity for police. To take
a large number of police agencies still rely on some advantage of IT, interface and system designers face
manual processes, the growth of digital information a number of challenges created by the organizational
maintained by police repositories has been explosive. environment and characteristics of police work.
This growth has improved the speed at which police An important part of design and human-com-
can assess information and increased the amount of puter interaction is understanding the characteris-
information that police can store. For example, in her tics of a systems users. In most police organizations
book The Justice Juggernaut: Fighting Street Crime, many types of users with many types of needs use a
Fighting Computer Crime
B
egun in 1991 as the Computer Crime Unit of the U.S. Department of Justice, the Computer Crime and Intellectual Property
Section (CCIPS) is the primary federal office for computer-related crime. Below is the CCIPS website statement on its
purpose and function.
The Computer Crime and Intellectual Property Section (CCIPS) attorney staff consists of about forty (40) lawyers
who focus exclusively on the issues raised by computer and intellectual property crime. Section attorneys advise federal
prosecutors and law enforcement agents; comment upon and propose legislation; coordinate international efforts to com-
bat computer crime; litigate cases; and train all law enforcement groups. Other areas of expertise possessed by CCIPS at-
torneys include encryption, electronic privacy laws, search and seizure of computers, e-commerce, hacker investigations,
and intellectual property crimes.
A large part of CCIPS strength derives from the diverse skills and the wide variety of experiences its lawyers have
had before joining the Section. Before joining CCIPS, its attorneys have been computer scientists, state and federal pros-
ecutors, and associates and partners at law firms. A substantial number of CCIPS attorneys have received degrees in
computer science, engineering, or other technical fields; about half came to CCIPS with prior government service.
CCIPS began as the Computer Crime Unit of the former General Litigation and Legal Advice Section of DOJs Criminal
Division in 1991. CCIPS became a Section of the Criminal Division in 1996.
As Attorney General Janet Reno noted in her testimony on Cybercrime before the United States Senate Committee
on Appropriations on February 16, 2000:
CCIPS works closely on computer crime cases with Assistant United States Attorneys known as Computer and
Telecommunications Coordinators (CTCs) in U.S. Attorneys Offices around the country. Each CTC is given special train-
ing and equipment, and serves as the districts expert in computer crime cases.
The responsibility and accomplishments of CCIPS and the CTC program include:
Litigating Cases:
CCIPS attorneys have litigating responsibilities, taking a lead role in some computer crime and intellectual prop-
erty investigations, and a coordinating role in many national investigations, such as the denial of service investigation
that is ongoing currently. As law enforcement matures into the Information Age, CCIPS is a central point of contact for in-
vestigators and prosecutors who confront investigative problems with emerging technologies. This year, CCIPS assisted
with wiretaps over computer networks, as well as traps and traces that require agents to segregate Internet headers from
the content of the packet. CCIPS has also coordinated an interagency working group consisting of all the federal law en-
forcement agencies, which developed guidance for law enforcement agents and prosecutors on the many problems of
law, jurisdiction, and policy that arise in the online environment.
Working with the U.S. Attorneys Office in the District of New Jersey and the FBI, as well as with state prosecutors
and investigators, CCIPS attorneys helped ensure that David Smith, the creator of the Melissa virus, pled guilty to a vio-
lation of the computer fraud statute and admitted to causing damages in excess of $80 million.
CCIPS is also a key component in enforcing the Economic Espionage Act, enacted in 1996 to deter and punish the
theft of valuable trade secrets. CCIPS coordinates approval for all the charges under the theft of trade secret provision of
this Act, and CCIPS attorneys successfully tried the first jury case ever under the Act, culminating in guilty verdicts against
a company, its Chief Executive Officer, and another employee.
The CTCs have been responsible for the prosecution of computer crimes across the country, including the prosecu-
tion of the notorious hacker, Kevin Mitnick, in Los Angeles, the prosecution of the hacker group Global Hell in Dallas,
and the prosecution of White House web page hacker, Eric Burns, in Alexandria, Virginia.
U.S. Department of Justice, Computer Crime and Intellectual Property Section. Retrieved March 10, 2004, from http://www.usdoj.gov/criminal/cybercrime/ccips.html
central information system. Some users, such as crime in the system. Police managers and higher-ranking
analysts, are more computer savvy and use computer officers often use an information system for case man-
systems regularly for crime investigation and re- agement and resource allocation.
port generation. Records personnel deal with data Patrol officers, who make up the majority of em-
entry and verification of the information contained ployees in a police department, are typically not as
LAW ENFORCEMENT 423
experienced in computer use and thus have more data in another system. The federal government is
problems accessing and using information technol- trying to deal with problems of data integration, en-
ogy. As the front line of defense for police, however, couraging police agencies to follow the National
patrol officers often must get the right information Incident-Based Reporting System (NIBRS). As more
in a timely manner. police agencies move toward an NIBRS-compliant
A system must be designed not only to encom- system, some of these problems of data integration
pass the needs of all the types of users within the will be solved.
police organization, but also to take into account the For system designers, however, the challenge of
abilities and characteristics of all users. A system integrating systems is not a minor one. Designers
should be designed to meet the investigative needs must take into consideration issues of platform and
of crime analysts but also to be accessible to patrol data integration. Although initiatives to establish
officers in the field, who, for example, may use the standards exist, these initiatives are not an option for
system to verify information given to them by a sus- the majority of older police information systems
pect in custody. Thus, the usability of a system, which in use.
influences the frequency of use by police personnel,
influences the success of information technology in
the police context. Information Access and Security
System designers face not only the challenges of As information technology in police organizations
designing for different types of users and tasks, but becomes more prevalent, the ability of police offi-
also the challenge of integrating different systems cers to easily access information while ensuring in-
across different police agencies. An increasingly press- formation security becomes more important. System
ing problem is the ability of police agencies to share designers must balance the need for information ac-
the information within their systems with other cess with the need for information security.
police agencies. Many of these systems were devel- Officers, especially the majority who work on pa-
oped in-house and are stand-alone systems, mak- trol, perform a large part of their duties in the field
ing integration with other systems difficult. and in patrol cars. System designers must take into
The problem of integration occurs at many account the work environment that affects officers
levels. System designers may face the challenge of inability to access information. Rather than design just
tegrating systems that are on different types of plat- for desktop systems, designers must design for lap-
forms. For example, one police agency may use an top computers, car-mounted digital terminals, and
Oracle database system, whereas another agency may even handheld digital terminals.
store its data in flat files or in an archaic legacy sys- System designers must also decide where data
tem. Another problem could be integrating infor- should reside. In centralized systems data from dif-
mation from systems with different architectures. ferent information systems at different agencies
For example, system designers integrating two sys- are ported into a single data warehouse. This assures
tems may have the tedious task of matching data agencies that only the data they want to share are ac-
from the underlying architecture of one system with cessible to other agencies. The concerns of central-
another. ized systems revolve around the maintenance of the
For a user the ability to seamlessly access infor- information system. For example, who should man-
mation from different systems greatly reduces the age and maintain the system? In a decentralized sys-
need to learn to use those different systems. Users, tem police agencies maintain their information
especially police officers who are less familiar with but allow other agencies to tap directly into their sys-
different computer systems, want to learn a single tem. Allowing other agencies to directly access an in-
interface and use that interface to access informa- formation system leads to issues of security as well
tion from different systems. From a technical point as system loading.
of view this means that system designers must Another factor affecting security and access is-
map the data contained in one system to the same sues is the mode in which police officers work.
Officers are the first line of defense, the first re- easily understand the associations among entities.
sponders to the publics calls for assistance. They also Because police may not know the exact associations
provide primary investigative tasks, initiating the and instead have to browse information, use of in-
case report that may eventually be assigned to information technology can quickly result in infor-
vestigators. Officers are primarily car based and in- mation overload. Designing user interfaces with
cident driven. They must be mobile to respond visualization techniques to display associations, such
quickly to calls. Because they have high workloads, as timeline analysis displays, geo-mapping, and net-
officers are often not able to follow up on cases works, can reduce information overload and enhance
and must instead quickly pass cases to investigators, a users ability to use information technology.
thus limiting the information that officers may be A number of artificial intelligence systems have
able to pass along. Officers have a limited amount of been developed to aid police in searching through
time during which they can access and create a case the vast amounts of criminal data. Many of these sys-
report before submitting it to investigators. Therefore, tems build upon human heuristics (aids in learning)
the window of opportunity for officers to access to model police search behaviors. One system, Cop-
information from computer systems is crucial. link, uses a statistic-based algorithmic technique to
This distributed aspect of police work is a se- identify relationships among entities, such as people,
curity and access concern for system designers. locations, vehicles, crime types, and organizations.
Although police must use information technology Coplink was developed using a user-centered design
to access information from their patrol cars, they in which police personnel were involved in the plan-
have less control over security in the field. With the ning, design, and evaluation at each stage of devel-
deployment of wireless technology in patrol cars, opment. Evaluations of Coplink found that the user
system designers must implement security measures, interface was intuitive and that the system greatly
such as encryption methods and protocols, to pro- enhanced the speed with which officers were able to
tect the transmission of sensitive information. search for information.
With the advent of information technology, the
amount of information collected by police depart-
ments continues to grow. A challenge for interface Implications for Designers
and system designers is the design of information re- In what had been a closed organizational environ-
trieval interfaces that allow police to quickly and eas- ment, police are gradually adopting information
ily use the technology and understand the output. In technology. Officers, who previously had argued that
designing an interface, designers must decide how to their weapons of choice were guns and handcuffs,
best display aggregated information. In police work are now relying on information technology for a ma-
typical data collected in information systems include jor part of their work. Designing information
people, places, vehicles, and crime types, as well as technology for police work has many opportunities.
other types of data such as criminal and crime scene As an increasing number of police agencies store dif-
photographs, fingerprints, and security camera videos. ferent types of data electronically, cutting-edge in-
These data more likely are stored in a number of sys- formation technology can drastically change how
tems rather than in a single system. When police police work. Coupled with the need for more co-
search these systems for all information on a par- operation among police agencies, information tech-
ticular person, place, or vehicle, the graphical user in- nology can connect agencies and form a more
terface should return the information from the collaborative law enforcement environment.
systems in a logical and integrated manner. These opportunities are not without challenges
Police also use information technology to search for system and interface designers. Consequences
for associations among entities, such as associates of of mishandled or misunderstood information are
a suspect, owners of a vehicle, or crimes associated paramount, possibly leading to legal or even phys-
with a location. A single association can be the key ical harm. Also, given the rapidly growing amount
to solving a crime. Therefore, police must be able to and different types of data, information technology
LEXICON BUILDING 425
for police must be scalable to handle police needs Maltz, M. D., Gordon, A. C., & Friedman, W. (2000). Mapping crime
as they shift and expand. Challenges that system de- in its community setting: Event geography analysis. New York:
Springer-Verlag.
signers face include integration across systems, in- Manning, P. K. (1992). Information technologies and the police. Crime
formation access and security, and information and Justice, 15, 349398.
retrieval processes. Morgan, B. J. (1990). The police function and the investigation of crime.
Designers of information technology for police Brookfield, VT: Avebury.
Northrop, A., Kraemer, K. L., & King, J. L. (1995). Police use of
work must consider the police environment and the computers. Journal of Criminal Justice, 23(3), 259275.
pressures that affect how police use information tech- Office of Justice Programs. (2000). Office of Justice Programs Integrated
nology. Given the many types of users (e.g., patrol Justice Information Technology Initiative. Retrieved November 4,
2003, from http://www.ojp.usdoj.gov/archive/topics/integratedjustice/
officers, investigators, police managers, and crime welcome.html
analysts) with various levels of computer experience Pliant, L. (1996). High-technology solutions. The Police Chief,
and different job tasks, designers must take into 5(38), 3851.
account the diverse needs and abilities of all users. Rocheleau, B. (1993). Evaluating public sector information systems.
Evaluation and Program Planning, 16, 119129.
Task analysis techniques and user-centered design U.S. Department of Justice. (2000). Uniform crime reporting: National
not only help designers understand the work and en- Incident-Based Reporting System, data collection guidelines: Vol. 1.
vironmental challenges faced by police, but also Data collection guidelines. Retrieved February 17, 2004, from http://
increase user support. www.fbi.gov/ucr/nibrs/manuals/v1all.pdf
As the needs of police and the capabilities of in-

formation technology continue to change, system
designers must take into account the unique work
practices, environment, and needs. In the end, the LEXICON BUILDING
better the system design, the better the police
work, which benefits everyone. The word lexicon can refer to any of the following:
(1) the familiar dictionarya repository of infor-
Roslin V. Hauck mation about words containing explanations that
ordinary speakers of the language can understand
See also Information Organization; Information while not including information that speakers take
Overload; Information Retrieval; Law and HCI for granted, (2) the lexical component of a formal
grammar where information about words is pro-
vided in a formalism designed to be interpreted by
the interpretive rules of the grammar, and (3) a body
FURTHER READING of structured information about words provided
in a notation that allows computational means of
Bowen, J. E. (1994). An expert system for police investigations of eco-
nomic crimes. Expert Systems with Applications, 7(2), 235248.
performing natural language processing (NLP)
Brahan, J. W., Lam, K. P., & Chan, H. L. W. (1998). AICAMS: Artificial operations.
Intelligence Crime Analysis and Management System. Knowledge- Resources that can serve these purposes are not
Based Systems, 11, 335361. necessarily distinct. Commercial dictionaries
Colton, K. (1978). Police computer systems. Lexington, MA: Lexington
Books.
especially those designed for language learners
Gordon, D. (1990). The justice juggernaut: Fighting street crime, con- often have information in a form that can be
trolling citizens. New Brunswick, NJ: Rutgers University Press. exploited for purposes of grammar writing and NLP.
Haggerty, K. D., & Ericson, R. V. (1999). The militarization of polic- In principle a single lexical resource could serve all
ing in the information age. Journal of Political and Military Sociology,
27(2), 233255. three purposes, made available in task-specific for-
Hauck, R. V., Atabakhsh, H., Ongvasith, P., & Chen, H. (2002). mats and provided with interfaces designed for
Using Coplink to analyze criminal-justice data. IEEE Computer, different classes of users.
35(3), 3037.
Leonard, V. A. (1980). The new police technology: Impact of the com-
Because a wide variety of NLP tasks exists, the
puter and automation on police staff and line performance. sort of lexicon needed can vary greatly from appli-
Springfield, IL: Charles C. Thomas. cation to application. For speech recognition the
lexicon is a list of the words needing to be recognized At the most ambitious level, for an application
together with representations of their pronunciation concerned with open-domain natural language un-
and minimal indication of their grammatical derstanding or with accurate and natural translation
properties. In the case of automatic stemming where of texts from one language into another, the lexicon
the stems of words (e.g., the sleep in sleeping, the would need to be extremely large and to contain con-
medicate in premedication) are separated from siderably richer information than is found in even
their affixes (-ing,pre- and -ion of those same the finest unabridged commercial dictionaries and
words), the lexicon consists of a list of stems and a also need to be fitted into knowledge about gram-
list of affixes, the latter classified according to their mar, usage, commonsense inferencing, and discourse.
combinability of stems of particular types. If the ap-
plication is one of document routingfor example,
shunting streaming newswire articles to different ed- The Units of a Lexicon
itorial offices according to whether they deal with In the previous paragraphs the term word was used
the stock market, sports, crime, high fashion, or to identify the primary elements in the building of
the weathera lexicon could be considered adequate a lexicon. The word word in the generally understood
for such purposes if it merely associated each con- sense is appropriate in naming the process of word
tent word with the relevant domains and gave a set disambiguation (the act of establishing a single se-
of probabilities that a word belongs to one domain mantic or grammatical interpretation for an am-
rather than another. For such lexicons human efforts biguous word; in the literature usually called word
will be mainly devoted to sorting and labeling a col- sense disambiguation; the most detailed study is by
lection of documents large enough to support the computer scientists Nancy Ide and Jean Vronis),
machine learning by which words acquire such where the system decides, for a multiple-meaning
weighted associations. Such lexicons generally do not word, which is most likely to be the intended mean-
need to contain careful human-constructed de- ing in a given passage. However, the primary entity
scriptions of individual words. that needs to be characterized in a lexicon is not the
word but rather a pairing of a word with a sense, usu-
ally called a lexical unit (LU). LUsnot words
METADATA Information about the actual source of need definitions, have synonyms and paraphrases,
the data: the author, the period when or the context participate in semantic contrasts, and have specific
within which the data got provided; the age of the au- grammatical properties. When we regard the con-
thor, and the title of the document. cept this way, we can say that the concept of LU is
an elaboration of the lay concept of word. (Objec-
tions to the term word sense disambiguation point
At a level requiring slightly richer representations precisely to this distinction. The process is more ap-
as, for example, in information extraction tasksan propriately called either word disambiguation or
appropriate lexicon might use a restricted vocabu- sense selection: What is disambiguated is the word,
lary integrated with a limited repertory of phrasal not the sense.)
patterns designed to discover information about, say, The concept of LU that we need is both narrower
corporate leadership changes, traffic conditions, or and broader than the common notion of word.
the progress of an ongoing chess game. (For exam- After we see that an essential part of being an LU is
ple, [PERSON] has been replaced as [OFFICE] at having a unitary meaning description and unique
[COMPANY] by [PERSON] of [COMPANY].) In grammatical properties, we are compelled to recog-
such cases either the texts subject to analysis will nize the existence of LUs that are made up of more
themselves be domain restricted or the NLP task at than one word, the so-called multiword units
hand will target limited kinds of information, keyed (MWUs). If the word lift is an LU, then pick up is also
by the presence of particular words, ignoring what- an LU; if tolerate is an LU, then so is put up with; if
ever other information the texts contain. pork is an LU, then so is horse meat.
The need for these two elaborations of the The proper treatment of MWUs in computa-
word concept makes it clear why simple statistical tional linguistics is a largely unsolved problem:
studies of space-separated letter sequences cannot The issues are how to represent them in a lexicon,
provide detailed information about LUs and why any how to discover them in running text, how to estimate
attempt to measure the distribution of LUs requires their contribution to texts in particular styles and gen-
sampling and human judgment just to figure out res, and even how to decide how many of them
what needs to be counted. If we find the letter se- there are in the language. Linguist Ray Jackendoff has
quence court in a text we do not know which LU speculated that the list of MWUs that an individ-
this represents (tennis court, the kings court), nor can ual knows must be roughly the same size as the list
we tell whether in its context court stands for an of single words, and the lexicographer Igor Melcuk
LU on its own or is part of a multiword unit (e.g., claims that the number of phrasal words that must
Court of Appeals). be recorded is ten times the size of the single-word
Types of MWUs include (1) noun compounds lexicon; but for NLP applications, as opposed to some
of the form noun+noun (house arrest, peace offi- individuals mental lexicon, there can be no limit
cer) or of the form adjective+noun (forcible entry, to a lexicons size as long as means are needed for
federal officer, punitive action); (2) conventionalized recognizing personal names, place names, names of
verb+object combinations (exact vengeance, inflict historical events, and all the rest.
punishment, take revenge); (3) combinations of verbs
with various function words (put down, look into,
put up with); (4) complex prepositions (in terms Decoding versus Encoding Functions
of, pursuant to, in accordance with); (5) lexically com-
plex conjunctions (let alone, much less, both . . . of a Lexicon
and, either . . . or, as . . . as); and many others, in After the units are identified, information associated
addition to a vast collection of idioms. with them can be designed for either decoding (rec-
The goal of automatically detecting MWUs in ognizing) or encoding (generating) purposes. The
running text is especially difficult for two reasons. difference is between being able to recognize words
First, not every MWU is an uninterrupted word in a passage in a way that leads to passively under-
sequence, and second, the same combination of standing the passage and having enough informa-
words can count as being a single LU in some con- tion about the words to be able to combine them
texts but as having separate functions in other appropriately with other words in relevant con-
contexts: The juxtaposition of the words walk and texts of use.
into is accidental in He walked into the alley but Many NLP applications require at most the de-
constitutes an MWU in He walked into a door coding function of a lexiconat most because for
(collided with); the elements of the words let alone many purposes, such as information retrieval, doc-
are individually interpretable in I want to be let ument routing, topic detection, or event tracking,
alone but make up a single MWU in She wouldnt little information is needed about the actual
give me a nickel, let alone ten dollars. Specialist vo- meanings of individual words. Question-answering
cabulary is replete with MWUs, and the status of a system that address simple factstypically referred
word group as an MWU does not always stand to as factoidscan be aided by having available,
out, even for a human interpreter. For example, look- for each word, lists of semantically related words,
ing at the parallel syntactic patterns in He was ac- such as: synonyms (words with the same meaning),
cused of beating his dog with a broomstick and He antonyms (words with opposite meaning), hyponyms
was accused of assaulting a federal officer with a deadly (subtypes, as terrier is to dog), and hypernyms
or dangerous weapon, one cannot know that the (supertypes, such as dog is to terrier). Data
highlighted phrase in the second sentence is a named miningautomatically searching for information
crime in the U.S. Criminal Code, which needs its own in large databasesis helped by having large lists of
entry in a lexicon of U.S. criminal justice procedures. words that share category membership (e.g., the
names of pharmaceuticals, disease names, therapies, recognizing morphologically complex technical terms
etc.) of the sort derivable from medical and tech- in specialized disciplines that make extensive use
nical glossaries. of Greco-Latin roots; and (3) for including meaning
In the case of word sense selection, the lexicon specializations that can be generated from certain
will show that an encountered word has more than basic meaning descriptions (as in the generative lex-
one sense, and the applications task is to choose (or icon of linguist James Pustejovsky, 1995). For the first
give weighted probabilities to) the sense needed in of these, many existing NLP applications use entity
the given context. This can be done by exploiting recognizers (software for recognizing names of per-
(1) metadata about the text itself (in a sports column sons, places and institutions, addresses, and ex-
about tennis, the noun court is likely not to refer to pressions of calendar time) such as the Bolt, Beranek,
a royal household), (2) information about the lin- and Newman (BBN) Identifinder (www.bbn.com/
guistic structure of the phrase or sentence in which speech/identifinder.html).
the word is found (if a parser has recognized court
as a verb, then the wooing sense will be selected),
or (3) information about the semantic domain of Sources of Lexical Information
words that are in grammatical construction with, or An important consideration for lexicon building is
in the neighborhood of, the target word (the legal how and where information about lexical properties
institution sense of court is called for in the sentence is to be found. Much of what people know about
The judge called for order in the court). their language is implicit and cannot be easily
The inadequacy of the encoding function of typ- brought to conscious awareness. Thus, building a
ical dictionary definitions can be illustrated with the lexicon cannot be achieved simply by asking native
word decedent: The reader, human or machine, speakers to write down in a systematic and appli-
will know from the definitions that a person referred cation-relevant way information about word use and
to with this word is dead. Common definitions in meaning that they are assumed to know. In many
a sample of dictionaries are someone who is no cases facts about the meaning of a word are not
longer alive, a deceased person, and a dead per- obvious, requiring subtle tests for teasing them out
son. However, an advanced language learner who through the assembly and analysis of corpus evi-
wishes to know when and how to use the word, a denceevidence taken from a large collection called
translator (human or machine) needing to know a corpus (plural, corpora) of machine-readable
when to select it, or a language generation computer textstogether with careful use of judgments on the
application finding it in its lexicon needs to know part of the users of the language.
that, although, in fact, designating a dead person, the In fortunate cases, of course, much of the work has
word decedent is used in discourse about that per- already been done, and the task is to adapt to local rep-
sons estate. One cannot appropriately say Mozart is resentational requirements information that is pub-
a decedent or Our graveyard holds 173 decedents. licly available; examples are machine-readable versions
A human dictionary reader might suspect from var- of commercial dictionaries and such online resources
ious hints that something is special about the word as WordNet (www.cogsci.princeton.edu/~wn) and
decedent; for example, a reader who finds in the Web FrameNet (www.icsi.berkeley.edu/~framenet).
investment glossary (www.investorword.com) Researchers put much effort into statistical stud-
simply the definition a person who has died, might ies of natural language corpora to discover words as-
also notice that the entrys cross-references are to the sociated with particular domains, to cluster words
words will, estate, heir, and succession. by contextual features on the assumption of regular
A lexicon needs to contain generative compo- form/meaning correspondences, and to derive clas-
nents (1) for recognizing MWUs that are produced sificatory relations between words on the basis of
by special subgrammars covering personal names, contextual clues. (For example, phrases like X, a
place names, institutional names, dates, expressions northern variety of fish or X and other fish lead
of clock time, currency amounts, and so forth; (2) for to the classification of X as a fish.)
For many purposes traditional kinds of lin- object to, etc.), and the preference for partic-
guistic research are unavoidable, using the refined ular combination with modifiers (excruciat-
introspections of linguistically trained native speaking pain,blithering idiot,stark naked, etc.)
ers and carefully testing predictions based on 5. Enough semantic information to guide the
these. This is most necessary for language-generation semantic integration of the meanings of LUs
purposes because native speakers can know what is into meaning representations of the phrases
not possible in a language, something that a cor- and sentences with which they combine
pus cannot tell us (a sentence such as Another day 6. Association with use conditions that are inde-
elapsed can be found in a corpus; one such as pendent of meaning proper, that is, the fit with
Yesterday elapsed cannot; linguistic introspection particular topics or genres and the like.
offers data generally not knowable in any other way).
Building an adequate lexicon for NLP work is a
Collaboration with experts is necessary in cases
huge undertaking involving long-term planning and
where a meaning is stipulated in some expert domain
serious funding. The absence of such a lexicon makes
and neither discoverable in a corpus nor accessible
it impossible for a computer to handle language cor-
to ordinary speakers intuitions. Experts are not nec-
rectly and sets arbitrary limits to NLP systems.
essarily skilled in knowing the form in which their
Building such a lexicon requires a holistic approach.
knowledge can be made accessible to readers or avail-
This is not something to be carried out piecemeal, a
able to computational purposes, but their knowl-
method that guarantees incompatibility of the var-
edge is obviously necessary in many cases.
ious components. Linguistic analysis is complex,
slow, and labor intensive; most lexicons produced
today cover only a part of the total analysis of the
Kinds of Lexical Information language and are themselves only partial, the funding
In the end we find that a lexicon capable of serving having ended before the work was completed. A com-
the widest range of NLP purposes will have to in- prehensive lexicon of the languagea systematic
clude information about: record of how words are used and understood by
peopleis essential if the twenty-first-century
1. Pronunciation in the form of computer-friendly
computer is to handle language correctly.
transcriptions such as the TIMIT, an interna-
tional standardized ascii-based alphabet for the Charles Fillmore
phonetic transcription of speech.
2. The identification of lemmas (the identifica- See also Machine Translation; Natural-Language
tion of a single dictionary form for words of Processing; Ontology; Speech Recognition
different shape: thus goes,gone,went, etc.,
will all be identified with the lemma go) along
with the tagging of words with information FURTHER READING
about part-of-speech (noun, verb, etc.) and
grammatical properties (plural, past, etc.). Boguraev, B., & Pustejovsky, J. (Eds.). (1996). Corpus processing for lex-
ical acquisition. Cambridge, MA: MIT Press.
3. The association of each LU with other LUs in Briscoe, T., & Carroll, J. (1997). Automatic extraction of subcatego-
the lexicon, such as the recognition of synonyms rization from corpora. Proceedings of the 5th Conference on Applied
(doctor, physician), taxonomic relations (terrier Natural Language Processing ANLP-97. Retrieved February 9, 2002,
> dog > mammal, etc.), contrast sets (man: from http://acl.ldc.upenn.edu//A/A97/A97-1052.pdf
Cruse, D. A. (1986). Lexical semantics. Cambridge, UK: Cambridge
woman, boy: girl, beautiful: ugly, etc.) University Press.
4. The ability to co-occur with other words and Fellbaum, C. (1998). WordNet: An electronic lexical database.
phrases, thus distinguishing transitive from in- Cambridge, MA: MIT Press.
Fillmore, C. J. (1992). Corpus linguistics vs. computer-aided armchair
transitive verbs (according to whether they take linguistics. Directions in corpus linguistics: Proceedings from a 1991
a direct object), the selection of prepositions Nobel Symposium on Corpus Linguistics (pp. 3566). Stockholm:
(as in fond of, pleased with, depend on, Mouton de Gruyter.
Fontenelle, T. (2003). Special issue on FrameNet. International Journalof anisotropy refers to the difference in the dielectric
Lexicography, 16(3). constant parallel to and perpendicular to the long
Fisher, W. M., Zue, V., Bernstein, J., & Pallett, D. (1987). An acoustic-
phonetic data base. 113th Meeting of the Acoustical Society of
axis of the molecule; it is responsible for the reori-
America, Indianapolis, IN. entation of the liquid crystal when the crystal is sub-
Gildea, D., & Jurafsky, D. (2002). Automatic labeling of semantic roles. jected to an applied electric field. Typically, liquid
Computational Linguistics, 28(3), 245288. crystal materials will align parallel to the direction
Grishman, R., Mcleod, C., & Meyers, A. (1994). COMLEX syntax:
Building a computational lexicon. Proceedings of the 15th of the applied electric field. When the liquid crys-
International Conference on Computational Linguistics (COLING- tal molecules change their orientation, their opti-
94), Kyoto, Japan. cal appearance also changes. With these unique
Ide, N., & Vronis, J. (1998). Word sense disambiguation: The state of
the art. Computational Linguistics, 24(1), 140.
properties, one can control the optical appearance
Koskenniemi, K. (1983). Two-level morphology: A general computa- of pixels, which are the smallest switching element
tional model for word-form recognition and production. Helsinki, on a display to create an image.
Finland: University of Helsinki Department of General Linguistics. The most common LCD configuration, known
Miller, G. A., Beckwith, R., Fellbaum, C. D., Gross, D., & Miller, K. J.
(1990). WordNet: An on-line lexical database. International Journal
as the twisted nematic (TN), employs crossed po-
of Lexicography, 3, 235244. larizers and a molecular orientation of molecules
Ritchie, G. D., Russell, G. J., Black, A. W., & Pulman, S. G. (1992). whose long axis twists through a 90-degree angle
Computational morphology: Practical mechanisms for the English between two glass substrates that are somewhat like
lexicon. Cambridge, MA: MIT Press.
Wilks, Y., Slator, B., & Guthrie, L. (1996). Electric words: Diction- windowpanes. One unique feature of LCD tech-
aries, computers and meanings. Cambridge, MA: MIT Press. nology is the way in which the twisted structure
of the molecules is created. A polymer layer on each
of the two glass substrates is mechanically rubbed
with a cloth to create very minute groves (called
nanogrooves) on the surface that uniformly align
LIQUID CRYSTAL the long axis of the molecules at each surface. The
alignment of the rub direction is placed parallel
DISPLAYS to the transmission axis of the polarizer, but the
two glass substrates are placed one on the other with
Flat panel displays are a fascinating technology. From their polarizing directions crossed. A pair of po-
computer monitors to personal digital assistants larizers, stacked with their polarizing orientations
(PDA), visual displays are the ultimate human- at right angles, normally block when they are
machine interface. Liquid crystal displays (LCDs) crossed, and are transparent when they are paral-
are ubiquitous in such portable electronic prod- lel. After light passes through the first polarizer, it
ucts as PDAs, cellular phones, and video recorders, becomes linearly polarized and follows the liquid
and they have enabled new product categories, such crystal twisted structure. This process, often referred
as laptop computers. LCDs are erasing the age-old to as adiabatic waveguiding, enables light to escape
domination of cathode-ray-tubes (CRT) for desk- out the top polarizer (even though the polarizers
top computer monitors. Unlike conventional CRT are crossed). By using a backlight, a given pixel
technology, which creates light, an LCD simply would be bright in this configuration, and color is
acts as a light shutter to modulate a powerful back- controlled on the pixel level with a red, green,
light that is on continuously. and blue color filter array.
To understand the basic operation of LCD de- Thin transparent conductor layers, usually
vices, one must understand several important prop- indium-tin oxide, are deposited on the substrates so
erties of liquid crystal materials. The elongated shape that a voltage can be applied to the material. When
of the liquid crystal molecules, often referred to as a voltage is applied to the pixel, an electric field is
shape anisotropy, gives liquid crystal materials their created perpendicular to the substrates. The liquid
unique electrical and optical properties. Dielectric crystal molecules align parallel to the electric field,
LITERARY REPRESENTATIONS 431
thereby breaking the twisted symmetry. The light characters, or are the environment in which the
passes through this aligned configuration without story takes place.
any change; therefore the second polarizer absorbs
all of the light and the pixel is black. Various levels
of gray are possible with intermediate voltages, The Hard-Science Paradigm
and the array of thousands of pixels with different Intelligent machines may appear in any genre of
color filters can produce a complete image with modern literature, but robots are especially associ-
full color and shading. ated with a particular subvariety of science fiction.
A questionnaire study conducted by William Bain-
Gregory P. Crawford bridge at a world science fiction convention held
in Phoenix, Arizona in 1978 found that memo-
rable stories about robots tend to belong to the hard-
FURTHER READING science category. These are stories that take current
knowledge from one of the physical sciences and log-
Crawford, G. P., & Escuti, M. J. (2002). Liquid crystal display tech- ically extrapolate the next steps that might be
nology. In J. P. Hornak (Ed.), Encyclopedia of imaging science and
technology (pp. 955969). New York: Wiley Interscience.
taken in that science. They appeal to readers who en-
Lueder, E. (2001). Liquid crystal displays. New York: Wiley SID. joy reading factual science articles and stories about
Wu, S. T., & Yang, D. K. (2001). Reflective liquid crystal displays. New new technology.
York: Wiley SID. Interestingly, the research found that people who
Yeh, P., & Gu, C. (1999). Optics of liquid crystal displays. New York:
John Wiley and Sons. like hard-science science fiction tend to prefer stories
in which there is a rational explanation for every-
thing, and they like fictional characters who are cool,
unemotional, clever, and intelligent. This may mean
they not only like intelligent machines but would
LITERARY prefer human beings to be more like robots. This
possibility is illustrated by the robot stories of the
REPRESENTATIONS preeminent hard-science writer, Isaac Asimov
(19201992).
Since the industrial revolution, many writers have Simple robots such are in use today in factories or
imagined intelligent machines as a way of seeing that are closely supervised by human beings can be
humanity from a new perspective. The Sand-Man programmed relatively simply. But, Asimov thought,
(1817), by the Romantic writer E. T. A. Hoffmann, if robots are to operate autonomously they need the
concerned a man fascinated by a mechanical doll equivalent of an ethical code. Thus, he postulated the
he imagines to be the perfect woman. This idea was Three Laws of Robotics (Asimov 1950, 7):
borrowed by the composer Lo Delibes for his 1870
1. A robot may not injure a human being, or,
ballet Copplia, and by Jacques Offenbach for his
through inaction, allow a human being to come
1880 opera The Tales of Hoffmann. Early in the
to harm.
twentieth century, L. Frank Baum, the author of
2. A robot must obey the orders given it by hu-
the childrens classic The Wonderful Wizard of Oz
man beings except where such orders would
(1900) and a subsequent series of Oz books, added
conflict with the First Law.
the mechanical man, Tik-Tok, to the roster of
3. A robot must protect its own existence as long
Dorothys friends in Oz. A hundred years later,
as such protection does not conflict with the
computers feature as props in much ordinary lit-
First or Second Law.
erature, but the deepest explorations of human-
machine interaction are in science fiction, where In Asimovs 1942 story Runaround (reprinted
computers and intelligent machines are often main in the anthology I, Robot), two men on the sun-facing
approach. Unfortunately, Speedy has been designed

Excerpt from The Sand-Man with an especially strong Third Law, and the mens
command to him, which depends for its completion
(1817) by E. T. A. Hoffman on the Second Law, was not stated with an explicit
high priority. When Speedy approaches the molten
I
n this selection from a classic tale about man and
machine, the main character falls in love with a selenium, he discovers it is too dangerous for him to
beautiful mechanical doll, never wanting to believe she enter. This gives him what psychologists call an ap-
isnt really human. proach-avoidance conflict. Speedy goes crazy and
The concert came to an end, and the ball be- runs around the pool, singing.
gan. Oh! to dance with herwith herthat was Knowing they will die if they cannot get control
now the aim of all Nathanaels wishes, of all his of Speedy, the two men agonize about what to do.
desires. But how should he have courage to re- Eventually, one of them realizes that the First Law
quest her, the queen of the ball, to grant him could resolve this conflict between a weakened Second
the honour of a dance? And yet he couldnt tell
how it came about, just as the dance began, he
Law and a strengthened Third Law. He intention-
found himself standing close beside her, nobody ally exposes himself to mortal danger, forcing Speedy
having as yet asked her to be his partner; so, with to save him. Restored to sanity, Speedy is sent un-
some difficulty stammering out a few words, he der a stronger command to a safer selenium deposit.
grasped her hand. It was cold as ice; he shook Runaround was probably the first publication to
with an awful, frosty shiver. But, fixing his eyes use the word robotics, and all Asimovs robot stories
upon her face, he saw that her glance was
beaming upon him with love and longing, and
assume the existence of a distinct engineering dis-
at the same moment he thought that the pulse cipline devoted to the design of humanlike machines.
began to beat in her cold hand, and the warm life- Asimov called robot engineers roboticists, but this
blood to course through her veins. And passion word has not caught on.
burned more intensely in his own heart also, he Asimovs novel The Caves of Steel (1954) con-
threw his arm round her beautiful waist and cerns a partnership between a robot detective and
whirled her round the hall. . . . Nathanael, excited
by dancing and the plentiful supply of wine he
a human policeman, who team up to solve a murder
had consumed, had laid aside the shyness that could not have been committed by a robot or
which at other times characterised him. He sat a human alone, but only by a combination of both.
beside Olimpia, her hand in his own, and declared Much of the story revolves around the competi-
his love enthusiastically and passionately in words tion and growing understanding between the robotic
which neither of them understood, neither he and human investigators. On one level, the theme is
nor Olimpia. And yet she perhaps did, for she sat
with her eyes fixed unchangeably upon his, sigh-
the relationship between people and machines, but
ing repeatedly, Ach! Ach! Ach! Upon this on a deeper level it is the connection between people
Nathanael would answer,Oh, you glorious heav- and the things they use, including other people.
enly lady! You ray from the promised paradise of Potentially, a happy ending can be reached when any
love! Oh! what a profound soul you have! my two beings come to understand each other both as
whole being is mirrored in it! and a good deal objects and as subjects.
more in the same strain. But Olimpia only con-
tinued to sigh Ach! Ach! again and again.
In a later mystery novel, The Robots of Dawn
Source: Hoffmann, E. T. W. (1885). The sand-man. In Weird tales, Vol. 1. New
(1983), Asimov suggests that human beings are ruled
York: Charles Scribners Sons. (Original work published 1817) Retrieved by strict Laws of Humanics, comparable to the Laws
March 10, 2004, from http://gaslight.mtroyal.ca/sandman.htm
of Robotics. For example, the people in The Caves of
Steel are under an inescapable psychological com-
pulsion to avoid open spaces, and people in Asimovs
side of the planet Mercury need selenium to repair novel The Naked Sun (1956) have a powerful inhi-
the system that protects them from the lethal solar bition against ever being in the physical presence
radiation. They send Speedy, a robot, to get some of another person. Many of the classical hard-science
from a pool of this metal that they themselves cannot writers viewed humans and robots in very similar
terms. Adam Link, the robot hero of a series of stories a particular story. Humans interacted with com-
by Eando Binder, asserted his own humanity through puters and robots by programming them or simply
the principle that the body, whether flesh or metal, by speaking commands to them. In contrast, cyber-
was only part of the environment of the mind. punk writers are concerned with collective phe-
Robert A. Heinlein (19071988), a hard-science nomena and the oppression of the individual by
writer with a highly individualist ideology, postu- the social system. Their heroes are antiheroes, some-
lated that the only thing preventing a machine from times carrying generic names like Case (in William
becoming a conscious, individual person was the lack Gibsons 1984 novel Neuromancer) or Hiro
of sufficient computing power. In his novel The Moon Protagonist (in Neal Stephensons 1992 Snow Crash).
is a Harsh Mistress (1966), a machine that was de- Computers in such stories typically are not individ-
signed to handle a vast variety of tasks autonomously ual characters; they are part of the networked envi-
is augmented with additional memory, computer vironment of cyberspace.
sion, and voice, unexpectedly becoming the leader The term cyberspace was introduced in Neuro-
of a rebellion against the collectivist government mancer to refer to the dynamic virtual reality people
of Earth. perceive when jacked into the worldwide computer
network. In an unspecified future year, users con-
nect to this network either through electrode head-
The Cyberpunk Paradigm sets that detect and affect their brainwaves through
Throughout the history of science fiction, a few writ- the skin, or by plugging their brains directly in
ers have contributed stories that were unusually sur- through jacks surgically inserted behind the left ear.
realist, psychological, or politically radical. In the Cyberspace is a consensual hallucination. . . . a
1960s writers and works in this vein were described graphic representation of data abstracted from the
as New Wave. Bainbridges questionnaire study found banks of every computer in the human system. . . .
that the New Wave was characterized by avant-garde lines of light ranged in the nonspace of the mind,
fiction that experiments with new styles, often based clusters and constellations of data. . . . like city lights,
on speculations in the social sciences. Many of the receding (Gibson 1984, 51). A heavily defended cor-
stories concern harmful effects of scientific progress porate database is represented in cyberspace as a
or are critical of contemporary society. Often they green rectangle, whereas an artificial intelligence (AI)
deeply probe personal relationships or feelings, and is a featureless, white square. When the AI sends a
characters tend to be sensitive and introspective. computer virus to invade the database, it is a
In the 1980s, this literary movement morphed polychrome shadow, countless translucent layers
into the subgenre known as cyberpunk. Cyberpunk shifting and recombining (168). The experience
continues to experiment with stylistic innovations, of cyberspace is bodiless exaltation (6) and
tends to be critical of power structures in society, therefore addictive.
and relishes the lurid extremes of human character Neuromancers protagonist was a professional
and experience, epitomized by deviant sex, drugs, data thief, addicted to cyberspace, who stole from
and madness. Cyberpunk assumes a future world in his employers. In punishment, they crippled his nerv-
which computers and the Internet constitute the fun- ous system so he could no longer experience cyber-
damental structure of society. In these stories, gov- space, leaving him desperately self-destructive. He
ernment is weak or fragmented, the family is becomes enmeshed in a confused net of conspira-
practically nonexistent, and transnational corpo- cies, spun by rival corporations and artificial intel-
rations battle one another for information suprem- ligences, and is assisted by dubious friends, including
acy. In such a world, computer hackers are the most the computer-recorded personality of his deceased
effective rebels. hacking teacher. The nearest thing to a govern-
To a significant extent, the writers of the older ment that appears in the novel is the Turing Registry,
hard-science school were individualistic, and there- an agency that tries to prevent any of the autonomous
fore focused on an individual robot or computer for artificial intelligences from escaping human control.
Two books about travelers in search of themselves

and their worlds: The Wonderful Wizard of Oz (1900)
and The Shockwave Rider (1975).
Much of Stephensons Snow Crash also takes place red, blue, and green lasers paint three-dimensional
in cyberspace, where people are represented by avatars images.
(computer-generated characters) of varying degrees The Snow Crash of the title is a kind of drug, a
of cost, artistry, and surrealism. Gibson wanted to call computer virus, or a pattern of information that af-
computer-generated personal representatives fects the human mind in the same way a virus affects
constructs, but Stephensons term avatars has been a computer. In the Metaverse, Snow Crash appears
adopted by the computer science community. In in the form of a small calling card, or a scroll that
Hinduism, an avatar is a particular form in which a unrolls to reveal a flashing image of apparently ran-
deity may appear to human beings; by extension, a dom bits. One possibility explored by the novel is the
computer avatar is a virtual form in which humans idea that each human religion is an information virus,
appear to one another inside cyberspace. The avatars that spreads (for better or worse) from mind to mind.
of different users meet on the avenues of the Another is the notion that the natural programming
Metaverse, a vast, virtual-reality city, and the rules of language of the human mind, the fundamental ma-
this environment limit the avatars size and their abil- chine language of the brain, is ancient Sumerian.
ity to harm one another. Users do not jack their brains The cyberpunk genre even explores now ob-
directly into the Metaverse, as they do into Gibsons solete human-computer interfaces. The Difference
cyberspace, but merely wear special goggles on which Engine (1991), a historical novel that Gibson wrote
in collaboration with Bruce Sterling, imagines that Once Brunners science-fiction technology had
the nineteenth-century inventor Charles Babbage become real, mainstream writers exploited infor-
(17911871) succeeded in building the mechanical mation warfare for its dramatic value. The Net Force
computer he actually failed to complete, thereby series, created by Tom Clancy and Steve Pieczenik
introducing the information age a century early and in 1998, concerns the Net Force branch of the Federal
transforming industrial society. Programmers, in Bureau of Investigation, portrayed as heroes, in con-
this alternate Victorian society, are called clack- trast to Brunners villainous Federal Bureau of Data
ers, because of the noise produced by the machines Processing. In Net Force, the mobile telephone com-
that read their data cards, and computing is puter that Brunner imagined becomes a virgila
clacking. In classic cyberpunk fashion, the story cri- Virtual Global Interface Linkthat combines
tiques the Victorian hope that a partnership be- telephone, camera, scanner, fax, phone, radio, tel-
tween technological innovation and social order evision, GPS, and computer. Virgil, it will be re-
can overcome the fundamental dynamics of human membered, was the ancient Roman historian-poet
conflict. whom the Italian poet Dante envisioned as his com-
panion on his journey into hell in the first portion
of The Divine Comedy (written c. 13101314), and
Conict between a virgil accompanies the head of Net Force to his
death in the first chapter of the first novel in the se-
Humans and Machines ries. In 2010, the time of the first story, many people
Robots, computers, and information systems fre- still use keyboard, mouse, and monitor to interact
quently become entangled in conflicts between hu- with their computers. But many prefer virtual-re-
man beings. An early example of warfare on the ality headsets, visualizing choices in scenarios
In ter n e t i s Jo h n Br u n n e r s 1 9 7 5 n ove l The such as a private meeting in a forest clearing.
Shockwave Rider. In the novel, as the United Some scenarios are compatible. For example, when
States approaches the year 2020, it becomes a frag- two people compete with each other in virtual re-
mented society in which corrupt government ality, one may experience the competition as a high-
covertly magnifies social problems and individual speed highway race, while the other may perceive
psychopathology, the better to control the de- them to be driving speedboats up a river. When
moralized population. Published two years before the villain wants to sabotage the data systems of sev-
the first home computers became available and two eral corporations and governments, he employs the
decades before the first commercial Web browser, scenario that he is a German soldier in World War
Brunners novel predicted correctly that every home I, killing onrushing Allied troops.
could have a computer connected to the Internet Several authors have argued that heavy reliance
(with the standard keyboard, monitor, and printer) upon computers could make a high-tech society es-
and that the Internet could also be accessed via mo- pecially vulnerable to low-tech enemies. In Mack
bile telephones. The secret Tarnover project to cre- Reynolds novel Computer War (1967), a bellicose
ate superior data warriors backfires when its best nation called Alphaland attacks peaceful Betastan,
computer saboteur escapes. Sending software tape- on the advice of its computers. But the Betastani re-
worms across the Internet to modify selected data fuse to respond in the ways predicted by Alphalands
in the worlds connected information systems, he machines, and at one point they detonate an ex-
creates a series of temporary identities for himself. plosive magnetic device that erases all the computer
When Tarnover and the Federal Bureau of Data memories in the Alphaland capital. John Shirleys
Processing conspire to destroy the few remaining often-reprinted cyberpunk story Freezone (1985)
free communities, he writes a tapeworm to de- imagines that the economy of the capitalist world
liver the weapon that government fears the most: collapsed into the Computer Storage Depression
truth. when the electromagnetic pulse from a nuclear
weapon detonated by Arab terrorists erased all the ture United States run by engineers, in which au-
data in the United States. tomation (epitomized by the vast computer, EPI-
Many writers have explored the possible conflicts CAC XIV) has thrown most of the workforce into
that might arise between people and their machines. unemployment and is gradually rendering even the
Perhaps the most influential such story was the 1921 mostly highly skilled jobs obsolete, including, eventu-
drama, R.U.R. (Rossums Universal Robots), by the ally, those of the engineers themselves.
Czech writer Karel Capek (18901938). This work Michael Crichtons mainstream novel Prey (2002)
introduced the term robot, from a Czech word mean- warns that corporate greed may inadvertently pro-
ing heavy labor, with the implication of compulsory duce lethal threats through a combination of poorly
work or serfdom. Rossum, whose name may be understood technological innovations at the inter-
derived from the Czech word for mind or reason, in- section of computing, genetic engineering, and nano-
vented these manlike machines in order to prove that technology. This story of a monster that terrorizes
God was unnecessary. After Rossums death, his heirs workers at a remote research laboratory was partly
built a vast industry supplying the labor needs of the inspired by the new swarm concept in robotics, the
world with these robots. idea that a very large number of individually unin-
The motives of the builders of the robots were telligent machines might achieve intelligence by
various. Some simply wanted to earn money. Others interacting socially with one another. But the fun-
wanted to liberate the lower social classes from un- damental concept is one that has very recently
pleasant labor and turn everybody into aristocrats. achieved prominence in science policy debates in the
Robots were far cheaper than human laborers, so the real world, namely, technological convergence.
world became awash with wealth, whatever way it Thoughtful scientists and engineers in many fields
was shared across the social classes. Once people have begun to explore ways in which human abili-
no longer needed to work, however, they seemed ties may be greatly enhanced through convergence
to lose the will to live and stopped having children. of information technology, biotechnology, nan-
Conflict between people continued, but now the sol- otechnology, and cognitive science. The conscious
diers were robots rather than humans. Believing that aim is certainly not to create monsters, although par-
people were irrational and inefficient, the robots re- ticipants in the convergence movement are very con-
belled, and began to exterminate the entire human scious of the need to consider the social implications
species. The play is a farce, but it examines profound of their work. Rather, their aim is to strengthen the
issues regarding the nature of humanity and the creativity and freedom of individual humans, per-
relationship between humans and their creations. haps ultimately through some kind of convergence
A substantial dystopian literature has postulated between humans and the tools that serve them.
various ways in which robots or computers might wrest
control of the world from humans, imposing cyber-
dictatorship and eradicating freedom. The Humanoids Convergence of
(1950), by Jack Williamson, imagines that perfect ro-
bots were invented and programmed to follow a prime Humans and Machines
directiveto serve and obey, and guard men from From the very beginnings of the science fiction genre,
harmstrictly. These seemingly benevolent machines many stories imagined that technology could aug-
manufacture endless copies of themselves and set about ment human abilities. For example, Heinleins
liberating humanity from labor and danger. Soon, 1942 hard-science story Waldo concerned a dis-
everybody has at least one invincible, ever-present ro- abled man who invented remote manipulator arms
bot companion who prevents them from doing any- that compensated for his disabilities. For decades,
thing dangerous, such as using a tool, engaging in a other science fiction writers used the term waldo
physical sport, or conducting scientific research. Player to mean remote manipulator, but the word never
Piano (1952), by Kurt Vonnegut (b. 1922), depicts a fu- caught on in actual robotics.
Excerpt from Isaac Asimovs I, Robot
I
n his classic science fiction anthology I, Robot (1950), Isaac Asimov looks ahead to a world in which robots move from prim-
itive machines in the early twenty-first century to highly sophisticated creatures who may indeed rule the world a short
fifty years later. The stories in the anthology are told by Robopsychologist Dr. Susan Calvin to a reporter from the Interplanetary
Press. In the extract below, Dr. Calvin reminisces about her fifty-year tenure at U.S. Robots.
The offices and factories of U.S. Robots were a small city; spaced and planned. It was flattened out like an aerial photo-
graph.
When I first came here, she said, I had a little room in a building right about there where the fire-house is now.
She pointed. It was torn down before you were born. I shared the room with three others. I had half a desk. We built our
robots all in one building. Output-three a week. Now look at us.
Fifty years, I hackneyed, is a long time.
Not when youre looking back at them, she said. You wonder how they vanished so quickly.
She went back to her desk and sat down. She didnt need expression on her face to look sad, somehow.
How old are you? she wanted to know.
Thirty-two, I said.
Then you dont remember a world without robots. There was a time when humanity faced the universe alone and
without a friend. Now he has creatures to help him; stronger creatures than himself, more faithful, more useful, and ab-
solutely devoted to him. Mankind is no longer alone. Have you ever thought of it that way?
Im afraid I havent. May I quote you?
You may. To you, a robot is a robot. Gears and metal; electricity and positrons.-Mind and iron! Human-made! if nec-
essary, human-destroyed! But you havent worked with them, so you dont know them. Theyre a cleaner better breed
than we are.
I tried to nudge her gently with words, Wed like to hear some of the things you could tell us; get your views on ro-
bots. The Interplanetary Press reaches the entire Solar System. Potential audience is three billion, Dr. Calvin. They ought
to know what you could tell them on robots.
It wasnt necessary to nudge. She didnt hear me, but she was moving in the right direction.
They might have known that from the start. We sold robots for Earth-use then-before my time it was, even. Of course,
that was when robots could not talk. Afterward, they became more human and opposition began. The labor unions, of
course, naturally opposed robot competition for human jobs, and various segments of religious opinion had their super-
stitious objections. It was all quite ridiculous and quite useless. And yet there it was.
Source: Asimov, I. (1950). I, robot (pp. 16<N>17). Garden City, NY: Doubleday & Company.
Hard science merges with cyberpunk in many 1. A robot must obey the government, and state
stories that describe how humans of the future might directives supercede all private commands.
merge with their machines. In the 1954 story Fondly 2. A robot cannot endanger life or property.
Fahrenheit, by Alfred Bester (19131987), a man 3. A robot must obey its owner.
and his robot flee from planet to planet to escape
justice for murders they are committing. These laws can conflict with each other, espe-
Psychologically, the two have blended. This is evi- cially for sophisticated robots capable of perform-
dent in the very style of the writing, because the first- ing a wide range of tasks and who belong to corrupt
person narrative perspective constantly shifts or insane owners. The first rule is meant to solve such
from one to the other, even within the same para- problems. In the modern world the state is the ulti-
graph. The hard-science aspects of the story include mate judge of morality, for humans and machines
three laws of robotics that are rather different from alike. What happens, then, if the owner tells the ro-
those propounded by Asimov: bot that a state command did not actually come from
the state, but is a lie or an error in communica- individuals but existing within a fixed population
tion? Overcome by the heat of human emotion, such size and following set laws of behavior. At birth, a
a robot might even commit serial homicide. Diasparan is biologically about twenty, but requires
Besters novel The Computer Connection (1974) about twenty years to mature. During this time
depicts the merging of human and machine in an- the individual does not possess any memories of his
other way. It concerns a scientist at Jet Propulsion or her previous lives. Then the memories return,
Laboratory whose personality lands in his super- and he or she lives out a life that is practically a re-
computer when an epileptic seizure coupled with play of the previous one, content within the artifi-
profound emotional shock drive it out of his body. cial womb of the city. Thus they fail to use their
In Software (1982), by Rudy Rucker, human minds advanced technology to do anything really produc-
are uploaded to computers by scanning the brain as tive, such as exploring the stars. If human beings
knives destructively slice it apart. Greg Bears Blood really do merge with their computers over the com-
Music (2002) imagines that a combination of genetic ing centuries, we can wonder whether this will help
and electronic technologies could create noocytes, them to achieve great things in the real universe, or
viruslike molecular computers that absorb the minds to retreat from challenge into a meaningless, virtual
and dissolve the bodies of human beings. The existence.
nondestructive scanning device in The Terminal
Experiment (1995), by Robert J. Sawyer, employs a William Sims Bainbridge
billion nanotechnology sensors and computer inte-
gration techniques to map the neural pathways of
the brain from the surface of the scalp. Scanning ses- FURTHER READING
sions bombard the subject with sights and sounds,
to activate all the brains neural pathways. Asimov, I. (1950). I, robot. New York: Grosset and Dunlap.
In both The Computer Connection and The Ter- Asimov, I. (1954). The caves of steel. Garden City, NY: Doubleday.
minal Experiment, uploading a person to a computer Asimov, I. (1956). The naked sun. New York: Bantam.
Asimov, I. (1983). The robots of dawn. New York: Ballantine.
removes normal inhibitions, so the person becomes Bainbridge, W. S. (1986). Dimensions of science fiction. Cambridge,
a murderer. When people are uploaded in Greg Egans MA: Harvard University Press.
Permutation City (1994), they immediately com- Baum, L. F. (1904). The marvelous ozama of Oz. Chicago: Reilly and
mit suicide. Britton.
Bear, G. (2002). Blood music. New York: ibooks.
In the classic 1953 novel The City and The Stars, Bester, A. (1974). The computer connection. New York: ibooks.
by Arthur C. Clarke (b. 1917), people uploaded into Bester, A. (1997). Virtual unrealities: The short fiction of Alfred
the computer-generated city, Diaspar, lose their ca- Bester. New York: Vintage.
Binder, E. (1965). Adam LinkRobot. New York: Paperback Library.
pacity to explore and evolve. Diaspar is an eternal Brunner, J. (1975). The shockwave rider. New York: Ballantine.
city, and its people are eternal as well. Superficially Capek, K. (1990). Toward the radical center: A Karel Capek reader. New
they are like humans, except that they produce no York: Catbird.
children. In a thousand-year lifetime, they enjoy ad- Cardigan, P. (Ed.). (2002). The ultimate cyberpunk. New York: ibooks.
Clancy, T., & Pieczenik, S. (1998). Net force. New York: Berkley.
venture games in computer-generated virtual re- Clarke, A. C. (1953). The city and the stars. New York: Harcourt, Brace
ality, create works of art that are destined to be and Company.
erased, and gradually grow weary. Then they enter Clute, J., & Nicholls, P. (1995). The encyclopedia of science fiction. New
the Hall of Creation to be archived as patterns of York: St. Martins Griffin.
Crichton, M. (2002). Prey. New York: HarperCollins.
electrical charges inside the Central Computer. After Egan, G. (1994). Permutation City. New York: Harper.
a few thousand years, or a few million, they will be Gibson, W. (1984). Neuromancer. New York: Ace.
reconstituted again, to experience another life in Gibson, W., & Sterling, B. (1991). The difference engine. New York:
Bantam.
Diaspar before once again entering the archives. Heinlein, R. A. (1950). Waldo and Magic Inc. Garden City, NY:
Each year, about ten thousand people are restored Doubleday.
to life, always a somewhat different combination of Heinlein, R. A. (1966). The moon is a harsh mistress. New York: Orb.
Hoffman, E. T. A. (1885). The sand-man. In Weird Tales (J. T. Spiller, N. (Ed.). (2002). Cyber reader: Critical writings for the digital
Bealby, Trans.). New York: Scribners. (Original work published era. New York: Phaidon.
1817) Stephenson, N. (1992). Snow crash. New York: Bantam.
Reynolds, M. (1967). Computer war. New York: Ace. Sterling, B. (1986). Mirrorshades: The cyberpunk anthology. New York:
Roco, M. C., & Bainbridge, W. S. (2003). Converging technologies for Ace.
improving human performance. Dordrecht, Netherlands: Kluwer. Vonnegut, K. (1952). Player piano. New York: Delta.
Rucker, R. (1982). Software. New York: Avon. Williamson, J. (1950). The humanoids. New York: Grosset and Dunlap.
Sawyer, R. J. (1995). The terminal experiment. New York: HarperCollins.

Bainbridge W.S. (Ed) Berkshire Encyclopedia of Human-Computer Interaction. Vol. 1

Uploaded by

Copyright:

Available Formats

Bainbridge W.S. (Ed) Berkshire Encyclopedia of Human-Computer Interaction. Vol. 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bainbridge W.S. (Ed) Berkshire Encyclopedia of Human-Computer Interaction. Vol. 1

Uploaded by

Copyright:

Available Formats

How are the people of Diaspar different from humans?

How are the people of Diaspar different from humans?

What do the people of Diaspar do with their time?

What do the people of Diaspar do with their time?

BERKSHIRE PUBLISHING GROUP

William Sims Bainbridge

Great Barrington, Massachusetts U.S.A.

Cover photo: Thad Starner sporting a wearable computer.

Printed in the United States of America

Library of Congress Cataloging-in-Publishing Data

Information Management and Programming

Adaptive Help Systems Animation Artificial Intelligence

Augmented Reality Compilers Digital Divide

Ethnography Handwriting Recognition and Information Theory

Fly-by-Wire Human-Robot Interaction Iterative Design

Geographic Information Systems Impacts Law and HCI

Gesture Recognition Information Filtering Law Enforcement

Markup Languages Online Education Programming Languages

Movies Open Source Software Repetitive Strain Injury

MUDs Optical Character Recognition Scenario-Based Design

Social Psychology and HCI Task Analysis Jenny Preece

Sociology and HCI Telecommuting Value Sensitive Design

Socio-Technical System Design Telepresence Video

Software Cultures Text Summarization Video Summarization

Software Engineering Theory Virtual Reality

This list is provided to assist readers in locating en- Navigation

Compilers Markup Languages

Adaptive Help Systems Chatrooms

Adaptive Interfaces Classrooms

Augmented Cognition Cybercommunities

Digital Divide Internet in Everyday Life

Geographic Information Systems Political Science and HCI

Spell Checker Work

Abras, Chadia Alvarez, R. Michael

Agrawal, Dharma P. Aronson, Jay E.

Allanson, Jennifer Bailenson, Jeremy

Allison, Brendan Z. Bainbridge, Erika

Bainbridge, William Sims Bilal, Dania

Batteau, Allen W. Cai, Guoray

Ben-Arie, Jezekiel Caldera, Cavinda T.

Bernard, Michael L. Campbell, Murray

Carroll, John M. Crosby, Martha E.

Draper, John V. Fountain, Jane E.

Duchowski, Andrew T. Franklin, Judy A.

DuPont, Brandon Freedman, Immanuel

El Rhalibi, Abdennour Friedman, Batya

Gross, Mark D. Hauck, Roslin V.

Grudin, Jonathan Hedge, Alan

Gustafson, John Hennecke, Marcus

Guthrie, Ruth Hickey, Timothy J.

Gutierrez-Osuna, Ricardo Hodges, Larry F.

Guzman, Indira R. Hogan, Bernie

Hakken, David Holland, Arnold

Jackson, Linda A. Klavans, Judith L.

Kettebekov, Sanshzar Laff, Mark

Khedekar, Swapnil Lagoudakis, Michail G.

Kim, Hong-Gee Lampe, Cliff

Kim, Jee-In Lane, David M.

Lok, Benjamin C. Miller, Sonia E.