Kurosu (2013) HCI User and Contexts of Use

Masaaki Kurosu (Ed.
Human-Computer
LNCS 8006
Interaction
Users and Contexts of Use
15th International Conference, HCI International 2013
Las Vegas, NV, USA, July 2013
Proceedings, Part III
123
Lecture Notes in Computer Science 8006
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Alfred Kobsa
University of California, Irvine, CA, USA
Friedemann Mattern
ETH Zurich, Switzerland
John C. Mitchell
Stanford University, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
Oscar Nierstrasz
University of Bern, Switzerland
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Germany
Madhu Sudan
Microsoft Research, Cambridge, MA, USA
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbruecken, Germany
Masaaki Kurosu (Ed.)
Human-Computer
Interaction
Users and Contexts of Use
15th International Conference, HCI International 2013

Las Vegas, NV, USA, July 21-26, 2013
Proceedings, Part III
13
Volume Editor
Masaaki Kurosu
The Open University of Japan
2-11 Wakaba, Mihama-ku, Chiba-shi 261-8586, Japan
E-mail: [email protected]
ISSN 0302-9743 e-ISSN 1611-3349

ISBN 978-3-642-39264-1 e-ISBN 978-3-642-39265-8
DOI 10.1007/978-3-642-39265-8
Springer Heidelberg Dordrecht London New York
Library of Congress Control Number: 2013941394
CR Subject Classification (1998): H.5, H.3, H.4, K.3, K.6
LNCS Sublibrary: SL 3 – Information Systems and Application, incl. Internet/Web

and HCI
© Springer-Verlag Berlin Heidelberg 2013
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and
executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication
or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location,
in its current version, and permission for use must always be obtained from Springer. Permissions for use
may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution
under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of publication,
neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or
omissions that may be made. The publisher makes no warranty, express or implied, with respect to the
material contained herein.
Typesetting: Camera-ready by author, data conversion by Scientific Publishing Services, Chennai, India
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
Foreword
The 15th International Conference on Human–Computer Interaction, HCI In-

ternational 2013, was held in Las Vegas, Nevada, USA, 21–26 July 2013, incor-
porating 12 conferences / thematic areas:
Thematic areas:
• Human–Computer Interaction
• Human Interface and the Management of Information
Affiliated conferences:
• 10th International Conference on Engineering Psychology and Cognitive

Ergonomics
• 7th International Conference on Universal Access in Human–Computer
Interaction
• 5th International Conference on Virtual, Augmented and Mixed Reality
• 5th International Conference on Cross-Cultural Design
• 5th International Conference on Online Communities and Social Computing
• 7th International Conference on Augmented Cognition
• 4th International Conference on Digital Human Modeling and Applications
in Health, Safety, Ergonomics and Risk Management
• 2nd International Conference on Design, User Experience and Usability
• 1st International Conference on Distributed, Ambient and Pervasive Inter-
actions
• 1st International Conference on Human Aspects of Information Security,
Privacy and Trust
A total of 5210 individuals from academia, research institutes, industry and gov-
ernmental agencies from 70 countries submitted contributions, and 1666 papers
and 303 posters were included in the program. These papers address the latest
research and development efforts and highlight the human aspects of design and
use of computing systems. The papers accepted for presentation thoroughly cover
the entire field of Human–Computer Interaction, addressing major advances in
knowledge and effective use of computers in a variety of application areas.
This volume, edited by Masaaki Kurosu, contains papers focusing on the
thematic area of Human–Computer Interaction, and addressing the following
major topics:
• Identity, Privacy and Trust

• User Studies
• Interaction for Society and Community
• HCI for Business and Innovation
VI Foreword
The remaining volumes of the HCI International 2013 proceedings are:

• Volume 1, LNCS 8004, Human–Computer Interaction: Human-Centred De-
sign Approaches, Methods, Tools and Environments (Part I), edited by
Masaaki Kurosu
• Volume 2, LNCS 8005, Human–Computer Interaction: Applications and Ser-
vices (Part II), edited by Masaaki Kurosu
• Volume 4, LNCS 8007, Human–Computer Interaction: Interaction Modali-
ties and Techniques (Part IV), edited by Masaaki Kurosu
• Volume 5, LNCS 8008, Human–Computer Interaction: Towards Intelligent
and Implicit Interaction (Part V), edited by Masaaki Kurosu
• Volume 6, LNCS 8009, Universal Access in Human–Computer Interaction:
Design Methods, Tools and Interaction Techniques for eInclusion (Part I),
edited by Constantine Stephanidis and Margherita Antona
User and Context Diversity (Part II), edited by Constantine Stephanidis and
Margherita Antona
Applications and Services for Quality of Life (Part III), edited by Constan-
tine Stephanidis and Margherita Antona
• Volume 9, LNCS 8012, Design, User Experience, and Usability: Design Phi-
losophy, Methods and Tools (Part I), edited by Aaron Marcus
• Volume 10, LNCS 8013, Design, User Experience, and Usability: Health,
Learning, Playing, Cultural, and Cross-Cultural User Experience (Part II),
edited by Aaron Marcus
• Volume 11, LNCS 8014, Design, User Experience, and Usability: User Ex-
perience in Novel Technological Environments (Part III), edited by Aaron
Marcus
• Volume 12, LNCS 8015, Design, User Experience, and Usability: Web, Mobile
and Product Design (Part IV), edited by Aaron Marcus
• Volume 13, LNCS 8016, Human Interface and the Management of Informa-
tion: Information and Interaction Design (Part I), edited by Sakae Yamamoto
tion: Information and Interaction for Health, Safety, Mobility and Complex
Environments (Part II), edited by Sakae Yamamoto
tion: Information and Interaction for Learning, Culture, Collaboration and
Business (Part III), edited by Sakae Yamamoto
• Volume 16, LNAI 8019, Engineering Psychology and Cognitive Ergonomics:
Understanding Human Cognition (Part I), edited by Don Harris
• Volume 17, LNAI 8020, Engineering Psychology and Cognitive Ergonomics:
Applications and Services (Part II), edited by Don Harris
• Volume 18, LNCS 8021, Virtual, Augmented and Mixed Reality: Designing
and Developing Augmented and Virtual Environments (Part I), edited by
Randall Shumaker
• Volume 19, LNCS 8022, Virtual, Augmented and Mixed Reality: Systems
and Applications (Part II), edited by Randall Shumaker
Foreword VII
• Volume 20, LNCS 8023, Cross-Cultural Design: Methods, Practice and Case
Studies (Part I), edited by P.L. Patrick Rau
• Volume 21, LNCS 8024, Cross-Cultural Design: Cultural Differences in Ev-
eryday Life (Part II), edited by P.L. Patrick Rau
• Volume 22, LNCS 8025, Digital Human Modeling and Applications in Health,
Safety, Ergonomics and Risk Management: Healthcare and Safety of the En-
vironment and Transport (Part I), edited by Vincent G. Duffy
• Volume 23, LNCS 8026, Digital Human Modeling and Applications in Health,
Safety, Ergonomics and Risk Management: Human Body Modeling and Er-
gonomics (Part II), edited by Vincent G. Duffy
• Volume 24, LNAI 8027, Foundations of Augmented Cognition, edited by
Dylan D. Schmorrow and Cali M. Fidopiastis
• Volume 25, LNCS 8028, Distributed, Ambient and Pervasive Interactions,
edited by Norbert Streitz and Constantine Stephanidis
• Volume 26, LNCS 8029, Online Communities and Social Computing, edited
by A. Ant Ozok and Panayiotis Zaphiris
• Volume 27, LNCS 8030, Human Aspects of Information Security, Privacy
and Trust, edited by Louis Marinos and Ioannis Askoxylakis
• Volume 28, CCIS 373, HCI International 2013 Posters Proceedings (Part I),
edited by Constantine Stephanidis
• Volume 29, CCIS 374, HCI International 2013 Posters Proceedings (Part II),
edited by Constantine Stephanidis
I would like to thank the Program Chairs and the members of the Program
Boards of all affiliated conferences and thematic areas, listed below, for their
contribution to the highest scientific quality and the overall success of the HCI
International 2013 conference.
This conference could not have been possible without the continuous sup-
port and advice of the Founding Chair and Conference Scientific Advisor, Prof.
Gavriel Salvendy, as well as the dedicated work and outstanding efforts of the
Communications Chair and Editor of HCI International News, Abbas Moallem.
I would also like to thank for their contribution towards the smooth organi-
zation of the HCI International 2013 Conference the members of the Human–
Computer Interaction Laboratory of ICS-FORTH, and in particular George
Paparoulis, Maria Pitsoulaki, Stavroula Ntoa, Maria Bouhli and George Kapnas.
May 2013 Constantine Stephanidis

General Chair, HCI International 2013
Organization
Human–Computer Interaction
Program Chair: Masaaki Kurosu, Japan

Jose Abdelnour-Nocera, UK Kyungdoh Kim, South Korea
Sebastiano Bagnara, Italy Heidi Krömker, Germany
Simone Barbosa, Brazil Chen Ling, USA
Tomas Berns, Sweden Yan Liu, USA
Nigel Bevan, UK Zhengjie Liu, P.R. China
Simone Borsci, UK Loı̈c Martı́nez Normand, Spain
Apala Lahiri Chavan, India Chang S. Nam, USA
Sherry Chen, Taiwan Naoko Okuizumi, Japan
Kevin Clark, USA Noriko Osaka, Japan
Torkil Clemmensen, Denmark Philippe Palanque, France
Xiaowen Fang, USA Hans Persson, Sweden
Shin’ichi Fukuzumi, Japan Ling Rothrock, USA
Vicki Hanson, UK Naoki Sakakibara, Japan
Ayako Hashizume, Japan Dominique Scapin, France
Anzai Hiroyuki, Italy Guangfeng Song, USA
Sheue-Ling Hwang, Taiwan Sanjay Tripathi, India
Wonil Hwang, South Korea Chui Yin Wong, Malaysia
Minna Isomursu, Finland Toshiki Yamaoka, Japan
Yong Gu Ji, South Korea Kazuhiko Yamazaki, Japan
Esther Jun, USA Ryoji Yoshitake, Japan
Mitsuhiko Karashima, Japan Silvia Zimmermann, Switzerland
Human Interface and the Management of Information
Program Chair: Sakae Yamamoto, Japan
Hans-Jorg Bullinger, Germany Mark Lehto, USA

Alan Chan, Hong Kong Hiroyuki Miki, Japan
Gilsoo Cho, South Korea Hirohiko Mori, Japan
Jon R. Gunderson, USA Fiona Fui-Hoon Nah, USA
Shin’ichi Fukuzumi, Japan Shogo Nishida, Japan
Michitaka Hirose, Japan Robert Proctor, USA
Jhilmil Jain, USA Youngho Rhee, South Korea
Yasufumi Kume, Japan Katsunori Shimohara, Japan
X Organization
Michale Smith, USA Kim-Phuong Vu, USA

Tsutomu Tabe, Japan Tomio Watanabe, Japan
Hiroshi Tsuji, Japan Hidekazu Yoshikawa, Japan
Engineering Psychology and Cognitive Ergonomics
Program Chair: Don Harris, UK

Guy Andre Boy, USA Paul Salmon, Australia
Joakim Dahlman, Sweden Axel Schulte, Germany
Trevor Dobbins, UK Siraj Shaikh, UK
Mike Feary, USA Sarah C. Sharples, UK
Shan Fu, P.R. China Anthony Smoker, UK
Michaela Heese, Austria Neville A. Stanton, UK
Hung-Sying Jing, Taiwan Alex Stedmon, UK
Wen-Chin Li, Taiwan Xianghong Sun, P.R. China
Mark A. Neerincx, The Netherlands Andrew Thatcher, South Africa
Jan M. Noyes, UK Matthew J.W. Thomas, Australia
Taezoon Park, Singapore Rolf Zon, The Netherlands
Universal Access in Human–Computer Interaction
Program Chairs: Constantine Stephanidis, Greece,

and Margherita Antona, Greece
Julio Abascal, Spain Georgios Kouroupetroglou, Greece

Ray Adams, UK Patrick Langdon, UK
Gisela Susanne Bahr, USA Seongil Lee, Korea
Margit Betke, USA Ana Isabel B.B. Paraguay, Brazil
Christian Bühler, Germany Helen Petrie, UK
Stefan Carmien, Spain Michael Pieper, Germany
Jerzy Charytonowicz, Poland Enrico Pontelli, USA
Carlos Duarte, Portugal Jaime Sanchez, Chile
Pier Luigi Emiliani, Italy Anthony Savidis, Greece
Qin Gao, P.R. China Christian Stary, Austria
Andrina Granić, Croatia Hirotada Ueda, Japan
Andreas Holzinger, Austria Gerhard Weber, Germany
Josette Jones, USA Harald Weber, Germany
Simeon Keates, UK
Organization XI
Virtual, Augmented and Mixed Reality
Program Chair: Randall Shumaker, USA
Waymon Armstrong, USA Mark Livingston, USA

Juan Cendan, USA Michael Macedonia, USA
Rudy Darken, USA Gordon Mair, UK
Cali M. Fidopiastis, USA Jose San Martin, Spain
Charles Hughes, USA Jacquelyn Morie, USA
David Kaber, USA Albert “Skip” Rizzo, USA
Hirokazu Kato, Japan Kay Stanney, USA
Denis Laurendeau, Canada Christopher Stapleton, USA
Fotis Liarokapis, UK Gregory Welch, USA
Cross-Cultural Design
Program Chair: P.L. Patrick Rau, P.R. China

Pilsung Choe, P.R. China Sheau-Farn Max Liang, Taiwan
Henry Been-Lirn Duh, Singapore Liang Ma, P.R. China
Vanessa Evers, The Netherlands Alexander Mädche, Germany
Paul Fu, USA Katsuhiko Ogawa, Japan
Zhiyong Fu, P.R. China Tom Plocher, USA
Fu Guo, P.R. China Kerstin Röse, Germany
Sung H. Han, Korea Supriya Singh, Australia
Toshikazu Kato, Japan Hsiu-Ping Yueh, Taiwan
Dyi-Yih Michael Lin, Taiwan Liang (Leon) Zeng, USA
Rungtai Lin, Taiwan Chen Zhao, USA
Online Communities and Social Computing
Program Chairs: A. Ant Ozok, USA,

and Panayiotis Zaphiris, Cyprus
Areej Al-Wabil, Saudi Arabia Niki Lambropoulos, Greece
Leonelo Almeida, Brazil Effie Law, Switzerland
Bjørn Andersen, Norway Soo Ling Lim, UK
Chee Siang Ang, UK Fernando Loizides, Cyprus
Aneesha Bakharia, Australia Gabriele Meiselwitz, USA
Ania Bobrowicz, UK Anthony Norcio, USA
Paul Cairns, UK Elaine Raybourn, USA
Farzin Deravi, UK Panote Siriaraya, UK
Andri Ioannou, Cyprus David Stuart, UK
Slava Kisilevich, Germany June Wei, USA
XII Organization
Augmented Cognition
Program Chairs: Dylan D. Schmorrow, USA,

and Cali M. Fidopiastis, USA
Robert Arrabito, Canada Santosh Mathan, USA
Richard Backs, USA Rob Matthews, Australia
Chris Berka, USA Dennis McBride, USA
Joseph Cohn, USA Jeff Morrison, USA
Martha E. Crosby, USA Mark A. Neerincx, The Netherlands
Julie Drexler, USA Denise Nicholson, USA
Ivy Estabrooke, USA Banu Onaral, USA
Chris Forsythe, USA Lee Sciarini, USA
Wai Tat Fu, USA Kay Stanney, USA
Rodolphe Gentili, USA Roy Stripling, USA
Marc Grootjen, The Netherlands Rob Taylor, UK
Jefferson Grubb, USA Karl van Orden, USA
Ming Hou, Canada
Digital Human Modeling and Applications in Health,

Safety, Ergonomics and Risk Management
Program Chair: Vincent G. Duffy, USA and Russia

Karim Abdel-Malek, USA Kang Li, USA
Giuseppe Andreoni, Italy Tim Marler, USA
Daniel Carruth, USA Michelle Robertson, USA
Eliza Yingzi Du, USA Matthias Rötting, Germany
Enda Fallon, Ireland Peter Vink, The Netherlands
Afzal Godil, USA Mao-Jiun Wang, Taiwan
Ravindra Goonetilleke, Hong Kong Xuguang Wang, France
Bo Hoege, Germany Jingzhou (James) Yang, USA
Waldemar Karwowski, USA Xiugan Yuan, P.R. China
Zhizhong Li, P.R. China Gülcin Yücel Hoge, Germany
Design, User Experience, and Usability
Program Chair: Aaron Marcus, USA

Sisira Adikari, Australia Ana Boa-Ventura, USA
Ronald Baecker, Canada Jan Brejcha, Czech Republic
Arne Berger, Germany Lorenzo Cantoni, Switzerland
Jamie Blustein, Canada Maximilian Eibl, Germany
Organization XIII
Anthony Faiola, USA Jennifer McGinn, USA

Emilie Gould, USA Francisco Rebelo, Portugal
Zelda Harrison, USA Michael Renner, Switzerland
Rüdiger Heimgärtner, Germany Kerem Rızvanoğlu, Turkey
Brigitte Herrmann, Germany Marcelo Soares, Brazil
Steffen Hess, Germany Christian Sturm, Germany
Kaleem Khan, Canada Michele Visciola, Italy
Distributed, Ambient and Pervasive Interactions
Program Chairs: Norbert Streitz, Germany,

and Constantine Stephanidis, Greece
Emile Aarts, The Netherlands Christian Müller-Tomfelde, Australia
Adnan Abu-Dayya, Qatar Fabio Paternó, Italy
Juan Carlos Augusto, UK Gilles Privat, France
Boris de Ruyter, The Netherlands Harald Reiterer, Germany
Anind Dey, USA Carsten Röcker, Germany
Dimitris Grammenos, Greece Reiner Wichert, Germany
Nuno M. Guimaraes, Portugal Woontack Woo, South Korea
Shin’ichi Konomi, Japan Xenophon Zabulis, Greece
Carsten Magerkurth, Switzerland
Human Aspects of Information Security, Privacy and Trust
Program Chairs: Louis Marinos, ENISA EU,

and Ioannis Askoxylakis, Greece
Claudio Agostino Ardagna, Italy Ronald Leenes, The Netherlands

Zinaida Benenson, Germany Javier Lopez, Spain
Daniele Catteddu, Italy Steve Marsh, Canada
Raoul Chiesa, Italy Gregorio Martinez, Spain
Bryan Cline, USA Emilio Mordini, Italy
Sadie Creese, UK Yuko Murayama, Japan
Jorge Cuellar, Germany Masakatsu Nishigaki, Japan
Marc Dacier, USA Aljosa Pasic, Spain
Dieter Gollmann, Germany Milan Petković, The Netherlands
Kirstie Hawkey, Canada Joachim Posegga, Germany
Jaap-Henk Hoepman, The Netherlands Jean-Jacques Quisquater, Belgium
Cagatay Karabat, Turkey Damien Sauveron, France
Angelos Keromytis, USA George Spanoudakis, UK
Ayako Komatsu, Japan Kerry-Lynn Thomson, South Africa
XIV Organization
Julien Touzeau, France Claire Vishik, UK

Theo Tryfonas, UK Melanie Volkamer, Germany
João Vilela, Portugal
External Reviewers
Maysoon Abulkhair, Saudi Arabia Vassilis Kouroumalis, Greece

Ilia Adami, Greece Stephanie Lackey, USA
Vishal Barot, UK Janelle LaMarche, USA
Stephan Böhm, Germany Asterios Leonidis, Greece
Vassilis Charissis, UK Nickolas Macchiarella, USA
Francisco Cipolla-Ficarra, Spain George Margetis, Greece
Maria De Marsico, Italy Matthew Marraffino, USA
Marc Fabri, UK Joseph Mercado, USA
David Fonseca, Spain Claudia Mont’Alvão, Brazil
Linda Harley, USA Yoichi Motomura, Japan
Yasushi Ikei, Japan Karsten Nebe, Germany
Wei Ji, USA Stavroula Ntoa, Greece
Nouf Khashman, Canada Martin Osen, Austria
John Killilea, USA Stephen Prior, UK
Iosif Klironomos, Greece Farid Shirazi, Canada
Ute Klotz, Switzerland Jan Stelovsky, USA
Maria Korozi, Greece Sarah Swierenga, USA
Kentaro Kotani, Japan
HCI International 2014
The 16th International Conference on Human–Computer Interaction, HCI

International 2014, will be held jointly with the affiliated conferences in the
summer of 2014. It will cover a broad spectrum of themes related to Human–
Computer Interaction, including theoretical issues, methods, tools, processes and
case studies in HCI design, as well as novel interaction techniques, interfaces
and applications. The proceedings will be published by Springer. More infor-
mation about the topics, as well as the venue and dates of the conference,
will be announced through the HCI International Conference series website:
http://www.hci-international.org/
General Chair
Professor Constantine Stephanidis
University of Crete and ICS-FORTH
Heraklion, Crete, Greece
Email: [email protected]
Table of Contents – Part III
Identity, Privacy and Trust

SP-CIP: A Secure and Privacy Aware Patient Centric Information
Portal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Subrata Acharya
Identity Management through “Profiles”: Prototyping an Online

Information Segregation Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Julio Angulo and Erik Wästlund
An Empirical Investigation of Similarity-Driven Trust Dynamics in a

Social Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Yugo Hayashi, Victor Kryssanov, and Hitoshi Ogawa
The di.me User Interface: Concepts for Sharing Personal Information

via Multiple Identities in a Decentralized Social Network . . . . . . . . . . . . . . 29
Fabian Hermann, Andreas Schuller, Simon Thiel,
Christian Knecht, and Simon Scerri
Trust and Privacy in the di.me Userware . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Marcel Heupel, Mohamed Bourimi, and Doğan Kesdoğan
From Living Space to Urban Quarter: Acceptance of ICT Monitoring

Solutions in an Ageing Society . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Simon Himmel, Martina Ziefle, and Katrin Arning
The Impact of Explanation Dialogues on Human-Computer Trust . . . . . . 59

Florian Nothdurft, Tobias Heinroth, and Wolfgang Minker
Trustworthy and Inclusive Identity Management for Applications in

Social Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Till Halbach Røssvoll and Lothar Fritsch
Towards a Design Guideline of Visual Cryptography on Stereoscopic

Displays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Shih-Lung Tsai and Chao-Hua Wen
User Studies
Interface Design for Minimizing Loss of Context in In-situ Remote
Robot Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Jong-gil Ahn and Gerard Jounghyun Kim
XVIII Table of Contents – Part III
Assessing the Impact of Latency and Jitter on the Perceived Quality of

Call of Duty Modern Warfare 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Rahul Amin, France Jackson, Juan Gilbert, Jim Martin, and
Terry Shaw
The Effects of Negative Interaction Feedback in a Web Navigation

Assistant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Marcelo G. Armentano and Analı́a A. Amandi
Automatic Layout Generation for Digital Photo Albums:

A User Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Francine B. Bergmann, Isabel H. Manssour, Milene S. Silveira, and
João Batista S. de Oliveira
Icon Design for Older Users of Project Management Software . . . . . . . . . . 127

Christina Bröhl, Jennifer Bützler, Nicole Jochems, and
Christopher M. Schlick
Search Strategies in Hypermedia Navigation and Spatial Abilities:

A Comparison with Physical Navigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Aurélie Brouwers
Data Visualisation, User Experience and Context: A Case Study from

Fantasy Sport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Rob Euman and José Abdelnour-Nocera
A Pedestrian Navigation Method for User’s Safe and Easy

Wayfinding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Hiroshi Furukawa and Yutaka Nakamura
The Effect of Information Quantity on Cbp Interface in the Advanced

Nuclear Power Plant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
Min-Chih Hsieh and Sheue-Ling Hwang
Development of Communication Support System at Mealtimes Using

Tabletop Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Junko Itou, Shizuka Takao, and Jun Munemori
Improving User Performance in Conditional Probability Problems with

Computer-Generated Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Vince Kellen, Susy Chan, and Xiaowen Fang
Establishing a Cognitive Map of Public Place for Blind and Visual

Impaired by Using IVEO Hands-On Learning System . . . . . . . . . . . . . . . . . 193
Qing-Wen Lin, Sheue-Ling Hwang, and Jan-Li Wang
An Application of the Ballistic Movement Method for Evaluating

Computer Mice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Ray F. Lin, Ching-Wen Chung, Yi-Chien Tsai, and Chi-Yu Huang
Table of Contents – Part III XIX
AugmentedBacklight: Expansion of LCD Backlights Using Lighting

Methods in the Real World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Maho Oki, Koji Tsukada, and Itiro Siio
An Exploratory Study to Understand Knowledge-Sharing in

Data-Intensive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Jongsoon Park and Joseph L. Gabbard
Interaction of the Elderly Viewer with Additional Multimedia Content

to Support the Appreciation of Television Programs . . . . . . . . . . . . . . . . . . 227
Kamila Rodrigues, Vânia Paula de Almeida Neris, and
Cesar Augusto Camillo Teixeira
Exploring Children’s Attitudes towards Static and Moving Humanoid

Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Fang-Wu Tung and Tsen-Yao Chang
Interaction for Society and Community

Mobile Money Services in Uganda: Design Gaps and
Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Rehema Baguma
A Static and Dynamic Recommendations System for Best Practice

Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Pierfrancesco Bellini, Ivan Bruno, Paolo Nesi, and Michela Paolucci
Connecting Electric Vehicles and Green Energy . . . . . . . . . . . . . . . . . . . . . . 269

Peter Dempster
SOCIETY: A Social Reading Application to Join Education and Social

Network Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Elena Guercio, Fabio Luciano Mondin, Maurizio Belluati, and
Lucia Longo
The Evaluation of a Voting Web Based Application . . . . . . . . . . . . . . . . . . 285

Linda Harley, Keith Kline, Jerry Ray, Carrie Bell,
Andrew Baranak, Chandler Price, Matthew Hung, and
Brad Fain
A Refuge Location Prediction System for When a Tsunami Has

Occurred . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Akhiro Kawabe, Tomoko Izumi, and Yoshio Nakatani
Further Benefit of a Kind of Inconvenience for Social Information

Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Hiroshi Kawakami
XX Table of Contents – Part III
The Present Condition and Problems for Elderly People Participating

in Communities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
Noriyuki Kobayashi, Shohei Yamane, Kotaro Ohori, and
Akihiko Obata
Applying to Twitter Networks of a Community Extraction Method
Using Intersection Graph and Semantic Analysis . . . . . . . . . . . . . . . . . . . . . 314
Toshiya Kuramochi, Naoki Okada, Kyohei Tanikawa,
Yoshinori Hijikata, and Shogo Nishida
Search Engine Accessibility for Low-Literate Users . . . . . . . . . . . . . . . . . . . 324
Débora Maurmo Modesto, Simone Bacellar Leal Ferreira, and
Aline Silva Alves
Identifying and Representing Elements of Local Contexts in Namibia . . . 332
Kasper Rodil, Kasper Løvborg Jensen, Matthias Rehm, and
Heike Winschiers-Theophilus
A Framework for Community-Oriented Mobile Interaction Design in
Emerging Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
Monica Sebillo, Genoveffa Tortora, Giuliana Vitiello,
Pasquale Di Giovanni, and Marco Romano
The Effects of (Social) Media on Revolutions – Perspectives from Egypt
and the Arab Spring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
Christian Sturm and Hossam Amer
Memotree: Using Online Social Networking to Strengthen Family
Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
Tsai-Hsuan Tsai, Yi-Lun Ho, Hsien-Tsung Chang, and Yu-Wen Li
HCI for Business and Innovation

It Was Nice with the Brick So Now I’ll Click: The Effects of Offline and
Online Experience, Perceived Benefits, and Trust on Dutch Consumers’
Online Repeat Purchase Intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
Ardion Beldad and Mariël Segers
Interaction between Enterprise Resource Planning Systems and
Organizational Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
Carlotta Herberhold
Reconfiguring the Corporate and Commons: Mobile and Online
Activism as a Form of Socio-technical Design . . . . . . . . . . . . . . . . . . . . . . . . 388
Constance Kampf
Social Media in the Product Development Process of the Automotive
Industry: A New Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
Andreas Klein and Götz Spiegel
Table of Contents – Part III XXI
Do All People Enjoy the Benefits from Technology Innovation? . . . . . . . . 402

Yoon Jeon Koh, Jae Heon Park, and Byung Do Chung
Semantically Integrated Business Applications for Enterprise Resource

Planning Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411
Patricia Kraft and Rainer Thome
Towards a Next Generation Universally Accessible ‘Online

Shopping-for-Apparel’ System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
Kasper Kristensen, Nanna Borum, Line G. Christensen,
Henrik W. Jepsen, Jacob Lam, Anthony L. Brooks, and
Eva Petersson Brooks
Digital Menu Boards as Influencer for Healthy Eating . . . . . . . . . . . . . . . . 428

Anicia Peters and Brian Mennecke
Reverse Business Innovations – The Impact of ERP Software Upgrades

on Organizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Stefanie Rauff and Andreas Hufgard
Repository-Based Implementation of Information Pyramid:

A Study Based on an ERP Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446
Hans-Jürgen Scheruhn, Daniel Ackermann, Roman Braun, and
Ulrich Förster
Agility Based on Stakeholder Interaction – Blending Organizational

Learning with Interactive BPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
Christian Stary, Werner Schmidt, and Albert Fleischmann
Analysis of Customer Preference through Unforced Natural Passive

Observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466
Terumasa Tajima, Yusuke Iida, and Toshikazu Kato
The Development of an Innovative Design Process for Eco-efficient

Green Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
Shiaw-Tsyr Uang and Cheng-Li Liu
Principled Ways of Finding, Analysing and Planning for Communicative

Overhead in Interaction Technology for Fashion Industry . . . . . . . . . . . . . . 484
Jason Yang, Sean Rintel, and Stephen Viller
Continuing On-Premise or Adopt On-Demand? An Empirical Study of

ERP Adoption in SMEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
Fan Zhao and Elias Kirche
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501

Part I
Identity, Privacy and Trust
SP-CIP: A Secure and Privacy Aware Patient Centric
Information Portal
Subrata Acharya
Department of Computer and Information Sciences, Towson University,

Maryland, United States
[email protected]
Abstract. There is an urgent need for the establishment of patient-centric and

quality aware solutions in today’s healthcare environment. To this effect, Clini-
cal Information Systems (CIP) play a powerful role to significantly influence
the quality and efficiency of health care delivery systems. However, most of the
current functional CIP are static [1] in design and typically represent and cater
to only a single healthcare entity. This leads to an inaccurate and stale view of
the patient status both in the care of the patient and also in the delivery of essen-
tial healthcare services. Moreover, it also delays patient care and leads to in-
conveniences in patient billing and the overall usability of the system. Added to
this there is an oversight of security and privacy policies and procedures in the
implementation of such systems. Thus, with the current pervasive nature of
technology in such healthcare organizations, this creates a critical risk of such
information systems to fall prey to data leaks and/or abuse during the access,
storage and transmission of such systems.
To address the above key barriers towards the development of the effective
CIP this research proposes SP-CIP: a standardized, security and privacy aware pa-
tient portal system. By utilizing the proposed standard design framework, the fea-
tures of patient portals can evolve from static to a more dynamic design, which is
inclusive of a user-centric, secure and proactive approach for patient care. These
features also provide flexibility to include a de-centralized mechanism to access,
store and transmit healthcare information over cloud based systems, and aims to
establish the expectations of an effective patient portal, leading to improved cost
and quality of patient care in today’s healthcare systems. The pilot evaluation of
the proposed systems at a high-end federal government Healthcare Information
and Management Systems Society (HIMSS) 6 health care organization (> 500 bed
capacity) and its associated partner organizations and clinical practices further
demonstrates the strengths of proposed system.
Keywords: Clinical Information Portal, Security, Privacy Aware, Dynamic.
1 Introduction
With the dramatic increase in healthcare costs over the past few years many initiatives
are being explored to address the associated challenges of providing acceptable quali-
ty with affordable cost of patient care. The current national administration's mandate
M. Kurosu (Ed.): Human-Computer Interaction, Part III, HCII 2013, LNCS 8006, pp. 3–9, 2013.
4 S. Acharya
to support, promote and foster electronic healthcare management systems is one such
initiative. Furthermore, according to the Health Information Technology for Econom-
ic and Clinical Health (HITECH) Act, enacted as part of the American Recovery and
Reinvestment Act of 2009, the goal is to promote the adoption and meaningful use of
health information technology across the healthcare industry [2]. For instance, an
estimated 30 billion dollars has been allocated for the use of electronic healthcare
records (EHRs) in the Health Information Exchange (HIE) across the country [3].
Additionally, there are federal mandated guidelines and incentives for individual
and group healthcare organization to adapt to electronic medium and to promote the
meaningful use of their healthcare systems [4]. In this regard, healthcare information
management and security is vital and imperative to maintain high quality of patient
care, maintain reasonable costs as well as provide data confidentiality and address and
mitigate the breach of personal data. To this effect, Clinical Information Portals
(CIP) could play a very critical role to provide efficient and affordable quality health-
care to patients across the spectrum [5].
In this paper, we will first discuss the background of CIP and their current limita-
tions with regards to today’s technology. In section 3 will discuss the assessment of
healthcare practitioners at a real-world Healthcare Information and Management Sys-
tems Society (HIMSS) 6 healthcare organization (> 500 beds) and 12 partner health-
care practices. To address the challenges we present the proposed SP-CIP Model in
Section 4. In section 5 we discuss the pilot implementation of the proposed model and
finally, conclude the paper.
2 Background and Current Limitations

A true clinical patient portal allows healthcare providers, patients and also insurance
providers’ access to electronic health information of the patient, typically through a
web interface. In its current shape, patients can access their appointment schedules,
current medications, test results, etc. Healthcare providers, can access patient health
information to provide in-time and efficient care. This also allows out of network care
providers to help their peers in collaborating and providing the necessary care to the
patient as and when needed. CIP also allow patients to update contact information,
provide their annual history information and send emails to clinicians and prepare
them for a visit in a timely manner. It is also helpful for patient families to keep track
of patient progress and monitor the health improvement of their near and dear ones.
The insurance providers also connect to the CIP to provide quick decisions and timely
billing to facilitate patient care and provide better coordination of care between
healthcare providers and patients. In its current implementations the CIP do not ad-
dress their design goals and hence it leads to all parties (healthcare providers, patients
and insurance providers) to fall back on manual mechanisms of care, which defeats
the purpose of establishment of CIP.
The current design of CIP is static and limited in their functionalities. This makes
the use of such systems obsolete and creates delay in the overall care of the patient.
We are also alarmed that there exists no consistent definition of the term portal itself
SP-CIP: A Secure and Privacy Aware Patient Centric Information Portal 5
amongst the healthcare industry as a whole [6]. This creates additional inter-
operability issues when patient care is across distributed Health Information Ex-
changes (a typical scenario).
As part of the survey of both publically available and commercial CIP we found
that there is no specific mapping of a particular type of information portal to the im-
plementation environment. Also, with the surge of various technological means (such
as desktops, laptops, PDAs, mobile phones, etc.) of access of the CIP there are severe
constraints to the usability of such systems along with the issues of cost and security
leading to vast discrepancies in the use of CIP to provide quality and cost effective
patient care. Typically, CIP could operate in three modes, a) Server, b) Cloud and or
c) Hybrid. Also, the critical security and privacy issues of the current portal design
has led to non-use of these systems (more than 95% of the time), and users reverting
back to manual operation.
Moreover, since the effective use of patient portals is over the current Internet,
and since it operates on sensitive healthcare data; there are instance of significant
cyber-attacks to such systems, and hence require substantial efforts to ensure they are
secured appropriately. Our efforts in designing SP-CIP is to address the above said
challenges and provide a flexible, secure and privacy aware model for the overall goal
of effective healthcare care management to reduce cost and improve quality of patient
care.
3 Assessment of Healthcare Practitioners and Target Patients
The goal of the assessment was to conduct an organization wide survey of healthcare
practitioners and patients to understand the current limitations of CIP. We conducted
our anonymous assessment on a selected uniformly distributed group of 50 healthcare
providers (total 300 daily operational providers for the organization) and 30 in hospit-
al care patients for the HIMMS 6 healthcare organization ( > 500 beds).
Our goal was to evaluate the CIP design from the healthcare provider perspective
based on the type of CIP implemented (sever, cloud based or hybrid), care location
(in-hospital or out-patient care), interoperability, internetworking with other health-
care organizations, security and privacy of healthcare information. We also had me-
chanisms in place to measure per patient average time for care in the current target
system. We realize that this is just the initiation of the assessment process and we will
have other metrics to evaluate as we conduct periodic assessment and depending on
the changing operational dynamics of the healthcare organization. Based on the feed-
back from the providers, we summarize our assessment as follows:
─ The current server based design did not support the use of mobile technological
devices used by the healthcare providers (physicians, physician assistants, nurses,
lab technicians, etc). Due to the usability and security risks, the providers preferred
and conducted manual (telephonic, in person) operation and disregarded the CIP.
─ We noted that the attitudes of physicians favor server due to perception that their
data is more “secure” on a server. There were also concerns of resource limitation
in server base designs.
6 S. Acharya
─ Cloud base designs had the issue of HIPAA compliance and maintenance of the
cloud, though the providers realized the advantages of real time access, flexible,
low cost and resource availability over the cloud.
─ The current design was not interoperable leading to healthcare providers reverting
back to manual means to provide patient care.
─ The associated clinical practices and other partner healthcare organizations did not
connect in real time with the CIP of the central hospital causing delays in patient
care for patient transfers, specialist consults, etc.
─ Security and privacy metrics were not part of the CIP design. Also, there was lack
of security assessment in the use and access of the CIP. This was the primary rea-
son of lack of trust and non-usability of the system by a majority of the healthcare
providers.
Next, our primary goal in assessing the patient use of the CIP was to understand the
usability of the system and evaluate if the system aids their care by improving by
addressing time and cost factors. Based on the feedback from the patients, we sum-
marize our assessment as follows:
─ Ignorance of the presence of such a system was the primary finding of the
assessment.
─ The usability and the system and the interactive nature were inefficient to cater to
the patient care requirements, leading to patients reverting to walk-ins and
telephonic interactions for all practical purposes.
We did conduct one-to-one discussions with a few insurance providers and collected
statistics which also supported the above inferences on the current design of the CIP,
which in effect rendered them obsolete.
To address all the above limitations, we have proposed the design of a secure and
privacy aware portal system, SP-CIP model which we discuss in the next section. We
successfully implemented the model in the pilot healthcare organization and it is cur-
rently in successful operation in the system.
4 SP-CIP Model
To address the above critical barriers towards the development of the effective CIP,
this paper presents a standardized, secure and privacy aware patient portal system,
SP-CIP. By utilizing the proposed design framework, the features of patient portals
can evolve from static to a more dynamic design, which fits better in a real-time
dynamic situation and leads to effective care of the patient. Additionally, it also in-
corporates a user-centric design to cater to user specific needs and requirements in the
design of such CIP. Furthermore, the framework also incorporates secure design and
privacy aware policies and procedures as part of the foundation and develops a more
proactive approach towards patient care. Along with these features, the proposed
framework also provides flexibility to include a de-centralized (distributed) mechan-
ism to access, store and transmit healthcare information over cloud based systems,
and aims to establish the expectations of an effective patient portal, leading to reduced
cost and improved quality of patient care in future healthcare systems.
The SP-CIP model is based on a hybrid design (both server and/or cloud based).
We have incorporated the technologies and secure protocols used in current secure
web and wired and wireless communication as part of the design. A typical CIP node
is represented as in Figure [1]. The salient functionalities are summarized as follows:
─ Implementation of a secure multi-tier architecture that incorporates secure proto-
cols, security measures, policies and procedures. The fundamental property of CIA
(Confidentiality, Integrity and Availability) is verified for every information flow
in and out for the system.
─ The design isolates the web, application, and healthcare information servers behind
multiple layers of secure Intrusion Detection and Prevention systems (IDS/IPS).
There is also real-time monitoring and mitigation of developing vulnerabilities in
the system. Our plan also includes periodic stress test to test the security and avail-
ability of the model.
─ There is also secure mechanism of provisioning patient accounts on the CIP.
SP-CIP also provides proper authentication, access control, and ensures the secure
storage of the passwords. Standards for ensuring resets and adhering to providing
security of Least Privilege in the design and access of the system.
─ Implementing a proactive incident recognition and response program. Proactively
monitoring for suspicious events, service interruptions, code errors, and general
utilization issues.
─ Hierarchical tiered design to ensure secure information flow between central
hospital, peer healthcare practices, patient client devices and insurance providers
systems. Adherence to current secure protocol standards and proactive mechanism
to detect and mitigate potential anomalous conditions is mandated in the current
design of SP-CIP.
─ Adequate mechanisms for accountability via logging, log analysis and mitigation.
─ Appropriate redundancy and recovery operations in place to ensure worst care
operation and real-time availability of the system.
─ Federal mandated compliance via periodic audits to ensure the appropriate opera-
tion of the proposed SP-CIP model.
Fig. 1. SP-CIP Node

8 S. Acharya
In the next section we will discuss the integration and pilot of the proposed SP-CIP
model at the target HIMMS 6 healthcare organization. We also present the overall
mechanism to incorporate the design of secure and privacy aware CIP as part of the
standardization efforts to ensure security and privacy aware information flow in the
overall healthcare organization.
Phase 1
Information Security Abstract Assessment
(Policy, Procedure, Architecture)
Healthcare Healthcare
Information Security Information Security
Questionairre Assessment
Federal Regulations Healthcare Provider Healthcare Provider

(HIPAA) Survey Analysis
Information
Information Technology
Technology Architectural
Architecture Review Assessment
Federal Guidelines/
Recommendations Onsite Visits
(Interviews,
(NIST, HHS)
Inspections)
Healthcare Information Comprehensive Organization

Security Guideline Assessment and Roadmap
IT Industry
Best Practices
(HISG) (COAR)
Penetration &
Vulernability
Testing
Organizational Application
Feedback & Requests Review &
Testing
Phase 2
Healthcare Entity Information Security Technical Review
Training Manuals
Documentation
Maintain Implement (Administrative,
Routine Security (Policy,
Periodic Documentation COAR Functional,
Awareness Procedure)
Reassessment (Policy, Technical)
Training
Procedure)
HIPAA
Revise COAR & Compliance
HISG
Phase 3
Post‐Compliance Information Security Compliance Implementation
Fig. 2. Standardized framework for SP-CIP inclusion
5 Pilot Evaluation
Our proposed SP-CIP model has been currently piloted at a state-of-art federal gov-
ernment HIMSS 6 health care organization (> 500 bed capacity) and its associated 12
partner organizations and clinical practices. Figure 2 demonstrate the overall standar-
dization framework which incorporated the SP-CIP design for the healthcare organi-
zation. The phases include the initial assessment and evaluation, the technical review
and implementation and the final compliance evaluation and periodic monitoring and
enhancements.
The overwhelmingly positive feedback from both the in-hospital patients and
healthcare providers further demonstrates the strengths of the proposed system. Some
of the measures for patient satisfaction were based on time to care, response to care,
follow-ups on care and reminders, etc. The healthcare provider measures were based
on overall improvement in patient average, ease of care and effective response to care
situations, etc. Currently, we are in the process of conducting a comprehensive evalu-
ation of the proposed model over the distributed network of the pilot healthcare
organization and its partner institutions to demonstrate the improvement in cost reduc-
tion and effective patient care services. We also plan to design a checklist for laying
out the “best practices” for effective CIP design, implementation and operation in
collaboration with regulatory bodies at the federal government and share the proposed
framework with other healthcare organization in and around the nation.
6 Conclusion
In this paper we propose SP-CIP as a vehicle to provide cost effective and quality
aware patient care. Our model addresses all the current limitations and challenges by
incorporating the federally mandated polices and the secure and privacy aware proto-
cols and practices over the current WWW usage. We have been successful in provid-
ing improvements in cost and quality in our pilot organization and associated 12
partners. As future research directions we plan to incorporate the model in various
national distributed healthcare networks and demonstrate the strength of the model.
References
1. iHealthBeat Patient Portal: Not Just another Static Text Field Form (2013),
http://www.ihealthbeat.org/perspectives/2013/patient-portal-
not-just-another-static-textfield-form.aspx (last accessed February 2013)
2. United States. Health Information Technology (HealthIT). Meaningful Use of EHR (2013),
http://www.healthit.gov/policy-researchers-
implementers/meaningful-use (last accessed February 2013)
3. Healthcare Information Security. $30 Billion: Money well spent (2013),
http://www.healthcareinfosecurity.com/blogs/
30-billion-money-well-spent-p-1185 (last accessed February 2013)
4. United States. Center of Medicare and Medicaid Services (CMS). EHR Incentive Program
(2013), http://www.cms.gov/Regulations-andGuidance/Legislation/
EHRIncentivePrograms/Meaningful_Use.html (last accessed February 2013)
5. Patient Portal. Patient Portals (2013),
http://en.wikipedia.org/wiki/Patient_portal (last accessed February
2013)
6. Acharya, S., McSweeney-Feld, M.H.: Clinical Information Portals: Classification of Termi-
nology and Design Implications. In: Academy of Business Research Fall 2012 International
Conference, Atlantic City (September 2012)
Identity Management through “Profiles”:
Prototyping an Online Information Segregation
Service
Julio Angulo and Erik Wästlund
Karlstad University,
Universitetsgatan 2, 651 88 Karlstad, Sweden
{julio.angulo,erik.wastlund}@kau.se
Abstract. Whereas in real everyday life individuals have an intuitive ap-

proach at deciding which information to disseminate to others, in the dig-
ital world it becomes difficult to keep control over the information that is
distributed to different online services. In this paper we present the design
of a user interface for a system that can help users decide which pieces of
information to distribute to which type of service providers by allowing
them to segregate their information attributes into various personalized
profiles. Iterative usability evaluations showed that users understand and
appreciate the possibility to segregate information, and revealed possible
improvements, implications and limitations of such an interface.
Keywords: Usability, identity management, privacy preferences, partial

identities, audience segregation, digital transactions.
1 Introduction
In everyday life individuals are frequently and naturally playing different social
roles, as family members, professionals, hobbyist, activist, etc. Typically, people
do not reveal all of their personal information to all of their different social
groups, but rather they inadvertently and intuitively select the information that
is suitable to share with members of a certain group depending on the context
of a situation. Such phenomenon was first referred in the 1950’s as audience
segregation by sociologist Erving Goffman [15].
However, it can be claimed that in their ‘digital lives’ Internet users do not yet
have the appropriate tools that help them manage their partial digital identities
which let them segregate the information they distribute to different Internet ser-
vices. For this reason, users tend to present similar identifiable attributes to many
different service providers and to send more attributes than are actually needed
to complete a transaction. Besides, nowadays users create different accounts with
different services, which makes is hard for them to maintain and control which in-
formation is shared with whom. All these traces that users leave on the Internet
could lead to higher probabilities of user impersonation, identity theft, profiling,
and linkability attacks. A method for empowering users with control over their
own personal information is needed as a way to minimize these risks.

c Springer-Verlag Berlin Heidelberg 2013
Identity Management through “Profiles” 11
Therefore, in our work we try to tackle the challenge of designing an interface

for what we refer to as information segregation, or the act of encapsulating
different pieces of personal identifiable information in order to present only those
attributes to deliberately chosen online service providers.
In this paper we first look at some work related to audience segregation and the
usability of Identity Management (IdM) systems. We then present an overview of
the process of design for a system supporting the idea of information segregation
along with the performed usability evaluations. Finally we list some implications
and conclusions about our work.
2 Related Work
Attention has been given to the challenge of letting users select the audiences
to with whom they wish to share the content in social network sites (SNS),
seen for instance in [6] [7] [12] [16] [18] [20] [29]. Google+ is a SNS that makes
its audience segregation features explicit through the use of so called circles. It
has been argued that Google+’s users have a “clear understanding of circles,
using them to target information to those most interested in it” [31], and it can
be said that the circles’ interface offers further desired properties of interactive
systems, such as consistency, playability, pliability, learnability, affordance, and
others, as suggested, for instance, by Löwgren & Stolterman [21] and other design
heuristics.
However, there is a distinction between the act of segregating audiences in
SNSs, which creates a tension between the “desire for controlling our own in-
formation and the desire for unplanned social interaction” [17], and the act
of distributing personal information for receiving commercial online services or
products, which is motivated by a need or desire of the service or product being
requested or the experience it provides.
For this reason, some attempts have been done at allowing users to act under
different identities while communicating online. These include the efforts from
Mozilla’s Persona1 and Google’s Multiple Chrome Users2 , as well as the obsolete
Microsoft’s CardSpace and Firefox’s plugin “Sxipper”. Nevertheless, these ex-
isting systems are either at a proof-of-concept stage, have limited functionality,
acting mainly as role-based access controlled or password management systems,
or are hard to understand and therefore hard to adopt by regular users.
Regarding the usability of IdM systems, the research done by Jøsang et al.
[19] discusses different models for IdM and suggest a user-centric approach for
the management of user identities. Similarly, Eap et al. [13] recognize the need
to provide users with more control over their identities distributed over different
service providers, stating that IdM systems should assist users in their adoption
of identity management practices. Moreover, a prototype called DRIM (Dres-
den Identity Management) [10] tried to integrate identity management concepts
on Internet browsers, and subsequent work carried out as part of the PRIME
1
http://www.mozilla.org/en-US/persona/
2
http://www.chromium.org/user-experience/multi-profiles
12 J. Angulo and E. Wästlund
project3 has also realized the need for interfaces that support the notion of
partial digital identities [23]. Furthermore, Dhamija & Dusseault [26] list seven
concrete observations taken from their experience dealing with the design and
analysis of security systems, trying to inform the reader why such systems often
fail and ways to improve them. Alpár et al. [1] present some of the security,
privacy and usability issues encountered in current IdM systems and propose
recommendations for their improvement in each of these areas.
3 Conceiving an Interface for Information Segregation

3.1 Assumptions and Requirements
While conceiving an interface of a system that supported the idea of informa-
tion segregation, we made assumptions about the architecture and the role that
different parties would play in this envisioned system. For one, we assumed that
an architecture is in place similar to the one suggested in the European projects
PRIME and PrimeLife [8]. This implies that the system could be equipped with
anonymous credentials technology (such as the one suggested by IBM’s IdeMix
[9]), and with a privacy policy matching engine (similar to the Platform for Pri-
vacy Preferences (P3P) [27] or the PrimeLife Policy Language (PPL) [5]) which
can empower users to provide informed consent at the moment of releasing per-
sonal information.
However, in contrast to the architecture proposed in PrimeLife, it is assumed
that the users’ personal information is not stored locally on their devices, but
instead it is stored centrally in a secure and privacy-friendly manner, for example
in a cloud service provider.
It is also assumed that the proposed system can come in contact with different
identity providers that are able to provide users with certified proofs of their
identity. Similarly, online retailers would be able to offer their customers the
possibility to complete a digital transaction using the certified or non-certified
attributes contained within the customers’ partial identities.
3.2 Design Approach

By following an iterative process of design, we created a series of sketches and
lo-fi prototypes during three initial iteration cycles. The aim was to evaluate the
use of different segregating metaphors, terminologies used, users’ understanding
of the concept of information segregation, as well as the intuitiveness of the
interactive elements, look-and-feel of different design layouts and other visual
aspects of the interface. Early evaluations actually revealed that participants
consistently used the word profiles to refer to partial identities; therefore, we
decided to give the name of “Profiles” to the IdM service being prototyped.
3
PRIME - Privacy and Identity Management for Europe. www.prime-project.eu
Fig. 1. A view of the “Profiles” prototype allowing users to manage their pieces of
information at the top and their profiles at the bottom
Final Design Proposal. The final design pro-

posal of the interface (depicted in Figure 1) can
be briefly described as being divided into two
panels. The top panel has a prominent label that
reads ‘Your information box ” conveying to users
that it is a place where they can add, remove and
Fig. 2. The look of a piece of
manipulate their different pieces of information. information
Figure 2 represents an example of the look-and-
feel of a piece of information, where users can see the attribute value of the
information (the actual credit card number), the type of information (credit
card), whether it is certified or not, and by which third party it has been cer-
tified. When users want to add a piece of information a dialog pops-up, where
users can choose to either add non-certified attributes by typing them manually
(Figure 3(a)) or to download certified attributes from a trusted identity provider
(Figure 3(b)). As mentioned earlier, it is assumed that a list of trusted identity
providers is already populated.
The bottom panel of the interface enables users to create and manage their
different partial identities, or profiles. Results from the evaluations of earlier
iterations indicated that users had a hard time giving relevant names to their
profiles, therefore the interface provides a set of predefined profiles as a way to
help first time users to get started and understand the purpose of the system.
Profiles can be created and populated by clicking on the “create a new profile”
icon, or by dragging one or many pieces of information onto them.
(a) Adding non-certified attributes manually (b) Downloading certified attributes from a
trusted identity provider
Fig. 3. Dialogs allowing users to add pieces of information
Every time pieces of information are added, users are given the option of
selecting their preferences for the purposes for which service providers can use
that information. For instance, users can specify that their credit card infor-
mation should only be used for payment purposes. In this way, users can be
informed if service providers respect their wishes during a digital transaction as
explained in [3] [11]. Previous studies have shown the difficulty of users when
stating their privacy preferences [11] [14] and to appropriately set privacy set-
tings that would match their expectations to the reality of the protection of their
privacy [22]. Given these observations, we realized through the different design
iterations that the burden of setting such preference should be removed from
users whenever possible, providing them with privacy-friendly default values as
recommended, for instance, by data protection authorities. More importantly,
the interface should promote the act of setting privacy preferences only in a mo-
ment that is relevant to the action at hand (as also suggested in [2]), for instance
when information is added to a profile or when it is going to be distributed to
service providers.
Fig. 4. Dialog allowing users to set the preferences of usage for the information pieces
they add to a profile
3.3 Usability Evaluations
The initial design iterations were evaluated using different methods, such as
eye-tracking technology, cognitive walkthroughs, thinking aloud protocols and
questionnaires. Also, expert evaluations and feedback from professionals in the
fields of e-commerce and identity management were taken into account. Results
from these evaluations allowed us to identify possible improvements to the in-
terface, and indicated that participants understood and appreciated the purpose
behind information segregation.
For the final design iteration two main evaluation activities were carried out:
usability tests using an interactive prototype and a focus group session. During
the usability tests, all participants (n = 12) first answered a pre-questionnaire
about their Internet habits and familiarity with existing audience segregation
features in popular SNSs. Then, in the form of a cognitive walkthrough [24] they
were asked to perform a series of tasks using an interactive prototype represent-
ing a scenario in which they were supposed to setup the “Profiles” system in order
to make a purchase with an online service provider. The participants opinions
and interactions with the prototype were noted as they went through each task,
and they were also asked whether they considered a task to be easy or difficult
to accomplish. At the end, participants answered a post-questionnaire including
questions about the usability of the program, their understanding of its informa-
tion segregation features, and PET-USES Likert-scale statements which aim to
measure the secondary goals of users when interacting with privacy enhancing
technologies, as presented in [30].
During the focus group session participants (n = 30) were first shown a
demonstration of the “Profiles” interface and its privacy features. In order to
encourage discussions and minimize ‘group think’ [28] participants were divided
into smaller groups of 5 to 8 people. They were asked to discuss possible uses
and improvements of the proposed interface with the other group members.
Summary of Results. From 12 participants that completed the cognitive walk-

trough phase, 8 had a very good intuitive idea of the purpose of the prototype,
stating for example that “the top lets you write pieces of your information, the
bottom lets you create a profile with some of the pieces you wrote already in
the top.” Adding non-certified pieces of information was understood easier than
downloading certified information, probably because users’ unfamiliarity with
the concept of certificates. However, after having downloaded certified informa-
tion the first time, 11 out of 12 participants understood it well the second time
they were asked to add certified information into the system. Observations also
revealed that users were able to set privacy preferences easily, however in the
post-questionnaire they reported this task as being the hardest to accomplish,
suggesting a paradoxical view of privacy settings.
Participants from the focus groups session expressed, among other things,
their concerns about centralizing all pieces of information in a single program:
“if a hacker gets into the program, then they will have all that information from
me given by the Police, the Bank, etc.”. Moreover, they discussed the usage of
such a system that would fit their current situation and address their needs at
hand. For instance, they envisioned a scenario where having a series of profiles
could facilitate the way a user can look for different job positions that might
require different information.
The complete list of tasks that participants were asked to go through, the
instructions given during the focus group session and a detailed account of the
results of the evaluations can be found in the technical report presented in [4].
3.4 Implications
Based on the process of design and the results from the evaluations, the following
points present some of the major considerations to be taken when designing an
IdM system like the one suggested hereby.
Progressive and Contextual Formation of Partial Identities. Since the

early stages of design it became obvious that one of the major challenges would
be to conceive an interface that would encourage users to start interacting with
such a technology and adopt it for their routinary digital transactions. Thus, it
is important to keep in mind that the users’ first interactions with the system
should be effortless, intuitive and perceived as useful at a moment that is relevant
to what they are trying to achieve (e.g., purchasing a product through an online
service). As stated in [26], “identity management is not a goal in itself”, thus
forcing users to, for instance, populate the system with many personal attributes
so that it can work effectively during a purchase, will most likely discourage users
to continue using it.
Instead, a progressive approach should be adopted in which users start by
forming a partial identity in a simple way, gradually adding pieces of information
as they are needed and creating identities at their own pace. The system should
resemble to some extent the way identities are formed in real life and the contexts
in which real identities evolve [1], consisting of progressive steps and determined
by relevant life events, such as getting married, changing jobs, or opening a new
bank account.
Maintaining Different Profiles. Creating different profiles could have the

disadvantage that those profiles have to be maintained over time, which can
become a burdensome task if users need to continuously monitor and update
the information and the online services that they have associated with each
profile.
A centralized solution would allow users to maintain their profiles across dif-
ferent devices over the air, which could provide convenience and further promote
the gradual development of these partial identities. In this way, users would be
able to access their profiles whenever and wherever they wish to do so, contin-
uously evolving their identities ‘on the fly’ under relevant contexts of use in a
progressive manner (a similar approach has been suggested during the PRIME
and PrimeLife projects as presented in [3]).
Setting Privacy Preferences. Designing interfaces for letting users set their
privacy preferences proved to be a difficult challenge. It was not until the final de-
sign suggestion that we thought we had come up with a simple enough mechanism
to achieve this (seen in Figure 4). However, results from the post-questionnaire in-
dicated that users still perceived this as a difficult step to complete.
What we have learned is that a system handling privacy preferences should re-
lieve users from having to set those preferences from scratch, and instead a good
set of privacy-friendly preferences, as defined by a trusted authority, should al-
ready be selected by default. The option to set or modify those preferences should
not be made a priority, but should be made available within a context that a user
will understand; for instance, at the moment of having to disclose information
to a service provider, where she could, if she is interested or concerned, specify
the purposes of usage of her data and other conditions.
4 Concluding Remarks
We have presented a design proposal for a system that enables users to group
their information pieces into self-defined partial identities, or profiles. We refer
to this act as information segregation. The suggested approach allows users to
define the preferences for the data attributes contained within each profile, and to
use those profiles at the moment of contacting specific applications or groups of
online services with certain similarities, thus helping users protect their privacy
by having a clearer approach to control their data and minimizing the personal
information they disclose under certain application contexts.
We are aware of some of the limitations of the suggested interface. For one,
difficulty of visualizing large amounts of pieces of information and managing
more profiles than can be displayed on different screen screen sizes has not been
fully considered. The level of complication can escalate when the formation of
sub-profiles is taken into account (e.g., having a health profile that can be sub-
divided into pharmaceutical services, health clinic consultations, health clinic
administration, etc.). Moreover, the users’ understanding on how their informa-
tion flows, where is it located (remotely or locally) and how it is handled by the
IdM system is still unexplored (e.g., do users understand, or care about, what
happens when they delete a piece of information?). Similarly, the possible steps
to be taken to populate the system with trusted identity providers has not been
considered. Additionally, we are aware of the privacy and security consequences
that can arise from having all users’ data in a centralized remote location; how-
ever, following a privacy-friendly architecture as presented in [25] can ensure
users’ privacy even towards the cloud service which stores the users’ data.
Despite its limitations, we see this study as an exploratory approach towards
useful privacy-friendly IdM systems. As part of future work we are working
on creating scenarios where specific profiles might be employed (e.g., e-health
or e-banking profiles) during a digital transaction. Also, we plan to work on
adapting the interface to touch-screen devices of different sizes, and study the
users’ mental models of the location of their data as well as of the data flows
between devices and entities involved in a digital transaction. Moreover, in future

iterations the mechanisms for specifying privacy preferences will be extended
to not only include the purposes of data use, but also data retention periods,
deletion obligations and other similar conditions.
Acknowledgments. This work is funded by a Google Research Award and the

Swedish Knowledge Foundation (KK-stiftelsen) through the U-PrIM project.
We would like to thank our colleagues Martin Ortlieb, Stephan Micklitz, Peter
Gullberg, Simone Fischer-Hübner and Tobias Pulls.
References
1. Alpár, G., Hoepman, J.H., Siljee, J.: The identity crisis. Security, privacy and usabil-
ity issues in identity management. Computer Research Repository (CoRR) (2011)
2. Angulo, J., Fischer-Hübner, S., Pulls, T., König, U.: HCI for Policy Display and
Administration. In: PrimeLife - Privacy and Identity Management for Life in Eu-
rope, ch. 14, p. 261. Springer (June 2011)
3. Angulo, J., Fischer-Hübner, S., Pulls, T., Wästlund, E.: Towards usable pri-
vacy policy display & management. Information Management & Computer Se-
curity 20(1), 4–17 (2012)
4. Angulo, J., Wästlund, E.: Identity Management for online transactions - Using
“Profiles” to segregate personal information. Tech. rep., Karlstad University, Karl-
stad, Sweden (April 2012),
http://www.is.kau.se/julioangulo/angulo/publications/2012/
2012 IdentityManagementForOnlineTransactions.pdf
5. Ardagna, C.A., Bussard, L., Di, S.D.C., Neven, G., Paraboschi, S., Pedrini, E.,
Preiss, S., Raggett, D., Samarati, P., Trabelsi, S., Verdicchio, M.: Primelife policy
language. In: Proceedings of the W3C Workshop on Access Control Application
Scenarios, Luxembourg (November 2009)
6. van den Berg, B., Pötzsch, S., Leenes, R., Borcea-Pfitzmann, K., Beato, F.: Privacy
in social software. In: Camenisch, J., Fischer-Hübner, S., Rannenberg, K. (eds.)
Privacy and Identity Management for Life, pp. 33–60. Springer, Heidelberg (2011)
7. van den Berg, B., Leenes, R.E.: Audience Segregation in Social Network Sites.
In: Proceedings for the Second IEEE International Conference on Social Comput-
ing/Second IEEE International Conference on Privacy, Security, Risk and Trust,
pp. 1111–1117. SocialCom/PASSAT, SSRN, Minneapolis, USA (2010)
8. Camenisch, J., Fischer-Hübner, S., Rannenberg, K.: PrimeLife - Privacy and Iden-
tity Management for Life in Europe, 1st edn., vol. 14. Springer (June 2011)
9. Camenisch, J., van Herreweghen, E.: Design and implementation of the idemix
anonymous credential system. In: Proceedings of the 9th ACM Conference on Com-
puter and Communications Security, pp. 21–30. ACM (2002)
10. Clauß, S., Kriegelstein, T.: Datenschutzfreundliches identitätsmanagement. Daten-
schutz und Datensicherheit 27(5) (2003),
http://dblp.uni-trier.de/db/journals/dud/dud27.html#ClaussK03
11. Cranor, L.F., Guduru, P., Arjula, M.: User interfaces for privacy agents. ACM
Trans. Comput.-Hum. Interact. 13(2), 135–178 (2006)
12. DiMicco, J.M., Millen, D.R.: Identity management: multiple presentations of self in
facebook. In: Proceedings of the 2007 International ACM Conference on Supporting
Group Work, GROUP 2007, pp. 383–386. ACM, Sanibel Island (2007)
13. Eap, T.M., Hatala, M., Gasevic, D.: Enabling user control with personal iden-
tity management. In: IEEE International Conference on Services Computing, SCC
2007, pp. 60–67. IEEE, Salt Lake City (2007)
14. Fischer-Hübner, S., Pettersson, J., Bergmann, M., Hansen, M., Pearson, S., Mont,
M.: Human-Computer Interaction. In: Camenisch, J., Leenes, R., Sommer, D. (eds.)
Digital Privacy. LNCS, vol. 6545, pp. 569–595. Springer, Heidelberg (2011)
15. Goffman, E.: The presentation of self in everyday life. Doubleday (1959)
16. Gonçalves, J.: Groupster: Narrowcasting on Social Networking Sites. Master’s the-
sis, Madeira Interactive Technologies Institute, University of Madeira (2011)
17. Grimmelmann, J.: Saving Facebook. Iowa Law Review 94(4), 1137–1206 (2009)
18. Gross, R., Acquisti, A.: Information revelation and privacy in online social net-
works. In: Proceedings of the 2005 ACM Workshop on Privacy in the Electronic
Society, WPES 2005, pp. 71–80. ACM, New York (2005)
19. Jøsang, A., Zomai, M.A., Suriadi, S.: Usability and privacy in identity management
architectures. In: ACSW Frontiers, pp. 143–152 (2007)
20. Kairam, S., Brzozowski, M., Huffaker, D., Chi, E.: Talking in circles: selective
sharing in google+. In: Proceedings of the SIGCHI Conference on Human Factors
in Computing Systems, CHI 2012, pp. 1065–1074. ACM, New York (2012)
21. Löwgren, J., Stolterman, E.: Thoughtful Interaction Design: A Design Perspective
on Information Technology. MIT Press (2007)
22. Madejski, M., Johnson, M., Bellovin, S.M.: The failure of online social network
privacy settings. Tech. rep., Columbia University (2011),
http://www.futureofprivacy.org/wp-content/uploads/2011/07/
TheFailureofOnlineSocialNetworkPrivacySettings.pdf
23. Pettersson, J.S., Fischer-Hübner, S., Danielsson, N., Nilsson, J., Bergmann, M.,
Clauss, S., Kriegelstein, T., Krasemann, H.: Making PRIME usable. In: Proceed-
ings of the 2005 Symposium on Usable Privacy and Security, SOUPS 2005, pp.
53–64. ACM, New York (2005)
24. Polson, P.G., Lewis, C., Rieman, J., Wharton, C.: Cognitive walkthroughs: a
method for theory-based evaluation of user interfaces. International Journal of
Man-Machine Studies 36(5), 741–773 (1992)
25. Pulls, T.: Privacy-friendly cloud storage for the data track. In: Jøsang, A., Carlsson,
B. (eds.) NordSec 2012. LNCS, vol. 7617, pp. 231–246. Springer, Heidelberg (2012)
26. Dhamija, R., Dusseault, L.: The Seven Flaws of Identity Management: Usablity
and Security Challenges. IEEE Security & Privacy Magazine 6(2), 24–29 (2008)
27. Reagle, J., Cranor, L.F.: The platform for privacy preferences. Commun.
ACM 42(2), 48–55 (1999)
28. Rubin, J., Chisnell, D.: Handbook of usability testing : how to plan, design, and
conduct effective tests. Wiley Publ., Indianapolis, Ind. (2008)
29. Tootoonchian, A., Saroiu, S., Ganjali, Y., Wolman, A.: Lockr: better privacy for
social networks. In: Liebeherr, J., Ventre, G., Biersack, E.W., Keshav, S. (eds.)
CoNEXT, pp. 169–180. ACM (2009)
30. Wästlund, E., Wolkerstorfer, P., Köffel, C.: PET-USES: Privacy-enhancing tech-
nology – users’ self-estimation scale. In: Bezzi, M., Duquenoy, P., Fischer-Hübner,
S., Hansen, M., Zhang, G. (eds.) IFIP AICT 320. IFIP AICT, vol. 320, pp. 266–274.
Springer, Heidelberg (2010)
31. Watson, J., Besmer, A., Lipford, H.R.: +your circles: sharing behavior on google+.
In: Proceedings of the Symposium on Usable Privacy and Security, SOUPS 2012,
pp. 12:1–12:9. ACM, New York (2012)
An Empirical Investigation of Similarity-Driven Trust
Dynamics in a Social Network
Yugo Hayashi1, Victor Kryssanov2, and Hitoshi Ogawa2

1
Faculty of Library, Information and Media Science, Tsukuba University
1-2, Kasuga, Tsukuba, Ibaraki 305-8550, Japan
[email protected]
2
Faculty of Information Science and Engineering, Ritsumeikan University
1-1-1 Nojihigashi, Kusatsu 525-8577, Japan
{kvvictor,ogawa}@is.ritsumei.ac.jp
Abstract. Presently, people often create and keep lists of other people with
similar preferences for hobbies, such as books, movies, music, and food in
online social network service systems. Recent studies in recommender systems
have shown that the user’s data can be used to recommend items based on other
users’ preferences (e.g. as implemented in amazon.com). To make such
systems more effective, there is a need to understand the mechanism of human
trust formation. The goal of this study is to develop cognitive models describing
the trust formation in social networks. This paper presents results of a
controlled experiment conducted to collect human behavior data through a
series of trust evaluation tasks.
Keywords: Social Cognition, Social networks, Trust Dynamics, Recommender

Systems, Human-Computer Interaction, Conversational Agents.
1 Introduction
Studies in recommender systems have suggested that a user’s data can be used to
recommend this user items, based on other users’ preferences. Although, a relevant
question comes up: Can one predict how people may rely on and/or trust opinions of
others in their own decision-making? To develop efficient and effective recommender
systems, there is a need to understand the mechanism of human trust formation. The
paper proposes a new experimental design where computer agents are used to capture
the dynamics of trust formation through interactions in a social network.
1.1 Similarity and Trust in a Social Network

Studies in social psychology showed that trust is often related with similarity [1].
Studies in human and computer interaction, on the other hand, showed that users tend
to prefer recommendations from friends rather than from computer systems [2]. An
experiment conducted by [3] showed that the profile similarity may be related to the
An Empirical Investigation of Similarity-Driven Trust Dynamics in a Social Network 21
ways the users determine whether to trust other users when solving an on-line
selection task. The study provided implications for how to predict trust, and how the
corresponding model would be incorporated into related algorithms, such as
collaborative filtering algorithms used in recommender systems. The study focused on
the relationship between similarity and trust but the proposed model does not address
the dynamics of user similarity in an online-social network.
Given this background, two relevant questions arise: 1) how could one develop a
behavioral model of the similarity and trust development for the application in
recommender systems, using information extended from social networks, and 2) can
one simply use similarity as an index of trust? The presented study investigates these
questions in an open-ended environment, where individuals interact through time.
1.2 Recommendations by Other Users in a Social Network

In social network studies, there were attempts to develop recommender systems and
algorithms based on the users’ similarity and trust (e.g. [4], [5], and [6]). A popular
method to provide the user with adequate recommendations is the so-called
collaborative filtering. Collaborative filtering allows for recommending items to
users, based on their similarity and/or on the similarity of the items the users prefer to
other items. A more advanced method for recommending items to the users is the
Bayesian filtering algorithm. Using the latter algorithm, it is often possible to predict
the dynamics of human trust, based on the past results of trust/distrust and
similar/dissimilar evaluations. There were, however, few studies that would try to
validate the Bayesian framework with real human data. As many studies in
psychophysics showed, psychological states related to trust may differ due to
differences in physical world. To address this problem, the presented study focuses on
an experimental situation where an individual perceives a single dissimilar physical
state while interacting with other users.
1.3 This Studies Objective
This study thus aims to explore how human participants react towards the recovery of
trust throughout the interaction process. It investigates differences between a
simulation model currently used in many recommender systems and the
corresponding human performance observed in an experiment. The commonly used
Naïve Bayes model is validated with empirical data as a correct model of the human
trust dynamics. Data generated with the Bayes model is used as the criterion for
comparison with empirical human data for understanding parameters of the trust
development in a social network. The following three issues are examined in the
study:
1. How similar/dissimilar preferences affect trust dynamics in an (online) social
network?
22 Y. Hayashi, V. Kryssanov, and H. Ogawa
2. Assuming that after some time, similar preferences may contribute to the
development of a high trust and a relationship, then when one perceives a
dissimilar state, what is the immediate dynamics of trust?
3. How would the empirical data differ with the corresponding normative behavioral
data obtained through simulation with the classical Naïve Bayes model?
2 Method
A web-based system has been developed for conducting a controlled experiment to

understand cognitive aspects of trust in a social network. In the proposed
experimental paradigm, a multi-conversational agent was used as an imaginary actor
playing the role of a member of the social network. This experimental method helped
us understand specific stimulatory responses, as observed in trust dynamics, when an
unexpected incidence occurs in the social network.
2.1 Procedures
In the experiment, participants (who are all members of one social network) evaluated
opinions (preferences) of other participants for a topic provided. More specifically,
they watched a video-clip of a short cartoon, using smart phones, and shared their
impressions about the cartoon with other participants through a web-page. The
procedure was as follows: 1) Participants first watch the video, using smart phones, 2)
They evaluate how interesting they felt about it, 3) They see the other users’
evaluation results and, finally, 4) They evaluate how strong they ‘trust’ the other users
(see Fig. 1).
Fig. 1. The experimental procedure

The procedure was continued for about 10 trials (the exact number of trials depended
on the experiment). During the task, the participants watched 10 different video clips in
a predefined order, evaluated their impressions, and shared the evaluation results with
each user in every trial. Results from users were manipulated by computer agents (i.e.
by the hypothesized users), and were controlled to change over time. This methodology
is adopted from a previous study [7]. One of the hypothesized users’ evaluations was
adjusted to be always almost the same as the participant’s, excepting for one trial where
the participant experienced a “shock” event due to completely different evaluations
learned from the other users. Also, participants were asked to do a secondary evaluation
by evaluating the other users, based on how strong they “trusted” the other users’
opinions. Details of the experiment are discussed in the following sections.
Phase 1. First, participants accessed the Web-based application using their mobile
devices. In beginning of the task, all participants received the following instruction:
“You are randomly connected to four persons in this experiment, who are also now
watching the cartoon and giving their evaluations.” Actually, there were, however, no
other real users but computer conversational agents responding appropriately to each
participant. This instruction made participants to think that they interact with
members of the social network in real time. The participant starts the task by rating on
a collection of short films by accessing You Tube (http://www.youtube.com).
Episodes of a short cartoon “Tom and Gerry” were used as the stimuli. After the
participant finished watching one clip he or she evaluated it, based on three questions
(see Fig. 2). Each evaluation was on a ten-point lickert-scale, and the three questions
were: “how good was the content”, “how good was the character”, and “how good
was the scenario” (as translated from Japanese).
Fig. 2. Screenshots from the ratings phase
Phase 2. After the participant finished rating, he or she can see how other users rated
the episode. All other users’ ratings were generated by computer agents. These latter
ratings will be called hypothesized users’ “rating profiles”. Fig. 3 shows an example
of the rating profiles.
Fig. 3. Screenshots from the rating phase
Two types of the results were shown: (1) the sum of ratings by the participant and
(2) the sum of the rating profiles by the hypothesized users. Let us denote the
participant’s sum of the ratings of the episode in trial i as Si, the hypothesized user’s
sum of rating profiles as Ua,Ub, Uc, Ud, and the absolute difference between any two
ratings as ß. For each episode i, ß was randomly selected from the following three
categories by the computer agents:
─ Small variation : 0 ≤ ß ≤ 2,
─ Medium variation: 3 ≤ ß ≤ 5,
─ Large variation: 6 ≤ ß ≤ 9.
To investigate how similar/dissimilar preferences affect the trust dynamics, one of the
four hypothesized members labeled “User A” was always adjusted to generate small
variations. Two other hypothesized members (labeled “User B” and “User C”) always
generated large variations, and one such member generated medium variations (“User
D”). If people trust more to the users with similar ratings, it would be natural to
expect that a subject would develop higher trust towards “User A” than to the other
users. It also appears natural to expect that this tendency would increase over time.
Table 1. Rating profiles of users and trials
Hypothetical user Trial i ß

Ua 1, 2, 4, 6, 7, 8, 9, and 10 Small variations
3, 5 Large variations
Ub All trials Large variations
Uc All trials Medium variations
Ud All trials Large variations
User A was manipulated almost in the same manner as the participant behaved,
excepting for trial 3 and 5. In that trial, User A response, Ua, was adjusted to change
from Small variations to Large variations. Through this adjustment, the participant
experienced a shock, when the result of the evaluation changed dramatically
compared to other trials. Ub and Ud were adjusted to change always in Large
variations, and Uc was adjusted to change always in Medium variations. The data of
the secondary evaluation of trust was collected and used as the main data for the
analysis.
Phase 3. In this phase, the participants were asked to rate how much they trusted the
hypothetical users, based on what they observed in the rating profiles. Participants
evaluated four hypothetical users on a ten-point lickert scale. After finishing this
activity, they proceeded to the next trial, and watched a new video clip. This cycle
continued until the participants finished 10 trials.
2.2 Participants and Condition

Two experiments were conducted to investigate the cognitive aspects of trust in a
preference evaluation task. In both experiments, all the participants were students
enrolled in a psychology class. The participants were asked to participate in the
experiments for course credits. The experiments took place in the classroom and the
participants were told to use their smartphones in the experiments. In the experiment,
twenty-seven participants (male = 7, female= 20) made evaluations in ten trials.
3 Results
3.1 Overall Result
To investigate the trust dynamics for a longer period of time, Experiment 2 was
conducted. Results obtained in this experiment were then analyzed, using a 4 (users) x
10 (trials) within-subjects factorial design. Fig. 4 shows the results of the trust
evaluations for the hypothetical users. The vertical axis gives the average of the trust
evaluations, and the horizontal axis represents the trials.
Fig. 4. Results of evaluations

The ANOVA analysis revealed that there was an interaction between the two
factors (F(26,702)=4.768, p < .01). Simple main effects in the users detected
several differences among trials. Differences were found in all trials
(F(3,936)=19.753, p < .01; F(3,936)=27.758, p < .01; F(3,936)=6.624, p < .01;
F(3,936)=29.880, p < .01; F(3,936)=4.430, p < .01; F(3,936)=25.505, p < .01;
F(3,936)=24.200, p < .01; F(3,936)=35.359, p < .01; F(3,936)=40.421, p < .01;
F(3,936)=33.805, p < .01). Multiple comparisons, using Ryan’s method showed
that User A was evaluated higher than users B, C and D in trial 1 (p < .01; p < .01;
p < .01). These results were consistent for trials 1, 2, 4, 6, 7, 8, 9, and 10. The
result of the two experiments has, therefore, demonstrated that the ratings of trust
become higher when the participants observe a similarity to their own opinions in
the shared opinions about the cartoon. This finding generally confirms that similar
preferences enhance trust. The results obtained also showed that the ratings of trust
dropped as the participants experienced the shock of non-similar evaluations. This
confirms that similarity of preferences and trust are strongly correlated, and this
correlation can be enhanced over time.
3.2 Simulation
A further analysis was conducted, using the Bayesian model for a comparison with
the data. Evaluations towards User A were used for investigating how trust recovers
after experiencing the shock. In the analysis, evaluations of User A were binary
coded to either, “trust” (1) or “no trust” (0). After the second trial, when
participants rated lower than in the previous trial, the evaluating were coded as “not
trust”. When participants rated higher or the same, compared to the previous trial,
the evaluations were coded as “trust”. The ratio of selecting “trust”/”not trust” was
calculated for each trial. Using this coding scheme, all the data was then compared
with the Bayesian model. For the analysis, prior probability of perceiving a similar
/dissimilar evaluation in trial i was determined as Hi, and the probability of
generating trust was P(Hi|D). The following equation specifies the Naïve Bayesian
model used:
|
| (1)
| |
i 1, 2. (2)
Fig. 5 shows the empirical data in comparison with simulation results. The vertical
axis gives the probability of trust, and the horizontal axis represents the trials. The
results demonstrate that the Naïve Bayes model fails to accurately predict the speed of
recovering from distrust to trust but still qualitatively is in a good agreement with the
observed empirical dynamics.
Fig. 5. Results of the simulation
4 Discussion and Conclusions
The trust dynamics observed in the experiments revealed that when one perceives a
dissimilar state, the subjective perception of trust decreases temporary, but then
almost immediately recovers upon a positive experience. This fact is interesting in the
view of understanding the users of various social network service systems, who tend
to ignore social conflicts and keep developing their trust in respect to a particular
member, while ignoring negative experience of interactions with that member. This
phenomenon would be attributed to so-called conformation bias. Several studies have
recently shown that people are likely to become biased to (mis)trust others’ opinions
in an online environment (e.g. [8]). It follows from the results obtained in our study
that such bias on trust may rapidly develop in social network service systems. This
finding would have implications for design of recommender systems based on
information extracted from social networks. In future work, we plan to investigate in
detail the relationships connecting trust and similarity to find ways for controlling the
conformation bias.
References
1. Byrne, D.: The attraction paradigm. Academic Press, New York (1971)
2. Sinha, R., Swearingen, K.: Comparing recommendations made by online systems and
friends. In: Proceedings of the DELOS-NSF Workshop on Personalization and
Recommender Systems in Digital Libraries. Dublin, Ireland (2001)
3. Golbeck, J.: Trust and Nuanced Profile Similarity in Online Social Networks. ACM
Transactions on the Web 3(4-12), 1–32 (2009)
4. Walter, E.F., Battiston, S., Schweitzer, F.: A model of a trust-based recommendation system
on a social network. Journal of Autonomous Agents and Multi-Agent Systems 16(1), 57–74
(2007)
5. Ziegler, C.N., Lausen, G.: Spreading activation models for trust propagation. In:
Proceedings of the IEEE International Conference on e-Technology, e-Commerce, and
e-Service, pp. 83–97. IEEE Computer Society Press, Los Alamitos (2004)
6. Ziegler, C.N., Golbeck, J.: Investigating Correlations of Trust and Interest Similarity.
Decision Support Systems 42(2), 460–475 (2006)
7. Hayashi, Y.: The effect of "Maverick": A study of Group Dynamics on Breakthrough in
Collaborative Problem solving. In: Proceedings of the 34th Annual Conference of the
Cognitive Science Society, pp. 444–449. Lawrence Erlbaum Associates, Hillsdale (2012)
8. Joinson, A.N.: Understanding the psychology of internet behavior: virtual worlds, real lives.
Palgrave, Basingstoke (2003)
The di.me User Interface: Concepts for Sharing Personal
Information via Multiple Identities in a Decentralized
Social Network
Fabian Hermann1, Andreas Schuller1, Simon Thiel1,

Christian Knecht1, and Simon Scerri2
1
Fraunhofer-Institute for Industrial Engineering, Stuttgart, Germany
{firstname.lastname}@iao.fraunhofer.de
2
Digital Enterprise Research Institute, National University of Ireland, Galway, Ireland
[email protected]
Abstract. The di.me userware visualizes vast personal information from vari-
ous sources and allows for sharing them in a decentralized social network. Mul-
tiple identities can be used to avoid unintended linkability when communicating
to other users or external systems. The di.me user interface for that is presented
in this paper. A user-centered information- and view-structure for items of
the personal information sphere and concepts for multiple identities used for
sharing are discussed.
Keywords: user interface design, decentralized social networking, personal

identity management, security and privacy.
1 Introduction
Personal information is widely used for social networking and collaboration in busi-
ness and private life. The project digital.me1 researches on intelligent management
of personal information and its use for social networking. A developed framework
and system prototype – the di.me userware – shall integrate the personal information
sphere by crawling and synchronizing data from various sources, e.g. files from
personal storages and devices, profiles or contacts from external social networks. In
order to achieve a high user control on personal data, the system architecture is con-
ceptualized as a decentralized social network (Scerri et al., 2011), which allows to run
personal nodes of the di.me userware either as self-hosted server or as user account
on a provided server. Compared to o other decentralized social networks (Thiel et al.,
2012), di.me follows an extended approach by providing intelligent features like
context-aware recommendations, trust advisory (Heupel et al, 2012., in press), or rule-
based automation of user tasks (Scerri et al., in press).
The personal information in di.me is represented by a set of ontologies (Scerri
et al., 2012). Ontological knowledge is used to power the di.me semantic core in
1
www.di.me-project.eu
30 F. Hermann et al.
providing the automatic integration of distributed personal information, its improved

management, as well as the recommendations and user advisory.
A specific approach of di.me is to provide multiple, partial identities (Pfitzmann,
Hansen, 2010), that can be used to communicate or share information with other
di.me users or external systems. For the design multiple-identity-systems, it is an
important guideline (Hansen et al. 2008) to avoid linkability: Partial identities shall
not be linkable to the same sender if not intended. In single-identity-systems, linkabil-
ity risks are high as they support just one communication identifier, even if pseudo-
nymized. I users want to have distinct identities in single-identity-systems, they
must fall back to multiple user-accounts in the system (which e.g. Skype suggests2).
In contrast, multiple-identity-systems link partial identities to a root identity. To tech-
nically enable continuity in communication, communication identifiers have to be
used. In di.me, these identifiers are the “service-accounts” (for a discussion of ser-
vice-accounts and anonymity requirements on the network level see Bourimi et al.,
2012). Identities and claims in di.me are unproven, i.e. di.me does not act as an identi-
ty-provider. However, the integration of proven identities with anonymous credential
systems was demonstrated (Schwarte et al, in press).
The di.me user interface (UI) is realized as web UI and a mobile Android client.
The major challenge for the UI conceptualization and design presented in this paper
was to find a flexible model that enables interaction the personal information sphere
and to use them in a decentralized social network. A central research question was
how to interact securely with the partial identities, and which concepts or metaphors
for that shall be used in the UI.
2 UI Concepts for Personal Information and Identity
As social systems are highly developed, for many aspects relevant UI design patterns
or established design solutions can be applied to achieve a self-explanatory UI. A
prevalent pattern used in social systems are groups of contacts (see e.g. Crumlish,
Malone, 2009) which resemble closed networks of people and allow the user to
describe the relationship with individual contacts. This pattern was prominently mar-
keted as added-value feature by Google+ 3 (“circles”) in 2011 (Simonite, 2011).
Another example are “aspects” in the Diaspora network4. Groups strongly interplay
with personal identity management, as groups can be used to control disclosure of
partial information. The concept “group” may be simply a category of contacts of an
individual user only or can be shared with others (Schümmer, Lukosch, 2007), e.g. for
user management on centralized platforms. The group concept in OpenSocial5, e.g. is
a shared o centrally managed group for discussion forums.
The identity shown to others, often are called the “profile” (Crumlish, Malone,
2009), being the complete identity (Pfitzmann, Hanssen, 2010), i.e. the superset of all
2
https://support.skype.com/en/faq/fa829/
3
plus.google.com
4
diasporaproject.org
5
opensocial.org
The di.me User Interface: Concepts for Sharing Personal Information 31
personal attributes which may be separated by privacy options. For partial identities
(sets of attributes) that can be shown to others, different metaphors are used. In Mi-
crosoft CardSpace “cards” are shown which include attributes, potentially a certifi-
cate, and a recipient-specific, non-global identifier (Al-Sinani et al., 2010). The UI
guides the user to select the same card for a recipient each time (Steuer et al., 2010).
In the identity management of some providers of OpenID6, attribute sets are presented
as “personas”. Users can manage personas, e.g. “work” with different attributes like
e.g. nickname, or e-mail address. The term “persona” is also used by the Mozilla
identity management 7 of web site credentials. “Identity” stands for attributes at-
tached to the global identifier in the OpenID UI, probably indicating that the global
identifier is the superordinate concept to “personas”. A conflicting meaning of “iden-
tity” is used in a Mozilla Thunderbird extension8: Here, the attribute sets are called
“identities”, and they may be used within one email account. These and other exam-
ples may indicate that up to now, no common understanding and use of identity-
related concepts has evolved within the UI-design for end-user applications.
Non-global communication identifiers are needed for continuity in communicating
in multi-identity-systems. We consider them as less known to end-users, as many end-
user-systems are single-identity-systems (even if offering partial sets of attribute).
This may increase the difficulties of users with building up mental models of identity
management systems (Sun et al, 2011). However, one case of multiple identifiers
which is prevalent in end-user-systems are multiple email accounts. Email-clients like
Microsoft Outlook or Mozilla Thunderbird offer selection of the account used for a
message by a “From”-field in the UI.
3 The di.me UI for Personal Information and Identities
The di.me UI enables the user to interact with the personal information sphere and
communicate via multiple identities. For that, a UI object model based on the seman-
tic knowledge representations was developed and a view structure which focused on a
web UI but also considered the adaption to the Android client.
They resulted from a user-centered design process with two major concept ver-
sions, each tested by a usability test: A first concept version was realized through a
wireframe mockup without any visual design. It was tested in the first usability test.
21 participants from potential di.me user groups (private end-users and business users
working with social system e.g. as public relations managers) worked in moderated
sessions on predefined tasks. The method was a formative evaluation (Lewis, 2012).
The qualitative results of observation, think aloud, and interviews were aggregated to
categories considering the underlying causes of user problems (Dumas, 2003). Addi-
tional summative evaluation measures were used. Based on these results and further
technical requirements, a second concept version was realized through a partly func-
tional HTML-prototype with accurate visual design, and a functional Android client.
6
openid.net
7
mozilla.org/persona
8
addons.mozilla.org/thunderbird/addon/flexible-identity
It was tested in the second usability test with 20 participants, mainly private end-
users. The methodology used was similar to the first test.
In this section, we present the UI concepts for personal information and identity,
and qualitative test results related to that.
3.1 UI Objects
The object model defines the scope of information and the terminology applied in the
UI, and by that plays a key role for the perception and usability of the system by the
end-user. A specific challenge was to find object definitions that are appropriate to
convey not only the proprietary functionality but also the information synchronized
from external services. Besides the core concepts presented below, additional con-
cepts were e.g. “Message”, referring to direct communication messages to a contact or
group, “Service” referring to the external systems that are being connected to a di.me
personal node, or “Situation” which refers to a high-level description of the user’s
context for the context-related features.
The UI objects interplay with the semantic knowledge representation: They can be
mapped directly to one of the domain ontologies, e.g. databoxes (see below) to the
Privacy Preference Ontology (PPO) (Sacco & Breslin, 2012), or personal profiles to
the Nepomuk Contact Ontology9 (NCO). Therefore, a major role of the UI is to ab-
stract the complex ontological knowledge under a simple and intuitive tool that can
still enable users to take full advantage of the semantic core. In addition, the di.me UI
also allows people to personalize some of the ontologies across the UI. For example,
profiles can be customized by allowing users to extend NCO attributes and personal-
ize how they describe themselves as an individual.
Objects for Personal Information Items. Personal information in the di.me domains
cover two distinct types of concepts: Concepts that refer to personal information
which identifies and describes the user as an individual; and concepts for user item
collections, consisting of personal information items (such as documents, images,
etc.). In the object model, this distinction is reflected in by the following concepts:
─ Profile card: As central UI object representing personal identities the term “profile
card” together with a card-visualization is used. It contains arbitrary numbers and
selections of attributes describing the user, such as names, forms of contact, im-
ages, birth date, etc. Attributes are also modeled objects, as the user shall be able to
re-use in several profile cards. Multiple profile cards may be used, potentially with
different pseudonyms and non-overlapping sets of attributes. In the main naviga-
tion, the list of profile cards is labeled with “My Profile”.
Each profile card is either associated with information synchronized from a
particular service (e.g. personal profile information from LinkedIn), or can be set
up and extended freely by the user.
9
http://www.semanticdesktop.org/ontologies/nco/
─ Databox: A “databox“ refers to a virtual collection of common information items

that are typically shared by users. Databoxes can be used to categorize information
items, and customized databoxes for individuals and groups can be created.
─ Data: “Data“ refers to a single informational item coming from the user’s personal
information sphere that can be identified in a file. This includes documents (e.g.
PDF, doc, xls) or images.
Test results. In the usability tests, these concepts and terms were presented and dis-
cussed with the test participants. While no particular issues arose for ‘profile card’
and their attributes, the original wording “Information Sphere” for the list of data had
to be changed because it was not self-explaining. It was renamed to the term “My
Data” which, however, in the second usability study still was confused with the pro-
file cards by some test participants. A suggested optimization was “My documents”.
The term “databox” was self-explaining for most test participants. However, based
on the information provided in the tests, the uni-directional sharing in databoxes was
not clearly evident to half of the participants.
Objects for the Social Sphere. The following pair of UI objects describes the social
sphere:
─ Person: The object “person” refers to a contact of the user. Depending on the use
scenario, the personal di.me nodes will retrieve persons from already existing data-
bases, e.g. by synchronizing the user’s personal data (e.g. email contacts, or con-
tacts known through a service). The term “People” is additionally used as label in
the UI for the main view.
─ Group: Groups can either refer to a number of known persons manually defined by
the user, a group of persons associated to an external service, or a group of persons
suggested by the system depending on the current context (e.g. a group of nearby-
people). A person can be assigned to many groups. Whether the popular concept
“group” is sufficient to structure a user’s social sphere appropriately was an impor-
tant design question. For the di.me userware, the concept should cover as well fine-
grained sets of persons, e.g. the attendees of a meeting, groups like “friends”, but
also high-level concepts like “private lifesphere”.
These objects form not only the system’s information structure but also organize the
access rights management: Databoxes, data, and profile cards can be configured to be
shared with groups or individual persons.
Test results. As expected, these terms were self-explaining to the users. When asked
how they would organize personal groups, few test participants suggested hierarchical
groups for higher level groups like “business”. The statement of most participants,
however, was that they would use only few groups like “friends”, “colleagues” etc.
Only in some cases would assign people to multiple groups.
The service-provided groups (shown by a group of Facebook-friends in the test sce-
narios) were not expected together with the other groups by half of the participants.
These service-provided groups should be clearly identifiable as such and also be access-
ible via a list of connected services.
3.2 View Structure and Navigation Concept

The view structure has to visualize the UI objects and their relations in a flexible na-
vigation. For that, the design approach of an object-oriented UI structure was adopted:
Navigation and views are designed along the data UI objects, the relations amongst
them, their attributes etc.
As navigation concept for main views, a typical tabbed view was chosen for the
web UI (figure 1). The tabs form the main navigation for the main views and are of-
fered at the top of the di.me use interface as a fixed header.
A first main view “People” (selected in figure 1) combines lists for the objects
people and groups. Similarly, the data objects and the databoxes are shown in the
combined view “My Data”. This view type shows the full list of all elements on the
top, and the expandable list of categories (groups or databoxes) in the bottom. The
layout allows for a drag-and-drop interaction for assigning elements to categories. A
second view type shows flat lists of elements, e.g. as thumbnails. This type is used for
the main view “My Profile”, which displays the different profile cards. Further main
views were offered for messaging, system settings, or⎯only in some of the tested
concepts⎯specific visualizations (a timeline, a configurable dashboard).
One central UI element that is being shown in the main views is the properties bar
(figure 1, right margin column). It shows attributes of the currently selected object,
general information (e.g. file name, file size, file path), but also the di.me access
rights, related to the awareness of shared information (e.g. a list of files or databoxes
that a person has access to). In addition, the properties bar can contain shortcuts to
certain actions related to the object at hand.
For details of an element, different sub-pages for object details were developed. In
addition to the page-oriented views, modal dialogues are used for actions like assign-
ing items (e.g. data) to a category-object (e.g. databox), or sharing information to a
person or a group via a particular identity (see section 4.3).
Test results. Within the first versions of the concept, two views for the data and data-
boxes were offered, whereby the view type for ‘data” was focused on flexible search-
ing and filtering, and the view “databoxes” on configuration of access rights. This
separation, however, was criticized by the majority of the test participants. In fact,
many participants spontaneously recognized the similar relation of people to groups
and data to databoxes. Therefore they suggested similar layouts for both views.
Some insights were also gathered on the priority of views: The test particapants
regarded the connection to different external services as one major functionality be-
cause this supports an easy overview of contacts and data in different systems and
storages. For that, for users experienced with multiple communities the most impor-
tant view was the list of people and groups, together with a direct visualization who is
known via which service.
The di.me Usser Interface: Concepts for Sharing Personal Information 35
The properties bar (figure 1, right column) as common element for the main vieews
was introduced based on th he first usability test. Test participants strongly demandeed a
n about which information is accessed by a person, who has
stable place for information
access to what personal infoormation item by which service. Again, in particular pow wer
users working with multiple systems stressed this advantage. Also the requiremennt of
getting a quick overview on o people or groups can be realized if the properties bar
shows aggregated informatiion on multiple selected persons, groups, or data items.
Fig. 1. Web UI: techn

nical prototype with optimizations after the usability tests
3.3 Sharing via Multiple Identities
When shaping the presentaation and interaction with identities in the di.me userware,
the design decision was tak ken to offer only the profile cards for selecting the parrtial
identity if it is needed for communicating or sharing with others. This implies that the
non-global identifier, the service-account,
s is linked to the profile card: Selectinng a
profile card implicitly seleccts the service-account. As a result, the selection of a ppro-
file card decides on linkab bility and unlinkability: All information sent via a proofile
card is linkable to this id dentity. Other options were considered, e.g. to bind the
service-account to a group p. However, this would require the user to select serviice-
accounts for groups, to deefine groups by considering identities and linkability for
them, and to manage sharin ng rights for profile cards in addition. With the chosen ccon-
cept, the service-account iss not explicitly shown in the UI and the user only hass to
work with profile cards.
The selection of a profile card is shown in sharing dialogues (see figure 2a for the
mobile client version): To share e.g. a photo, the user chooses an action “Share to ...”.
In an opened dialogue, the recipients and profile card to share ‘via” can be selected.
(a) (b)
Fig. 2. Dialogues for the mobile UI: (a) “share resource” in the tested functional UI, (b)
“Share” in a scribble with optimizations from the second usability test
Test results. In the second usability test, this interaction sequence and the underlying
considerations were tested and discussed in detail with the participants. The general
metaphor to change the identity used for a communication action by selecting the
profile card was easily understood and not questioned by the participants. Having
tried out the interaction flow, the participants took for granted that shared information
would be linkable to the profile and to other information shared via the same profile.
Few participants mentioned that the content of a selected profile card must easily
be checked so that the user can be sure what is being shared. In a wireframe done
after the test (figure 2b) a preview of the profile card’s information is added for that.
While the role of the service-accounts as non-global identifiers could not be deeply
discussed with most participants, the participants assumed that the recipient cannot
link this communication to other identities.
A further aspect was partly difficult to understand: The majority of participants
suspected that a message was sent to Facebook if a profile card was selected that is
synchronized with a Facebook-Profile. In the interview, the participants explained
that the wording “via” (see figure 2, a) suggested this. Even if such a functionality is
planned, the wording should point more to the fact that a sender-identity is selected.
So, this could be changed to a label “From” as used for email-accounts (scribble in
figure 2 b). As soon the function of pushing content via third services will be in place,
this may additionally be indicated in the UI.
4 Summary
As an important result of the user-centered design, the information and UI structure

was developed which represents the system domains of personal information, identi-
ties, and social sphere. After the user testing, the key UI concepts and view structure
can be considered as stable.
However, further investigations addressing the interplay of the information struc-
ture and access rights are necessary: Specific cases like adding a person to a group
will potentially change access rights (e.g. if the group has access rights the person did
not own so far). For such cases, the awareness of the user must be ensured which
requires additional information in the UI. An approach here is developed by the man-
agement of contact’s trust in the di.me network (Heupel et al., in press).
The results of the user studies point into the direction that the profile cards con-
nected to service-accounts as communication identifiers worked well with respect to
the understanding of unlinkable identities. However, some implications of this design
approach were not deeply investigated so far. E.g., if information is shared to a group,
different profile cards may be known to its members. If a single card is chosen for
sharing, it would be newly disclosed to a part of the group members. In this case, a
linkability warning could be presented. A further requirement not investigated in the
studies is the case of intended linkability: The user may want to inform a recipient
that two profile cards are coming from the same sender. While users always may in-
form the recipient ‘manually’, the system should also support linking two profile
cards.
Acknowledgments. This work is funded by the European Commission under the

Seventh Framework Program FP7/2007-2013 (digital.me – ICT-257787).
References
1. Al-Sinani, H.S., Alrodhan, W.A., Mitchell, C.J.: CardSpace-liberty integration for Card-
Space users. In: IDTRUST 2010, Proceedings of the 9th Symposium on Identity and Trust
on the Internet, pp. 12–25. ACM, New York (2010)
2. Crumlish, C., Malone, E.: Designing Social Interfaces: Principles, Patterns, and Practices
Improving the User Experience. O’Reilly, Sebastopol (2009)
3. Dumas, J.: User-based evaluations. In: Jacko, J.A., Sears, A. (eds.) The Human-Computer
Interaction Handbook, pp. 1093–1117. Lawrence Erlbaum Associates, Mahwah (2003)
4. Heupel, M., Bourimi, M., Kesdoğan, D.: Trust and Privacy in the di.me Userware. In:
Kurosu, M. (ed.) Human-Computer Interaction, Part III, HCII 2013. LNCS, vol. 8006,
pp. 39–48. Springer, Heidelberg (2013)
5. Heupel, M., Fischer, L., Kesdogan, D., Bourimi, M., Scerri, S., Hermann, F., Giménez, R.:
Towards robust multidimensional trust for the digital.me userware by considering situa-
tional context. In: 2012 5th New Technologies, Mobility and Security (NTMS), pp. 1–6
(2012)
6. Lewis, J.R.: Usability Testing. In: Salvendy, G. (ed.) Handbook of Human Factors and
Ergonomics, 4th edn. Wiley (2012)
7. Pfitzmann, A., Hansen, M.: A terminology for talking about privacy by data minimization:
Anonymity, Unlinkability, Undetectability, Unobservability, Pseudonymity, and Identity
Management (Version v0.34) (2010), http://dud.inf.tu-dresden.de/
Anon_Terminology.shtml (retrieved)
8. Sacco, O., Breslin, J.G.: PPO & PPM2.0: Extending the Privacy Preference Framework to
provide finer-grained access control for the Web of Data. In: I-Semantics 2012 - 8th Inter-
national Conference on Semantic Systems (2012)
9. Scerri, S., Schuller, A., Rivera, I., Attard, J., Debattista, J., Valla, M., Hermann, F.,
Handschuh, S.: Interacting with a context-aware personal information sharing system.
In: Kurosu, M. (ed.) Human-Computer Interaction, Part V, HCII 2013. LNCS, vol. 8008,
10. Schwarte, P., Bourimi, M., Heupel, M., Kesdogan, D., Giménez, R., Wrobel, S., Thiel, S.:
Multilaterally secure communication anonymity in decentralized social networking. In:
Proceeding of 10th International Conference on Information Technology: New Genera-
tions, ITNG (in press, 2013)
11. Scerri, S., Gimenez, R., Hermann, F., Bourimi, M., Thiel, S.: digital.me – Towards an in-
tegrated Personal Information Sphere. In: Proceedings on the Federated Social Web Sum-
mit 2011 (2011)
12. Scerri, S., Cortis, K., Rivera, I., Hermann, F., Bourimi, M.: di.me: Context-Aware, Priva-
cy-Sensitive Management of the Integrated Personal Information Sphere. In: EU Project
Networking Track, 9th Extended Semantic Web Conference, ESWC 2012 (2012),
http://data.semanticweb.org/conference/eswc/2012/paper/
project-networking/372 (retrieved)
13. Schümmer, Lukosch: Patterns for Computer-Mediated Interaction. John Wiley & Sons,
Chichester (2007)
14. Simonite, T.: Google Invites Everyone to Be Friends on Its Social Network (2011),
http://www.technologyreview.com/view/506796/algorithm-
predicts-circles-of-friends-using-contacts-data/ (retrieved)
15. Steuer, J.K., Fernando, R., Bertino, E.: Privacy preserving identity attribute verification in
windows cardspace. In: DIM 2010, Proceedings of the 6th ACM Workshop on Digital
identity Management, pp. 13–16. ACM, New York (2010)
16. Sun, S.-T., Pospisil, E., Muslukhov, I., Dindar, N., Hawkey, K., Beznosov, K.: OpenID-
enabled browser: towards usable and secure web single sign-on. In: CHI EA 2011, CHI
2011 Extended Abstracts on Human Factors in Computing Systems, pp. 1291–1296.
ACM, New York (2011)
17. Thiel, S., Bourimi, M., Giménez, R., Scerri, S., Schuller, A., Valla, M., Fra, C., Wrobel, S.,
Hermann, F.: A Requirements-Driven Approach Towards Decentralized Social Networks.
In: Park, J.J., Leung, V.C., Wang, C.-L., Shon, T. (eds.) Future Information Technology,
Application, and Service. LNEE, vol. 164, pp. 709–718. Springer, Heidelberg (2012)
Trust and Privacy in the di.me Userware
Marcel Heupel, Mohamed Bourimi, and Doğan Kesdoğan
Institute for Information Systems, University of Siegen, Germany

[email protected]
Abstract. People in online social networks are constantly sharing infor-

mation with different audiences and it becomes more and more difficult
for them to keep track of distributed information. Further, due to the
complexity of the digital landscape, it is a constant risk to unintention-
ally share information to the wrong people, possibly leading to a loss
of reputation. The European research project di.me is concerned with
the development of a userware, empowering end-users to keep track of
their digital footprints, with an intelligent user interfaces (UI) and smart
advisory. In this paper we present how we calculate persons trust and
inform the privacy of resources shared among persons. We show the con-
cepts for trust and privacy advisory in the di.me userware and address
problems, we were confronted within the design and evaluation process
and how we tackled them. In this respect we specifically address change
requirements (i.e. trust model and UI improvements) we conducted after
an evaluation and user trials with a first prototype.
Keywords: Trust, privacy, user interface, privacy advisory, online social

networks, di.me.
1 Introduction
In todays online social life people are constantly sharing information. It is not
easy for end-users to keep track of all their information, distributed over different
online social networks (OSNs). When in the beginning of OSNs some years ago
most information posted was public, the privacy awareness of people raised over
the last years. Nowadays, for most applications it is common practice to configure
the visibility of posted information with the help of security settings or privacy
preferences in the user interface (UI).
The strong interplay of security and usability, more precisely, the fact that
usability is an important prerequisite for secure systems, is getting more and
more attention recently (see e.g., [1]) and it was already mentioned by Kerckhoff
in 1883 [2] in his six design principles for military ciphers. This is even more
crucial as user experience (UX) and usability have also consequences for pri-
vacy and trust, as it is well known that usability and UX are important factors
for trust [3]. A recent study of Madejski et al. [4] reported, almost all people
make errors when setting their privacy preferences in OSNs. This leads in the
end to information being shared with wrong audiences without intention and can
cause serious damage to the personal reputation in extreme cases. The European

40 M. Heupel, M. Bourimi, and D. Kesdoğan
research project di.me is developing a tool, integrating all information from con-
nected OSNs in a single-user controlled information sphere. Key functionalities
in this respect are the provision of an privacy-preserving intelligent UI and the
provision of smart advisory.
In this paper we will present our approach providing smart trust and privacy
advisory when sharing information in OSNs. The approach consists of a trust
metric, bringing together the privacy of information and trust of persons, which
was already presented in [5] and several UI concepts in order to adequately
present the advisory (based on that metric) to the user in an intuitive and
non-intrusive way. The first version of the metric, as well as corresponding UI
concepts, has been evaluated in first prototypes of the di.me userware. Main
focus of this paper is the presentation of the current status of work as well as
addressed change requirements (CRs) we identified in the evaluation.
The paper is structured as follows: In Section 2 we elaborate key concepts for
trust and privacy in di.me and put them in relation to related work. In Section
3 we provide some essential background information about the di.me userware.
In Section 4 we present our identified CRs, which we address in our approach,
discussed in Section 5. Finally we conclude the paper in Section 6.
2 Definition of Trust in di.me and Related Work

Trust and privacy advisory is one of the key functionalities of the di.me userware,
so it was of major importance to find a common understanding for those terms,
and to develop adequate design concepts for the UI to prevent misunderstandings
on the side of the end-users. Since it is not easy to give a short definition of trust,
it is even harder to present this in an intuitive way to the user. The concept
of trust is difficult to perceive, and the vast amount of definitions make it even
harder. Therefore, we will expound here the concept of trust as we use it in di.me.
In the past, there has been a lot of research in the field of trust, and there are
also a lot of different definitions. When talking about trust, many people think
about trust establishment with unknown parties (e.g., [6] and many more). This
is strongly related to reputation, since to the decision to trust an entity or not
is based on the public opinion. There are a lot of papers concerned with trust
in social networks. Usually they talk about trust in the OSN provider itself or
about trust in unknown entities (e.g., when adding new contacts etc.), which is
also more going in the direction of reputation (to trust someome because of good
reputation, see also [7]).
The di.me userware deals with user-centric information, contact management,
and it supports the user by interacting with already existing contacts in different
OSNs. Since the establishment of new connections is of minor relevance in the
context of online social networks1. Therefore, there is no concept of reputation
in di.me, not even a concept for friends-of-friends [8]. Consequently the concept
of trust in di.me differs from the previously mentioned definitions. In general the
1
We assume most people do not consider reputation to decide if they connect to a
person in a social network or not.
Trust and Privacy in the di.me Userware 41
di.me concept of trust can be defined like the following, which is very similar to
Josangs [7] definition of Reliability trust. To trust someone means that the other
party behaves as expected with a specific probability. This is of course always
connected to a certain context. In di.me we are more or less only moving in the
context of information disclosure, which means basically that trusted persons
will give private information with a very low probability to (untrusted) third
parties. Trust in di.me tries to measure the personal direct trust of the user to
each of his contacts. It is formed by the interaction between individuals (e.g.,
communication, sharing) and can consequently also be seen as measure for the
strength of relationship. Sharing private information to a person is an expression
of trust. As described in [5], we compute the trust based on previous interac-
tions, especially the disclosure of information. The more private the disclosed
information is, and the smaller the audience, the higher is the resulting trust
value. Related to privacy in OSNs many works deal with supporting the user in
privacy preferences (see e.g., [4,9,10]), which are rather static. Our approach, of
giving dynamic trust based privacy advisory on runtime, is rather novel and, to
our best knowledge, not covered by other works.
3 Background Information
3.1 The di.me Project
The di.me project targets the integration of personal information in a per-

sonal information sphere, in order to give users full control over their data and
allow intelligent support and advisory [11]. The di.me userware operates on
user-controlled servers, the so called di.me Personal Servers (PSs) and can be
accessed either by a Web-based UI or an Android application. It aggregates
information, gathered from connected services, like social networks or special
services for e.g, processing location information. The PSs’ form a decentralized
network and also implement State-Of-The-Art network security and anonymity
mechanisms, like e.g., Tor, StartTLS or OAuth in order to ensure maximum
protection of users’ privacy and their data.
Collected information is stored in a semantic repository enabling intelligent
advisory and advanced techniques like e.g., semantic matching of profiles [12].
3.2 The di.me TrustEngine
The component in the di.me userware, processing trust and privacy in order to
calculate respective advisory, is called TrustEngine. It is located on the PS and
accesses information stored in the semantic core as well as in live context data
in order to trigger warnings and advisory, shown in the UI. The TrustEngine is
integrated in all interaction flows concerned with the manipulation and especially
the disclosure of information (e.g., share a profile to a group, send a file to a
contact, etc.). Figure 1 depicts a simplified architecture of the di.me userware
in order to illustrate the integration of the TrustEngine into the core of the
Fig. 1. The TrustEngine in the architecture of the di.me userware
system. The core concept to calculate trust and trigger privacy warnings and
advisory was already presented in Heupel et al. 2012 [5]. The general idea is
that all information stored within the PS should be classified regarding the
privacy of the content. To classify information we us a scale from 0 to 1, where
0 is considered public information and 1 secret2 The trust value for contacts can
then be calculated based on the information shared with them. The disclosure
of very private information to a person is an implicit expression of trust. This
approach would of course require a short learning phase before it has enough
confidence to detect privacy flaws. Once it is calibrated, it is easy to detect
if private information is shared to untrusted persons. Further it can detect a
possibly wrong classification of information when information marked as private,
is shared to a lot of people, it is very likely that the classification is wrong. Besides
the automatic calculation of trust values, it is also possible to set it manually in
the UI, overriding the calculated value.
3.3 First Prototype and Evaluation

The first prototype was presented to the visitors of a summer school event in
Segovia in July 2012. The visitors had the possibility to test the application on
a booth or even download it to their own Android phones and were asked to fill
out a questionnaire. Additionally a focus group was selected, which got a special
presentation and tried the di.me userware on several provided devices. Figure 4
shows screenshots of the first prototype.Most interesting in the evaluation of the
first prototype, concerning trust and privacy advisory, was to get an impression
if the test users understood the general concept, if they thought it was some-
thing useful and if the concepts are presented in an intuitive and non-intrusive
way in the UI3 . Especially the members of the focus group (12 students with
2
A value of 1 should here really be understood as top secret, e.g., like a private
encryption key, which is usually never communicated to anyone.
3
The interested reader can find more information about the overall UI concepts in [13].
Fig. 2. Answers to one of the questions about privacy advisory in the questionnaire
technological majors) were really interested in the feature of privacy advisory

and considered it an extremely useful feature (see also Fig. 2 for some of the
evaluation results).
4 Requirements Gathering
Besides the general usefulness of the features, we asked also more concrete ques-
tions about selected features in the UI, if they are usable and understandable. By
doing this we identified several issues in our concepts4 , from which we address
the main critical ones in the context of this paper. In the following we will shortly
discuss the change requirements (CR1-CR3) we deducted from those issues.
A major critical point of the first prototype was concerned with the general
trust model. There is only one trust value for each person and only only privacy
value for each sharable information item. This model turned out to be not cov-
ering all different use-cases. The di.me userware is intended to cover all different
lifespheres of the user, this means business as well as private. We identified that
trust and privacy can be interpreted slightly different, depending on the life-
sphere5 . Especially when dealing with intersection of different lifespheres (like
e.g., business and private) this can lead to situation where an unintended in-
formation disclosure is not detected. An example for such a situation could be:
The user shares accidentally a highly confidential document to his grandmother.
Obviously this case will not be detected. Therefore we need to extend the model
to detect information being shared to a context, it does not belong to (CR1).
The second important issue we identified, was concerned with the presentation
of the trust and privacy values in the UI and how they can be manipulated. The
presentation of trust and privacy values in the UI is no trivial task, since it can
be easily misinterpreted. As shown in Fig. 4, the privacy value of files, as well
4
Developed internally by involving selected di.me consortium partners.
5
The privacy value of confidential business documents can mean something else than
for private photos.
(a) Privacy slider (b) Trust slider
Fig. 3. Screenshots of trust and privacy slider in the first prototype
as the trust value for persons are represented as a continuous slider. The label
next to the slider was showing the value as a percentage value, depending on
the position. This turned out not to be very intuitive. The users were not sure
what exactly the setting to a high or low trust means, respectively for the privacy
value. It turned out to be easy to misunderstand, if a low trust in someone would
mean just no trust (e.g., due to a lack of information) or distrust (meaning to
expect someone to behave to the users’ disadvantage). Therefore we needed to
redesign the UI elements representing trust and privacy values, in order to make
the concept really clear to the user (CR2).
The calculation of trust advisory takes place on the server and is triggered
by all calls involving manipulation of data and access rights. Therefore, the
TrustEngine is involved if information is shared to a person or group, but also
when adding persons to a group or documents to a databox. For all those actions
a HTTP request to the PS is necessary, which can produce a lot of communica-
tion overhead and lead to a bad UX in the end. Therefore the third identified
requirement is to optimize the communication flow for trust advisory (CR3).
5 Approach
5.1 Extending the TrustEngine (CR1)
One of the main critical points that has been identified in the evaluation of the
first prototype was, that the trust metric we used might not be applicable in all
use-cases, especially when dealing with intersection of different lifespheres (e.g.,
business and private). To overcome this problem, we analyzed the problem and
extended our approach from [5] respectively. Our solution envisages to use the
groups6 in the systems to identify lifesphere borders7 .
Fig. 4. Illustration of two overlapping groups with common files and persons
We assume that users usually tend to group the contacts and thereby au-
tomatically align them to their lifespheres. To foster this behavior we decided
to have 3 default groups in the system, for family, friends and business con-
tacts. On top of this, there are automatic created groups for contacts, imported
from existing social networks (e.g., all LinkedIn contacts) and automatic ad-hoc
groups that are created based on location context information, if contacts are
nearby for a certain time (e.g., sitting in the same room during a meeting). In
order to identify information passing over lifesphere borders without the users
intention, the first step is the identification of such borders. Naturally not every
group represents a separate lifesphere (Fig. 4 illustrates two overlapping groups
with some common files and contacts). In oder to identify a possible threat, we
compute the pairwise distance of two groups, the one the information is already
related to8 and the recipient group. Information becomes ”attached” to a group
by being shared to members of this group.
In order to calculate the distance between two groups, we consider the two
most relevant factors: common contacts and common files. Since we are inter-
ested in the difference, and not what both groups have in common, we take the
symmetric difference (), of both sets (defined for sets Y, Z in (1)). In a coordi-
nate system both values are orthogonal to each other, so the distance between
two groups (= distance to the origin) can be calculated like shown in (2).
Y Z = (Y \ Z) ∪ (Y \ Z) (1)
6
A group in di.me is only for ordering contacts on user side, and should not be confused
with “discussion groups”. They can be compared to the ”circles” of Google+ or the
user defined friend lists of Facebook.
7
A border in this context e.g., between business and private lifesphere is weak in most
cases, since there can be colleagues who are also friends or family.
8
Information becomes related to a group, by being shared to members of those groups
before.
2 2
|AB| |CD|
GD = + (2)
|A ∪ B| |C ∪ D|
With
√ this mechanism we can calculate a distance between two groups, where
2 is resembling the maximum distance (both groups are completely disjunct).
Combined with the privacy value, the new distance can now be used to trigger
additional warnings, not covered by the previous approach.
5.2 Reworking UI Concepts (CR2)
As a representation of the trust value in the first prototype, we used a continu-

ous slider and labels showing a percentage value depending on the position. This
turned out not to be very intuitive. The test users were not sure what exactly the
setting of a trust of e.g., 50% means and the same for privacy. Therefore we con-
solidated the user feedback and reworked our UI concepts together with UI design
experts. Figure 5 (a, b) shows some screenshots of the improved UI for the An-
droid prototype. We left the continuous slider, but the label is no longer showing
a percentage value. Instead we use simple text labels like ”private” or ”public” for
the privacy of files and ”trusted”, ”untrusted” for trust. Further, the labels are
changing their color depending on the restrictiveness (red→private/untrusted,
yellow→medium trust/privacy, green→public/trusted). To reduce the arbitrari-
ness of setted values (e.g., how should someone decide if a file is 64% or 67%
private?), we introduced 3 fixed steps on each slider (left, middle, right) and
thereby drastically reduced complexity and uncertainty of the user setting the
values.
(a) Privacy slider (b) Trust slider (c) Trust awareness
Fig. 5. New trust and privacy concepts in the Android UI

5.3 Optimizing the Interaction Flow (CR3)
In the first version of the clients the TrustEngine was completely integrated
in the server, checking all manipulations of data for possible trust and privacy
issues. If such an issue was detected, a warning will be sent to the UI instead
of an acknowledgment of a successful performed manipulation. The user had
then different options to dissolve the issue (e.g., by removing untrusted persons
from the recipients) or to ignore the warning and share anyway, which would
lead to an adaption of trust values of the recipients (like described before). This
approach did work well, but in practical tests we discovered some potential to be
optimized as we identified in section 4. Therefore we changed the API a bit, and
moved the logic of advisory calculation for file sharing to the client. By doing
this we were able to improve the user experience due to a recognizable reduced
communication overhead. We were also able to provide additional awareness
features in the UI, like e.g., a realtime indicator showing if there is a possible
trust issue when adding people to a group of recipients. Figure 5(c) shows two
of the new colored textfields, indicating the trust status.
6 Conclusions
In this paper we presented how trust and privacy are addressed in di.me and
discussed selected issues identified in first evaluations and user trials of the di.me
userware related to them. We also presented improvements to the general model,
the UI and the interaction design in order to solve those issues. Thereby we were
able to improve the warning mechanism to identify information, shared with the
wrong audience, made the UI more intuitive and increased the response time for
showing privacy advisory in the client UI.
Since the di.me project follows an agile approach merging best-practices from
the security, usability and HCI community [14], we will continuously conduct
further evaluations and improve the system accordingly. Especially the new con-
cepts introduced in this paper will be analyzed and evaluated again, with end-
users and experts. Besides this, we will further extend the privacy advisory by
including analysis of microposts with NLP and including live context informa-
tion going a step further as we already proposed in [15]. Another target for future
work is the improvement of initialization of trust and privacy values in order to
reduce the learning phase as well as the need for manual settings to a minimum.
Acknowledgements. The work carried out in order to write this paper was
supported by the Seventh Framework Program of the European Union, (FP7/
2007- 2013), in the digital.me project under grant agreement no. 257787. Special
thanks goes to our consortium partners from Fraunhofer IAO for their valuable
contributions.
References
1. Cranor, L.F., Garfunkel, S.: Security and Usability: Designing Secure Systems That
People Can Use. O’Reiley (2005)
2. Kerckhoffs, A.: La cryptographie militaire. Journal des Sciences Militaires IX, 5–38
(1883)
3. Corritore, C.L., Kracher, B., Wiedenbeck, S.: On-line trust: concepts, evolving
themes, a model. International Journal of Human-Computer Studies 58(6), 737–758
(2003)
4. Madejski, M., Johnson, M., Bellovin, S.M.: A study of privacy settings errors in an
online social network. In: International Conference on Pervasive Computing and
Communications Workshops (PERCOM), pp. 340–345. IEEE (2012)
5. Heupel, M., Fischer, L., Kesdogan, D., Bourimi, M., Scerri, S., Hermann, F.,
Gimenez, R.: Context-Aware, Trust-Based Access Control for the di.me User-
ware. In: 5th International Conference on New Technologies, Mobility and Security
(NTMS 2012), pp. 1–6 (2012)
6. Aberer, K., Despotovic, Z.: Managing trust in a peer-2-peer information system. In:
Proceedings of the Tenth International Conference on Information and Knowledge
Management, pp. 310–317. ACM (2001)
7. Jøsang, A., Ismail, R., Boyd, C.: A survey of trust and reputation systems for
online service provision. Decision Support Systems 43(2), 618–644 (2007)
8. Heupel, M., Bourimi, M., Scerri, S., Kesdogan, D.: Privacy-preserving concepts
for supporting recommendations in decentralized OSNs. In: Proceedings of the 4th
International Workshop on Modeling Social Media, MSM 2013. ACM, New York
(to appear, 2013)
9. Fang, L., Kim, H., LeFevre, K., Tami, A.: A privacy recommendation wizard for
users of social networking sites. In: Proceedings of the 17th ACM Conference on
Computer and Communications Security, pp. 630–632. ACM (2010)
10. Chaytor, R., Brown, E., Wareham, T.: Privacy advisors for personal information
management. In: SIGIR Workshop on Personal Information Management, Seattle,
Washington, pp. 28–31 (2006)
11. Thiel, S., Bourimi, M., Giménez, R., Scerri, S., Schuller, A., Valla, M., Wrobel, S.,
Frà, C., Herman, F.: A requirements-driven approach towards decentralized social
networks. In: Proceedings of the International Workshop on Social Computing,
Network, and Services (2012)
12. Cortis, K., Scerri, S., Rivera, I., Handschuh, S.: Discovering semantic equivalence
of people behind online profiles. In: Proceedings of the 5th International Workshop
on Resource Discovery, RED 2012 (2012)
13. Hermann, F., Schuller, A., Thiel, S., Knecht, C., Scerri, S.: The di.me user interface:
Concepts for sharing personal information via multiple identities in a decentralized
social network. In: Kurosu, M. (ed.) Human-Computer Interaction, Part I, HCII
2013. LNCS, vol. 8006, pp. 29–38. Springer, Heidelberg (2013)
14. Bourimi, M., Barth, T., Haake, J.M., Ueberschär, B., Kesdogan, D.: AFFINE for
enforcing earlier consideration of nFRs and human factors when building socio-
technical systems following agile methodologies. In: Forbrig, P. (ed.) HCSE 2010.
LNCS, vol. 6409, pp. 182–189. Springer, Heidelberg (2010)
15. Bourimi, M., Rivera, I., Scerri, S., Heupel, M., Cortis, K., Thiel, S.: Integrating multi-
source user data to enhance privacy in social interaction. In: Proceedings of the 13th
International Conference on Interaccion Persona-Ordenador, INTERACCION 2012,
pp. 51:1–51:7. ACM, New York (2012)
From Living Space to Urban Quarter: Acceptance of ICT
Monitoring Solutions in an Ageing Society
Simon Himmel, Martina Ziefle, and Katrin Arning
Human-Computer-Interaction Center, RWTH Aachen University, Germany

[email protected]
Abstract. The question of designing safe, secure, and human-centered urban

living environments is complex, as different and controversial needs from dif-
ferent sources have to be harmonized. The ongoing demographic change with
more and more older and frail persons living in urban environments raise the
question how modern technologies in the information and communication sec-
tor can be seamlessly integrated in specific urban spaces. Not only (medical)
safety and data security issues need to be considered but also technology accep-
tance and the fragile trade-off between the ubiquity of technologies on the one
and the perceived trust in and reliability of technologies on the other hand are of
pivotal importance. To understand users’ willingness to accept monitoring
technologies in different locations (ranging from home to public spaces) we
pursued an exploratory approach. 127 participants in a wide age range volun-
teered to take part in the questionnaire study. Results show that individuals
independently of age and gender are quite reluctant to accept such technologies
in general, even for medical reasons. It was ascertained that acceptance was im-
pacted by the type of technology and the space (domestic vs. public) and the
health status of the people.
Keywords: Technology Acceptance, ICT Health Monitoring, Living Spaces,

Urban Quarter, User Diversity.
1 Introduction
As a matter of fact the demographic change in most (western) societies raises consi-
derable challenges for urban environments with respect to providing humane and
human-centered living conditions. Characteristically, the requirements for designing
safe, secure, and human-centered urban living environments are highly complex. Dif-
ferent and sometimes controversial needs have to be taken into account. These re-
quirements include the harmonization of mobility aspects, the integration of safety
and security of (medical) technologies in the context of treatment and care up giving
sensitive consideration to the perceived intimacy, privacy and data security issues as
well as individuals’ fears of losing control. In addition, as communities might differ
regarding their economic wealth, also cost burden need to be respected. Politics and
communities take up the challenge and put effort into expanding conventional
and well-established solutions by investing in retirement homes and the education of
50 S. Himmel, M. Ziefle, and K. Arning
elderly care nurses [1]. For the next half of the decade all efforts cannot provide suit-
able solutions to the challenge of a numeral increase in older and care-needing per-
sons in contrast to the continuous decrease in younger and potential caring people [2].
While some age-induced issues like dementia or wound treatment require personal
assistance, a great number of older adults endangered by sudden breakdowns [3] are
at risk because they live alone and might not be found in case of emergency [4]. In
these cases different surveillance technologies could provide possible solutions, but
acceptance of these technologies, if researched at all, is desperately low.
While there is only restricted knowledge about the acceptance of integrated infor-
mation and communication technology (ICT) in domestic spaces [5],[6], it is well
known that public surveillance (CCTV) is controversially discussed especially
because of peoples’ fear of data privacy violation and the loss of control over the
protection of intimacy [7],[8]. Also, it was found that acceptance for technology dif-
fers depending on nationality [9] and culture [10].
So far, research does not sufficiently address the role of context factors that might
have a considerable impact on the extent to which persons would be willing to use
surveillance technologies in urban and private spaces.
2 Research Aims
The willingness of citizens to accept technologies in domestic and urban environ-
ments is a highly complex and also fragile phenomenon, which is impacted by a mul-
titude of environmental, spatial, and individual factors, driven by personal needs and
using motives. In order to gain a first insight into this complex interaction system, we
pursued an exploratory approach in which we selected the monitoring technology type
(camera, microphone and positioning system) within the context of medical monitor-
ing. As technology acceptance for or against the medical monitoring is considerably
influenced by the specifics of the location or space in which such technologies would
be installed, we also varied different spaces, ranging from home (sleeping and living
room, garden) to public spaces (park, station, shopping mall). Understanding the
trade-off between the perceived usefulness of technologies in the different loca-
tions/spaces and the extent of wished control [11] and need of privacy [12], partici-
pants had to evaluate the extent of willingness to use the technical systems in the
different locations.
3 Methodology
In this section we describe the procedure of the questionnaire study, the variables and
the sample.
3.1 Questionnaire Study

To reach a broad user group, we applied the questionnaire technique, addressing users
of different health states, ages and gender. The survey was broadcasted online via the
From Living Space to Urban
U Quarter: Acceptance of ICT Monitoring Solutions 51
authors’ private networks. The questionnaire was designed to obtain informattion

about (1) demographic dataa (age, gender, health status) and (2) individual living ccon-
ditions (rural vs. urban distrricts, alone vs. family, distance to medical care).
The health condition of the participants was determined by four questions regaard-
ing chronic diseases, regullarly check-ups, dependency on medical technologies and
health care. People negatin ng all questions were defined as healthy, whereas peoople
agreeing to one or more of these questions were defined as ill. Then, participants had
to evaluate the extent of accceptance. (3) Different integrated surveillance technologgies
(microphone, camera, posittioning system) were to be evaluated in (4) various dom mes-
tic private spaces (living ro oom, bedroom, garden) as well as in public, urban spaaces
(local bar, city council, paark, train station, museum, church, bus stop). In orderr to
classify the different locatiions, living spaces were split in three dimensions: privvate
vs. public, intimate vs. distaant, and indoor vs. outdoor (see fig.1).
train station
park, bus stop
market
museum
mall city
local bar church council
private to public
garden
living
groom
bedroom
intimate to distant
Fig. 1. Claassifying regarded spaces in three dimensions
The participants had to judge their willingness to integrate each technology (thrree)
in each space (twelve) on a six point Likert scale (never, no, rather no, rather yes, yyes,
always).
3.2 Participants
127 participants, aged 19 to o 74 years, took part in this study (M=37.8, SD=17.9). 441%
of the sample was female, 59% male. To investigate if technology acceptance diff ffers
depending on age and tech hnology generation, the sample was split in age groups re-
ferring to three different tecchnology generations: The young group was aged betw ween
19 and 28 years, mainly uniiversity students (40%, M=22.8 SD=2.6, 49% female/ 551%
male), the middle-aged working generation between 29 and 59 years (30%, M=38.2,
SD=9.4, 26% female/ 74% male), and the older adults between 60 and 74 years (30%,
M=65.9, SD=4,3, 44% female/ 56% male). The overall educational level (40% uni-
versity entrance diploma, 47% university degree) lay high above average. 70% of the
participants were classified as healthy (see 3.1).
4 Results
In a first exploratory research step we correlated the demographic variables with the
acceptance items in order to identify the influence of the regarded user demographics
on the acceptance. To get a deeper insight in the acceptance of monitoring for each
living space and each technology we followed a two-way deductive approach. We
first analyzed the general acceptance of our three chosen surveillance technologies
(microphone, camera, positioning system) by summing up the acceptance ratings of
this technology in all spaces and then checked for significant effects of user diversity
(age, gender, health condition) (see 3.1). Then we analyzed the general acceptance of
surveillance in all different living and urban spaces (see 3.2). Due to space limitations
only the general acceptance of each technology in each space will be dealt with here.
Finally, we focused on the most significant correlations by analyzing how the accep-
tance of camera surveillance in all spaces is affected by the health condition of the
user (see 3.3). Q-Q plots and boxplots indicated that all our scales and items were
normally distributed. Data were statistically analyzed by MANOVA procedures and
ANOVAs. Significance level was set at 5%.
4.1 Acceptance of Surveillance Technologies: Overall Technology Acceptance
If summarizing the different locations and spaces under study, we find a low accep-
tance of all regarded technologies (camera, microphone, positioning) with no signifi-
cant differences between them (see fig. 2).
Fig. 2. Mean acceptance ratings and SD for different technology types
Obviously, participants were quite negative about the fact of being monitored at all
and no technology was generally accepted, with the lowest levels of acceptance for
the camera, followed by the microphone and the positioning system.
From Living Space to Urban Quarter: Acceptance of ICT Monitoring Solutions 53
Age effects on Technology acceptance. When bringing the age groups into focus,
one could see the tendencies of age influencing acceptance differing for each type of
technology (see fig. 3). The acceptance of integrated microphones for health reasons
marginally increased with age whereas cameras are least accepted by old people, also
compared to all other technologies. Camera and positioning system are both most
accepted by the middle-aged group. The smallest differences in acceptance were seen
with the positioning system showing immunity against the influence of user demo-
graphics, which also was a key result in previous research [6]. Though, it should be
noted that differences did not reach statistical significance corroborating that age is
not driving acceptance: Rather the reluctance of people towards accepting monitoring
systems in the medical context is high in all age groups.
Fig. 3. Mean acceptance with SD by technologies, split by age
Gender effects on Technology acceptance. Different from other studies in which

gender did impact the technical self-confidence [13] or interest in specific technolo-
gies [14], no significant gender effect could be detected for the acceptance of different
technologies for health surveillance purposes (fig. 4).
Fig. 4. Mean acceptance with SD by technologies, split by gender

Health condition effects on Technology acceptance. Significant health-effects were

revealed for microphone (F(1,122)=4.24; p<0.05) and camera (F(1,122)=4.76;
p<0.05) surveillance acceptance. Also, the positioning system showed similar but not
significant tendencies. Ill people conveyed lower acceptance for health surveillance
technologies (see fig. 5).
Fig. 5. Mean acceptance with SD by technologies, health-effect
4.2 Acceptance of All Surveillance Technologies in Different Spaces
Breaking down the summarized (general) acceptance of the three technologies into all
twelve researched spaces indicated their reciprocal interaction. The consideration of
different dimensions for the regarded spaces (see fig.1), seems to depend on the type
of technology and needs further research. All data for the whole sample, not consider-
ing user diversity, did not show statistical significant effects. The following results are
visualized (see fig.6) and discussed descriptively.
For camera surveillance the lowest acceptance was recognized when entering
private space (living-, bedroom, garden), decreasing from intimate to more distant
spaces. This decrease was not transferable to public spaces, where there were no sig-
nificant differences for private and distant spaces. However, the descriptive data
shows positive tendencies for accepting camera surveillance at the train station and
bus stop. Regarding all public spaces, a decreasing acceptance of cameras could ac-
tually be recognized from indoor to outdoor spaces. The museum as the most ac-
cepted place for surveillance plays a particular role – a possible effect might be the
acceptance of anti-theft devices. The acceptance of auditory surveillance increased
from indoor to outdoor spaces, not differing between private vs. public using contexts
nor showing different evaluations regarding intimate vs. distant spaces. Microphones’
usefulness for surveillance in outdoor spaces could have been regarded as poor for
health reasons.
The positioning system seems to be independently accepted for all spaces – with
exception of the museum, which breaks ranks shouting for further research. It is may-
be affected by acceptance of anti-theft motion detecting, whose functionality basically
is one kind of positioning system. This result also verifies former research.
Fig. 6. Mean acceptance with SD by technologies and spaces
4.3 Acceptance of Camera Surveillance and Effects of Health Status
The explorative research approach using correlation analysis indicated the most sig-
nificant effects on acceptance regarding the camera technology for each space (de-
pendent variables) and the attribute of being healthy or ill (independent variable). Age
and gender did not show any significant correlation effects on acceptance.
The acceptance patterns for healthy people can be orientated on the general
acceptance results for camera (compare fig.6 with fig. 7)).
Fig. 7. Mean acceptance with SD of camera surveillance in different spaces, health effect
Ill and frail persons show a significantly lower acceptance for camera surveillance
except in private indoor spaces (living- and bedroom, in which both groups do not
accept cameras) and the exceptional museum (where both groups would accept cam-
eras, if at all). The camera in garden (F(1,121)=5.08; p<0.05), local bar
(F(1,121)=8.8; p<0.05) and church (just failed p<0.05 significance level) are con-
gruently assessed: No one accepts cameras which is more distinct within ill people.
This different acceptance rating grows with changing dimensions (see fig. 1.) from
private to public spaces. The distinction in acceptance regarding mall F(1,120)=21.24;
p<0.00) and city council (F(1,120)=9.72; p<0.00) still increases wherever both are
still not accepted. Moving on from indoor to outdoor respective from intimate to more
distant spaces, the acceptance not only keeps drifting, observing park F(1,120)=10.26;
p<0.00), market F(1,121)=20.22; p<0.00), train station F=(1,121)=15.11; p<0.00) and
bus stop F=1,121)=23.7; p<0.00) – the tendencies even show a turning point for
healthy people, who barely could accept these technologies. We therefore can say for
the camera as one health monitoring solution and healthy people as regarded user
attribute the debated dimensions have a significant influence on acceptance.
5 Discussion and Outlook

In this study we undertook a first exploratory approach regarding the question if and
if so persons would accept medical monitoring technologies within urban living envi-
ronments. The need of understanding individual barriers and perceived benefits of
technologies integrated within urban environments is high, especially as increasingly
more and older people are living in urban communities, which might be bothered by
age-related frailness and which need medical monitoring to maintain mobility and
independency at older age.
Summarizing the findings of this exploratory study we see a lot of reluctance to ac-
cept medical monitoring in urban environments. Especially the camera is least accepted,
followed microphone and positioning systems. Though, acceptance strongly differs
depending on the specific location (especially for the camera and the microphone, to a
lesser extent for the positioning system). The differentiation between private and public
spaces has a strong impact on acceptance, especially for camera surveillance, while the
difference between indoor and outdoor spaces plays a minor role.
Referring to well established technology acceptance theories we could replicate the
strong effect of technology adaption: surveillance systems for public places (train
station, museum), which were first installed ages ago for crime prevention [15],
received the highest acceptance. Also, user diversity is a modulating factor of tech-
nology acceptance of ubiquitous surveillance technologies. When widening the focus
from private living space to public urban quarter, a higher technology acceptance for
healthy people was revealed. This might be contra-intuitive on a first sight as healthy
persons could see less advantages in health surveillance (technologies) in comparison
to ill people, which - according to common preconceptions – are naturally accepting
medical technologies, simply because they do not have alternatives [16]. However,
the sensitive understanding of illness and age require deeper insights into individual
attitudes, wishes, needs, and values. Possibly, ill people could be more sensitive to
health monitoring technologies, especially those depending on medical devices, as
they might fear stigmatizing and violations of intimacy requirements and data privacy
[12]. Future studies will have to find out the reasons of older persons to accept or
decline medical surveillance technologies. Still more important we will have to ex-
plore possible conditions under which persons would conditionally accept medical
technologies implemented within the urban living environment.
Acknowledgements. Thanks to Julia van Heek and Barbara Zaunbrecher for research
assistance. This work was funded by the Excellence initiative of German state and
federal governments.
References
1. Beard, J.R., Biggs, S., Bloom, D.E., Fried, L.P., Hogan, P., Kalache, A., Olshansky, S.J.:
Global Population Ageing: Peril or Promise. World Economic Forum, Geneva (2011)
2. Valverde, J.L.: 2050: A Changing Europe. Demographic Crisis and Baby Friend Policies.
Pharmaceuticals Policy and Law 9(1-2), 270 (2007)
3. Moylan, K.C., Binder, E.F.: Falls in OlderAdults: Risk Assessment, Management and Pre-
vention. The American Journal of Medicine 120(6), 493.e1-493.e6 (2006),
doi:10.1016/j.amjmed.2006.07.022
4. Gurley, R.J., Lum, N., Sande, M., Lo, B., Katz, M.H.: Persons Found in Their Homes
Helpless or Dead. N. Engl. J. Med. 334, 1710–1716 (1996)
5. Osvald, F., Wahl, H.-W.: Dimensions of the Meaning of Home in Later Life. In: Rowles,
G.D., Chaudhry, H. (eds.) Home and Identity in Later Life: International Perspectives, pp.
21–45. Springer, New York (2005)
6. Ziefle, M., Himmel, S., Wilkowska, W.: When your living space knows what you do: Ac-
ceptance of medical home monitoring by different technologies. In: Holzinger, A., Simon-
ic, K.-M. (eds.) USAB 2011. LNCS, vol. 7058, pp. 607–624. Springer, Heidelberg (2011)
7. Slobogin, C.: Public Privacy: Camera Surveillance of Public Places And The Right to
Anonymity. Mississippi Law Journal 72 (2002), doi:10.2139/ssrn.364600
8. Surette, R.: The thinking eye: Pros and cons of second generation CCTV surveillance sys-
tems. Policing: An International Journal of Police Strategies & Management 28(1), 152–173
(2005)
9. European Commission (EUR 22030), “Eu Research on Social Sciences and Humanities
2004: On the Threshold to Urban Panopticon? Analysing the Employment of CCTV in Eu-
ropean Cities and Assessing its Social and Political Impacts – URBANEYE” Final report.
Office for Official Publications of the European Communities, Luxembourg (2007)
10. Alagöz, F., Ziefle, M., Wilkowska, W., Valdez, A.C.: Openness to accept medical tech-
nology - A cultural view. In: Holzinger, A., Simonic, K.-M. (eds.) USAB 2011. LNCS,
vol. 7058, pp. 151–170. Springer, Heidelberg (2011)
11. Arning, K., Ziefle, M.: Different Perspectives on Technology Acceptance: The Role of
Technology Type and Age. In: Holzinger, A., Miesenberger, K. (eds.) USAB 2009. LNCS,
12. Wilkowska, W., Ziefle, M.: Privacy and Data Security in E-health: Requirements from Us-
ers’ Perspective. Health Informatics Journal 18(3) (2012)
13. Beier, G.: Locus of control when interacting with technology (Kontrollüberzeugungen im
Umgang mit Technik). Report Psychologie 24, 684–693 (1999)
14. Ziefle, M., Himmel, S., Holzinger, A.: How Usage Context shapes Evaluation and Adop-
tion in different Technologies. In: Rebelo, F., Soares, F. (eds.) 2nd International Confe-
rence on Human Factors and Ergonomics in Healthcare 2012,, Advances in Usability
Evaluation Part II, pp. 2812–2821. CRC Press (2012)
15. Harris, C., Jones, P., Hillier, D., Turner, D.: CCTV surveillance systems in town and city
centre management. Property Management 16(3), 160–165 (1998)
16. Ziefle, M., Schaar, A.K.: Gender differences in acceptance and attitudes towards an inva-
sive medical stent. Electronic Journal of Health Informatics 6(2), e13, 1–18 (2011)
The Impact of Explanation Dialogues
on Human-Computer Trust
Florian Nothdurft, Tobias Heinroth, and Wolfgang Minker
Institute of Communications Engineering

University of Ulm, Germany
{florian.nothdurft,tobias.heinroth,wolfgang.minker}@uni-ulm.de
Abstract. Maintaining and enhancing the willingness of a user to interact with a

technical system is crucial for human-computer interaction (HCI). Trust has
shown to be an important factor influencing the frequency and kind of usage. In
this paper we present our work on using explanations to maintain the trust rela-
tionship between human and computer. We conducted an experiment on how
different goals of explanations influence the bases of human-computer trust. We
present the results of the conducted study and outline what this means for the
design of future technical systems and in particular for the central dialogue
management component controlling the course and content of the HCI.
Keywords: Computer applications, Knowledge based systems, Cooperative

systems, Adaptive systems, Expert Systems.
1 Introduction
Advances in human-computer interaction based technology enable the vision of mo-

bile or ubiquitous technical systems accompanying users in their daily life. These
systems have the possibilities to serve as a personal assistant due to the potential long-
term relationship between a human and a technical system. Personal assistants should
have the potential to solve complex problems the user is faced with daily or solely and
which require significant interaction. However, this paradigm of interaction requires a
working relationship between the human and the technical system. Such a relationship
is characterized by a user's cooperativeness during interaction and his trust in the
technical system.
Human-computer trust has shown to be a crucial point in keeping the user moti-
vated and cooperative. The users' trust in a technical system will decrease if he does
not understand system actions or instructions (Muir, 1992). This may lead to a change
in the willingness to interact or in the worst case scenario to an abort in interaction
and use (Parasuraman & Riley, 1997). However, providing explanations can help to
prevent a decrease of trust (Glass, McGuinness, & Wolverton, 2008).
Similar to explanations in human-human interaction, explanations in human-
computer interaction pursue a certain goal. Explanations are given to clarify, change
or impart knowledge with the implicit idea to align and adapt the mental models of
60 F. Nothdurft, T. Heinroth, and W. Minker
the participating parties. In this case the users’ constructed mental model of the sys-
tem has to be adapted to the correct mental model (i.e. the designed behavior, know-
ledge and reasoning) of the system. This means, that not only the knowledge of the
user has to be adapted to match the system required knowledge, but the behavior of
the system has to be explained to the user in order to keep his transparency in the
system. For example, if the users’ believed reasoning process of the system does not
match the real reasoning process, the occurrence of not understandable or as wrong
perceived situations seems inevitable. The correction of the users’ mental model can
be done by providing explanations.
In order to pursue a particular objective an explanation goal (see Table 1 for a list-
ing of explanation goals) has to be selected. However, a complex explanation
can pursue more than one goal at a time. For example, one can think of a learning
explanation which contains a conceptual explanation as well. Additionally a simple
explanation does not necessarily have to pursue only one goal. The mapping of the
presented final explanation to the inherited explanation goal does not have to be a one
to one mapping. This means that the effect of a given explanation is not limited to the
originally intended goal, but may implicitly pursue other goals as well.
Table 1. The different goals an explanation can pursue
Goal of Explanation Description

Justification Explain the motives of the answer?
Transparency How was the systems answer reached?
Relevance Why is the answer a relevant answer?
Conceptualization Clarify the meaning of concepts
Learning Learn something about the domain
For our experiment we concentrated on justification and transparency explanations.

Justifications are the most obvious goal an explanation can pursue. The main idea of
this goal is to provide support for and increase confidence in given system advices or
actions. The goal of transparency is to increase the users understanding in how the
system works and reasons. This can help the user to change his perception of the sys-
tem from a black-box to a system the user can comprehend. By this, the user can build
a mental model of the system and its underlying reasoning processes. Selecting the
appropriate goal of explanation based on users' human-computer trust is an unprece-
dented approach because existing studies concentrate on trust as a one-dimensional
concept. However, Trust is multi-dimensional and consists of several bases. For hu-
man relationships, Mayer (Mayer, Davis, & Schoorman, 1995) defined three levels
that build the bases of trust: ability, integrity and benevolence.
For human-computer trust (HCT) Madsen and Gregor (Madsen & Gregor, 2000)
constructed a hierarchical model (see Fig. 1). They tried to separate trust into nine
basic constructs but eliminated four constructs because of representative or discrimin-
ative issues. This results in five basic constructs of trust, with two major components
(cognitive- and affect-based components) and expected relationships modeled be-
tween them. However, as Mayer already stated, the bases of trust are separable, yet
The Impaact of Explanation Dialogues on Human-Computer Trust 61
related to one another. All bases must be perceived high for the trustee to be deem med
trustworthy. If any of the bases
b does not fulfill this requirement, the overall trustw
wor-
thiness can suffer (Madsen & Gregor, 2000).
Fig. 1. In this for human-comp puter trust constructed model, personal attachment and faith bbuild
the bases for affect-based truust and perceived understandability, perceived technical com mpe-
tence and perceived reliability for cognition-based trust (Madsen & Gregor, 2000)
If we want to use explan nations to influence the human-computer trust relationsship

in a directed and not arbitrrary way, we need to find the most effective mappingg of
explanation goals to HCT bases (see Fig. 2). This means, that we have to idenntify
which goal of explanation influences which base of trust in the most effective w way.
Thereby, undirected strateggies to handle HCT issues can be changed into directed and
well-founded ones, substanttiating the choice and goal of explanation.
In our experiment we wanted
w to test how the different goals of explanation doo in-
fluence the bases of trust in unexpected, not understandable situations in hum man-
computer interaction (HCI). The main idea was to influence the HCT relationship iin a
negative way and to analyzze how different explanation goals can help to remedyy or
reduce occurring trust issuees to prevent the user from losing the willingness to interract
with the system.
Fig. 2. The main goal is to fin

nd the most effective mapping between goals of explanation and
the bases of human-computer trust
t
2 The Experiment
The setting of the experiment was a web-based simulation of a nuclear power plant
control room. The subjects had to accomplish several rounds of interaction in which
they had to solve pre-defined tasks. During those rounds they were assisted by a
virtual anthropomorphic assistant (Lang & Minker, 2012) which helped the user
proactively with the upcoming tasks. The user interface represented the controls of
the nuclear power plant by distributing control room functionalities over various tabs
(see Fig. 3).
Fig. 3. This is a screenshot of the experiment interaction interface. On the right we can see the
agent in sleep mode. Distributed over the tabs are the nuclear power plant functionalities. Be-
neath is the text describing the task the user has to accomplish. In this case the weather is get-
ting worse and the power plant has to be adjusted to provide more power.
However, during selected rounds, the simulation reacted unexpectedly. The main
idea behind this was to provoke a decrease in human-computer trust between the hu-
man and the machine. By this we wanted to test how different goals of explanation
can help to prevent or reduce the expected trust loss. In total 60 test persons took
part in the experiment. For each kind of explanation 20 persons were tested with an
evenly distributed number of males and females. However, due to incomplete data
only 48 valid subjects remained. The average age was 23.35 with the majority of
the participants being students. In order to measure the influences on the bases of
The Impact of Explanation Dialogues on Human-Computer Trust 63
human-computer trust we were using a translated version of the working alliance

inventory questionnaire modified by Madsen and Gregor (Madsen & Gregor, 2000)
for the measurement of human-computer trust. The original questionnaire (Horvath &
Greenberg, 1989) measures which trust and belief a therapist and patient have in each
other in achieving a desired outcome. The HCT questionnaire was adapted to our
needs and consisted of 15 items (three items for each base of trust).
In total the subjects had to complete seven tasks. For example, the nuclear power
plant had to be controlled to output a certain amount of power. However, three of the
seven tasks were interrupted by unexpected situations. For example, a water pump
was broken or some control rod elements were defect. These situations were meant to
be incongruent to the users' mental model of the system and therefore not understand-
able and unexpected. The course of the experimental design can be seen in Table 2.
Table 2. The Course of the experiment regarding unexpected situations and provided HCT-
questionnaires
Round 1 2 3 4 5 6 7
Error / unexpected x x x
Situation
HCT Questionnaire x x x x x
In the beginning we wanted the user to accustom to the system. Therefore, the first
questionnaire was presented after the second task. In the third, fifth and sixth round
the task was interrupted by an unexpected system error (see Fig. 3).
Fig. 4. Rounds with unexpected situations in the experiment: In the first study we examined
justification, transparency and no explanation
As mentioned before the system reaction was either augmented by a transparency

or justification explanation. The baseline was a group provided with no additional
explanation. These experiments allowed us to determine, whether our constructed
unexpected situations did influence the HCT negatively. Our hypothesis was that
both goals of explanation would perform better in terms of keeping trust than no ex-
planation at all. Additionally, we assumed that justification explanations would help
especially the bases of technical competence and understandability. For transparency
explanations we expected influences on the bases of understandability and reliability.
3 Results
The first problem we encountered was that the unexpected situations did not induce
the anticipated trust loss. In our opinion, this was either due to the too good interac-
tion and help of the system represented by the virtual agent or the introductory phase
was too short to build a trust relationship between man and machine.
Observing the data we did not find any significant differences between providing
the user with no explanations, justifications or transparency explanations. Especially
the development over time seemed rather arbitrary in terms of system errors influen-
cing the human-computer trust relationship. However, when analyzing the data we
found some gender differences (see Fig. 4).
Concerning the base of the perceived reliability we found a marginal significant
( 2.9, .08) difference when using transparency explanations (4.29 for males
to 3.49 for females). When providing justifications we got a significant difference
( 4.0, .05) concerning the perceived faith between males (3.2) and females
(4.12). When observing only female subjects we could prove a marginal significant
( 8.5, .09) difference between providing no explanation (3.9) compared to
providing justifications (4.57) for the base of reliability.
Fig. 5. Results for the different types of intervention divided on gender
Closer examination of the data revealed some further tendencies. However, due to
the study design there were not sufficient numbers of subjects to analyze the male to
female differences to draw valid conclusions. In the next chapter we will discuss the
results and mention some of the rather interesting tendencies which we hope to ad-
dress in more detail in future work and experiments.
As mentioned before the situations meant to influence HCT negatively did not
serve their purpose. Despite experiencing not understandable situations the help pro-
vided by the agent was sufficient to handle the occurring problem. The agent provided
a step-by-step tutorial on how to overcome the experienced problems. Therefore, in a
follow up experiment we want to separate the occurring system error from the task the
subject has to accomplish. Additionally, we plan to extend the introductory phase.
Every type of task the user has to execute later on, should be done in a comparable
way in the beginning. This way we hope to build a more complete users' mental
model of the system.
Taking a closer look at the development of perceived reliability and perceived

technical competence (see Fig. 5) when providing transparency explanation, we
observe that for male subjects the curve has rather negative tendencies. Compared
to that, females seem to be influenced only in a marginal way by transparency
explanations, at least in our experiment.
Providing justifications seemed to benefit the perceived reliability regardless
of gender. For perceived technical competence females seemed to benefit from justi-
fications (as well see Fig. 5). As we lost in total 10 female subjects to incomplete
questionnaires or quitting of the experiment (compared to 2 males), the number of
females per explanation goal was limited. However, the results give some evidence
that the gender aspect of explanation goals is worth more thorough investigation in
the context of our follow-up studies.
Fig. 6. The values of perceived technical competence on the left and perceived reliability on the
right as given by the HCT questionnaire
Despite the setback of not influencing the HCT negatively, we can state that differ-
ent goals of explanation do influence the bases of trust in a particular way. For exam-
ple, justifications do influence the perceived reliability of the user towards the system
(i.e. Fig. 5). Despite being not significant, this could be a first indicator that goals of
explanation influence particular bases of trust.
4 Discussion
Technical systems meant to accompany and assist the user in his daily life are prone
to the consequences of users’ loss of trust. Personal assistants rely on users’ coopera-
tiveness and motivation to interact with the system and share relevant information.
This is necessary to facilitate the adaptation and personalization of the technical sys-
tem to the individual user. Especially future technical systems will have to adapt to a
user’s capabilities, preferences, requirements, and current needs. As these continually

available assistants should appear competent, co-operative, and reliable to their users
the relationship between human and technical system has to be healthy and trustwor-
thy. If technical systems should change from helpers for simple, domain dependent
tasks to individualized continually available assistants for complex and wide-spread
domain independent tasks the dialogue between human and technical systems can be
a typical bottleneck. The accomplishment of complex tasks requiring extensive dialo-
gue depends on the users’ willingness and motivation to interact. Maintaining the
trustworthiness of the systems is, as mentioned earlier, one of the key factors to keep
the user willing to interact with the system. Continually available and assisting sys-
tems possess a bunch of relevant information on the user, which can facilitate the
preservation of human-computer trust. Gender, education, technical background and
history of interaction can help to recognize situations prone to HCT-decrease and to
select the probably most suitable intervention to remedy possible negative effects.
As presented in this paper, we found significant results as well as important evi-
dence that the loss of trust has to be handled in a directed, not arbitrary way. Diverse
types of explanations can influence the bases of trust in directed way. A well-founded
selection of the most effective type of explanation will help to handle occurring trust
issues in a non-arbitrary way. Additionally, the experiment indicates that females and
males seem to react in a different way to types of provided explanations. As this does
not have to be the only additional factor in the explanation selection, future research
has to take other factors as education or technical background into account as well.
Especially technical background is known to have an effect on explanations. The
content, complexity and form of the explanation have to be adapted to the level of
technical background and knowledge a user possesses. For example, expert and no-
vice users profit differently from the diverse goals of explanation (Lim, Dey, & Avra-
hami, 2009). In our opinion, the goals of explanation may influence the bases of trust
differently regarding the expertise as well.
However, there are still some missing pieces to be able to use these and upcoming re-
sults in a technical system for adaptation and individualization. Knowing how to react in
the probably most effective way to a decrease in a base of trust is important, but the
information which base of trust was influenced is essential as well. For this we need to
categorize the context information in situations of trust loss. If we can state with a cer-
tain probability, that a particular context along with unexpected situations influences a
specific base or bases of trust, we can react in an appropriate manner. This will enable
us to include the results in an existing architecture to handle human-computer trust is-
sues by providing explanations (Nothdurft, Bertrand, Lang, & Minker, 2012).
5 Conclusion and Future Work

In this paper we outlined why it is necessary to consider human-computer trust in
technical systems. Especially in systems meant to be personal, individualized
and continually available assistants, the long-term relationship between human and
technical system is highly relevant. Unexpected or not understandable situations have
shown to influence the users’ HCT negatively. Research has shown that explanations
can help to remedy negative effects occurring in these situations. However, the map-
ping between the different goals of explanations and the diverse components of hu-
man-computer trust remains unclear. Therefore, we presented an experiment on how
different goals of explanation influence particular bases of human-computer trust. We
found indication that indeed differences exist in the effects of explanation goals on the
bases of trust. Additionally, we did find some gender aspects, which seem to be worth
analyzing more extensively in follow-up experiments. Therefore, it seems worthy to
consider individual and situational parameters when providing explanations in critical
situations.
Acknowledgements. This work was supported by the Transregional Collaborative

Research Centre SFB/TRR-62 “Companion-Technology for Cognitive Technical
Systems” which is funded by the German Research Foundation (DFG). Additional
thanks to M. Attenberger, L. Knecht, N. Maucher, M. Partsch and F. Rueck for con-
ducting the study.
References
Glass, A., McGuinness, D.L., Wolverton, M.: Toward establishing trust in adaptive agents. In:
IUI, pp. 227–236. ACM (2008)
Horvath, A., Greenberg, L.: Development and validation of the Working Alliance Inventory.
Journal of Counseling Psychology, 223–233 (1989)
Jörg, C., Anders, K.-P.: Designing Explanation Aware Systems: The Quest for Explanation
Patterns. In: Explanation-Aware Computing – Papers from the 2007 AAAI Workshop, pp.
20–27. AAAI Press, Vancouver (2007)
Lang, H., Minker, W.: A collaborative web-based help-system. In: Proceedings of the 2nd
International Conference on Web Intelligence, Mining and Semantics, pp. 61–65. ACM,
Craiova (2012)
Lim, B.Y., Dey, A.K., Avrahami, D.: Why and why not explanations improve the intelligibility
of context-aware intelligent systems. In: Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems, pp. 2119–2128. ACM, Boston (2009)
Madsen, M., Gregor, S.: Measuring human-computer trust. In: 11th Australasian Conference on
Information Systems (2000)
Mayer, R.C., Davis, J.H., Schoorman, F.D.: An Integrative Model of Organizational Trust. The
Academy of Management Review, 709–734 (1995)
Muir, B.M.: Trust in automation: Part I. Theoretical issues in the study of trust. In: Ergonomics,
pp. 1905–1922 (1992)
Nothdurft, F., Bertrand, G., Lang, H., Minker, W.: Adaptive Explanation Architecture for
Maintaining Human-Computer Trust. In: 36th Annual IEEE Computer Software and Appli-
cations Conference (COMPSAC 2012), pp. 176–184. IEEE, Izmir (2012)
Parasuraman, R., Riley, V.: Humans and Automation: Use, Misuse, Disuse, Abuse. Human
Factors. The Journal of the Human Factors and Ergonomics Society, 230–253 (1997)
Sormo, F., Cassens, J.: Explanation goals in case-based reasoning. In: Proceedings of the
ECCBR 2004 Workshops, pp. 165–174 (2004)
Trustworthy and Inclusive Identity Management
for Applications in Social Media
Till Halbach Røssvoll and Lothar Fritsch
Norwegian Computing Center (Norsk Regnesentral)

Gaustadalléen 23, 0314 Oslo, Norway
{Till.Halbach.Rossvoll,Lothar.Fritsch}@nr.no
Abstract. We describe a prototype for inclusive and secure identity

management regarding a bill sharing application in social media. Be-
ginning with the principals of universal design, and involving groups of
users with impairments, we designed a set of alternative authentication
methods based on OpenID. This work explains the scenario and the par-
ticularities of designing a trust, security, and privacy infrastructure with
a high degree of usability for diverse user groups, and which is aligned
with the requirements from regulatory frameworks. The user trials show
that several authentication alternatives in multiple modalities are wel-
comed by impaired users, but many have restrictions when it comes to
payments in the context of social media.
Keywords: Authentication, Authorization, OpenID, Identity Manage-

ment, Social Media, Payment, Accessibility, Usability, E-Inclusion.
1 Introduction
Online payment applications naturally require a high level of trust by the user.
This applies in particular to payment services inside social media, which are, as
of today, typically viewed as being insecure and open, in contrast to for instance
secure and privacy respecting internet banking [1].
The e-Me project [2] focuses on this trust and aims at providing accessible,
multimodal, and adaptive authentication and authorization methods for social
media that are usable for all users. In an integrated social-payment application
connected to online banking, an OpenID provider has been developed by means
of inclusive-identity management methods. The provider is used for both the
social-media access control and the embedded payment service.
So far, the trust issue has been discussed in the HCI community broadly with
respect to usable privacy and security, risk and online trust, considering different
objects of trust, e.g., websites, companies, and individuals [3–7]. Furthermore,
trust has been discussed generally as a factor as a part of the user experience, e.g.,
[8, 9], and as part of trust evaluation strategies [10]. This work places the term
trust inside a particular case, a payment application for social media and links
it to identity management in terms of authentication and authorization. Parts
of this work have been presented at previous occasions: [11] briefly introduces

Trustworthy and Inclusive IDM 69
the PayShare application mentioned further below, and [12] extends the topic
with more details on the e-Me project and a detailed discussion of security and
privacy measures [13] of the PayShare application. The novel contribution of this
paper is an in-depth description of the accessible OpenID server and a thorough
discussion of trust aspects of the entire solution.
The work is structured as follows: After the problem description and a brief
overview of the provided solutions, trust considerations are discussed in detail.
Then, best practices for design for trust are presented, before giving a detailed
description of the OpenID server and the description of conducted user trials.
Finally, the conclusion is drawn.
This work is funded by the Research Council of Norway through the VER-
DIKT program under contract no. 201554.
2 Problem Analysis
The formulation of the objective and constraints of the prototypes was put for-
ward in the description of the Norwegian research project e-Me [2]. The solutions
should be suitable for real-life use, including the context of social media, be ap-
plicable to authentication and authorization likewise, they should be accessible
and offer a high degree of usability, and they should avoid to compromise privacy,
security [13], and to offend legal frameworks. Additional constraints regarding
the honoring of universal-design and legal frameworks are summarized in [14]
and [15], respectively.
Prior to any development, a number of key challenges was identified that had
been pointed out as open challenges in related research work [16]:
1. The majority of users is suffering from having to handle too many user names
and passwords for authentication.
2. Poor accessibility and usability compromise security and privacy and hereby
trust.
3. The majority of current authentication mechanisms is not accessible to users
with impairments.
4. Users have different requirements and preferences for privacy and security
in electronic solutions.
5. Users experience multiple authentication processes in case of frequent au-
thorization as cumbersome.
6. Authentication as used in social media can be applied to privacy and security
aware applications without a degradation of the level of security or privacy.
7. When using universal design rather than the legal framework or the estab-
lished traditions of information technology, and in designing for security and
privacy, their incorporation becomes substantially different both in terms of
legal compliance as well as on the level of software engineering, information
security engineering, and privacy by design.
– Provide security and privacy mechanisms that adapt to various skills;
70 T.H. Røssvoll and L. Fritsch
– Communication, such as privacy policies, instruction, and terms and

conditions must be understandable for persons with a variety of back-
grounds;
– Interactions should be rather deterministic, intuitive, and memorable;
– The security infrastructure should be open to special peripheral devices
such as audio readers, braille terminals, or interaction interfaces.
Many conflicts between the project’s goals turned up. The prioritization of
usability and inclusion over security methods created tension both on the
technical and on the legal level. By using particular flavors of accessible
IDM, at the same time security and DP might become weaker for the users
of these methods and the system as a whole. Prioritization between advanced
security policies and accessibility had to be done as well as the compilation of
a portfolio of alternative authentication methods with focus on accessibility,
not on equal security levels. As a consequence, the prototype can accommo-
date various skill levels, but at the price of very distinct levels of robustness
of the security measures. In turn, this approach will make the job of risk as-
sessment and security management of an information security management
systems, e.g. according to the ISO 27000 family, more challenging.
Answering these challenges, two main services were developed: An OpenID

provider, and an application to share bills among friends and to manage joint
payments inside social media, named PayShare. Combined, they have the fol-
lowing properties:
1. OpenID cuts down the numbers of service accounts to remember for the user.
In addition, the seamless authentication experience based on a persistent,
personally adapted authentication channel matching the users’ skills through
several services reduce complexity for the users.
2. Full-scale accessibility and a high degree of usability increase the user’s trust.
3. Improved accessibility by authentication adaptation in terms of several login
alternatives: password, pictures, sounds, pattern, personal questions, addi-
tional one-time password (two-factor over separate channel/SMS).
4. User defined threshold for the application of more frequent authentications.
5. Validity of a person’s authentication for a user defined time span.
6. OpenID as an authentication means to authorize payments in an financial
application inside a social medium.
In the subsequent sections, these contributions are discussed in detail.
3 Considerations Regarding User Trust

Trust is strongly influenced by a service’s security measures and privacy aware-
ness. Trustworthy online services require a certain standard regarding user identi-
fication, authentication, authorization, role management, and information
security and privacy [17]. This applies particularly to a financial application
with a virtual wallet like the bill sharing service PayShare, which can be treated
like an internet bank. As a precondition for trust creation, the security mecha-
nisms need to add to the total experience and not get into the way, as often felt
by users [18]. From our user groups we learned that identity management con-
cepts and especially authentication are the first major hindrance in participation
on social media and other services, as nearly all authentication channels poten-
tially exclude certain user groups [15]. We decided therefore to offer a variety of
alternative authentication channels.
The portfolio of authentication methods made the design of security and pri-
vacy properties more difficult. The major challenge is the variety of mechanisms
with distinct security properties and divergent privacy properties. Depending on
a user’s choice of mechanisms, the overall security, privacy, and trust framework
can differ as compared to other users, and other use cases. Mechanisms may
also be changed at any time due to a changing life situation or context. There-
fore, it can be difficult to establish a risk management system with static risk
assessment. On the other hand, when evaluating the total system security, the
strength of the weakest authentication channel must be assumed.
Further issues arise from the identity management unification. By using the
OpenID provider as a universal mechanism, PayShare establishes itself as a pow-
erful 3rd party with observation capabilities both towards the social media plat-
form and towards the payment system. In addition, an identification chain from
the social media to the PayShare application is established, which continues to
’friends’ and to an associated bank. Hence, pseudonymous social media partic-
ipation is no longer possible. Further issues in trustworthiness are found in the
underlying authentication mechanisms, where audio-visual information may leak
out to other persons, dependent on the usage situation. It appears that usability
and inclusion requirements are in strong tension with security, privacy, and data
protection regulation.
Concerning e-inclusion aspects, the requirement of universal design implies a
high degree of accessibility and usability of all involved parts of the solutions
[14]. The e-Me project considered target groups consisting of users with various
impairments, and elderly. Acknowledged impairments were cognitive challenges
such as dyslexia, dyscalculi, orientation, learning, and memory problems, sensory
challenges like sight and hearing reduction, and motor challenges like trembling
hands. Elderly users sometimes have a combination of impairments. However,
apart from these groups, the solutions were required to be universally designed,
i.e., to be be able to use by virtually all persons.
3.1 Design Measures for Trust

The following design recommendation have been developed and were applied
in PayShare to increase the user’s trust into the service. They are deliberately
held as generic as possible to make them applicable for electronic services in
general. They compound both security and privacy aspects, as well as usability
best practices.
– Require extra authorization before critical actions. Example: An extra check
box has to be marked before for instance claim deletion
– Show concise and comprehensive system messages that explain the general
context (what the user is about to do), the concrete task at hand, the require-
ments needed, and the concrete instructions. Ex.: Instructive messages like
“XY has sent you a payment claim”, “You are about to reject the claim”,
“You may need an OpenID address”, “Fill in all input fields below, then
press ..” are shown
– Show brief and comprehensive error messages with both concrete and gen-
eral help information, and directions to a human contact. Ex: “The system
couldn’t contact your OpenID provider. It could be caused by .. Please check
.. in your settings. If you do not succeed after several trials, please contact
.. (link to assistance)”
– Offer a dashboard view to ease overview gaining. Ex.: Showing the status
quo, history of events, link to Terms, and link to settings
– Offer multiple easy-to-find links to the profile settings
– Offer several easy-to-find links to Terms&Conditions
– Make all user settings non-mandatory. Ex.: It is not required by the user dur-
ing the service registration to specify the amount threshold for the additional
authentication; instead, the most secure default is chosen (here: additional
authentication is needed always), and the user can change this setting later
on
– Use safe defaults for all user profile settings
– Only expose particular profile settings on demand. Ex.: Link to the setting
“Change the authentication threshold” from the claim payment page, in
order for the user to get rid of the “An additional authorization is needed”
messages
– Offer several easy-to-find possibilities to delete the user/profile Ex.: Links to
the deletion of the user account from at least the settings and the Terms
– Make as many user actions as possible reversible. Ex.: Give particular events
a short time span where they can be reverted
– Offer multiple possibilities to delete user data. Ex.: It should be possible for
a user to delete all own user entered data
– Anonymize all user data that are impossible to delete. Ex.: In case of an
interest conflict regarding the deletion of data, undeletable data should be
anonymized. Data minimization is advisable.
– Only show information relevant in a specific situation. Ex.: Do not let the
user change the settings when nothing has be changed
– Offer an archive with previous events and actions, comparable to a system
log
– Offer a multitude of authentication methods. Ex.: Offer authentication that
accounts for sensor, motor, and cognition impairments.
– Hold the design of an OpenID server different from the design of the service
to illustrate the mechanisms invoked during an OpenID authentication
– Honor accessibility and usability standards. Ex.: As a minimum, follow the
HTML, CSS, WCAG 2, and WAI-ARIA recommendations.
– Run risk assessment concerning the actual strength of the authentication
mechanisms versus the value-at-risk in the connected applications. Loss of
personal information should be treated as one of the risks.
To wrap up, the user’s trust can first of all be created by empowering the user /
giving the user access. Second, user trust can be increased by adaptation, where
a service provider really “sees” the user and tailors the system according to
her needs and preferences. Third, trust can be increased by setting the user in
control in terms of informing the user about what is going on, letting the user
interact with the system (for verification purposes), and by making the service
as transparent, predictable, and reliable/credible as possible. Finally, trust can
be increased by smart user support, i.e., by helping the user in case of confusion,
insecurity, and system failure.
4 Verification of Considerations
As already mentioned, two prototypes were developed to verify the above con-
siderations: The social-media application PayShare (presented and discussed in
[12]), and an OpenID provider. The latter service is the key authentication party
and developed with a high degree of accessibility and usability. It is used for lo-
gin into the social media, registration with the payment application, and for the
authorization of payments. To cope with a variety of possible user impairments
and preferences, it offers authentication by means of six different login methods,
as illustrated in Figure 1 on the following page:
– Password memorizing,
– recognition a series of pictures,
– recognition of a series of sounds (see Figure 2 on page 75),
– pattern drawing,
– knowing the answer to a series of personal questions, and
– a PIN code calculator as a smartphone application.
A good password choice is supported by a password strength calculator detailing

the use of lower and capital letters, symbols, and digits. The visual choice is made
by picking a sequence (here: five) of pictures out of a set of pictures grouped in
categories, such as animals, clothes, and food. Upon login, five different sets of
images are presented to the user in sequence, who then has to identify the one
correct image from each set. The size of the set of elements (pictures) to pick
from and the number of elements the user has to chose can be set as a parame-
ter depending on considerations regarding the number of permutations (in case
the order matters) or combinations (when order does not matter) required by
a specific application. As an additional security mechanism, a user’s account is
suspended after four bad trials. The same principle applies to the audio choice,
though with sounds and music, and to personal question-answer pairs. With
the latter alternative, the user can either choose from the preset collection of
questions or formulate own questions. The pattern is drawn by mouse or key-
board on an 8x8 array of points, where between five and 34 points have to be
marked. Finally, the PIN code calculator presents a two-channel authentication
Fig. 1. Screenshot of six login alternatives shown during the OpenID registration
alternative in scenarios where this is required by the service provider (e.g., a

bank). It outputs a 6-digit PIN code, which also can be read out load to the
user, and which must be typed instead of the user’s password. The PIN code is
calculated based on a one-time password, which is sent to the phone from the
server, and a device identifier.
The user can fill additional data into the OpenID profile, such as birth date,
home address, etc. besides the mandatory full name and mail address. More im-
portantly, the prototype also shows that it is possible to personalize the profile
by a dedicated settings for the color scheme / contrast, which can easily be ex-
tended to other parameters such as font family, font size, and other parameters.
Upon authentication, a service may ask OpenID server for all parameters, but
it is up to the user to decide which parameters the OpenID server may share
with the application. In case a service supports the aforementioned accessibility
parameters, the service’s design is altered immediately according to the parame-
ter’s values when the user returns from the authentication process at the OpenID
server. This allows for a “specified once, personalize anywhere” strategy of elec-
tronic services. However, in turn, accumulated profile information in OpenID
attributes can reveal individual disabilities, and thereby create privacy issues.
Disability information might establish processing of medical information, which
is regulated strong with respect to data protection in many countries.
Fig. 2. Screenshot of sound based authentication
4.1 User Trials

In the user trials, eight participants tried out PayShare in combination with the
OpenID server. The participants were organized in four groups of two persons
and consisted mostly of elderly individuals, some with of them with visual im-
pairments, and some with dyslexia. The groups had first to generate an OpenID
account each before it was used for authentication and authorization with the
payment sharing application. In the beginning, they were instructed about the
situation (“you are NN, have been out eating with XY, you paid the entire bill,
and now you claim your money back”), and they were given the names of the
other virtual persons. All groups logged in to the given social networking site,
which looked very similar to Facebook to provide the proper setting, and after
that they were self-driven and only observed by the test leaders. One group filled
in a claim, the other groups were notified automatically and had to pay their
debt. This “game” quickly lead to new claims forth and back. Once the groups
were finished, the participants had to fill out a brief questionnaire.
The first finding is that the deployment of the OpenID server was well ac-
cepted, even though none of the participants was familiar with this concept.
Only two of the groups, those with sight impaired and dyslectic individuals,
chose images as authentication method (“useful”, “want that for my token cal-
culator”), the others went for the password alternative (“familiar”). Some of the
sight impaired tried out pattern authentication but soon gave up (“not accessible
for us”). It was also commented on the visibility of the pattern and images to
others as compared to the concealed password input field. This naturally applies
to audio and question-answer pairs as well.
The PayShare service was in general viewed as useful, but the majority had
restrictions to make when it comes to trust, mentioning that financial services
and social media were in their opinion not compatible. Some of the participants
read parts of the 2̃00-word Terms&Conditions, while nobody had a look at the
300-word privacy text. None of the participants visited their settings, underlining
the importance of secure and sensible defaults.
5 Conclusion
We presented a prototype for inclusive and trustworthy authentication and au-
thorization in the context of a bill sharing application in social media. The
OpenID server offers a high degree of universal design in terms of six different
login mechanisms, each of which based on different modalities. We also discussed
how security, privacy, and universal design can increase trust.
A number of key factors is vital to achieve a high degree of trust of the
user in the service: Accessibility, adaptation, usability, user control, information
availability, interaction, verifiability, transparency, predictability, and reliabil-
ity/credibility. Accessibility is the most crucial factor as it empowers users in
certain situations to use the respective service at all. The other factors increase
the feeling of control and thereby the user’s trust. The perception of increased
trust is not only applicable to users with impairments but rather all, as it is
widely recognized that e-inclusion measures for particular focus groups gener-
ally increase the service’s usability for everybody [19].
References
1. Johnson, M., Egelman, S., Bellovin, S.M.: Facebook and privacy: It’s complicated.
In: Symposium on Usable Privacy and Security (SOUPS), pp. 24–29 (2012)
2. e-Me Consortium: Inclusive Identity Management in New Social Media, VERDIKT
research project no. 201554, Research Council of Norway (2011),
http://www.nr.no/pages/dart/project_flyer_e-me
3. Hochheiser, H., Feng, J., Lazar, J.: Challenges in universally usable privacy and
security. In: Symposium on Usable Privacy and Security (SOUPS), vol. 2008 (2008)
4. Birge, C.: Enhancing research into usable privacy and security. In: Proceedings of
the 27th ACM International Conference on Design of Communication, pp. 221–226.
ACM (2009)
5. Dhamija, R., Dusseault, L.: The seven flaws of identity management: Usability and
security challenges. IEEE Security & Privacy 6(2), 24–29 (2008)
6. Karat, C., Brodie, C., Karat, J.: Usable privacy and security for personal informa-
tion management. Communications of the ACM 49(1), 56–57 (2006)
7. Cranor, L.F., Garfinkel, S.: Security and Usablity: Designing secure systems that
people can use. Theory in practice. O’Reilly, Sebastopol (2005)
8. Schade, A., Nielsen, J.: Trust and Credibility, 2nd edn. E-Commerce User Experi-
ence, vol. 9. Nielsen Norman Group (2000)
9. Corbitt, B., Thanasankit, T., Yi, H.: Trust and e-commerce: a study of consumer
perceptions. Electronic Commerce Research and Applications 2(3), 203–215 (2003)
10. Fritsch, L., Groven, A.-K., Schulz, T.: On the Internet of Things, Trust is Relative
(chapter 9). In: Wichert, R., Van Laerhoven, K., Gelissen, J. (eds.) AmI 2011.
CCIS, vol. 277, pp. 267–273. Springer, Heidelberg (2012)
11. Røssvoll, T.H.: Trust implications for universal design of social-networking appli-
cations. In: User-Centered Trust in Interactive Systems Workshop at NordiCHI
(2012)
12. Røssvoll, T.H., Fritsch, L.: Reducing the user burden of identity management:
A prototype based case study for a social-media payment application. In: Sixth
International Conference on Advances in Computer-Human Interactions, ACHI
(2013)
13. Fritsch, L.: Social media, e-id and privacy - background for the e-me project.
Technical Report DART/02/2011, Norsk Regnesentral (2011)
14. Fuglerud, K.S.: Universal design in ICT services, Trondheim, Norway, 244–267
(2009)
15. Fritsch, L., Fuglerud, K.S., Solheim, I.: Towards inclusive identity management.
Identity in the Information Society 3(3), 515–538 (2010)
16. Fuglerud, K.S., Røssvoll, T.H.: Usability and accessibility of personal identification
management systems in electronic services. In: Proceedings of eChallenges-2011.
IIMC International Information Management Corporation Ltd., Florence (2011)
17. Fritsch, L.: Privacy visualization requirements in the internet of things - a utrustit
fp7 ict project note. Technical report, Norsk Regnesentral (Norwegian Computing
Center) (2012)
18. Adams, A., Sasse, M.A.: Users are not the enemy: Why users compromize computer
security mechanisms and how to take remedial measure. Commun. ACM 42(12),
41–46 (1999)
19. Huber, W., Vitouch, P.: Usability and accessibility on the internet: Effects of ac-
cessible web design on usability. In: Miesenberger, K., Klaus, J., Zagler, W.L.,
Karshmer, A.I. (eds.) ICCHP 2008. LNCS, vol. 5105, pp. 482–489. Springer, Hei-
delberg (2008)
Towards a Design Guideline of Visual Cryptography
on Stereoscopic Displays
Shih-Lung Tsai1 and Chao-Hua Wen2

1
Chunghwa Picture Tubes LTD., Taoyuan, 334, Taiwan
[email protected]
2
Graduate Institute of Color and Illumination Technology,
National Taiwan University of Science and Technology, Taipei, 10607, Taiwan
[email protected]
Abstract. This paper proposed a new visual cryptography scheme with the ste-
reoscopic display which showed and accurately decrypted the hidden informa-
tion for gray images. Results indicated that contrast ratio and pixel disparity of
the decrypted stereo-image were key problems that would impact on the
perceived quality of the decrypted image. Next, this research performed a sub-
jective experiment of shifting pixels between both of left and right images to
investigate the disparity effects of decrypted information on a full HD stereo-
display with film-pattern-retarder technology. In addition, the effects of font
size and contrast ratio were addressed as well. Results revealed that the thre-
sholds of pixel disparity were between 2 and 7 pixels. To alphabets, the font
size of 50 points was lower boundary to show the decrypted information. To
numeric, the font size of 45 points was lower boundary over different contrast
ratios.
Keywords: Visual cryptography, Stereoscopic display, Design guideline.
1 Introduction
With the advanced development of science and network technologies, multimedia

information is usually delivered over the Internet conveniently. While using secure
images, the importance of secure issues cannot be overemphasized because hackers
can easy to steal secure information which they need in the Internet. Two popular
studies are considered, in general, information hiding is categorized in Watermarking
and secure sharing as Visual Cryptography (VC).
Visual Cryptography was proposed firstly by Naor and Shamir [1]. Information of
an original image is separated to two or more images. And people just only stacked all
encrypted images to decrypt information of original image with Human Visual Sys-
tem. In order to improve the quality of decrypted image, there are many algorithms
have been developed to improve the encryption for gray or color images. There are
various measures on the performance of many kinds of VC schemes, but rare studies
on exact stereo-reproduction in visual cryptography. This study provided a method to
Towards a Design Guideline of Visual Cryptography on Stereoscopic Displays 79
amend VC of encryption side and decryption side. Through a series of experiments,

visibility requirements of the decrypted information were addressed.
2 Method
Noar and Shamir proposed the (k, n) threshold scheme or k out of n threshold scheme
which illustrated a new paradigm in image sharing [1]. In this scheme a secrete image
is divided into n share images. With any k of the n shares the secret can be perfectly
reconstructed, while even complete knowledge of (k-1) shares reveals no information
about the secret image.
This research is expected to answer two questions, 1) what is reasonable disparity
in pixels for the decrypted secure image with wearing 3D glasses and 2) what is the
limited viewing condition to avert data revealing without wearing 3D glasses. Follow-
ing sections will describe the basic VC scheme, stereoscopic visual cryptography for
gray images and experimental design.
2.1 Basic VC Scheme for Binary Images
The (2, 2) VCS is illustrated to introduce the basic concepts of threshold visual secret
sharing schemes as shown in Fig. 1. During the encryption, every secret pixel is split
into two shares, and each share belongs to the corresponding share image. For in-
stance, a secret white pixel, the dealer randomly chooses one of the first two rows of
Fig. 1 to encode Share1 and Share2. The possibilities of the two encoding cases are
equally likely to occur independently whether the original pixel is black or white.
Thus, neither Share1 nor Share2 exposes any clue about the binary color of pixel. In
the decryption, the two corresponding shares are stacked together, using OR/AND, to
recover the secret pixel. Two shares of a white pixel are of the same while those of a
black pixel are complementary. Consequently a white pixel is recovered by a share
with the stacked result of half white sub-pixels and a black pixel is recovered by all
black. Using this basic scheme, the contrast ratio of the decrypted image is reduced
results from halving intensity of the white secret pixels.
Pixel Probability Share 1 Share 2 Result
1
p= ⊗ =
2
1
p= ⊗ =
2
1
p= ⊗ =
2
1
p= ⊗ =
2
Fig. 1. Encoding scheme for a binary pixel into two shares

80 S.-L. Tsai and C.-H. Wen
2.2 Stereoscopic Visual Cryptography for Gray Images

First we used error diffusion to decompose the gray-scale secret image. And then the
halftoning resultant image was processed by the encryption of a basic visual crypto-
graphy scheme [2-3]. A 23” LG® D2342P monitor with TriDef® 3D supports was
used in this study, the odd rows display the left view and the even rows display the
right view in 3D mode. Therefore, it is necessary to rearrange the two encryption
share images to match the display format as decrypting the secret images. The VC
algorithm for side-by-side file format of stereo polarized displays is described as
follows.
Encryption and decryption. Here uses an example of a full HD binary image

(1920×1080) to simplify the encryption procedure. First step is to down scale the
input binary image. Quarter down-scaling is performed in horizontal direction, e.g.,
1920×1080 pixels to 480×1080 pixels. Second is to create the encrypted shares. Based
on the basic binary VC, we apply diagonal shares to build a pair of Share1 and
Share2. Due to the basic VC scheme, the image size of the Share1 and Share2 are
extend to double in horizontal and vertical direction, e.g., 960×2160. After then, we
take out the odd row of Share1 and Share2. Third is to rearrange the configuration of
Share2. It found that the monitor with TriDef® 3D deals with the odd rows of Share1
to the left eye and the even rows of Share2 to the right eye. For next stage of decryp-
tion, it is necessary to switch Share2’s the first row to the second row, the third row to
the fourth row, and so forth. Fourth is to combine both share images (960×1080) into
a full HD encrypted image with side by side format, Share1 is on left side and Share2
is on right side. Fig. 2 illustrates the framework of the stereo visual cryptography
scheme for stereoscopic displays in this article. In decryption stage, the encrypted
side-by-side image can directly be played on the monitor in 3D mode, and then the
secret image can be decrypted clearly without wear glasses. Fig. 3(a)-(c) shows an
original secret image, an encrypted side-by-side image and a simulated resultant by
our stereoscopic visual cryptography scheme respectively.
Secret image
Quarter down-scaling in horizontal
Visual Cryptography
Share1 Share2
Registration
Combination to side-by-side format
Decryption on stereo-displays
Fig. 2. Framework of the stereo visual cryptography scheme for stereoscopic displays
Towards a Design Guid
deline of Visual Cryptography on Stereoscopic Displays 81
(a) secret image (b) two encryption shares (c) simulated decryption 3D
(Left/Right format) image
Fig. 3. An
n example of stereoscopic visual cryptography
Changing disparity (pixeel shift). For changing disparity, we just move pixelss in
horizontal of each share with
w same shift but opposite directions. It is known off in-
creasing disparity can enhaance depth perception cues. That is the main reason of de-
crypted secure image with h wear 3D glasses and limited viewing condition to avvert
data revealing. In this work ks, it is interesting about how many amount of disparity is a
suitable setting for VC. This study investigated six different shift pixels on disparrity,
such as 0, 4, 8, 16, 24 and 32 pixels. For example, Share1 (left image) is shifted six-
teen pixels on the right and d Share2 (right image) is shifted sixteen pixels on the lleft,
so the total disparity is thirtty-two pixels for side by side image. As the same way, we
can finish six disparities forr encryption side.
Minimum contrast ratio forf binary VC. In this article, we define contrast ratio us-
ing Michelson contrast acco
ording to the measurements during display profile creattion
®
via Eye One Pro by X-Ritee . The Michelson contrast is commonly used for patteerns
where both bright and dark features are equivalent and take up similar fractions off the
area. The Michelson contrast ratio CR is defined as Equation (1),
CR = ( I max − I min ) /( I max + I min ) (1)
Here Imax and Imin represen nting the highest and dark luminance. The denominaator
represents twice the average of the luminance. The luminance of white pixel and
black pixel were obtained from the color profile. The gamma values of red, greeen,
blue were 2.11, 2.25 and 2.28 respectively. In this article, there are five conttrast
ratios with gray level are in
nvestigated, including 100% (Imax = 1; Imin = 0), 98% (Immax =
0.791; Imin = 0.006), 91% (Imax = 0.609; Imin = 0.028), 73% (Imax = 0.452; Imiin =
0.0691), and 42% (Imax = 0.321; Imin = 0.13).
2.3 Experimental Desig

gn
Participants. Seven male anda six female students between 22 and 26 years of agee in
National Taiwan University y of Science and Technology were participated in the exxpe-
riment. All of them had a normal
n or corrected to normal visual acuity and passed the
stereopsis test according to RANDOT® Stereotests.
Stimuli. Five contents (1920 x 1080 pixels) included three logo icons, a series of
Cambria numbers with varied sizes range from 20 to 110 points, and a series of Times
New Roman alphabets with varied sizes range from 100 points to 10 points as shown
in Fig. 4. At decryption stage, this study investigated six different shift pixels of
disparity, such as 0, 4, 8, 16, 24 and 32 pixels. We used Michelson contrast ratio ac-
cording to the measurements as making the display profile. There were five levels of
contrast ratio 100%, 98%, 91%, 73% and 42% respectively. Therefore, there were a
total of 150 secret images as visual stimuli in this article.
Experimental task and procedure. The experimental task was a visual detection
task. The ambient condition is about average 700 lux. Viewing distance was 50 cm to
70 cm from the display dependent on observer’s comfortable viewing position. This
study asked participant to watch the contents started from disparity 0 to 32 pixels in
five levels of different contrasts from 42% to 100%, then to answer whether perceive
the depth clearly or not. At the beginning of the experiment visual acuity and stereop-
sis were assessed. The experimental procedures are listed as follows.
1. To adjust the monitor set and posture to allow view images at the optimal viewing
angle as comfortable as possible;
2. To show a series of a kind of a stimuli on the stereodisplay began to perceive dis-
parity can ascent from 0 pixels to 32 pixels until the subjects do not perceive depth
of the original image;
3. To show contents began to perceive disparity can descent 32 pixels to 0 pixels until
the subjects can perceive depth of the original image;
4. To record the disparity value, disparity range and calculate the threshold of dispari-
ty by font size and contrast ratio;
5. After each stimulus content, a short break could be taken in which the experimen-
ter changed the condition and prepared the experimental software and displays for
the next steps. All experimental lasted approximately 90 min.
Heart icon NTUST logo Smile icon Numbers Alphabet
Fig. 4. Five stimulus contents
3 Results and Discussion
In the following sections, the results about disparity threshold and contrast threshold
for 5 contents will be presented, followed by the discussion of relationships between
CR and font size.
Towards a Design Guid
deline of Visual Cryptography on Stereoscopic Displays 83
3.1 Disparity Threshold and Contrast Threshold

Here demonstrates the resu ults of disparity threshold which is the acceptable level or
range for applications of stereoscopic
s visual cryptography. Cormack et al. useed a
random-dot stereogram and d found a cube-root dependency of stereoacuity on conttrast
o contrast [4]. And the difference is probably due to the
at superathreshold levels of
different spatial-frequency content of the stimuli used in the experiment. Therefoore,
this study reexamined and also found the disparity threshold was an important coom-
parison among different typpes of content and contrast.
onship between contrast ratio and disparity threshold. Re-
Fig. 5 depicts the relatio
sults revealed that the higheest disparity threshold was occurred between CR 91% and
98% and the lowest dispariity threshold occurred at CR 42%. Additionally, we confi-
dently concluded that dispaarity threshold range between 2 and 7 pixels over differrent
contrast ratio. Due to shiftt two pixels with Share1 and Share2, we cannot see any
information on the stereo--display without glasses. Therefore, we need to take out
disparity threshold two pixxels downward. Fig. 6 illustrates the relationships betwween
different contents and dispparity threshold. In evidence, here is the highest dispaarity
threshold in the smile imag ge and the lowest disparity threshold in the Alphabet and
Numbers image.
12 95% C.I. 16 95% C.I.

Mean disparity, pixel
14
Mean disparity, pixel
10 12
8 10
8
6 6
4
4
2
2 0
0
42% 73% 91% 98% 100%
Contrast ratio Stimulus content
Fig. 5. Plot of the mean disparrity threshold as Fig. 6. Mean disparity pixel for 5 stim
muli
a function of the contrast ratio contents
3.2 Relationship betweeen Contrast Ratio and Font Size

Fig. 7 reveals the relationsh
hip between font size and contrast ratio in the alphabet ssize
and number character. The alphabet size is blue line with diamond marks; here is the
highest font in the 91% con ntrast ratio and the lowest font in the 100% contrast raatio.
The number character is reed line with square marks; here is the highest font in the
91% contrast ratio and the lowest
l font in the 100% contrast ratio.
This experiment tried to find out the disparity threshold of contrast ratio for difffer-
ent content. Results indicatted that disparity threshold range is between 2 and 7 pixxels
over different contrast rattios. For the minimum requirement of alphabet size for
decryption, the font size is 50 points upward (visual angle is about 0.95 degree) in the
different contrast ratio. For number characters, the font size is 45 points upward (vis-
ual angle is around 0.85 degree) in the different contrast ratio.
70 Alphabet Number
60
Mean font size (pt)
50
40
30
20
10
0
42% 73% 91% 98% 100%
Contrast ratio
Fig. 7. Mean font size (pt) for different contrast ratio
4 Conclusions
This study proposed a new visual cryptography scheme with the stereo polarized dis-
play which can rendering decrypted stereo gray images accurately. For stereo secret
images, this method only uses two encryption share images and the decryption can be
performed via a side-by-side file format of 3D displays.
Hereby the future works are summarized as follows. First, it is obvious that a lot of
time and effort have been dedicated to visual secret sharing using visual cryptogra-
phy. The trends have been identified within visual cryptography [5]: contrast im-
provement, share size improvement, wider range of suitable image types, efficiency
of VC schemes, ability to share multiple secrets. Finally, because emerging 3D TV, it
needs more research on how to use different stereoscopic display with optimal dispar-
ity to control suitable and comfortable content protruding.
References
1. Naor, M., Shamir, A.: Visual cryptography. In: De Santis, A. (ed.) EUROCRYPT 1994.
LNCS, vol. 950, pp. 1–12. Springer, Heidelberg (1995)
2. Lin, C.C., Tsai, W.H.: Visual cryptography for gray-level images by dithering techniques.
Pattern Recognition Letters 24(3), 349–358 (2003)
3. Wen, C.H.: Visual Cryptography on Color Video Displays. In: The 17th International Dis-
play Workshop, Fukuoka, Japan, pp. 2105–2108 (2010)
4. Cormack, L.K., Stevenson, S.B., Schor, C.M.: Interocular correlation, luminance contrast
and cyclopean processing. Vision Research 31, 2195–2207 (1991)
5. Weir, J., Yan, W.Q.: A comprehensive study of visual cryptography. Springer Transactions
on DHMMS 5, 1–10 (2010)
Part II
User Studies
Interface Design for Minimizing Loss of Context
in In-situ Remote Robot Control
Jong-gil Ahn and Gerard Jounghyun Kim
Digital Experience Laboratory,

Korea University, Seoul, Korea
{hide989,gjkim}@korea.ac.kr
Abstract. When controlling robot in-situ, the operator’s attention is often di-
vided between the scene and the interface. This often causes inefficiency in the
control performance. One possible solution to address this attention switch is to
employ a camera (or sensor) view (despite being at the site) in which critical
parts of the operating environment can be shown side-by-side with the control
interface so that the user is not distracted from the either. In addition, when the
user switches one’s attention away unavoidably and then back to the control in-
terface, the interface can be configured so that user can easily continue the task
at hand without momentary the loss of context. In this paper, we describe the
design of such an interface and investigate in the possible user attentive beha-
viors based on it. In particular, we present an experiment that compares three
variant forms of interactions: (1) Nominal (no camera view), (2) Fixed (using a
camera view and user not allowed to overlook into the scene), and (3) Free
(using a camera view but user is free to overlook into the scene). The three ap-
proaches represent different balances between information availability, inter-
face accessibility and the amount of attentional shift. Experiment results have
shown that all three interaction models exhibited similar task performance even
though the Fixed type induced much less attentional shift. However, the users
much preferred the Nominal and Free type. Users mostly ignored the camera
view, shifting one’s attention excessively into the operating scene, due to the
lack of visual quality, realistic scale and depth information of the camera view.
Keywords: Human-Robot Interaction, In-Situ Robot Control, Interface Design.
1 Introduction
In-situ remote robot control refers to a situation in which a robot is remotely con-
trolled using an interface device from a near distance (e.g. 2~20m). In-situ remote
control often occurs in situations where it is important for the operator to be present at
the operating site (e.g. safety reasons, requirement of human expertise, need for effec-
tive communication and interaction, slave robots with only limited secondary roles) or
when on-board or environment sensors are either limiting or not readily available for
complete situation monitoring. Typical examples include bomb diffusing robots, fire-
fighting (water gun) robots and robot secretaries.
88 J.-g. Ahn and G.J. Kim
Fig. 1. The three forms of interaction for in-situ robot control which uses the task/scenario
based hand-held interface: (1) Nominal (without camera view), (2) Fixed (using a camera view
and users not allowed to look over into the operating scene) and (3) Free (using a camera view
and but users free to look over into the operating scene).
This is a situation distinct from both close/direct interaction (that e.g. uses voice,
touch screen input or gesture) and completely sensor-based tele-operation. Like any
interfaces, important factors in the efficient and convenient in-situ remote robot con-
trol would include the availability of required information to make the right decision
(or situation awareness), access time to the interface for the needed actions and the
amount of any distraction or attentional shift between the robot scene and the control
interface.
One possible solution to address these factors collectively is to use a hand-held de-
vice (such as smart phones and iPad like devices) and at the same time provide a
camera/sensor view (from the hand-held device) in which critical parts of the operat-
ing environment can be shown side-by-side with the control interface. This way, the
user can minimize the times they have to look away (because the camera/sensor view
supplies the needed information) from the interface and see directly into the scene,
and thus react faster to make proper actions. The user may still be allowed to look
over and make direct visual confirmation, if necessary. Thus a provision is still
needed to minimize inefficiencies arising from this occasional switching of attention
and context.
Note that robots are becoming more versatile and may even require multitasking,
and as such also more difficult to control. In addition, hand-held media devices have
emerged as attractive and flexible platforms for remote control for their wide popular-
ity and ever-increasing computational and multimedia capabilities [4, 16].
Recently, there has been great amount of research in the field of Human Robot Inte-
raction (HRI), however, concentrated mostly in either direct interaction methods or
for tele-operated robots. On the other hand, relatively little research has been focused
on remote control interface design for on-site robots. Thus, there is not a clear guide-
line as how to properly integrate (or if one should), e.g. the camera view and organize
the hand-held interface for the best performance and highest usability.
In this paper, we have designed an interface for hand-held in-situ remote robot con-
trol that is organized around the task sequence and reconfigured continuously accor-
dingly so that user can be aided in maintaining the task context, and that provides a
camera/sensor view to minimize the attentional switch in the first place. Furthermore,
based on the interface, we compare three forms of interaction to study user attentive
Interface Design for Minimizing Loss of Context in In-situ Remote Robot Control 89
behavior. The three compared interactions represent different balances between in-
formation availability, interface accessibility and the amount of attentional shift.
This paper is organized follows. We first shortly review other related research.
Then, we describe the task-oriented reconfigurable hand-held robot interface with the
camera/sensor view. Section 4 will describe the details and report the results from the
user behavior experiment. Finally, we will conclude the paper with a discussion of the
results, a summary and directions for future work.
2 Related Works
2.1 Human-Robot Interaction

With the proliferation of robots, HRI has generated lots of interests among researchers
lately. Several guidelines and evaluation metrics have emerged for HRI from the HCI
and robotics communities over the years [1, 3, 6, 9, 13, 17]. Most guidelines are quite
abstract and general, and also seem to be oriented toward “direct and close” interac-
tion [2, 8, 11] or tele-operation control [10, 12, 14]. Despite relatively less number of
studies devoted to in-situ robot control and interaction, many such principles do apply
equally. One such guideline is to provide good awareness and less cognitive load [3].
More specifically, Clarkson et al. [1] maintained that natural cues may not be neces-
sary and clear provision of relevant interactive information was more important.
On the other hand, the most typical form of an “indirect” communication with a
robot would be through the use of a remote control. Again, we make a note that the
indirect approach has been used mostly for tele-robotics or multi-robot control
rather than for in-situ bystander driven control. Therefore, the ordinary remote control
tends to be small/compact, one way, function-oriented and difficult to use (e.g. but-
ton/code/mode based) [5, 7].
2.2 Human-Robot Interface

Based on these HRI principles, guidelines and other requirements, many different
forms of human-robot interfaces have been developed [3, 4, 5, 20, 23]. In particular,
interface design becomes even more important for efficient control of or interaction
with multi-robots or robots with multi-tasks (e.g. navigation, manipulation, interac-
tion, etc.) [19, 21, 25]. In this context, traditional interface devices such as joysticks,
joypads, wheels, switches and buttons are being replaced with a new breed that offers
rich visual feedback and mobility (e.g. smart phones, tablet PC [26]), gestural input
(e.g. Nintendo Wii [22]), and tangibility to compensate for the operator’s limited
working memory and induce more natural and easier interaction. Among the afore-
mentioned features, visual feedback (usually available through camera views) is one
of the most important factors in situation awareness and efficient tele-operation [24].
Zhu et al. studied the tele-operation and multi-tasking performance with regards to
different camera views and view control methods [19]. Nielson and Goodrich com-
pared navigation performance between when using camera and map views [23].
3 Interface Design
The robot to be controlled in-situ was originally designed for operator led guidance
tasks (e.g. at museums). It was based on a wheeled mobile robot (Adept Pioneer 3-
DX) and mounted with a computer, pan-tilt camera, speaker, ultrasonic sensors, and a
panning projector (Figure 2). We envisioned that the robot would be operated by a
human guide for series of sequential/concurrent tasks involving individual subtasks
such as aiming the projector to the wall, selecting and projecting information, control-
ling the camera to find landmarks (for spatial registration and localization), play-
ing/stopping narrations, positioning the robot, and navigating.
Fig. 2. The museum guidance robot (left) and the scenario based interface (right). The sequence
of tasks appear as buttons for the operator to simply activate (under normal circumstances). The
tasks are logically grouped. The Camera/Sensor view is in the lower middle (camera remained
blank for the Nominal condition).
Since there were relatively many tasks/subtasks to be managed and accomplished

as quickly as possible and in order to aid the user quickly regain one’s operating con-
text between the occasional view switches, we developed a “scenario” flow based
interface. It lays out, in the “Task Sequence View” (top part of the display in Figure
2), pre-configured buttons corresponding to a sequence of tasks to be accomplished
for the operator to simply activate. For example, in Figure 2, the eight buttons in the
top horizontal line prescribe “Move (robot) forward 1m,” “Rotate (robot) 90 degrees
to the right,” “Turn on projector,” “Play narration,” and so on. The layout is confi-
gured by interpreting an XML based task specification file. The buttons were color-
coded or highlighted to e.g. indicate their status (currently pending, already finished)
and types (navigation, camera control, projector control, etc.). Therefore, if the robot
was to assist a human guide e.g. in a corridor with a number of museum artifacts at
known locations, a complete and typical operational scenario can be prescribed and
reflected into the interface automatically. Normally, the user would simply use these
pre-configured scenario-based buttons to accomplish the given task. In addition the
task sequence view interface was used as a way to remind the user what to do without
having to memorize the order and specifics of the experimental tasks.
However, note that it was not possible to automatically infer whether a certain task,
despite being activated by the operator, really was carried out successfully or satisfac-
torily. For example, even if a “Move forward 1m” command was issued and executed,
the robot may fail to accomplish it due mechanical slippage. In fact, there may be
situations that deviate from the normal scenario flow (e.g. new obstacle to avoid). In
such a case, the user would need a way to go around the scenario based interface and
directly access the various task control interfaces (located in the lower left parts of the
display).
The Camera/Sensor view (lower middle part of the display) simply showed the
video imagery captured by the camera located in the back of the interface facing
toward the operating environment. Note that the camera view was not used for the
Nominal interaction model (switched off and remained blank).
The interface was implemented on an Apple iPad [18] which communicated the
user commands (with screen touch to various control buttons) to the on-board com-
puter which in turn controlled the robot (movement), projector (on/off, panning),
camera (pan and tilt), and voice output. In informal trials, the scenario based interface
vastly improved the performance and user satisfaction over the naively designed
“bare” interface (see Figure 3). No formal experiment was further conducted as this
result was quite expected and obvious.
Fig. 3. The naively designed original “bare” interface with simple control buttons
4 Experiment
4.1 Experimental Design and Three Interaction Models

The objective of the experiment was to study the user behavior model in regards to
one’s shifting of attention between the interface and the actual scene despite the
provision of scene view through the camera. Thus we tested three representative inte-
raction models (also see Figure 1) in terms of how the camera view may be utilized
(or not).
• “Nominal” type: The Camera view is not provided and the user inspects the operat-
ing scene directly.
• “Fixed” type: The hand-held device always aimed toward the target, and the cor-
responding Camera view is provided. The user is basically not allowed to look over
into the operating scene unless absolutely necessary. This represents the usual
desktop based tele-operation.
• “Free” type: The user is free to either aim the hand-held device into the scene and
use the Camera view or directly look into the operating scene.
The experiment was designed as a 3 x 1 repeated measure within subject, the lone
factor being the type of the interface employed. A series of robot control task (details
explained in the next section) was given to a user to be carried out by using the given
interaction model. The task performance (task completion time and number of clicks
made to the interface) was measured and a usability survey was filled out, with a par-
ticular focus on the matter of attentional shift. More procedural details are given in the
subsequent section.
Table 1. An example sequence of 3 composite tasks to be completed in an experimental session

and the experimental space for the user to carry out the sequences of tasks
Sequence Composite Task Description

Move forward to near Placemark 1
Aim the camera to a marker on the wall
1
Turn on the projector / narration (wait)
Turn off the projector
Move forward to Placemark 2 and
avoid the obstacle in the way
2
Turn on the projector / narration (wait)
Turn off the projector
Move forward to Placemark 3
Rotate the robot 45 degrees to the right
3 Rotate the projector 45 degrees to the left
Turn on the projector / narration
Move forward to finish line
4.2 Task
For a given interaction model and the associated interface (given in a balanced order),
a series of 3 composite tasks was asked to be completed as fast as possible. An exam-
ple sequence of the tasks is shown in Table 1 to be carried out in the experimental
space shown in Figure 4. To avoid any learning effects as much as possible, the tasks
were differentiated between different sessions with a similar degree of difficulty.
Interface Design for Minim
mizing Loss of Context in In-situ Remote Robot Control 93
Fig. 4. The experimental sp

pace for the user to carry out the sequences of tasks in Table 1
4.3 Detailed Proceduree

Eight paid subjects (all meen) participated in the experiment with the mean agee of
29.5. After collecting one’’s basic background information, the subject was brieefed
about the purpose of the experiment
e and instructions for the experimental taskk. A
short training was given fo or the subject to get familiarized to the interface and the
tasks to be completed.
Each subject went throu ugh 9 sessions (3 interfaces types x 3 repeat) lasting for
about one hour. Upon startting the experimental session, the subject was asked to use
the interface from a fixed position
p to carry out the task as specified and guided by the
Task Sequence View interfaace. A particular interface type (Nominal, Fixed, Free) w was
presented to each subject in
n a balanced fashion.
The task complete time and
a number of clicks were measured. In addition, the nuum-
ber of times of attentional shift
s was counted manually (e.g. looking or glancing aw way
from the interface into thee operating environment or vice versa). After the subjject
finished the three sessions, they were asked to fill out the aforementioned survey.
4.4 Results and Discusssion

Figures 5, 6 show the taskk performance, usability (ease of use and fatigue) and the
amounts of shifts in the linne of sight. Overall there were no statistically significcant
mpletion nor in the number of clicks made, even though the
differences in the task com
user made much less (with h statistically significant difference) attentional shifts w
with
the Fixed type. There was not any significant difference in the number of shifts be-
tween the Nominal and Freee type. Most users treated the Nominal and Free type too be
mostly equivalent.
K
We can deduce that the visual

v information from the direct look into the scene ccon-
tributed much more to com mpleting the task (which were mostly spatial in nature) inn an
efficient manner (even wh hen done excessively). Subjects reported, when using the
Fixed type, the need to look k over to gain more insight into the situation such as prooper
depth and scale information n (which could not be conveyed well with a small cam mera
view). Note that the users ofo the Fixed type was driven by the experiment to minim mize
attentional shift, yet it did not improve the task performance measures. It is interrest-
ing, despite what the userss felt or preferred, similar task performance was achieved
with the Fixed type. This wasw most apparent when the users were trying to avoidd an
obstacle with the robot.
Attentional shifts occurrred much less during relatively stable/simple task opeera-
tions (e.g. while the robot was moving with sufficient clear room, turning on/off the
projector, turning on the naarration). Subjects also had a great tendency to make vissual
confirmation of the task co ompletion (by directly looking into the operating envirron-
ment) and whether one follo owed the right task in the given sequence (by looking bback
into the interface). This gen nerally caused excessive shifting of the line of sight. U
Usa-
bility-wise, subjects mostly y grouped the Nominal and Free type together as beeing
more comfortable, easier to o use and less tiring than the Fixed type. This, in summaary,
the results indirectly indicate that the provision of the “high fidelity global” vview
is very important. For the Free type, the overhead in attentional shifts was comppen-
sated by the benefit of the global view for control performance, and reversely for the
Fixed type.
Fig. 5. Task performance reesults: task completion time (left) and no. of clicks made (rightt)
Fig. 6. Subjective ratings(easse of use (left) and fatigue (middle)) and number of attentioonal
shifts(right)
5 Conclusion
In this paper, we have presented a hand-held interface design considering the frequent
attentional shift and concern for loss of operational context. We also ran an experi-
ment to investigate the user behavior with regards to the use of camera/sensor view.
Our expectation that the attentional shift would play an important and significant role
in the efficient and fast task performance was rejected. Despite the much less number
of visual shifts, somewhat forced with the Fixed type model, the overall performance
was not different from the others. Users much preferred the Nominal and Free type
and mostly ignored the camera view despite having to shift one’s attention excessive-
ly, due to the lack of visual quality, realistic scale and depth information (which were
deemed essential in carrying out the mostly spatial robotic tasks). The study illustrates
the need to provide a high fidelity global view into the scene. For instance, in a re-
mote tele-operation situation, if a global view can be somehow provided (e.g. a third-
person point of view camera), the control performance can be improved. We believe
this result can serve as an important and useful guideline for designing control inter-
faces not only for in-situ robot control, but also for augmented reality and tele-
presence systems. It reaffirms the importance of scale, depth perception and realism
in robotic control.
Despite the ineffectiveness of the camera view in this work, we posit that cam-
era/sensor view that provides information that is not obtainable by the naked eyes will
still be important. Different results might be obtained if the camera view had better
quality e.g. higher resolution, stereoscopic rendering and larger size. Note that
this study focused on the user’s choice when both camera and direct imagery were
available. Further experiments are needed to investigate the effects toward carrying
out non-spatial tasks or toward the quality of the task accomplishment.
Acknowledgement. This work was supported by the National Research Foundation

of Korea(NRF) grant funded by the Korea government(MEST)(No 2012-0009232).
References
1. Clarkson, E., Arkin, R.: Applying Heuristic Evaluation to Human-Robot Interaction Sys-
tems. In: Proc. of FLAIRS (2007)
2. Drury, J., Scholtz, J., Yanco, H.: Awareness in Human-Robot Interactions. In: IEEE Inter-
national Conference on Systems, Man and Cybernetics, Washington, DC (October 2003)
3. Drury, J., Hestand, D., Yanco, H., Scholtz, J.: Design Guidelines for Improved Human-
Robot Interaction. In: Proc. of ACM CHI (2004)
4. Muller, F.: iPhone-Robot Command Interface: A Case Study. Semester-Thesis, Swiss Fed-
eral Inst. of Technology Zurich (2009)
5. Fong, T., Thorpe, C., Glass, B.: PdaDriver: A Handheld System for Remote Driving. In:
IEEE International Conference on Advanced Robotics (2003)
6. Goodrich, M., Olsen, D.: Seven Principles of Efficient Human Robot Interaction. In: IEEE
International Conference on Systems, Man, and Cybernetics, pp. 3943–3948 (2003)
7. Hruschak, P.: DIY: Make a Roomba Cell Phone Remote Control (2010),
http://www.gadgetell.com/2006/09/
diy-make-a-roomba-cell-phone-remote-control/
8. Nourbakhsh, I.: Course Notes for Principles of Human Robot Interaction. Unpublished
Work (2007)
9. Olsen Jr., D., Goodrich, M.: Metrics for Evaluating Human-Robot Interactions. In: Proc. of
PERMIS (2003)
10. Paulos, E., Canny, J.: Ubiquitous Tele-embodiment: Applications and Implications. Inter-
national Journal of Human-Computer Studies 46(6), 861–877 (1997)
11. Quigley, M., Goodrich, M.A., Beard, R.W.: Semi-autonomous Human-UAV Interfaces for
Fixed-wing mini-UAVs. In: Proc. of IROS (2004)
12. Scholtz, J., Bahrami, S.: Human-Robot Interaction: Development of an Evaluation Metho-
dology for the Bystander Role of Interaction. In: Proc. of IEEE Intl Conf. on Systems, Man
& Cybernetics, pp. 3212–3217 (2003)
13. Scholtz, J., Young, J., Drury, J., Yanco, H.: Evaluation of Human-Robot Interaction
Awareness in Search and Rescue. In: Proc. of IEEE Intl. Conf. on Robotics and Automa-
tion, pp. 2327–2332 (2004)
14. Steinfeld, A., Fong, T., Kaber, D., Lewis, M., Scholtz, J., Shultz, A.: Common Metrics for
Human-Robot Interaction. In: Proc. IEEE International Conference on Intelligent Robots
and Systems (2004)
15. Networkworld, CES: Battling AR helicopter controlled by iPhone (2012),
http://www.youtube.com/watch?v=TSv2ca-IECc
16. Gates, B.: A Robot in Every Home. Scientific American (January 2007)
17. Apple iPad 2 (2012), http://www.apple.com/ipad/specs
18. Zhu, D., Gedeon, T., Taylor, K.: Exploring Camera Viewpoint Control Models for a Multi-
Tasking Setting in Teleoperation. In: Proc. CHI 2011, pp. 53–62 (2011)
19. Kawamura, K., Peter, A., Johnson, C., Nilas, P., Thongchai, S.: Supervisory Control of
Mobile Robots using Sensory EgoSphere. In: Proc. ICRA 2001, pp. 523–529 (2001)
20. Grandall, J.W., Goodrich, M.A., Olsen Jr., D.R., Nielsen, C.W.: Validating Human-Robot
Interaction Schemes in Multitasking Environment. IEEE Transaction on Systems, Man and
Cybernetics, Part A: Systems and Humans 35(4) (July 2005)
21. Guo, C., Sharlin, E.: Exploring the Use of Tangible User Interfaces for Human-Robot Inte-
raction: A Comparative Study. In: Proc. SIGCHI (2008)
22. Nielson, C., Goodrich, M.: Comparing the Usefulness of Video and Map Information in
Navigation Tasks. In: Proc. ACM HRI, pp. 95–101 (2006)
23. Baker, M., Keyes, B., Yanco, H.: Improved Interfaces for Human-Robot Interaction in Ur-
ban Search and Rescue. In: Proc. IEEE International Conference on Systems, Man and
Cybernetics, pp. 2960–2965 (2004)
24. Chen, Y.C., Haas, E.C., Barnes, M.J.: Human Performance Issues and User Interface De-
sign for Teleoperated Robots. IEEE Trans. on SMC-Part C: Applications and Re-
views 37(2), 1231–1245 (2004)
25. Liang, C.K., Foo, C.H., Lee, Y.S.: Interactive Methods of Tele-operating a Single Un-
manned Ground Vehicle on a Small Screen Interface. In: Proc. ACM HRI, pp. 121–122
(2011)
Assessing the Impact of Latency and Jitter on the
Perceived Quality of Call of Duty Modern Warfare 2
Rahul Amin1, France Jackson1, Juan E. Gilbert1, Jim Martin1, and Terry Shaw2
1
Clemson University, Clemson, SC, USA
2
Cable Television Laboratories, Inc., Louisville, CO, USA
{ramin,fljacks,juan,jmarty}@clemson.edu, [email protected]
Abstract. Jane McGonigal stated in her 2010 TED Talk that humans spend 3
billion hours a week playing video games around the planet. Americans alone
devote 183 million hours per week to gaming. With numbers like these, it’s no
wonder why end user demands for bandwidth have increased exponentially and
the potential for network congestion is always present. We conduct a user study
that focuses on the question: “How much network impairment is acceptable be-
fore users are dissatisfied?” In particular, the main objective of our study is to
measure a gamer’s perceived Quality of Experience (QoE) for a real-time first
person shooter (FPS) online game Call of Duty Modern Warfare 2 in pres-
ence of varied levels of network congestion. We develop a Mean Opinion Score
(MOS) metric to determine each gamers’ QoE. We investigate the following
hypothesis: The gamers’ perceived QoE correlates to their skill level.
Keywords: Online Gaming, First Person Shooter Games, Network Impairment,

Quality of Experience.
1 Introduction
Resource allocation is the process by which network elements try to meet competing
demands that applications have for network resources. While broadband cable net-
works have steadily increased the size of the ‘pipe’ available to subscribers, end user
demand for bandwidth continues to grow at an exponential rate. Therefore, network
operators must continue to expand the capacity of their networks in order to avoid
congestion. However, it is impossible to avoid occasional network congestion. This
congestion is most likely to occur at prime time when most subscribers are active and
many different type of applications are simultaneously in use (with many subscribing
households using multiple applications simultaneously). The response by each appli-
cation to congestion differs widely. File transfer applications such as File Transfer
Protocol (FTP) or Peer-to-Peer (P2P) file sharing are typically the most flexible in
performance requirements. Applications that involve real-time end user interaction,
such as video conferencing, VoIP (voice over IP), or online games are less flexible as
they generally require low latency. Emerging web-based video streaming applications
are generally flexible; however, once performance thresholds are surpassed, perceived
98 R. Amin et al.
quality drops off quickly. Finally, over-the-top HD (high definition) voice and high
quality video broadcasts (e.g. live sportscasts) are the most sensitive to fluctuation in
bandwidth, latency and loss.
The research described in this paper focuses on network gaming applications. In
particular, we study real-time online games. The academic community has studied the
impact of latency and loss on real-time games such as first person shooter (FPS) or
action games [1]-[7]. The wide range of previous results confirms that mapping net-
work performance metrics such as latency and loss to an assessment of perceived
quality is extremely difficult. The challenge is in part caused by the fact that each
game responds to latency and loss is unique ways. Further, for a given game, specific
modes of operations and even game play scenarios (which change quite frequently)
will react differently to latency and loss. Perhaps the most problematic aspect of as-
sessing user’s quality of experience is due to the fact that a subjective assessment is a
measure of how the gaming experience meets (or fails to meet) a user’s expectation.
Accurately quantifying expectations is very difficult as a gamer’s expectation depends
on past history and personal factors.
Our research is motivated by the observation that much of the prior work described
in the literature has focused on objective studies. While subjective studies have been
considered, there have been limited results. This is primarily because a widely ac-
cepted utility function that maps objective metric results to a perceived quality
assessment has not been established. While the research described in this paper is
primarily exploratory in nature, we consider the following hypothesis: a gamer’s
perceived quality of experience correlates to his/her skill level.
We focus on the widely played FPS game Call of Duty Modern Warfare II. We en-
gage the student population at Clemson University to participate in a large scale
experimental study designed to explore the impacts of network impairments on the
perceived quality. We collect information from the participants including their self-
assessed gaming expertise, scores achieved during the testing session, and assessment
of perceived quality. Our results do support our hypothesis by clearly showing that
user’s level of experience defines and determines the sensitivity to latency. However,
we find ‘expert’ gamers who are presumably the most sensitive to latency are also the
most adept at compensating for impaired conditions.
This paper is organized as follows. Section 2 presents a literature survey and sum-
marizes relevant related work. Section 3 provides a description of our experimental
methodology. Section 4 documents our results and provides an analysis of our results.
Finally, Section 5 provides conclusions including a discussion of the limitations of
our study and possible next steps.
2 Background
The broad genre of multi-player, online games includes non-real-time games such as
board games or fantasy games as well as real-time games such as first person shooter,
Assessing the Impact of Latency and Jitter on the Perceived Quality of Call 99
sports, or real-time strategy. The work in [5] presents a useful taxonomy for classify-
ing real-time online games. The authors correlate the game’s relationship between
latency requirements and ‘precision’ (either movement or shooting) requirements.
Further attributes include the interaction model and the game perspective. The inte-
raction defines how a player interacts with the game world. The authors indicate the
interaction is either through an avatar or with an omnipresent model. The avatar mod-
el exists at a particular location in the game context and is only involved with local
interaction. The omnipresent model provides players a global view of the game. The
game perspective is tied to the interaction model as it defines how players view the
game world. For avatar interaction, such as in shooter games, a first or third person
shooter perspective is common. For omnipresent interactions, such as in role-playing
games, the perspective might be variable. A player might have access to aerial views
as well as first or third person views.
Because real-time online games are sensitive to network and system perfor-
mance, there has been a significant amount of prior academic research in this area.
Many of these studies share our goals and methods. Works such as [1]-[7] all seek to
understand the impact that loss and latency have on real-time online games. A com-
monly used evaluation approach is to emulate loss and latency in a controlled test-bed
environment. Either objective measures such as game results or subjective tests based
on Mean Opinion Scores (MOS) are used. The impacts of loss are arguably not as
significant as those caused by latency because most games are designed to tolerate
some amount of loss either by sending game state or control commands frequently or
by utilizing error correction. Therefore, much of the research presented in the litera-
ture has focused on the impacts of latency.
In [1], the authors found that the ability to carry out ‘precise’ movement and
shooting tasks is not significantly affected by loss rates or latency that are typical for
the FPS game Unreal. They did observe players scores (in terms of kills and deaths)
in some game modes drop as the latency increased beyond 100 milliseconds (ms). In
[2], the authors focus on the effects of latency on World of Warcraft. They find that
because the game is real time strategy (rather than first person shooter), a high level
of induced latency (well beyond 100 ms) is noticeable but has negligible effect on the
outcome. Other works have used methods involving subjective tests where gamers
were asked about their perceived quality. In [7], as round trip times exceed 150 ms,
the performance of Halo gamers decreased by about 50% while the perceived quality
was still high. Further, games with over 500 ms of delay and 4% loss were considered
unplayable. The work in [3] shows that in some scenarios (i.e., play situations of a
specific FPS game), even small amounts of latency and jitter (as small as 60 ms)
caused users to notice a drop in perceived quality.
The work examined in [5] attempted to group the games in prior studies based on
model (avatar or omnipresent), perspective (first or third person or varied), and genre
(FPS, Racing, Sports, RPT, RTS) to one of three levels of latency sensitivity. The
thresholds of induced latency for high, medium and low sensitivity settings were es-
timated to be 100 ms, 500 ms, and 1,000 ms. More recent work [6] has addressed this
100 R. Amin et al.
by exploring the effects of latency on classes of games. Using the open source third-
person shooter game BZFlag, the authors illustrate that the range of tolerated latency
can range by an order of magnitude (from less than 100 ms to 1.0 seconds) depending
on the required precision for game play.
The work in [8] provides a useful framework for describing and assessing game
quality of experience through the use of ‘perceptual view inconsistency’. The frame-
work assumes that 1) Quality of Experience (QoE) is based on an overall assessment
that is based on individual (and decomposable) perceptions; 2) Inconsistencies result
from the disconnect between the gamer’s view and the server’s (or global) view; 3)
An overall QoE assessment requires each objective inconsistency to be mapped to
corresponding subjective values. The authors refer to prior measurement work and
show that the degradation of various metrics of perceived quality, each of which are
specific to certain segments of specific games, can be modeled as a half normal
(Gaussian) curve.
Based on the wide range of results that exists in the related work, we draw the
following motivating conclusions:
• The impacts of latency and loss depend not just on the class of game, but also on
particular modes of play and on specific gaming contexts.
• Establishing the relationship between latency and loss network measures with per-
ceived quality is the most challenging issue that remains largely unsolved.
Our study develops a methodology for exploring the complex interaction between
network impairment, the impacts on FPS online games, and the quality of experience
perceived by end users (i.e., the gamers).
3 Methodology
3.1 Experimental Setup
Fig. 1 represents the experimental setup. For this setup, there are two sets of game-
players. The first set consists of two players, referred to as “Participants”, who play
on a controlled network environment. Looking at Fig. 1, we emulate network impair-
ment with the Linux netem capability (Please refer to this site for further information
on netem: http://www.linuxfoundation.org/collaborate/workgroups/networking/netem)
operating on our router. We add latency and jitter to packets arriving at the router
destined for the ‘congested participant’. The other participant is connected directly to
the university’s network and is referred to as the ‘uncongested participant’. The
second set of players, referred to as “Gamers”, play from their home or dormitory
room on campus during each session to maintain a controlled but realistic level of
background competitors. The location of the game host is an experimental parameter.
We designate either the congested participant or the uncongested participant as the
host.
Fig. 1. Experiment Setup
3.2 Experimental Procedure

Before each game session, the two participants are asked to fill out a pre-survey
questionnaire geared to determine their expertise level and to estimate their network
performance expectation based on their normal gameplay environment. Each game
session lasts for ten minutes where all participants and gamers play in the “Free for
All” game mode, and then the participants answer a post-survey questionnaire that
consists of four Likert Scale survey questions. The four post-survey questions deal
with each gamer’s perception on following experiences: (i) Gameplay Satisfaction (ii)
Gameplay Frustration (iii) Impact of Lag on Gameplay (iv) Likeliness to change net-
work service providers. A MOS metric is computed for each participant using a
weighted-sum approach where each of the four post-survey responses is given equal
importance. The MOS metric derived for each participant is averaged to assess overall
MOS (or QoE) for each network impairment setting. A detailed description of pre-
survey questionnaire, post-survey questionnaire and MOS calculation is provided
in [9].
3.3 Study Design

The study is divided into two phases: (i) Calibration phase (ii) Actual Experimenta-
tion phase. The ‘calibration’ phase involved a series of experiments designed to iden-
tify the set of network congestion settings that would be used in the actual study. The
impact of jitter is quite different depending on which participant serves as the host.
We conducted a pilot study in which we invited experienced Call of Duty Modern
Warfare 2 players to play on several combinations of latency and jitter settings. In all
experiments, 100 ms latency and a range of jitter in increments of 50 ms is added to
all packets sent or received by the congested user. The results of the pilot study, pre-
sented in Fig 2, suggested that when the uncongested participant serves as the host,
jitter in the range of [0, 300] ms is appropriate and when the congested participant
serves as the host, jitter in the range of [0, 150] ms is appropriate. Moreover, the
102 R. Amin et al.
functionality of netem was verified in the calibration phase of the study. A detailed
description of these calibration phase findings is presented in [9]. Using the network
congestion settings presented in Fig 2, a total of 56 experiments were conducted using
university students as test subjects to obtain 7 MOS results per network congestion
setting during the actual experimentations phase.
(a) (b)
Fig. 2. Pilot Study Results
4 Results and Analysis
We performed 56 total experiments, which resulted in 112 survey responses. Figs 3,

4, and 5 profile the participants. Fig 3 indicates that the participants are reflective of
the undergraduate population at Clemson University. The majority rely on wired ra-
ther than wireless connectivity. Fig 4 suggests that the participants represent a broad
range of FPS gaming experience, ranging from novice to expert. Finally, further anal-
ysis of the pre-survey responses suggests that 10% (about 10) of the participants were
complete novices to any online games and another 10% were experts with Call of
Duty Modern Warfare as shown in Fig 5. The rest of the participants had some level
of experience with FPS games.
Fig. 3. Participant Demographics

Fig. 4. Participant Internet Access Method
Fig. 5. Participant Gaming Experience
The results from actual experiments are shown in Fig 6(a) and 6(b) when the un-
congested participant and congested participant are game hosts respectively. Note that
only the results derived from post-survey responses of congested participants are pre-
sented in this study. The results derived from post-survey responses of uncongested
participants are random in nature as these participants do not play under any network
impairment and are presented in the extended version of this paper [9]. Fig 6(a) sug-
gests that as jitter approaches 250 ms, the QoE degrades to a level (MOS < 3) that
make the gamers’ experience unacceptable when the game host is the uncongested
participant. Fig 6(b) suggests that as jitter approaches 100 ms, the QoE degrades to a
level (MOS < 3) that makes the gamers’ experience unacceptable when the game host
is the congested user. We observe that the MOS in Fig 6(b) drops as jitter is increased
but not as definitively as in Fig 6(a). This happens because the congested participant
is the game host in Fig 6(b). For this setting, the command and control data that the
congested participant sends is actually avoiding the netem latency effects since the
server is located at the local host. We conjecture that the outbound traffic stream is
more sensitive to latency than the inbound traffic stream.
104 R. Amin et al.
(a) (b)
Fig. 6. Overall Actual Experimentation Results
(a) (b)
(c) (d)
Fig. 7. Results based on Participant’s Expertise Level
The results that compare the responses provided by experienced gamers vs. inex-
perienced gamers are presented in Figs 7(a)-(d). For both cases, when the game host
is uncongested or congested participant, the results support our hypothesis by clearly
suggesting that participant’s level of expertise defines and determines the sensitivity
to network congestion. The more experienced the participant (Figs 7(b) or 7(d)), the
less tolerant he/she is to network congestion compared to the less experienced partici-
pants (Figs 7(a) or 7(c)). Also, the ‘expert’ or highly experienced gamers significantly
berate even the lowest network congestion setting of 0 jitter/100 ms latency as seen in
Figs 7(b) and 7(d) and have a hard time quantifying the level of congestion when the
congestion settings are low.
5 Conclusions
We have studied the impact of latency and jitter on the perceived quality of the FPS
game Call of Duty Modern Warfare 2. Our experimental methodology involved col-
lege students (gamers) playing a closed session of “Free for all” game mode. A total
of seven gamers were involved in each game. Two gamers (referred to as participants)
were under observation, one of whom had his/her network impaired in a controlled
manner. The participants under observation provided a numeric assessment of their
satisfaction, level of frustration, impact of lag and likeliness to change network
service providers.
Our results suggest that, for the cases when the host is the uncongested user and
the user under study is the congested user, a random jitter that is in a range of [0, 250]
ms leads to a MOS less than 3. For the case when the host is the congested user and
the user under study is the congested user, a random jitter in the range [0, 100] ms
leads to a perceived MOS less than 3. Our original hypothesis was that the perceived
quality correlates to a gamer’s skill level. We did observe different thresholds of
tolerance of network impairment based on the level of experience and skill. The more
experienced the participant, the less tolerant he/she is to network congestion. The
‘expert’ or highly experienced gamers significantly berate even the lowest level of
network impairment (with constant 100 ms latency and no jitter). However, these
‘expert’ gamers who are presumably the most sensitive to latency/jitter are also the
most adept at compensating for impaired conditions as observed from their final game
scores (which are not presented in this paper).
Our methodology has several limitations. First the number of user’s in the study
was not sufficient. While most data points in the results curves were based on 7 sam-
ples, some results (in particular those shown in Figs 7(a)-(d)) are based on only 2-3
samples. Second, our methodology did not precondition the test subjects. In other
words, ideally, we would have a participant take our survey in his/her usual gaming
environment. This provides a familiar context with which a user can base his/her as-
sessment. In our next steps we would like to take a small set experienced Call of Duty
gamers and first train their expectations and game play on ‘ideal’ conditions. Then,
each gamer will periodically be asked their opinion allowing us to fully map study
participants to their perceived quality based on controlled network settings.
106 R. Amin et al.
References
1. Beigbeder, T., Coughlan, R., Lusher, C., Plunkett, J., Agu, E., Claypool, M.: The Effects of
Loss and Latency on User Performance in Unreal Tournament 2003. In: Proceedings of 3rd
ACM SIGCOMM Workshop on Network and System Support for Games, NetGames 2004
(August 2004)
2. Sheldon, N., Girard, E., Borg, S., Claypool, M., Agu, E.: The Effect of Latency on User
Performance in Warcraft III. In: Proceedings of ACM Network and System Support for
Games Workshop (NetGames), Redwood City, California, USA (May 2003)
3. Quax, P., Monsieurs, P., Lamotte, W., Vleeschauwer, D., Degrande, N.: Objective and Sub-
jective Evaluation of the Influence of Small Amounts of Delay and Jitter on a Recent First
Person Shooter Game. In: Proceedings of the SIGCOMM 2004 (August 2004)
4. Woundy, R., Lee, Y., Veiga, A., Williams, C.: Congestion Sensitivity of Real-time Residen-
tial Internet Applications. In: Proceedings of the SCTE Cable-Tec Technical Workshop
(November 2010)
5. Claypool, M., Claypool, K.: Latency and Player Actions in Online Games. ACM Commu-
nications of the ACM (November 2006)
6. Claypool, M., Claypool, K.: Latency Can Kill: Precision and Deadline in Online Games. In:
Proceedings of Multimedia Systems Conference (MMSys 2010) (February 2010)
7. Lang, T., Armitage, G.: A Ns2 Model for the Xbox System Link Game Halo. In: Proceed-
ings of the Australian Telecommunications Networks and Applications Conference (2003)
8. Chan, P., Zarki, M.: Perceptual View Inconsistency: An Objective Evaluation Framework
for Online Game Quality of Experience (QoE). In: ACM NetGames 2011 Workshop (Octo-
ber 2011)
9. Amin, R., Jackson, F., Gilbert, J., Martin, J., Shaw, T.: Assessing the Impact of Loss and
Latency on the Perceived Quality of Call of Duty Modern Warfare 2,
http://people.cs.clemson.edu/~jmarty/projects/GamingAssessments/
GamingStudyReportV6.0.pdf (last accessed January 3, 2013)
The Effects of Negative Interaction Feedback in a Web
Navigation Assistant
Marcelo G. Armentano and Analía A. Amandi
ISISTAN Research Institute (CONICET/UNICEN)

Campus Universitario, Paraje Arroyo Seco, Tandil (7000), Argentina
{marcelo.armentano,analia.amandi}@isistan.unicen.edu.ar
Abstract. Recommender systems are a common solution used to assist users in

searching and retrieving information on the web due to the benefits that can be
obtained from the evaluation and filtering of the vast amount of information
available. This article presents a user study on the feasibility of using negative
interaction, that is the absence of interaction with some items in a list of sugges-
tions, as implicit feedback used to improve the performance of a web navigation
assistant. Results showed an increment of 16.65% in the acceptance of the sug-
gestions provided by the assistant and an increment of 43.05% in the average
use of the suggestions window when using negative interaction with respect to
not using this feedback mechanism.
Keywords: Intelligent Agents, Web Navigation Assistance, Implicit Feedback.
1 Introduction
The Web is daily queried with multiple purposes, such as reading the news, research-
ing about a specific topic, planning vacations or searching for references to answer
specific questions. Search engines facilitate finding information quickly on any topic
of interest. Using a set of keywords as an input, search engines offer as a result a list
of Web links hopefully addressing the issues described by those keywords.
Search engines offer a quick access to the information indexed from the web, au-
tomating the classification of the contents to deliver the most relevant information to
each query. However, it is unusual that users are able find the information they intend
to in their first search because they are constantly overloaded with content and links.
This can be frustrating to users, who know that the information is somewhere availa-
ble but they are not able to find it. A usual problem with search engines is that many
users do not know how to “build” the search queries to obtain the information they are
looking for. The queries used in a search engine are usually short and ambiguous, and
different users might use the same query with completely different needs [1].
To address this problem we propose a web navigation assistant that recommends
possible interesting web pages to the user while navigating the web. The assistant
observes the user interaction with a web browser to obtain implicit indicators for the
subject of the user’s interests. Then, the assistant automatically performs new web
108 M.G. Armentano and A.A. Amandi
searches to obtain pages related to those recently visited by the user. The interaction
of the navigation assistant with the user is only through a small button indicating the
availability of suggestions. When the user presses this button the top ranked results
are shown in a window integrated with the web browser so that the user can continue
navigating as usual if he/she is not interested in the suggestions.
We performed a set of experiments with real users in which we compared the re-
sults obtained considering the impact of two mechanisms of implicit feedback: the
time spent on a page and the negative interaction. For these experiments we computed
the percentage of usage of the window presented by the assistant, and the percentage
of acceptance of the suggestions selected by the user. The results showed that consi-
dering both implicit mechanisms of feedback improve the performance of the assis-
tant. Particularly, considering the negative interaction feedback allows the agent to
modify his state when it has not properly inferred the user’s search intention.
The rest of this paper is organized as follows. Section 2 presents some related
work. Section 3 presents the web navigation assistant implemented, detailing its gen-
eral architecture and how it process the implicit user feedback to improve the user
profile for a searching session. Section 4 presents the user study conducted. Finally,
Section 5 presents our conclusions of the study.
2 Related Work
There have been many research efforts focused on systems that recommend pages to
web users. In this section we summarize some of those studies centered on obtaining
implicit feedback from the user.
Our web navigation assistant learns a short term user model by analyzing the text
of the pages that it assumes the user is interested in. Chi et al [2] obtains information
about the user from the text of the links he/she clicks. However, many times the text
of the links does not have relevant information about the linked page (e.g. “click
here”). Zhu et al. [3] build a user model by analyzing the user’s behavior in order to
build rules such as “any word which is present in three consecutive pages visited by
the user will be present in a page that is important to the user”. These rules, called
“abstract navigation patterns”, are generated according characteristics of web pages
visited by the user and not with any particular page. Matthijs and Radlinski [4] pre-
sented a personalization approach that combines the content and previously visited
websites to build a user interest profile using the users' complete browsing behavior.
This model is then used to re-rank web results. Hu et al. [5] also proposed a persona-
lization approach but restricted to searching web services. They obtain the user inter-
ests from both the search requests and the previously used services by extracting the
meaningful words from the service request and the used service description files.
The time spent in a page has been commonly used in many approaches as an indi-
cator of the user interest in any piece of information. Parsons et al. [4], for example,
found that there is a positive relationship between the time users spend watching an
item available for purchase and their intention to buy it. Although the time spent
watching an item depends also of external factors (such as the amount of visual
The Effects of Negative Interaction Feedback in a Web Navigation Assistant 109
details, images, distractors, etc.), it is believed that it is a potential indicator of the

user preferences.
Joachims et al. [7] studied the behavior of the users interacting with the results
provided by a search engine. The results obtained shown that users take decisions
from the descriptions provided by the search engine, by clicking on the links they are
interested. However, the user clicks are conditioned by the order in which the links
are shown. The user mainly interacted with the top ranked results (that is the links in
the top of the list) even when the descriptions were less relevant that results in lower
positions. Moreover, the user spent more time reading the descriptions of the top
ranked results, and descriptions at lower positions received less attention.
Our approach takes the findings of Joachims et al. [7] to define the concept of
negative interaction, that is to use the information of the search results skipped by
the user to improve the user profile related to his/her information needs in a search
session.
3 Web Navigation Assistance
WebHelper is a web navigation assistant we developed to observe the user behavior

while interacting with a web browser to detect his/her subject of interest and suggest
potentially interest pages. The objective of the assistant is to help the user to find web
pages that he/she might not be able to find with the keywords he/she is using.
The search intention of the user is inferred by processing the text of the web pages
he/she visit. After each visited page, the assistant updates the weights assigned to
different words that describe the search intention of the user in a session. The top
ranked words are then used as keywords to perform a new web search and present the
results obtained to the user in an independent window integrated with the browser.
This window is only shown when the assistant has enough information to infer user’s
the search intent. However, the user is able to close this window at any time and to
open it asking for suggestions.
3.1 General Architecture
WebHelper is implemented server-side, in the form of a web proxy. This way, the
user can start using the assistant simply configuring his/her web browser to use the
assistant’s proxy. When the user sets his/her browser to use this proxy, all requests are
intercepted and two processes are applied. The first process extracts the information
contained in the response pages and assigns different weights according to the context
of each extracted term. The second process injects in the HTML response the code
corresponding to the suggestion window, if appropriate.
The dataflow in WebHelper is shown in Fig 1. The user searches the internet using
a web browser. The WebProcessor module captures the response to the server request
and executes two processes: information extraction and information injection. The
information extraction process captures the content of the web page and passes it to
the WordProcessor module. If the assistant has any recommendation to present to the
user, the information injection process adds to the response the code corresponding to
the suggestions window to be presented to the user. On the other hand, the WordPro-
cessor module takes the page content received from the WebProcessor module and
updates the user profile. We describe this process in Section 3.2
Internet
WebProcessor
WordProcessor User
Profile
Fig. 1. WebHelper general architecture
3.2 Information Processing
When the WordProcessor component receives a new visited document, it first con-
verts it to plain text, detects its language1, removes stop words and applies the corres-
ponding Porter stemming algorithm [10].
The user profile keeps the set of most frequent words, each of them with an asso-
ciated weight. These weights vary according to the place in the page in which each
word appears, and vary after each visited page, either by adding or subtracting a new
constant value. Different constant values were defined for words appearing in head-
ings, normal text, links, text typed in input boxes, meta-information, etc.
On the other hand, words weights in the user profile are decremented for three dif-
ferent reasons: (1) the word does not appear in the current page; (2) the negative inte-
raction feedback is applied; (3) the user does not select any suggestion after a fixed
amount of visited pages. We give details about the last two items in Section 3.3.
After processing each visited page, the following situations can arise:
• There were no suggestions up to that moment, and no terms in the user profile
exceed the “suggest-me” threshold. In this case, nothing happens and no sugges-
tions are given to the user.
• There were no suggestions up to that moment, and two or more words exceed
the “suggest-me” threshold. In this case, the assistant use the words exceeding the
threshold and performs a web search using those words as keywords. The top
results of the search engine are presented as suggestions. In the current implemen-
tation of WebHelper, the Google® API is used for querying the web.
1
We currently work with English and Spanish
• The assistant was already showing suggestions, but after processing the informa-
tion of the new document the weights of all words are under the “suggest-me” thre-
shold. In this case, the suggestions window is not injected in the response to the
user.
• The assistant was already showing suggestions and after processing the informa-
tion of the new document the words whose weights exceed the “suggest-me” thre-
shold are the same that in the previous cycle. In this case, suggestions are rotated,
and a new set of web pages from the previous query are shown in the suggestions
window.
• The assistant was already showing suggestions and after processing the informa-
tion of the new document a new set of words weights exceed the “suggest-me”
threshold. In this case a new web search is performed using those new words as
keywords and the top results are presented as suggestions, replacing the previous
window.
Before querying the web, the Google® API is used to obtain suggestions for each
candidate keyword. This service provided by the Google® API, receives a word as
input and returns a set of related words as output. The suggested words for each key-
word and the keywords themselves are combined to query the search engine. This
step improved the performance of the assistant since the terms of in the user profile
are words trimmed by the stemming process. Using those stems to query the search
engine produced unexpected results.
3.3 Feedback Processing
To determine the search intention of the user in a session, the navigation assistant
considers two implicit feedback indicators: the time the user spends in a page and the
interaction with the suggestions window.
The time the user spends on a page have been extensively used as an indicator of
interest in a piece of information [4][8]. A web page is considered active by WebHel-
per when the user first access to it or when it is selected by using the browser’s tabs.
We consider two types of interactions with the suggestions window: positive and
negative interactions. A positive interaction is triggered when the user clicks on
link in the suggestions window. In this case, the content of the suggestion (title and
snippet) is used to improve the user profile for that session.
Moreover, at the same moment when the user shows interest in a suggestion, there
is also a negative interaction with respect to the items located above the one selected
by the user in the list of suggestions. We apply a heuristic based on statistical studies
that showed that lists are usually read from top to bottom [7]. This way when the user
selects an item in a list, he/she is not only indicating interest in the suggestion selected
but also is implicitly indicating that he/she is not interested on the previous items.
Assume for example that the assistant presented a list of suggestions <s1, s2, s3, s4,
s5, s6, s7> and the user clicked on s1, s3 and s5. We can assume that s3 is more rele-
vant that s2, since the user read the description of s2, but preferred clicking on s3.
For the same reason we can assume that s5 is more relevant that s2 and s4. We can
assume that s2 and s4 are not relevant to the user’s needs and we can use the informa-
tion they contain to update the user profile as negative information [9]. The sugges-
tions list can then be re-organized to eliminate the preceding recommendations.
Similarly, if the user has not interacted with the suggestions window after a certain
amount of visited pages, it is assumed that the suggestions are not interesting to the
user. Consequently, suggestions are rotated to show the user a new set of possible
interesting websites. Following the idea of negative interaction, when suggestions are
rotated the weights of the terms appearing in the hidden suggestions information are
decreased.
3.4 Suggestions Window
When the assistant has suggestions to make to the user, it creates a drop down win-
dow integrated to the page returned to the user. This window is situated on the right of
the returned page and contains a maximum of eight suggestions, consisting in a title
and a snippet (Fig. 2). This window can be hidden and re-opened at any time.
Fig. 2. Suggestions window while visiting a page

4 Experimental Evaluation
4.1 Experiment Setup

To validate the proposed approach, we tested our system with 32 volunteer users, both
male and female, with ages ranging 18 to 55, and different levels of expertise in
searching the web. Users were divided into groups to test four different configurations
of the system.
We asked each user to freely use the web to search for information about any sub-
ject of interest. The navigation assistant was slightly modified for these experiments:
after selecting a suggested site from the suggestions’ window, users were asked
whether the page they had just visited was interesting according to their search inten-
tions. This feedback was only logged to compute the performance of the assistant and
it was not used to modify its normal behavior since in a live scenario feedback is nev-
er asked to users.
The duration of the experiments was, on average, 25 minutes. Most users started
their session accessing Google’s search engine and entered the keywords they be-
lieved best expressed their information needs. Other users interested in news opened
their preferred online newspaper and start browsing through it. Finally, other users
visited pages they usually visit.
4.2 Metrics
Two metrics were used to evaluate the performance of our navigation assistance: the
percentage of use of the suggestions’ window and the percentage of acceptance of the
suggestions selected by the user.
To compute the average use of the suggestions’ window, we divided the number of
suggestion selected by the user by the number of windows shown to the user during
the experiment.
# suggestions _ selected
use _ of _ the _ assis tan t =
# suggestions _ shown
Likewise, the percentage of acceptance of the suggestions is computed dividing the
average number of accepted suggestions by the average number of suggestions se-
lected by the user.
# suggestions _ accepted
accep tan ce =
# suggestions _ selected
4.3 Time Spent on a Page

The first experiment explored whether considering the time the user spent on a page
any influence the performance of the assistant. In the first configuration of the naviga-
tion assistant, any page visited by the user was processed, without considering how
long the user remained in the same page. We use this first configuration as a base-line
for comparison with the other three configurations. In the second configuration, pages
were only considered for improving the user profile when the user spent more than a
predefined amount of time reading the page. For our experiments this threshold
was set to six seconds. In this configuration, the system took more time to generate
suggestions since not all information visited by the user is processed and, consequent-
ly, the weights assigned to individual terms increased more slowly than in the first
configuration.
Since different experiments took different durations, the average results were nor-
malized by the duration of the individual experiment. For example, nine suggestions
windows were shown to User 1 for the first configuration, and User 1 used the system
for 30 minutes, so we used a value of 0.3 windows per minute. Table 1 shows the
results obtained.
Table 1. Results for user tests with and without time processing
Suggestion Suggestions Suggestions Time for Duration Google

windows selected by accepted by showing the of the searches
shown the user the user first sug- experiment
gestions
Without 0.662 per 0.148 per 0.113 per 3.62 mi- 22.125 0.414
Time minute minute minute nutes minutes per
processing minute
With time 0.350 per 0.184 per 0.163 per 10.62 22.375 0.373
processing minute minute minute minutes minutes per
minute
From Table 1 we can compute that the average use of the assistant without consi-
dering the time spent on a page was 22.35% while the average acceptance of the sug-
gestions selected by the user was 76.35%. On the other hand, when considering the
time the user spent on a page, these values are increased to 52.57% and 88.59% re-
spectively.
4.4 Negative Interaction Feedback
The third configuration of the experiment was set up to consider negative interaction
feedback but did not consider the time spent on a page. Finally, the fourth configura-
tion used both feedback mechanisms together. Table 2 shows the results obtained. A
new column is added to show the number of times that the suggestions were rotated.
The average acceptance of the suggestions selected by the user was 87.1% (with an
average use of 26.59%) if we consider the negative interaction feedback alone.
However, we reach an average acceptance of 93% and an average use of 65.4% by
combining both feedback mechanisms.
Table 2. Results for user tests with and without time processing
Sugges- Sugges- Sugges- Time for Duration Google Rotation

tion tions tions showing of the searche of sug-
windows selected accepted the first experi- s gestions
shown by the by the sugges- ment
user user tions
Without 0.583 0.155 per 0.135 per 8.750 24.000 0.307 0.102 per
Time proper minute minute minutes minutes per minute
cessing minute minute
With time 0.327 0.214 per 0.199 per 10.500 22.625 0.140 0.072 per
proper minute minute minutes minutes per minute
cessing minute minute
4.5 Discussion
For the first configuration, while the acceptance of the suggestions selected by the
user was high (76.35%), the use of the suggestions window was quite low (22.35%).
This fact suggests that the information shown to the user in the suggestions’ window
might have not convinced the user to click on a suggestion. Since for this experiment
all visited pages were processed, the threshold for showing suggestions was reached
quickly, but the system was not able to properly infer the intention search of the user.
The second experiment solved this problem by increasing the use of the suggestions
window to 56.57% at the expense of the time that the system needed to give assis-
tance to the user.
On the other hand, by using the negative interaction feedback alone, we obtained
an acceptance of the suggestions similar to that we obtained when we used the time
processing alone. The average usage in this case was better than the first configura-
tion but lower than the second configuration. The time spent to show suggestions to
the user is lower than the second configuration, but higher than the first configuration.
This is due to the fact that the negative interaction lowers the weights of the terms that
are present in the unvisited documents. Then these weights trend to increase slowly.
By combining both feedback mechanisms we obtained a better general perfor-
mance of the assistant. Furthermore, although the time that took the system to show
the first suggestions window was similar to that in the second configuration, it
performed fewer queries to the search engine, achieving a better acceptance of the
suggested sites.
5 Conclusions
In this article we described a web navigation assistant that infers the search intention
of a user in order to provide him with suggestions of sites in which he/she might be
interested. The assistant uses implicit feedback mechanisms to improve a temporal
user profile that contains information about the subject of interest in a search session.
We presented a user study on the feasibility of using negative interaction, which is

the absence of interaction with some items in a list of suggestions, as implicit feed-
back used to improve the performance of the web navigation assistant.
The study allowed us to conclude that using both the time spent on a page and the
negative interaction feedback improves the performance of the assistant, obtaining an
increment of 16.65% in the acceptance of the suggestions provided and an increment
of 43.05% in the average use of the suggestions window.
References
1. Hu, R., Dou, W., Liu, X.F., Liu, J.: Personalized Searching for Web Service Using User
Interests. In: Proceedings of IEEE Ninth International Conference on the Dependable, Au-
tonomic and Secure Computing (DASC), pp. 156–163 (2011)
2. Chi, E., Pirolli, P., Chen, K., Pitkow, J.: Using information scent to model user information
needs and actions and the Web. In: Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems, pp. 490–497. ACM, New York (2001)
3. Zhu, T., Greiner, R., Häubl, G.: Learning a model of a web user’s interests. In: Brusi-
lovsky, P., Corbett, A.T., de Rosis, F. (eds.) UM 2003. LNCS, vol. 2702, pp. 65–75.
4. Matthijs, N., Radlinski, F.: Personalizing web search using long term browsing history. In:
Proceedings of the Fourth ACM International Conference on Web Search and Data Mining
(WSDM 2011), pp. 25–34. ACM, New York (2011)
5. Hu, R., Wanchun, D., Liu, X., Liu, J.: Personalized Searching for Web Service Using User
Interests. In: Proceedings of the 2011 IEEE Ninth International Conference on Dependa-
ble, Autonomic and Secure Computing (DASC 2011), pp. 156–163. IEEE Computer
Society, Washington, DC (2011)
6. Parsons, J., Ralph, P., Gallagher, K.: Using Viewing Time to Infer User Preference in Re-
commender Systems. In: Proceedings of the AAAI Workshop on Semantic Web Persona-
lization held in conjunction with the 9th National Conference on Artificial Intelligence
(2004)
7. Joachims, T., Granka, L., Pan, B., Hembrooke, H., Gay, G.: Accurately interpreting click-
through data as implicit feedback. In: Proceedings of the 28th Annual International ACM
SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2005),
pp. 154–161 (2005)
8. Konstan, J., Miller, B., Maltz, D., Herlocker, J., Gordon, L., Riedl, J.: GroupLens: apply-
ing collaborative filtering to Usenet news. Commununications of ACM 40(3), 77–87
(1997)
9. Kurosu, M., Ookawa, Y.: Effects of Negative Information on Acquiring Procedural Know-
ledge. In: Proceedings of the International Conference on Computers in Education (ICCE
2002), p. 1371 (2002)
10. Porter, M.: An Algorithm for Suffix Stripping. Program 14(3), 130–137 (1980)
Automatic Layout Generation for Digital Photo Albums:
A User Study
Francine B. Bergmann, Isabel H. Manssour, Milene S. Silveira,

and João Batista S. de Oliveira
Faculdade de Informática, Pontifícia Universidade Católica do Rio Grande do Sul

Avenida Ipiranga 6681, 90619-900, Porto Alegre, RS, Brazil
[email protected],
{isabel.manssour,milene.silveira,joao.souza}@pucrs.br
Abstract. The low cost and ubiquitousness of digital cameras allow ever larger
collections of photos. Although storage is not usually a problem, selection for
presentation requires time and effort. Studies show the existence of user groups
that never tried to create digital photo albums or were discouraged by the diffi-
culty of this task. Trying to minimize the effort to create albums, we developed
a prototype called Twingle. Using Twingle the creation of an album requires
dragging pictures into a page and the program is responsible for positioning and
resizing. To investigate whether such an automatic layout generation could be
an alternative to users, we made a study exploring Twingle with users from
several ages and professional areas. The related works, the performed user
study and its results are discussed here.
Keywords: digital photo albums, automatic layout generation.
1 Introduction
Before the age of digital cameras, several considerations limited the amount of photos
taken by most people and affected the practices adopted for their organization and
presentation: a small number of photos were taken and developed to be organized as
photo albums. The access to digital cameras favored the growth of photo collections
[2,8,11], and these much larger collections demand new tools and practices for their
organization.
It was observed [4] that even though the most accepted way of viewing pictures are
the printed albums, most people do not create them. Some never tried, and some of
those who tried were discouraged by the difficulty of creating an album [4]. Users of
commercial tools consider the creation of digital photo albums (DPA) a costly and
time consuming task [20], since these tools require manual edition of the pages by
placing pictures, text, backgrounds and decorative elements [17].
In this context, we believe that the automatic layout generation of DPAs (or part of
them) could benefit users who have difficulties to perform this task or do not want to
spend time at it. To test this, it is necessary to explore and understand the behavior of
118 F.B. Bergmann et al.
users of such tools and find out the needs that these tools should address. In this con-
text, we developed a prototype called Twingle that automates the process of DPA
production using automatic layout generation. Twingle tries to minimize the effort to
create albums, since it only requires dragging pictures into a page and the program is
responsible for positioning and resizing them.
In order to investigate whether such an automatic layout generation could be an al-
ternative to users, we made a study with people from several professional areas and
ages exploring Twingle. The results and a discussion about our findings, as well as
the related works, are presented in the following sections.
2 Related Works
Most people are photographed or receive photos at some point in life [18]. In this
context, several authors [4,6,7,14,16] tried to understand how people deal with their
photo collections and those studies show an evolution in the practices adopted during
the transition from non-digital to digital photos.
Photos organized in an album usually tell a story and people feel obliged to organ-
ize their pictures for family and friends. When they do not do that, they say that
“ought to” or “definitely planned to” [18]. Families usually select and arrange their
favorite photos into albums to present them personally to friends and relatives, but the
production of photo albums is considered complex and time consuming [4]. The same
study also stated that few people spent time organizing their digital photos [4], but
they usually "throw" their photos into a folder similarly to a shoe box where pictures
are often stored [14].
Another study shows that most people organized their digital photos into directo-
ries according to the convention "date/event" or similar [7]. To do this, they use inter-
faces similar to Windows Explorer, rejecting photo management software. A similar
study [1] shows that people place images into directories and feel more satisfied with
the organization of their digital photos than the non-digital, as they had "all in one
place", giving an illusion of organization, since the digital photos do not take up phys-
ical space.
To ease the organization and creation of photo albums, several studies [1,8,
11,13,14,16] have proposed methods and tools to manage large collections of images,
integrating resources such as timestamps, GPS data and face recognition information
as a way of clustering photos into collections. The goal is to speed up and ease the
user experience in time-consuming and tedious tasks [1].
In addition, to understand how people deal with their photo collections, several
works [1,5,6,9,16,21,22] focused on the creation of DPAs, and tools [10,19] have
been developed commercially to the same purpose. DPAs (or photobooks) are typical-
ly produced for printing. The result is usually a file to be sent to the owner of the
software, which produces a high-quality printing of the album. These tools (available
online or in desktop versions) allow for the free edition of the layout and use tem-
plates [10,19] to organize the images on the page. Others [3,9,20] provide solutions
for the generation of collages, solve problems of unfilled spaces in album pages, and
Automatic Layout Generation for Digital Photo Albums: A User Study 119
generate DPAs with only a few initial selections required. There are still other author-
ing systems that automate user tasks [5,22] like image grouping, background selec-
tion, automatic page layout and other functions.
3 User Study
To investigate whether a simple interactive tool would ease the production of photo
albums, helping to assuage the lack of time and knowledge of the users, a prototype
called Twingle - characterized by the automatic organization of the pictures and texts
in photo albums - was offered to a group of people.
The study had two phases: the first explores the first version of Twingle. In
this phase, several enhancement suggestions were given by users to improve their
experience with the production of DPAs. The second phase was performed with an
improved version of Twingle, containing the most relevant functionalities suggested.
Subsequent sections describe Twingle and the two phases of the study.
3.1 What Is Twingle?
Twingle is a prototype that tries to reduce the effort to create photo albums: the user
does not need to position or resize images on the page and the layout is done automat-
ically. It was implemented using HTML/CSS and JavaScript, allowing its use in any
web browser supporting these languages, and communicates with a server responsible
for the layout algorithm.
The creation of an album requires only dragging images into a page, as the pro-
gram on the server is responsible for positioning and resizing images using an algo-
rithm for automatic layout generation [12]. Briefly, the algorithm receives a set of
rectangular items to be placed on the page and estimates the amount of area needed by
each item, using a divide and conquer approach to put them all on the page. To each
added item, the procedure is performed again to position it in a visually pleasant way,
in which images and their assigned regions should have similar aspect ratios. Moreo-
ver, the page needs to be homogeneously covered, that is, items should be evenly
placed on the page and not cluttered on some region, not leaving excessive empty
space and filling the entire page as much as possible.
There are several tools and systems mentioned on Section 2 that may be similar to
Twingle, however none of them used an automatic layout generation for DPAs com-
posed of texts and images selected by the user. Works that provide an algorithm for
automatic layout generation [3,9,20] do not support text inclusion, whereas some
commercial tools allow the creation of DPAs, but only using templates [10,19]. Oth-
ers [1,5,22] have different purposes, automating the task of selecting photos or allow-
ing a different way of interaction [6] or organizing the pictures [16]. Thus, Twingle
was designed with a specific goal to investigate the reasons behind the difficulty of
producing and presenting DPAs.
3.2 First Phase
In the first phase, besides dragging images to compose a page, Twingle offered the
following features: start a new album, add new pages, view the album in full size,
generate a PDF file for printing, as well as delete an image from the page, add a cap-
tion to a picture or shuffle the images to obtain a new page layout. Figure 1 shows a
sample album.
Fig. 1. Page generated by Twingle as images and captions are inserted
Objective.
To study whether a simpler tool for page design would be well received by users. A
prototype was tested to produce DPAs just by dragging and dropping images to a
page, and layout was produced automatically.
Description.
This study was conducted with 15 users from several professional areas, such as com-
puter science, graphic design, architecture, management and psychology, among oth-
ers, and ages between 18 and 60 years.
A pre-interview showed that not all users like to make photo albums, and some of
those who like sometimes thought the task was too laborious and time consuming.
Thus, these users were organized into three groups (five users each), according to
their willingness to create DPAs: Group 1, users that like to make photo albums;
Group 2, users that would make photo albums if they knew how or if it were “practic-
al”; and, Group 3, users that do not like/do not wish to make photo albums.
To split the users into such groups, their answers to the following four questions
were analyzed: Generally, how do you present photos to your friends/family? Do you
like to make photo albums? Do you make traditional photo albums (print and place
photos in an album)? What is your opinion about making DPAs?
Each user had the opportunity to explore the software and produce albums with it.
As the purpose of the study was not to evaluate the tool but to allow users to evaluate
a simpler process of automatic creation of DPAs, they were free to ask questions to
the observer during the study.
Thereafter, an interview was made with open questions (the same for all groups)
and one variable question according to the specific group. The main idea was to iden-
tify, during the experience with Twingle, the users’ behavior and needs regarding
automatic tools for creating DPAs.
Discussion.
Each user was identified by a group number (1 to 3) and a letter (A to E). Among the
15 users, 9 showed previous discouragement about the creation of photo albums due
to time and work involved, including some of the users who like to make photo al-
bums, as the case of user 1D, which considers commercial tools too laborious for
people not used to designing documents: “The matter of design is troublesome. As I
am not a designer, it took me a month doing an album with another tool, changing the
orientation and photos on it.”.
Considering these users - who like to make albums - three of them are used to tools
for DPA generation. When asked if they would make an album using Twingle, they
said they possibly would, but more features need to be available. On the other hand,
those who had never used a tool to create albums said that they could use Twingle, as
mentioned by user 1C: “I really liked, it is very practical. It's easier than the pro-
grams I normally use to print photos, such as Word. I tried to organize everything
into Word or printed directly from Paint, Photoshop.”.
Among those in group 2, 2C uses a tool for DPA generation, and 2B did not know
these tools; the others knew but did not use them. After the experience of using Twin-
gle, only 2A had restrictions about its use, saying that he would use it only if addi-
tional resources were available.
In group 3, only two users previously used a tool for DPA generation: 3A and 3D
indicated that it is a laborious task and 3A also highlighted that when using the auto-
matic layout, the results were unsatisfactory: “I take a long time, until today I used
only one automatic tool, but the result was disastrous.”.
After using Twingle, when asked about its use for generating albums, 3D said he
would not use it because he prefers to have full control over the final layout: “I am a
user who likes to be in control (...). I like to adjust the size of the photos or place of
the caption (...) I like to move items to let it the way I want (...).”. User 3A says he
would make those without restrictions, and the others would require more features.
When asked whether the results of Twingle could be shown to friends and rela-
tives, only two users had restrictions to the final layout: 1A and 1B, which usually
make albums. User 1A indicates: “As it is today I don’t think so, it lacks some fea-
tures.”, and user 1B said that he preferred to show the photos in full screen. Among
the main reasons for presenting the results of Twingle for friends and family, the prac-
tical organization of the photos and the pleasing results were mentioned by users from
the three groups.
When asked whether they would print the results, there was a trend to use the result
in the computer instead of printing albums, and users mentioned emailing or publish-
ing on the internet. Some of the reasons for not printing are the absence of resources
they would like and the lack of knowledge about places for printing a PDF file with
photo quality.
122 F.B. Bergmann et al..
User comments show thaat for better results some features are especially importtant.
Among the most mentioned
d functionalities, were:
• change the position of a picture while keeping the general layout;

• resizing a specific photo;
• choose the page orientatiion (portrait or landscape);
• have background or ornaament effects available;
• be able to place general texts
t (not captions) on the page;
• free control of the spacin
ng between images;
• preview the results on th
he screen, including video and interaction.
3.3 Second Phase

To the second phase, Twiingle was refined according to the results from the ffirst
phase. The most relevant features
f demanded by users were added: upload (workking
both for photos and backgroounds), addition of general text/headlines, page orientatiion,
background insertion and innteractive possibilities to resize or change the position oof a
photo, and delete a whole page.
p Figure 2 shows a DPA created with the new versiion,
containing a headline, severral photos, text areas and captions.
.
Fig. 2. DPA
A album generated by new version of Twingle
Objective.
This phase was performed to verify if there was any difference in the results coom-
pared to the ones from the first phase and to check if the implementation of the nnew
features had turned the exp
perience of creating photo albums more enjoyable and lless
complex.
Description.
This phase was performed 8 months after the first one. Participants belong to the saame
professional and age groupss mentioned previously.
To determine whether thhere was an improvement of the user experience, we callled
new participants, as well as
a volunteers from each group from the previous researrch.
The methodology used an nd the classification of the new volunteers into grooups
remained the same from the previous phase. We had the participation of 8 people
from the first phase, and 6 new volunteers, totaling 14 participants: five users at
Group 1, five at Group 2 and four at Group 3.
The new participants received the same questionnaire of the previous phase ac-
cording to their classification into the groups, and the old ones answered a new list of
questions, among which we mention: What did you think of the new features of
Twingle? On what occasions would you use this tool?
Discussion.
For analysis and discussion of the results, the 8 volunteers from the first phase were
identified by their previous codes (A to E) and the 6 new users were numbered in
sequence (F to G).
Regarding to the new functionalities, 6 from the 8 participants of the first phase
acknowledged that both the use experience and the results provided by the new ver-
sion were better, and 11 from the 14 users said it was “quite good”.
Among the 6 new users, 3 never used photo album software and believe such work
is tiresome and time-consuming. When Twingle was used for the first time they ob-
served that the process may be quick and practical, and showed interest in using such
a tool to organize their pictures for family and friends. Participant 2F said: “I did not
know those tools, I had no idea how easy it was to produce an album. It is more inter-
esting to see the pictures like this, as I use to organize them by event and it would be
nice to put them all on the same page.”. 2G talks about Twingle as an image organiz-
er: “It's a very simple way I have to organize my photos. I found it very practical, fast,
and creating the album was an easy task.”.
Four participants suggested a new feature for Twingle: to generate pages automati-
cally without the user having to drag the images individually, but just select a folder
on their computer and let the program take care of producing the album. User 3A
says: “Since this is an automatic tool, it is interesting that I could insert all my photos
and the program returns the pages of the album.”.
Among the 4 users from the third group, those that dislike or do not consider pro-
ducing albums, only one continued disliking after using the new version of Twingle.
The reason was that this user does not enjoy taking pictures or manipulating them
digitally, and does not even publish them in social networks. That user said that he
would print the pages and send them to friends and relatives, but only if it was really
necessary and if the results were pleasant.
On the other hand, people used to full-featured album tools missed more capabili-
ties. User 1B said: “It is far from what I want from album software, as this layout
is not enough. I wish to place pictures in different positions and change their sizes
freely.”.
Regarding the new features, the possibility to use backgrounds was valued by us-
ers, and 7 from the 14 users wished to add titles to the top of the page, to be handled
differently from the current model (where texts are placed in a box and positioned as
images) and to be manipulated outside the reach of the automatic layout algorithm.
Suggestions for changes among the participants concerned page margins, text
fonts, and captions, as well as their color and alignment. Swapping the position of two
images was also suggested. When asked about the lack of features of the tool, only 2
from the 14 users found that Twingle needed more functionalities. The participant 3G
says: “I would use [the tool] to print photos, and for doing that I think the features it
has are enough.”.
3.4 Final Analysis
The study found a tendency of people to increase the use of digital content and, there-
fore, a need for ways to publish photo albums in digital format, both as PDF and
HTML files.
Some users were receptive to the idea of a tool for making albums in a fast and
simple way. In the first phase, 11 from the 15 users who do not use similar tools be-
gan to consider the possibility of using them, and respectively 10 from the 14 users of
the second phase, especially because of the increased speed given by the automatic
layout.
In the second phase 70% of the participants, who are frequent users of social
networks, were motivated to use Twingle if there was an application for Facebook,
revealing an interesting proposal to be investigated.
With respect to functionality, the amount of suggestions for improvement on the
second phase fell considerably compared to the first version. This leads us to believe
that the options present in the second version of Twingle could be in accordance with
the requirements of some users, making the user experience more usable and satisfac-
tory, and providing a more pleasing result.
Even with the evolution of Twingle, people who like to create photo albums con-
tinued feeling the lack of editing features, listing several new options to change the
tool and making Twingle deviate from its initial purpose. In this case, perhaps Twin-
gle’s layout algorithm may become an add-on to programs already on the market, as
an alternative capable of providing a solution for those who do not have time or
patience to spend on making photo albums.
The items described in this final analysis highlight the potential of automated tools
such as Twingle. While some people who enjoy all possible resources present in com-
mercial tools, there are also those who need or prefer to design a DPA quickly and
easily. Thus, it is essential to have an option to facilitate the construction of DPA while
providing a pleasant result that does not require further edition. These tools should be
capable of generating pages without the user having to manually select the photos, but
only indicating a folder containing images he considers interesting to the album to be
produced. The main intention is to reduce labor and time required to create DPAs.
Therefore, through Twingle, we notice that there is room for tools for different user
profiles, from those who enjoy more powerful features for editing, to those who just
want to point to a folder on the computer and have their album generated.
4 Conclusions and Future Work
In this work a literature review showed that it is not uncommon for people to be dis-
couraged to create photo albums because they consider it a costly and time consuming
task. Thus, we investigated, through a user study, if the automatic generation of DPA
would reduce the effort.
The results of the study show the potential of the automatic layout generation in
DPA applications. While some people enjoy controlling all possible resources of
commercial tools to edit the layout of DPAs, others prefer a faster and easier alterna-
tive, producing DPAs through less options and reduced effort. Thus, it is essential to
have an option to facilitate the construction of DPA while providing a pleasant result
that does not require further (or too much) edition. The main aim is to reduce issues
such as excessive labor and time pointed out by the people who need or are used to
create photo albums.
All users were enthusiastic about the possibility of generating automatic layout, but
they highlight features that a tool like this should provide: change the position of a
picture while keeping the general layout; photo resizing; choose the page orientation;
provide background ornament effects; allow the placement of general texts on the
page; provide free control of the spacing between pictures and the possibility of
inserting several photos at once.
After the user study, it was possible to verify that users were receptive to the idea
of a tool for making albums in a fast and simple way. More than 71% of them who do
not use similar tools began to consider the possibility of using them, especially
because of the increased speed and good results given by the automatic layout.
Currently we are developing a version of Twingle for Facebook, and exploring the
potential of its automatic layout generation for other applications, such as picture
panels and portfolios. In addition, we intend to investigate the use of tools like Twin-
gle in tablets, to identify alternative features made possible by the different forms of
interaction available on these devices.
Acknowledgements. The results presented in this paper were obtained in cooperation

with Hewlett-Packard Brazil Ltda., using the Brazilian Informatics Law (Law nº
8.248 of 1991).
References
1. Boll, S., Sandhaus, P., Scherp, A., Westermann, U.: Semantics, content, and structure of
many for the creation of personal photo albums. In: Proceedings of the 15th international
conference on Multimedia (MULTIMEDIA 2007), pp. 641–650. ACM, New York (2007)
2. Cunningham, S., Masoodian, M.: Identifying personal photo digital library features. In:
Proceedings of the 7th ACM/IEEE-CS Joint Conference on Digital Libraries (JCDL 2007),
pp. 400–401. ACM, New York (2007)
3. Diakopoulos, N., Essa, I.: Mediating photo collage authoring. In: Proceedings of the 18th
Annual ACM Symposium on User Interface Software and Technology (UIST 2005), pp.
183–186. ACM, New York (2005)
4. Frohlich, D., Kuchinsky, A., Pering, C., Don, A., Ariss, S.: Requirements for photoware.
In: Proceedings of the 2002 ACM Conference on Computer Supported Cooperative Work
(CSCW 2002), pp. 166–175. ACM, New York (2002)
5. Gao, Y., Atkins, C., Cheatle, P., Xiao, J., Zhang, X., Chao, H., Wu, P., Tretter, D., Slatter,
D., Carter, A., Penny, P., Willis, C.: MagicPhotobook: designer inspired, user perfected
photo albums. In: Proceedings of the 17th ACM International Conference on Multimedia
(MM 2009), pp. 979–980. ACM, New York (2009)
6. Jin, Y.K., Choi, S., Chung, A., Lee, I.M., Kim, M.C., Woo, J.: GIA: design of a gesture-
based interaction photo album. Personal Ubiquitous Comput. 8(3-4), 227–233 (2004)
7. Kirk, D., Sellen, A., Rother, C., Wood, K.: Understanding photowork. In: Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems (CHI 2006), pp. 761–770.
ACM, New York (2006)
8. Lee, B.N., Chen, W., Chang, E.Y.: Fotofiti: web service for photo management. In: Pro-
ceedings of the 14th Annual ACM International Conference on Multimedia
(MULTIMEDIA 2006), pp. 485–486. ACM, New York (2006)
9. Li, X., Takayama, Y., Shimizu, N., Mori, T.: An Aesthetic Solution for Photo Collection
Page Layout. In: International Conference on Knowledge and Systems Engineering (KSE
2009), pp. 33–38 (2009)
10. MyPublisher, http://mypublisher.com
11. O’Hare, N., Gurrin, C., Lee, H., Murphy, N., Smeaton, A.F., Jones, G.J.F.: My digital pho-
tos: where and when? In: Proceedings of the 13th Annual ACM International Conference
on Multimedia (MULTIMEDIA 2005), pp. 261–262. ACM, New York (2005)
12. Oliveira, J.B.: Two algorithms for automatic page layout and possible applications. Multi-
media Tools Appl. 43(3), 275–301 (2009)
13. PicasaWeb, http://picasaweb.google.com
14. Platt, J.C.: AutoAlbum: clustering digital photographs using probabilistic model merging.
In: Proceedings IEEE Workshop on Content-based Access of Image and Video Libraries,
pp. 96–100 (2000)
15. Platt, J.C., Czerwinski, M., Field, B.A.: PhotoTOC: automatic clustering for browsing per-
sonal photographs. In: Proceedings of the 2003 Joint Conference of the Fourth Internation-
al Conference on Information, Communications and Signal Processing, 2003 and the
Fourth Pacific Rim Conference on Multimedia, vol. 1, pp. 6–10 (2003)
16. Rodden, K., Wood, K.R.: How do people manage their digital photographs? In: Proceed-
ings of the SIGCHI Conference on Human Factors in Computing Systems (CHI 2003),
pp. 409–416. ACM, New York (2003)
17. Sandhaus, P., Thieme, S., Boll, S.: Processes of photobook production. Special Issue of
Multimedia Systems Journal on Canonical Processes of Media Production (2008)
18. Sarvas, R., Mäntylä, M., Turpeinen, M.: Human-Centric Design of Future Print Media. In:
Proceedings of the PulPaper 2007 Conference, Helsinki, Finland (2007)
19. Snapfish, http://www.snapfish.com
20. Wang, W., Teo, P., Muzzolini, R.: Automatic digital photobook making system. In: Pro-
ceedings of the SPIE., vol. 7540, USA (2010)
21. West, D., Quigley, A., Kay, J.: MEMENTO: a digital-physical scrapbook for memory
sharing. Personal Ubiquitous Computing 11(4), 313–328 (2007)
22. Xiao, J., Lyons, N., Atkins, C.B., Gao, Y., Chao, H., Zhang, X.: iPhotobook: creating pho-
to books on mobile devices. In: Proceedings of the International Conference on Multime-
dia (MM 2010), pp. 1551–1554. ACM, New York (2010)
Icon Design for Older Users of Project Management
Software
Christina Bröhl, Jennifer Bützler, Nicole Jochems, and Christopher M. Schlick
Institute of Industrial Engineering and Ergonomics, RWTH Aachen University

{c.broehl,j.buetzler,n.jochems,c.schlick}@iaw.rwth-aachen.de
Abstract. Working in projects is an important part of many jobs in service in-

dustry. Due to their knowledge and experience project planning is often accom-
plished by older employees. Therefore, and with regard to the demographic
change an integration of the needs of older employees into the work environ-
ment is required. Common to most IT-based systems, including project man-
agement software, is the use of icons. To investigate different icon sets in
project management software, regarding age related differences, two studies
were conducted. The first study aimed at investigating two different icon sets
regarding execution times and eye movements in an applied setting. The second
study consisted of a questionnaire where subjects had to map different icons to
their corresponding functions and had to compare these icons regarding their in-
tuitiveness. Results revealed that older users profit from an icon design which is
low in complexity but no impact by different icon designs was found for
younger users.
Keywords: Usability, icon design, human-computer interaction, project man-

agement, eye-tracking, age differences.
1 Introduction
Well-organized project planning and subsequent execution of the project plan is no-
wadays a central predictor for the success of a project and thus a main competition
factor for companies. Since projects are often linked to high risks as they are highly
complex and constrained by tight time schedules [1], the planning process is often
supported by computers. Therefore, the number of software products which support
project planning is growing. These project management software applications offer
different functions (e.g. project scheduling, cost planning, resource assignment plan-
ning, quality management, documentation) and differ in their visualization and plan-
ning methods (e.g. network diagram, Gantt chart). To navigate through the system
and to generate different functions most commercial off-the-shelf project management
software applications use icons. An icon is a small picture or a graphical symbol,
which provides an illustration of a function or file on the system. Such an icon should
transmit information about its underlying function in an abbreviated, simplified form
and should not depend on letters or words [2]. Especially software tools which are
128 C. Bröhl et al.
unfamiliar to the user benefit from a well designed icon structure as the ease of use is
facilitated which enhances the acceptance and further use of the tool.
Since project management places many demands on a person, like professional,
methodological, expertise and life experience, it is often accomplished by older
employees. Due to age-related changes, the interaction with computers is often partic-
ularly straining for older employees. These changes include decreases of fluid cogni-
tive abilities like abstract reasoning [3], mental speed and short term memory storage
[4], perceptual impairments like vision [5] as well as impairments in performing com-
plex motor activities [6] which may lead to longer movement times [7] and difficul-
ties in performing computer mouse tasks [8]. Although different people experience
these impairments at different points in time and to different degrees, older people in
general must deal with these functional limitations [9].
The aim of the current studies was an age differentiated usability evaluation of two
different icon sets within a project management software application. The two icon
sets that were used differed in their appearance: Icon set A included very simple icons
which only consisted of three colors and were designed flat, whereas Icon set B in-
cluded icons that had a more realistic appearance in that the icons were more colorful
and had a glossy surface that was supposed to simulate three-dimensional objects.
Within the first study (study I) participants generated typical project management
tasks with both icon sets and results were analyzed regarding execution times and eye
movements. The second study (study II) consisted of an online questionnaire where at
first subjects had to map icons of both sets to their corresponding function and at
second had to give a rating of icon intuitiviness.
2 Literature Review
2.1 Icon Design for Aging Users
Interfaces using icons aim at reducing complexity for the user as the mental workload
is decreased when the icons are designed properly [10]. Icons represent information
for the user graphically because they express semantic information not through
the use of words but via nonverbal cues. Because of impairments in cognitive and
perceptual functions which can go along with the aging process, older users may have
difficulties in information processing [11] and hence in interpreting icons. Until now
little has been reported in the research literature about icon design for the elderly.
Mertens et al. [12] investigated the graphical presentation of icons and it could be
shown that among elderly, the use of photos compared to pictograms or clip art leads
to a higher recognition rate. However, the study was conducted using a hardcopy
questionnaire and not a real user interface. McDougall, de Bruijn and Curry [13]
found out that two parameters which play a critical role when studying icon design
are the concreteness and the complexity of icons. Concreteness in this context means
how closely the icon resembles items in the real world and complexity refers to
amount of visual detail depicted in the icon. Schröder and Ziefle [11] investigated the
effects of icon concreteness and complexity on semantic transparency in younger and
Icon Design for Older Users of Project Management Software 129
older users of mobile devices. Results showed that concreteness of an icon has a sig-
nificant effect on subject’s confirmatory response per item and that this effect does
not vanish with familiarity of the icon. For icon complexity there was no significant
difference in subject’s confirmatory responses but subjects had higher reaction times
when complex icons needed to be processed instead of simple ones. Regarding age
related differences an effect for the processing time was found but no effect of the
icon design so that the authors recommend a uniform design approach.
To build on existing research by exploring age-related differences in the effects of
concreteness and complexity two different icon sets were investigated in this study
which differed in the level of complexity and concreteness. Current research was
extended by using icons of a commericial off-the-shelf project management software
and studying the usability of these icons in an applied setting and in a questionnaire.
2.2 Eye-Tracking Metrics
Research has shown that eye-tracking measures can be used to detect usability prob-
lems in various contexts [14-16]. A number of eye movement metrics have been used
to measure usability [16-17]. The mean fixation duration is one eye movement para-
meter that is often analyzed in usability studies. According to Goldberg and Kotval
[14] longer fixation durations are an indicator for the difficulty to extract information
from a display.
Concerning age differences in mean fixation durations most studies report higher
mean fixation durations for older subjects. When analyzing navigational behavior on
web pages, for example, Fukuda and Bubb [18] found that subjects aged between 62
and 74 years had longer fixation durations than younger participants aged between 17
and 29 years. Moreover, Hill et al. [19] investigated computer expertise when using
the web within older subjects (70-93 years) and found that older novices had signifi-
cant higher mean fixation durations than older experts.
Studies from other fields also address the effect of aging on eye-tracking metrics.
In reading studies higher fixation durations were often found for older subjects
[20-21]. However, there are also few studies that report no age differences regarding
the fixation durations. Veiel et al. [22], for example, investigated age differences in
the perception of visual stimuli and found no age difference regarding the fixation
durations.
In line with these studies we used the mean fixation durations as dependent varia-
ble of the eye movements. Because the icons were arranged in the menu part of the
software we furthermore analyzed the relative amount of fixation durations in menu
parts.
3 Method
Dependent variables in study I were the execution time, the mean fixation durations
and the relative amount of fixation durations in the menu area. Study II was analyzed
regarding errors made in a mapping task and regarding ratings in intuitiveness for the
two different icon sets as dependent
d variables. Icons were selected from an existting
commercial project manageement software and differed regarding their complexity and
concreteness. Table 1 showws the icons of both sets used in this study. Icon set A ccon-
sisted of very simply icons which were low in complexity and concreteness since the
icons only consisted of threee colors and were designed in bi-dimensional space. IIcon
set B included icons that haad a more realistic appearance in that the icons were m
more
colorful and had a glossy surface that was supposed to simulate three-dimensioonal
space and was therefore rateed as being high in complexity and concreteness.
Tab
ble 1. Icon sets investigated in the study
Meaning Icon set A Icon set B

Save
Reload
Undo
Redo
Settings
Progress evvaluation
Insert activ
vity
Delete
Upgrade
Degrade
Collective task
Type of tassk
Critical
Done
Started
Relationship
Zoom out
Zoom in
3.1 Participants
Altogether 10 subjects partticipated in study I and a sample of 19 subjects in studyy II.

Subjects were divided in tw
wo age groups. In study I age ranged from 29 to 37 yearrs in
the younger age group (meean=32.00, SD=3.08) and from 43 to 60 in the older age
group (mean=52.40, SD=7 7.13). In study II subjects in the younger age group w were
o (mean=30.55, SD=4.80) and between 42 and 56 in the
between 21 and 38 years old
older age group (mean=49.25, SD=5.06). All participants had experience with projject
work and project managem ment software but were not familiar with the design of the
icons used in the study.
3.2 Apparatus
The experiment was conducted at a 22”-inch LCD TFT-widescreen-monitor. Eye
movements were measured during the task using a Tobii X120 eye tracking system.
Fixation durations were measured according to data typically found in the literature
for usability studies, e.g. [16-17]. With regard to these studies the mimimum time for
a fixation was set to be at least 100ms and the eye-in-head position had to be in a
threshold of dispersion of about 2°.
3.3 Procedure and Task
Study I was conducted in a laboratory where subjects were seated in front of the mon-
itor with a viewing distance of 500 mm. After calibrating the eye-tracking system, the
subjects executed the tasks using the software with icon set A. This procedure was
repeated for the software with icon set B as well. The representation of the icon set
was permutated. The task consisted of typical steps (subtasks) when developing a
project plan. Table 2 gives an overview of the nine sequential subtasks.
Table 2. Specification of the task
Subtask 1 Generate a new project plan

Subtask 2 Create five new activities and change duration of the
activities
Subtask 3 Insert two new activities, change duration and level of
the activities
Subtask 4 Insert four new activities, change duration and level of
the activities
Subtask 5 Link activities
Subtask 6 Insert new milestone and link it to activity
Subtask 9 Assign responsible person
Study II was conducted using a questionnaire which could be completed online. In

the first part of the questionnaire subjects had to map icons to their underlying func-
tion. Therefore, icons were arranged in blocks of 6 icons plus one icon which had no
corresponding function serving as control item to make possible contradictions visi-
ble. In the second part of the questionnaire subjects had to rate icons according to
their intuitiveness. Thereby, a function was given and subjects could choose between
three answer options: Icon of set A, icon of set B and as a third option subjects could
indicate whether the icon of set A or set B makes no difference to them.
3.4 Statistical Analysis
To analyze the data of study I statistically repeated measures analysis of variance was
used. The age of the participants served as a between-group factor. Study II was ana-
lyzed using two-way ANOVA with age group and icon set as independent variables.
Furthermore, a correlation analysis was used to investigate whether there is a relation-
ship in the rating of intuitiveness and the performance in the mapping task. Because
data for the correlation analysis were not measured at the interval level, correlation
coefficients were calculated using Spearman’s rho. According to Cohen & Cohen [23]
effect sizes can be classified into small (r=.10), medium (r=.30) and high (r=0.50).
The level of significance was set to α = 0.05.
4 Results
4.1 Execution Times
To analyze the execution time similar subtasks were analyzed by combining the data.
The analysis of the execution times resulted in significant main effect for the age
group for subtasks 2-4 “Insert two new activities, change duration and level of the
activities” (F(1,8)=11,224; p=0,010) and for subtasks 6-8 “Insert new milestone and
link it to activity” (F(1,8)=7,008; p=0,029). In both tasks participants in the older age
group needed significantly more time to execute the tasks than participants in the
younger age group. For the main effect of the icon set as well as for the interaction
effect between age group and icon set no significant effects were found.
4.2 Mean Fixation Durations and Relative Amount of Fixation Durations in

Menu Parts
Mean fixation durations and the relative amount of the fixation durations in the menu
part could not be analyzed with regard to age differences because data of the older age
group were insufficient. This may be attributed to the problem of droopy eyelids.
Droopy eyelids of participants when using eye-tracking metrics cause data loss be-
cause the eyelids or the eyelashes cover the pupil in specific gaze directions. The
problem of droopy eyelids is a matter of individual differences and it grows with the
age of participants [24]. Consequently, this might have been the case with data col-
lected in this study.
Regarding the mean fixation durations no significant effect was found for the icon
set when analyzing the graphical user interface as a whole. In order to investigate for
which icon set more time was spent fixating it, the relative amount of fixation dura-
tions in the menu part where the icons are arranged was analyzed. Data revealed no
significant effect when looking at the differences between the two icon sets when the
mean fixation durations in the menu part were analyzed.
4.3 Questionnaire
The first part of the questionnaire was analyzed by measuring the performance in the
mapping task for each icon set. Results indicate that subjects in the older age group
made more errors with icon set B compared to subjects in the younger age group
(F(1,34)=8.816; p=0,005). No age effect was found for icon set A.
When analyzing the second part of the questionnaire scores revealed that there was
no significant difference in the intuitiveness ratings for one icon set or the other. To
analyze whether intuitiveness ratings of the subjects are linked to the errors they made
in the first part of the questionnaire a correlation analysis was conducted. According
to the classification by Cohen and Cohen [23] a medium to high effect size was found
for icons which were rated as being low in intuitiveness and errors made in the map-
ping task r=.429.
5 Discussion
The aim of the study was an age-differentiated analysis of two icon sets. The investi-
gation of the execution times resulted in a significant main effect for the age group
for two of the subtasks. In both tasks participants in the older age group needed sig-
nificantly more time to execute the tasks than participants in the younger age group.
This effect may be attributed to declines in perceptual and cognitive processes [3,4]
like the hand eye coordination which may have an influence in this study because
participants performed the tasks with a computer mouse. Regarding the execution
time no effects were found for the icon set. So the complexity and the concreteness of
the two investigated icon sets make no difference in accomplishing the subtasks used
in this study. However, a tendency for the older age group was found in the scores to
have shorter execution times when working with icon set A.
Results of eye-tracking data analyzed by means of fixation durations of the inter-
face as a whole and the relative amount of fixation durations in menu parts did not
differ for the two icon sets. As the fixation duration is assumed to be an indicator of
the difficulty to extract information from a display it can be concluded that the design
of the icons used in this study has no effect on the ability to extract information.
Unfortunately, we were not able to investigate eye-tracking data with regard to age
differences because we had difficulties in collecting data from older subjects. In fu-
ture research we are trying to improve the measuring procedure so that an age diffe-
rentiated analysis will be possible.
When analyzing the questionnaire results revealed that older subjects had more dif-
ficulties in the mapping task when working with icon set B. Together with the tenden-
cy for the older age group to have shorter execution times when working with icon set
A this could lead to the conclusion that the older age group performs better with icons
that are not high in complexity and concreteness. An explanation for this finding
might be that the older age group already used to work with rather simple icons in the
past since they grew up with early computers which used such an icon design in com-
parison to the younger age group, which showed no tendency for one icon set or the
other. Regarding the intuitiveness scores results showed that fewer errors were made
in the mapping task with the icon set which was rated as being higher in intuitiveness.
This finding is in accordance with existing literature [2] and shows that icon design
places a crucial role in human-computer interaction.
Acknowledgements. The research was funded by the German Research Foundation

according to the transfer project (SCHL 1805/6-1).
References
1. Shtub, A., Bard, J.F., Globerson, S.: Project Management: Processes, Methodologies, and
Economics. Pearson Prentice Hall, Upper Saddle River (2005)
2. Horton, W.: Das Icon-Buch: Entwurf und Gestaltung visueller Symbole und Zeichen. Ad-
dison-Wesley, Bonn (1994)
3. Rodrigue, K.M., Kennedy, K.M.: The cognitive consequences of structural changes to the
aging brain. In: Schaie, K.W., Willis, S.L. (eds.) Handbook of the Psychology of Aging,
7th edn., pp. 73–91. Elsevier, San Diego (2011)
4. Anstey, K.J., Low, L.F.: Normal cognitive changes in aging. Australian Family Physi-
cian 33, 783–787 (2004)
5. Goldberg, E.B.: Sensation & Perception. Thomson Wadsworth, Canada (2007)
6. Newell, K.M., Vaillancourt, D.E., Sosnoff, J.J.: Aging, complexity, and motor perfor-
mance. In: Birren, J.E., Schaie, K.W. (eds.) Handbook of the Psychology of Aging, 6th
edn., pp. 163–182. Elsevier, San Diego (2006)
7. Cerella, J.: Aging and information processing rate. In: Birren, J.E., Schaie, K.W. (eds.)
Handbook of the Psychology of Aging, pp. 201–221. Academic Press, New York (1990)
8. Smith, M.W., Sharit, J., Czaja, S.J.: Aging, motor control, and the performance of comput-
er mouse tasks. Human Factors 41, 389–396 (1999)
9. Pijukkana, K., Sahachaisaeree, N.: Graphical Design and Functional Perception on Tech-
nology-Driven Products: Case Study on Mobile Usage of the Elderly. Procedia - Social
and Behavioral Sciences 42, 264–270 (2010)
10. Lodding, K.: Iconic interfacing. Graph. Appl. 3, 11–20 (1983)
11. Schröder, S., Ziefle, M.: Effects of icon concreteness and complexity on semantic transpa-
rency: Younger vs. Older users. In: Miesenberger, K., Klaus, J., Zagler, W.L., Karshmer,
A.I. (eds.) ICCHP 2008. LNCS, vol. 5105, pp. 90–97. Springer, Heidelberg (2008)
12. Mertens, A., Brandl, C., Przybysz, P., Koch-Körfges, D., Schlick, C.: Design recommen-
dations for the creation of icons for the elderly. In: Soares, M., Jacobs, K. (eds.) Work: A
Journal of Prevention, Assessment and Rehabilitation - IEA 2012: 18th World congress on
Ergonomics - Designing a Sustainable Future, Amsterdam, p. 41 (2012)
13. McDougall, S.J., de Bruijn, O., Curry, M.B.: Exploring the effects of icon characteristics
on user performance: the role of icon concreteness, complexity, and distinctiveness. Jour-
nal of Experimental Psychology Applied 6(4), 291–306 (2000)
14. Goldberg, J., Kotval, X.P.: Computer interface evaluation using eye movements: methods
and constructs. International Journal of Industrial Ergonomics 24(6), 631–645 (1999)
15. Cowen, L., Ball, L.J., Delin, J.: An eye movement analysis of webpage usability. In:
Faulkner, X., Finlay, J., Detienne, F. (eds.) People and Computers XVI - Memorable yet
Invisible, pp. 317–335. Springer, London (2002)
16. Jacob, R.J.K., Karn, K.S.: Eye tracking in human–computer interaction and usability re-
search: Ready to deliver the promises. In: Hyona, J., Radach, R., Deubel, H. (eds.) The
Mind’s Eye: Cognitive and Applied Aspects of Eye Movement Research, pp. 573–605.
Elsevier (2003)
17. Poole, A., Ball, L.J.: Eye tracking in HCI and usability research: Current status and future
prospects. In: Gahoui, C. (ed.) Encyclopedia of Human-Computer-Interaction, pp. 211–219.
Idea Group Reference, Hersey (2005)
18. Fukuda, R., Bubb, H.: Eye tracking study on web-use: Comparison between younger
and elderly users in case of search task with electronic timetable service. PsychNology
Journal 1(3), 202–228 (2003)
19. Hill, R., Dickinson, A., Arnott, J., Gregor, P., McIver, L.: Older web users’ eye move-
ments: experience counts. In: Proc of the SIGCHI Conference on Human Factors in Com-
puting Systems, pp. 1151–1160. ACM, Vancouver (2011)
20. Kliegl, R., Grabner, E., Rolfs, M., Engbert, R.: Length, frequency, and predictability ef-
fects of words on eye movements in reading. European Journal of Cognitive Psycholo-
gy 16(1-2), 262–284 (2004)
21. Rayner, K., Castelhano, M., Yang, J.: Preview benefit during eye fixations in reading for
older and younger readers. Psychology and Aging 25(3), 714–718 (2010)
22. Veiel, L.L., Storandt, M., Abrams, R.A.: Visual search for change in older adults. Psy-
chology and Aging 21(4), 754–762 (2006)
23. Cohen, J., Cohen, P.: Applied multiple regression/correlation for the behavioral sciences.
Erlbaum, Hillsdale (1983)
24. Holmqvist, K., Nyström, M., Andersson, R., Dewhurst, R., Jarodzka, H., van de Weijer, J.:
Eye tracking: A comprehensive guide to methods and measures. Oxford University Press,
Oxford (2011)
Search Strategies in Hypermedia Navigation and Spatial
Abilities: A Comparison with Physical Navigation
Aurélie Brouwers
Research Fellow F.R.S-F.N.R.S, Université catholique de Louvain, Belgium

[email protected]
Abstract. This article focuses on spatial abilities mobilized during hypertext

navigation. Based on the evidence that spatial cognition plays a central role in
navigation, we present an experiment involving information search tasks both in
physical environment and in hypermedia environment. We investigate how us-
ers make use of their spatial abilities to search information in hypermedia, by
comparing their performances in hypermedia navigation and physical naviga-
tion. As data collection and analysis are still in progress, we present preliminary
results based on available data.
Keywords: hypermedia, navigation, visuospatial abilities, spatial cognition.
1 Introduction
The term “navigation” has been used for decades to describe the interaction between a
user and a hypertext system [1]. The use of this spatial metaphor is backed by at least
two kinds of research results. On the one hand, a number of studies [2]–[4] have
uncovered correlations between subject performance in hypertext navigation and per-
formance in standardized tests of visuospatial cognitive abilities. On the other hand,
hypertext users have been shown to use spatial metaphors extensively and systemati-
cally when they speak of their interaction with hypertext [5], [6], indicating that these
metaphors play a role in the way users think about this interaction.
This article focuses on the role of spatial cognition and specifically visuospatial ab-
ilities in information search in a hypertext environment. We investigate how users
make use of their spatial abilities to search information in hypermedia, by comparing
their performances in hypermedia navigation and physical navigation. First, we will
review relevant research in spatial cognition and information search. Next, we will
present an experiment requiring subjects to interact with a physical environment and a
hypertext environment. As data collection and analysis are still in progress, we will
present preliminary results based on available date.
2 Spatial Cognition and Cognitive Mapping

The concept of cognitive map is central to the field of spatial cognition research. Sie-
gel and White distinguished between landmarks, routes and survey representations as
Search Strategies in Hypermedia Navigation and Spatial Abilities 137
components of cognitive maps. Landmarks are features of the environment that are
saillant to the subject. Routes connect different landmarks. Survey knowledge orga-
nizes routes into configurations that provide the subject with an overview of their
environment. Siegel and White described a “Main Sequence” for the development of
spatial knowledge that goes “from landmarks, to route-maps, to survey-maps” as “a
process of going from association to structure, and of deriving simultaneity from suc-
cessively” [7]. As Siegel and White [7] pointed, cognitive maps are not very much
“map-like”: they tend to be fragmented and are prone to distortion. Hence, we prefer
to speak about spatial mental representations.
Moeser [8] compared the navigation of student nurses who has been working in
hospital for three years and the navigation of naive subjects who studied the hospital
map. He showed that the naive subjects performed significantly better on objective
measures of cognitive mapping (distance estimation, drawing plan, etc.) than did the
nurses. He concluded that the extended and intensive experience of an environment
doesn’t systematically provide the individual with survey knowledge of it. Each per-
son tends to construct mental representations that include relevant elements for their
own use of the environment. For these student nurses, landmarks and route know-
ledge seemed sufficient to navigate their professional environment.
In hypermedia research, the concept of cognitive map has been used to describe the
user’s mental representation of the hyperdocument’s organization [9]–[12]. Following
Siegel and White’s model, several studies in hypermedia research (e.g. Sedig et
al.[13]; Kim & Hirtle [11]; Edwards & Hardman [14]) relied on the view that land-
marks, routes and survey knowledge are acquired successively. This implies that
survey knowledge is necessarily the most advanced form of spatial knowledge. Kim
& Hirtle [11] and Edwards and Hardman [14] emphasized that subjects with survey
knowledge are rarely disoriented in hypermedia.
Spatial mental representations of the environment can be constructed either
through primary learning (i.e. by observing the actual environment directly) or
through secondary learning (i.e. by means of an external presentation, such as a map)
[15]. Secondary learning may represent an alternative way of acquiring spatial know-
ledge (compared to Siegel and White’s “Main Sequence”), as survey knowledge
may be acquired directly from a map, without being derived from route knowledge
acquired through navigation. While this distinction is relevant to physical environ-
ments, its application to hypermedia environments is problematic, as the organization
of hyperdocuments only exists in representations, be they external (e.g. concept maps,
or the system’s interface itself) or internal (the mental representation constructed by
the user). Hence, the very notion of primary learning seems void in the context of
hypermedia. However, the notion of learning routes from navigating (which can in
turn be elaborated into survey knowledge) as opposed to learning survey configura-
tions from a site map (which in turn can be converted into specific routes from one
page to another), stays relevant in this context.
People differ in their spatial abilities. Research by Goldin and Thordyke [16]
showed that the difference between “poor cognitive mappers” (subjects with lower
spatial abilities) and “good cognitive mappers” (subjects with higher spatial abilities)
138 A. Brouwers
may lie in their ability to construct spatial mental representations of their environ-
ment, both from navigation and from maps. However, when forced to study a map so
as to construct a reliable representation of a given environment, “poor cognitive map-
pers” navigated as efficiently as “good cognitive mappers” suggesting that “when
utilizing equally accurate knowledge, poor cognitive mappers can navigate as well as
good cognitive mappers”[16].
Spatial visualization abilities of hypermedia users had been shown to predict their
navigation performance by previous studies [2]–[4]. Spatial visualization can be de-
fined as an internal spatial ability. Such internal abilities, requiring “a purely mental
effort to obtain the correct answer”, were distinguished by Dahlbäck et al. [4] from
external spatial abilities that involve the active manipulation of physical objects (such
as the arrangement of block in a pattern defined by a picture of arranged blocks),
which did not correlate with performance in hypertext information search.
Carroll [17] showed that the spatial visualization factor assessed by these tests in-
volves the cognitive processing required to mentally encode and manipulate spatial
shapes. He also noted that the successful completion of tests such as VZ-1 and VZ-2
[18] requires a mental representation in three dimensions. This would explain his
observation that individuals who perform well in task in 2-D environments tend to be
also performing well in tasks in 3-D environments.
Interestingly, the tests used to assess spatial visualization abilities actually require
subjects to process external representations (pictures) into mental representations they
need to manipulate internally in order to complete the task. Constructing a reliable
mental model from an external representation (as opposed to using a mental model to
act in the real world, or to using an external representation to act in the real world).
Research on the use of concept maps to represent hypermedia structure has yielded
ambiguous results. Dee-Lucas & Larkin [19] showed that a structured overview of
hypertext had beneficial effects on the memorization on the document’s textual con-
tents. Vörös et al. [20] showed that concept maps helped subjects with poor spatial
abilities to better remember the hypermedia structure after their navigation. However,
their study did not test whether subjects with more accurate cognitive maps of the
hypertext navigated more efficiently (i.e. performed better on search tasks). In Nils-
son & Mayer’s [21] studies, subjects who used a hyperdocument with a non-clickable
map performed better than subjects who used the same hyperdocument without the
map at first, but the comparison inverted over time, showing the map could become
cumbersome during navigation. Scott & Schwartz [22] showed that while processing
a hypertext concept map generated additional cognitive load, the extra effort paid off
when the spatial relationships depicted by the map matched the semantic relationship
of the hypertext’s contents.
3 Research Question and Hypothesis
Our research question is the following how do spatial visualization abilities come to
play in hypermedia navigation?
Spatial visualization abilities could play a role (1) in the construction of a mental
model of the hypertext’s organization, (2) in the use of this mental model to make
navigational choices, or (3) in the use of the hypertext interface and navigation tools
(as an external representation of its organization) to make navigational choices.
Based on our literature review, we hypothesize that spatial visualization abilities
mainly play a role during the construction of the hypertext mental model. To test this
hypothesis, we set up an experiment requiring subjects to perform a series of informa-
tion search tasks in a physical and a hypertext environment. Our experimental proto-
col allows us to proceed with intra-subject comparisons of navigation behaviors in
hypertext and in real life, as well as inter-subjects comparisons based on their level of
spatial visualization abilities (cf. infra). It also allows us to clarify the role of map
usage during hypertext navigation.
4 Method
4.1 Fields of Observation
Our experimental protocol includes tasks to be completed in a large-scale physical

environment and a hypermedia environment. The Planckendael (Belgium) zoo was
chosen as the physical environment, as it met all of our requirements. In addition to
making different animals available for public display, the zoo includes and organizes
information (about animals) spatially (e.g. in the form of posters presented at each
animal’s enclosure). It clearly bounded space corresponds to a bounded informational
space. In this context, information search tasks can be designed, involving the
processing of both spatial and semantic information by subjects. Also, it is an outdoor
environment, which allows GPS tracking of our subjects.
Fig. 1. Home page

140 A. Brouwers
Fig. 2. Animal page
We designed a forty-five-page hypermedia environment on wildlife, including in-

formation on different animal species that are not visible at the zoo. Each animal is
presented on a separate page. Pages presenting an animal species are grouped accord-
ing to the continent on which the species lives. This hierarchical structure mimics that
of the zoo, which dedicates a part of its space to each continent.
4.2 Participants
Students enrolled at the Université Catholique de Louvain (Belgium) are currently

being recruited as voluntary participants in our experiment. Ten subjects have com-
pleted the experiment so far. Twenty more subjects will be tested in the near future.
4.3 Individual Variables
Subjects are tested with respect to two types of cognitive abilities. On the one hand,
we use VZ-1 and VZ-2 tests from the kit of factor-referenced cognitive tests [18] to
assess our subjects spatial visualization abilities.
On the other hand, we test our subjects with a discourse comprehension test [23].
As information search in hypermedia involves the processing of both semantic and
spatial information [24], we use this test to control our subjects verbal abilities.
4.4 Tasks
Subjects are required to perform five information search tasks in the physical
environment, and seven information search tasks in the hypertext environment.
For each task subjects are instructed to answer a specific question about an animal
species, the answer of which is located at the animal’s enclosure in the zoo, or on
the animal’s page in the hypertext. The questions used in the two environments are
different.
4.5 Procedure
Prior to the navigation, half of the subjects (randomly selected) are asked to study the
map of the environment they are about to interact with (subjects either study both
maps, or none). Subjects are tested on their memorized map, and the study is repeated
until they learned the map perfectly. According to our hypothesis, subjects with low
visual-spatial abilities who studied the map should perform similarly to with high
visual-spatial abilities, as visuospatial abilities are supposed to play a decisive role in
the construction of the mental model of the environment.
During the navigation, subjects are provided with the task questions one at a time.
When the subject thinks to have found the answer, he notifies the experimenter. If the
answer is correct, he is given the second task question. If not, the subject has to con-
tinue searching. While searching subjects are asked to verbalize their navigation in-
tentions before executing them. In physical environment, subjects are asked to
estimate the direction of the searched information. During navigation, subjects have
the opportunity to consult a map. The hypertext map can be accessed though a tab
(preventing concurrent navigation and map viewing) and in the zoo, subjects are re-
quired to stop whenever they want to view the map. The aim is to force subjects to
verbalize their use of the map.
In each condition (map-study vs. no-map-study), half the subjects navigate the zoo
prior to the hypertext, and the other half navigate the hypertext prior to the zoo, in
order to control the training effect of one navigation on the other.
Finally, subjects are submitted to the spatial and verbal ability tests after the two
navigation sessions, in order to avoid a sense of demotivation due to poor perfor-
mance in these tests.
4.6 Data Collection
In the zoo, navigation is recorded using a Looxcie, a mini camera equipped with a
microphone. A GPS tracker, Map My run, an android application, records the routes.
Hypermedia navigation tasks are recorded using TechSmith Morae, a software suite
that captures screen and webcam activity, as well as mouse and keyboard input (e.g.
clicks, menu selection, etc.).
142 A. Brouwers
5 Preliminary Results
At the time of writing, we have tested ten subjects and have started to analyze the
collected data for nine of them. Our current analysis focuses on the influence of
the study of the map on navigation behavior and efficiency. Table 1 presents some of
the data that were collected for these subjects.
Table 1. First data
Zoo Hypertext
Sub- Condition VZ-1+ Total Total Map Total Map

ject VZ-2 search time distance views search time views
score
1 no-map- 1 01:08:00 5,57 km 7 00:28:38 0
study
2 no-map- 6 00:57:35 3,59 km 12 00:11:40 10
study
3 no-map- 12 00:45:00 6,49 km 8 00:27:57 0
study
4 no-map- 16 00:59:21 4,60 km 6 00:09:58 0
study
5 no-map- 17 01:12:00 4,13 km 19 00:30:00 0
study
6 map-study 2 01:10:00 5,50 km 12 00:12:18 0
7 map-study 12 00:38:42 2,54 km 4 00:11:30 7
8 map-study 19 01:04:00 6,68 km 8 00:09:09 0
9 map-study 22 00:21:08 3,54 km 7 00:18:57 0
Contrary to previous research results, scores to the spatial visualization tests did
not correlate significantly with time spent navigating in either of the environments, or
with distance traveled in the zoo (using Pearson’s correlation coefficient), although
this may be due to the low number of subjects (N=9) in our current dataset.
Our current observations for the physical environment are not consistent with
previous studies [16]. Indeed, subjects who studied the map of the zoo are not neces-
sarily more efficient than others (Meanmap-study = 48:28; Meanno-map-study = 1:00:23).
However, subjects who studied the site map prior to navigation completed their tasks
more quickly than the other subjects (Meanmap-study = 12:58; Meanno-map-study = 21:39).
We observed that the subject with the lowest results in VZ-1 and VZ-2 test in the
map-study condition was able to use her survey knowledge in the zoo, as she was able
to accurately estimate the direction of the searched item. However, she was unable to
convert her survey knowledge into a proper route plan, as the geodesic distance be-
tween her current location and her planned destination did not correspond to an actual
path in the zoo. In this case, a survey vision of the environment proved to be of little
help to our subject. In the hypertext environment, she successively browsed the
main nodes of the interface (the continent pages) to gain visual access to the different
links (routes) to specific animal pages, until she identified a potential target node. In
this case, she used her memory of the overall (survey) organizing principle of the
hypertext to structure her use of the interface (as an external representation of the
system’s structure).
Finally, the majority of our subjects did not use the sitemap tab while navigating.
When asked why they didn’t, most subjects replied that they feared the concept map
would confuse them, as they felt sitemaps are generally useless.
6 Conclusions and Future Work
The experiment presented in this paper aims at better understanding the role of vi-
suospatial abilities in hypertext navigation. To do this, we compare the information
search strategies developed by subjects in a physical environment and in a hypertext.
Observation and analysis are still in progress. Nevertheless, preliminary results in-
dicate that the study of the map before hypertext navigation does make navigation
easier for individuals with low visual-spatial abilities. This suggests that the visuospa-
tial abilities may play a more decisive role in the construction of the mental represen-
tation of the environment than in the use of this representation during navigation.
Future qualitative analyses of our data will attempt to identify and compare strate-
gies for navigating our two environments. Specifically, we intend to identify (1)
the type of information subjects use to make navigation choices (route vs. survey
knowledge; their mental model vs. the map vs. cues in the environment) and (2) the
cognitive processes they perform during navigation to use this information, e.g. coor-
dinating the map with the territory, or converting survey knowledge into a route plan.
As far as the effect of spatial abilities is concerned, we intend to observe whether
subjects with low spatial abilities develop only route knowledge, whether studying the
map prior to navigation allows subjects with low spatial abilities to develop survey
knowledge, and whether they are able to use it to navigate the environment.
As part of our analyses, we will diversify the indicators we use to assess navigation
efficiency. We will use the efficiency metric proposed by Smith (1996) for hypertext
navigation, which combines three kinds of indicators: a measure of the redundancy
(repeated visits to the same nodes) of navigation, a ratio between the number of nodes
required to complete a task and the number of different nodes visited, and an indicator
of the successful completion of the task. We are working on an adaptation of this
metric to the physical environment, using the traveled distance instead of the number
of visited hypertext nodes.
We are hopeful that our work will yield recommendations for hypertext designers
on how to provide users with low spatial abilities with survey information they can
easily convert into route plans.
References
1. Parunak, H.V.D.: Hypermedia topologies and user navigation. In: HYPERTEXT 1989,
Proceedings of the Second Annual ACM Conference on Hypertext, pp. 43–50. ACM,
New-York (1989)
144 A. Brouwers
2. Blustein, J., Ahmed, I., Parvaiz, H., Fu, C.-L., Wang, C., Hu, Y.: Impact of spatial visuali-
zation aptitude on WWW navigation. In: Proceedings of the Workshop on Cognition and
the Web 2008, Granada, pp. 185–192 (2008)
3. Campagnoni, F.R., Ehrlich, K.: Information retrieval using a hypertext-based help system.
ACM Transactions on Information Systems 7(3), 271–291 (1989)
4. Dahlbäck, N., Höök, K., Sjölinder, M.: Spatial Cognition in the Mind and in the World the
Case of Hypermedia Navigation. In: Proceedings of the Eighteenth Annual Conference of
the Cognitive Science Society, Mahwah, pp. 195–200 (1996)
5. Maglio, P.P., Matlock, T.: The Conceptual Structure of Information Space. In: Höök, K.,
Benyon, D., Munro, A.J. (eds.) Designing Information Spaces: The Social Navigation
Approach, pp. 385–404. Springer, New-York (2003)
6. Collard, A.-S., Fastrez, P.: A model of the role of conceptual metaphors in hypermedia
comprehension. In: Proceedings of CICOM: Communication Sciences International Con-
gress on Communication, Cognition and Media, pp. 241–255 (2009)
7. Siegel, A.W., White, S.: The development of spatial representations of large-scale envi-
ronments. In: Advances in Child Development and Behavior. Academic Press, New York
(1975)
8. Moeser, S.D.: Cognitive Mapping in a Complex Building. Environment and Beha-
vior 20(1), 21–49 (1988)
9. Dillon, A., McKnight, C., Richardson, J.: Navigation in hypertext: a critical review of the
concept, Amsterdam, New York, pp. 587–592 (1990)
10. Dillon, A., McKnight, C., Richardson, J.: Space - The final chapter or Why physical repre-
sentations are not semantic intentions. In: McKnight, C., Dillon, A., Richardson, J. (eds.)
Hypertext. A Psychological Perspective, pp. 169–191. Ellis Horwood Ltd. (1993)
11. Kim, H., Hirtle, S.C.: Spatial metaphors and disorientation in hypertext browsing. Beha-
viour & Information Technology 14, 239–250 (1995)
12. Shum, S.: Real and Virtual Spaces: Mapping from Spatial Cognition to Hypertext. Hyper-
media 2(2), 133–158 (1990)
13. Sedig, K., Rowhani, S., Liang, H.-N.: Designing interfaces that support formation of
cognitive maps of transitional processes: an empirical study. Interacting with Comput-
ers 17(4), 419–452 (2005)
14. Edwards, D.M., Hardman, L.: Lost in Hyperspace”. Cognitive mapping and navigation in a
hypertext environment. In: McAleese, R. (ed.) Hypertext: Theory into Practice, ch. 7, pp.
105–125. Blackwell Scientific Publications Ltd., Oxford (1989)
15. McDonald, T.P., Pellegrino, J.W.: Psychological perspectives on spatial cognition. In:
Gärling, T., Golledge, R.G. (eds.) Behavior and Environment. Psychological and Geo-
graphical Approaches, pp. 47–82. North Holland, Amsterdam (1993)
16. Goldin, S.E., Thorndyke, P.W.: Spatial Learning and reasoning skill. Rand Corporation
(1983)
17. Carroll, J.B.: Human cognitive abilities: a survey of factor-analytic studies. Cambridge
University Press (1993)
18. Ekstrom, R.B., French, J.W., Harmon, H.H.: Kit of Factor-referenced Cognitive Tests.
Educational Testing Service, Princeton (1976)
19. Dee-Lucas, D.: Effects of overview structure on study strategies and text representations
for instructional hypertext. In: Rouet, J.-F., Levonen, J.J., Dillon, A., Spiro, R.J. (eds.)
Hypertext and Cognition, pp. 73–108. Lawrence Erlbaum Associates Inc., Mahwah (1996)
20. Vörös, Z., Rouet, J.-F., Plèh, C.: Content maps help low spatial capacity users memorize
link structures in hypertext. In: Proceedings of the Workshop on Cognition and the Web
2008, Granada, pp. 185–192 (2008)
21. Nilsson, R.M., Mayer, R.E.: The effects of graphic organizers giving cues to the structure
of a hypertext document on users navigation strategies and performance. International
Journal of Human-Computer Studies 57, 1–26 (2002)
22. Scott, B.M., Schwartz, N.H.: Navigational spatial displays: The role of metacognition as
cognitive load. Learning and Instruction 17(1), 89–105 (2007)
23. Boutard, C., Claire, I., Gretchanovsky, L.: Le vol du P.C.: évaluation fonctionnelle de la
lecture chez les jeunes de 11 à 18 ans. Ortho edn. (2006)
24. Dillon, A.: Spatial-Semantics. How users derive shape from information space. Journal of
the American Society for Information Science 51, 521–528 (2000)
Data Visualisation, User Experience and Context:
A Case Study from Fantasy Sport
Rob Euman and José Abdelnour-Nocera
Sociotechnical Centre for Internationalisation and User Experience

University of West London, Ealing, W55RF, 69121 United Kingdom
[email protected], [email protected]
Abstract. Fantasy Football is a rapidly growing online social game. As users

become more sophisticated and technology advances, the amount of data that is
available to inform users’ decision making is growing rapidly. Representing this
data in an informative and engaging way can be a challenge but data visualisa-
tion offers many ways to achieve this. This paper focuses on the design of
interactive solutions that are measured against existing products by way of a
comparative evaluation. In order to study the impact on user performance, effi-
ciency and accuracy are measured for clearly defined tasks carried out on each
design. The user experience is measured to understand the satisfaction and per-
ceived ease of use of each visualisation system. This study will be useful to
validate or challenge existing principles of data visualisation design and percep-
tion as well as offering suggestions for improving fantasy football products. The
study will also serve as a case study to support further research into data visua-
lisation evaluation methods. The paper concludes by discussing the findings and
possible areas for further research and design.
Keywords: HCI, social games, data visualisation, design, evaluation.
1 Introduction
Fantasy sports users are becoming increasingly sophisticated in the way they consume
and analyse sports data. However, the ways in which this data is represented has
not evolved as much as in other fields. There may be opportunities for the
fantasy sports world to learn from domains such as finance, science and healthcare by
using data visualisation techniques to improve the performance and experience of
its users.
The design rationale for this study follows the hypothesis that data visualisation
can improve the both the performance and experience of fantasy football users. In
challenging this assumption, empirical research was conducted involving real fantasy
football users to understand their current experience and ways in which this could be
improved. The research was conducted over a limited time period and focused on the
English Premier League season of 2011/12.
Data Visualisation, User Experience and Context: A Case Study from Fantasy Sport 147
A comparative study was conducted which required a solid understanding of exist-

ing fantasy football games that fairly represents the existing fantasy football expe-
rience and comparing this with new systems that were designed and prototyped as
part of the research process. Although the prototypes evaluated represent only a subset
of a fully functioning product, they were adequate as concepts and included enough
functionality to test and challenge the hypothesis.
2 Existing Literature
2.1 Data Visualisation

Information Visualisation is the accepted term for the field of graphical communica-
tion that data visualisation sits within. Underpinning the design solutions and data
analysis in this study are well established principles of human perception and statis-
tical graphics. Clevelend & McGill [1] wrote a foundational paper in this field where
visual elements such as colour, size and shape were tested to measure their effective-
ness in communicating quantitative information. In addition to this, Bertin’s Semiolo-
gy of Graphics [2] looks at the language of visualisation techniques and Ware’s
Information Visualisation [3] looks at the subject from the perspective of perceptual
science. Almost all research on this subject will reference the work of Edward Tufte
who’s books [4, 5] are ground breaking in their approach to information presentation
and visual literacy.
The field of interactive data visualisation is broad and continuously evolving due to
advances in technology. Ben Shneiderman has been responsible for many milestones
in this research field. The paper Visual Information Seeking [6] is often referenced as
one of the first papers to consider visualisations beyond static representations of data
and promoted the benefits of interactivity. In the paper The Eyes Have It, Shneider-
man [7] looks at a taxonomy system for interactive visualisation techniques that are
categorised by data type.
2.2 Data Visualisation Evaluation
As demand for new ways of visualising data increases, so does the need for standards
and guidelines to support their creation. There is a growing body of work that focuses
purely on data visualisation evaluation and the challenges this field faces that are not
common in standard interactive systems evaluation [8]. Plaisant [9] addresses these
challenges and recommends improving access to repositories of data, toolkits and
development tools and case studies that could encourage greater investment in the
field. Carpendale [10] discusses the trade-off between generalisability, precision and
realism and recommends a mixed methods approach to evaluations which includes
both qualitative and quantitative research. Lam et al [11] recommend evaluation ap-
proaches specific to seven specific scenarios of which Evaluating User Performance
and Evaluating User Experience are key to this study.
148 R. Euman and J. Abd
delnour-Nocera
3 Methodology
3.1 Evaluation Approaach

The main challenge in evalluating data visualisation systems lies in the fact that iit is
difficult to replicate the con
nditions in which these systems are used. A standard usaabil-
ity evaluation may involve a controlled or informal study of a user interacting witth a
particular system. Tasks arre often brief and satisfaction, efficiency and effectivenness
can all be measured withiin the test. Systems that involve large amounts of ddata,
whether visualised or not, are
a usually used in a more exploratory way and discoverry is
a key part of the experiencce. It may take some time before the user completes a ttask
such as finding a specific piece
p of information amongst the data and this can happen
long after use of the visualiisation system. Furthermore, a visualisation system’s efffec-
tiveness may lie in answerin ng questions you didn’t know you had [9].
In order to avoid some of these common pitfalls of data visualisation evaluatiion,
this research has used a co ombination of methods that are supported by previous re-
search and case studies. As there is no universally applicable methodology for eeva-
luating data visualisation sy ystems, care has been taken to focus on scenarios that are
relevant to the context of thhis study. This study focuses on two scenarios in challeeng-
ing the above hypothesis. These are Evaluating User Performance and Evaluatting
User Experience [13].
3.2 Research Design

In this study three ways off visualising fantasy football data in an interactive conttext
are compared in the form of o interactive prototypes. These prototypes represent onlly a
subset of a fantasy footballl game as the study only focuses on the visual represennta-
tion of player data. Takin ng this approach is also beneficial as a fully horizonntal
prototype containing unneccessary functionality would distract the user and takee up
unnecessary design and dev velopment time.
Prototype A (Fig. 1) is a table that reflects the conventional format of existing ffan-
tasy football products. This prototype served as a benchmark to measure the other ttwo
data visualisation techniquees against. Prototype B uses the same table format but w with
the player points data fromm each match of the season represented in a graphical ffor-
mat. Prototype C uses the same dataset as A and B but it is presented as a treem map.
The comparison involved carrying
c out a simultaneous study of each interface ussing
real fantasy football users carrying out tasks reflective of their normal pattern of use.
Table with raw data Table with inline charts Treemap

Prototype A Prototype B Prototype C
Fig. 1. Three prototypes to be evaluated

3.3 Identification of Variables

To maximise the accuracy of this evaluation, each of the variables has been identified
(Table 1). Data visualisation techniques, represented by the three prototypes, are the
independent variables that have been managed throughout the evaluation. The depen-
dant variables of user performance and experience were measured in relation to the
independent variable. There were a number of controlled variables that did not change
throughout the evaluation. Where it was not possible to control these variables, it is
acknowledged and considered in the analysis of the results.
Table 1. Independent, Dependant and Controlled Variables
Independent Dependant Controlled
Data visualisation tech- User Performance: Task com- Participants: Domain

nique: Table with raw pletion time, task error rates. knowledge, technical ability.
data, table with inline User experience: Perceived ease Evaluation: Tasks, questions,
charts, treemap. of use, satisfaction. device, environment.
3.4 Evaluation Methods

In order to make valid judgments on the three prototypes they were measured against
specific metrics (Table 2) through a variety of evaluation methods. The mixed
methods approach helped capture both objective and subjective information in
order to measure the impact each visualisation technique has on user performance and
experience.
Table 2. Metrics to be Measured Throughout the Evaluation
User Performance User Experience
Completion time: The efficiency of the Ease of use: The participants’ perceived
system as participants complete the tasks. ease of use.
Error rates: The effectiveness and toler- Satisfaction: Whether the system was
ance of the system. enjoyable to use and likelihood of adoption.
Quantitative Methods
Timed Tasks. These are scripted tasks that ask participants to interact with each proto-
type in a controlled way. Each task reflected that of a real session such as making
selections and comparisons based on fantasy player points and prices. The tasks were
ordered in such a way that avoided bias toward either prototype or participant.
Error Rates. As part of the controlled experiment, the amount of tasks that are failed
or abandoned was also measured. Errors are highlighted as significant issues when
measuring the performance of the system.
150 R. Euman and J. Abdelnour-Nocera
Qualitative Methods
Observations. Behaviours were observed as the participants performed each task.
These are brief, subjective and unobtrusive with the participant being asked to focus
on completing the task instead of discussing their actions at the same time.
Voice Mapping. This is a technique for measuring participants’ attitudes towards the
system. It is particularly useful when making comparisons between systems or mea-
suring the effect of design improvements over time. The participants were asked to
plot a mark on an axis (measuring satisfaction and perceived ease of use.
3.5 Prototype Design

Tables. Although the table format reflects existing fantasy football interfaces, there
are still a number of variables that need to be controlled such as sorting, filtering and
visual design. There are 531 rows, one for each Premier League player. Columns con-
tain alphabetic data including player name, club and position as well as numerical
data such as total player points, average player points, player value and games played.
The only difference between the two tables is the representation of player points per
match. Prototype A lists points as numerical values in a column. Prototype B
represents this data both as numerical data and as a bar chart representing each of
the 38 matches of the football season (although few players actually made this many
appearances).
531 rows of data would have a negative impact on cognitive and page load so it is
important for the user to be able to control the amount and order of this information.
The default view shows 10 players while a dropdown allows the user to change this to
50 or 100 players. The column headers also allow the user to sort the data in the table.
Columns containing individual match player points have had this functionality re-
moved as users are only interested in the cumulative player point totals. However,
position, team, player value and points are all categories that users like to sort by.
There were opportunities to improve on the presentation of these tables by applying
established design principles of information design. However, this would introduce too
many variables to the evaluation and distance both prototype A and B from the exist-
ing fantasy football product table design.
Treemaps. Treemaps use colour, size and proximity to represent relevant attributes of
the data. This is represented in the form of a 2 dimensional display of nested rectan-
gles. Within each rectangle is a collection of tiled rectangles with varying sizes and
colours. The information represented by the colour and size of these rectangles can
vary depending on domain and dataset. For the purpose of this study I have chosen to
focus on the typical financial use of treemaps where the size of the rectangle
represents volume or value while the colour of the rectangle represents percentage
change or stock performance.
The Squarified treemap [12] format lends itself well to compare, interact with and
perceive structure. While the Squarified algorithm is limited in its ability to order the
rectangles, This approach was chosen to create the Fantasy Football treemap as the
order of player values is less relevant to the size and colour of the rectangles.
Ultimately fantasy football users are looking for information on the performance of
individual players. However, the routes users take to access this information can dif-
fer. Some users are looking for the best value players in a specific position, others are
looking for the most consistent players of a specific team and others may simply be
looking for a specific player. Therefore, it is important to use the hierarchy to allow
users to make sense of the data format and access the information they need.
The player performance data in the fantasy football treemap is represented by a
graduation of colour from green for positive values to red for negative values. To pro-
vide contrast and to avoid undesirable colour blends, white is used as a midpoint.
Colours are displayed not on actual points scored but on each player’s points in
relation to the average point score. This has been done as most players’ scores are a
positive amount (i.e., the highest score is 2451 and the lowest is -15) which means the
treemap will display predominantly green rectangles which will only use around 50%
of the colour scale. This will make comparisons more difficult. By assigning colours
for player points in relation to the average player point figure there should be an even
split of red and green rectangles.
While colour is an effective technique for conveying information, around 10% of
the male population suffers from some form of colour blindness. For this reason, an
option to change the treemap to use the colours yellow and blue to represent player
points should be included in a final product.
3.6 Evaluation
The tasks used in the evaluation were chosen carefully to insure accurate and consis-
tent data could be captured and to reflect the common tasks of the participants’ normal
usage. It was also important that each participant did not carry out the same task on
more than one prototype. This would skew the results as, once a participant had found
the information via one prototype, the same task on another prototype would be one of
navigation instead of discovery.
Nine tasks were chosen which could be categorised into three groups; overview
level tasks, team level tasks and position level tasks. The order of the tasks within
each category was consistently varied per user to minimise any distortion as a result
of familiarity with any given prototype. However, the lower level overview tasks were
performed by each user first to mitigate any learning curve. Learnability and memo-
rability will not be measured as part of this evaluation as participants were not
expected to conduct tasks of significant complexity.
4 Findings
4.1 Task Completion Times

The task competition times reveal some interesting patterns relating to the efficiency
of each of the prototypes (Fig. 2). It was expected that the three overview tasks that
were carried out first would be easier, but there is a surprising drop in completion
times for the final three. This could be due to the participants increasing familiarity
with each of the systems. What is clearer is that Prototype C is the most efficient of
the three systems.
Fig. 2. A voice map measuring ease of use and satisfaction
4.2 Error Rates

There were only three errors made throughout the evaluation. While this is positive,
the numbers are too small to draw concrete conclusions from. They do however raise
issues that should be addressed in a next phase of design.
Two of the errors were made on Prototype C during the same task where partici-
pants were asked to select the player with the most points from the teams overview. In
this view, two players (Robin van Persie and Wayne Rooney) stand out as the highest
scorers because of the solid green colour of the rectangles but, as their points are so
similar, it is difficult to differentiate between them. Further to this, the incorrect selec-
tion, Wayne Rooney, appears above Robin van Persie so is the first player the user
will see. This lack of control of the treemap layout should be seen as a weakness and
clearly demonstrates that this technique is less appropriate as a static representation of
data. The error on Prototype A involved the user giving up after 45 seconds which
would have been longer than any of the other tasks took to complete.
4.3 Key Observations

The observations that were made were brief in order to not disrupt the tasks that were
being carried out. The observations were analysed and categorised into themes that
highlight some recurring issues and behaviours throughout the evaluation.
Although all participants found the table sorting functionality of Prototypes A and
B easily, three asked to be reminded of the multi-column sorting functionality.
The participants that did master this functionality seemed to get there through trial
and error, clicking a few times before consistently being able to perform the function
correctly.
All participants seemed to find scrolling through the data in Prototypes A and B
quite laborious with four confirming this in the later discussion. In this context,
Tufte’s principle of presenting data-rich displays is less relevant. Although the tables
have a large amount of information up front, there is a limit to how much you can
display in one view, particularly on the average web browser.
The treemap supports the principle that it is easier to recognise than to recall. Using
both table prototypes, users needed to remember data to make comparisons instead of
being able to instantly compare relationships through size or colour. The charts in
Prototype B are useful for analysing form throughout the season; they are less effec-
tive for making easy comparisons based on total player points.
4.4 Ease of Use and Satisfaction

The voice map in Figure 3 highlights some interesting insights into the participants’
attitudes to the three visualisation systems. As a measure of satisfaction it seems Pro-
totype B was significantly more successful than Prototype A. Although they compared
similarly when measured by task completion times, it seems the participants reflective
experience of prototype B was more positive because of the additional graphical re-
presentation of player points. This is emphasised by that fact that there was no signifi-
cant difference in the perceived ease of use.
Fig. 3. A voice map measuring ease of use and satisfaction
There is also only a slight uplift in perceived ease of use for Prototype C. Although
tasks carried out were completed in significantly less time than Prototypes A and B,
the users still felt there was a degree of complexity to it. However, Prototype C per-
formed better on satisfaction which could support the theory that a satisfying expe-
riences make users more tolerant of functional complexity [13, 14].
4.5 Recommendations for Further Design

Although each prototype gave opportunities to sort or navigate through the informa-
tion, there is a case for offering more effective ways of filtering unwanted information
out early. Sliders are a very common interaction design component that enable users
to select a value from any given scale. They can have a single control to define a
specific value or two to define a range of values. Alternatively, conversational filter-

ing is a technique where filters are positioned in a sentence that is formed in the way
that it would be spoken. This can be a more intuitive way of accessing information as
it supports the user’s mental model of accessing and processing information.
One of the potential challenges with the treemap is that, while it aids comparisons,
size and colour are not effective ways of comparing when there is minimal variation in
the data. Although the squarified treemap algorithm was created to place nodes of a
similar scale together, clear differentiation isn’t always guaranteed. Other techniques
for organising hierarchical information that help further remove uncertainty should be
considered.
One such technique is the ‘Slopegraph’ as created by Edward Tufte [4]. The slope-
graph is a way of comparing the scale of changes over time for a list of variables.
Using this technique, the variables of team performance and team spending could be
ranked in ascending order and linked by lines, the angle of the lines represent the scale
of differentiation. The weight of the lines would represent the scale of team spending
with colour illustrating whether the differentiation is negative or positive.
This technique provides an interesting and concise alternative to the treemap
layout. Where the treemap’s reliance on colour and size to represent information could
lead to ambiguity and confusion, the slopegraphs benefits of combining this with the
angle and order of connecting lines is clear. It could be safe to assume that the two
errors committed in the evaluation when comparing the player points of Robin van
Persie and Wayne Rooney could have been avoided using this technique.
Another way to make interactive visualisation more intuitive is by introducing ani-
mated transitions as the user manipulates and navigates the data. Animations are a
useful way of facilitating the perception of changes when transitioning between data
graphics. This would be particularly relevant for the treemap which uses an interac-
tion model of zooming and displaying information on demand.
5 Conclusions
The results of the evaluation prove, with a relative degree of certainty, that data visua-
lisation can improve the performance and experience of fantasy football users. While
this validates many of the of the established principles that informed the design, this
report also proves that these principles should be used with sensitivity to the context
of use. A visualisation system targeted at search and discovery as championed by Ben
Shneiderman [6] will not lend itself well to high-resolution data analysis. Equally,
data-rich displays of information will not perform well when deep exploration of data
is required.
These findings give further weight to the theory that a positive user experience can
help appease functional complexity. As well as reducing task times, the treemap
is designed to enhance the user’s emotional engagement instantly making sense
of otherwise complex information by highlighting relationships and encouraging
comparisons.
To take this research further, it would be useful to conduct a longitudinal study
measuring the impact of data visualisation techniques over a longer period of
time. This could be conducted as part of a comparative evaluation using real users
competing against each other in their natural context of use for the duration of a foot-
ball season. User performance in this case would therefore be measured by competi-
tive advantage through exploration and discovery as opposed to operational efficiency
and user experience would be measured by adoption and engagement metrics.
One participant remarked that he felt empowered when using the treemap system.
If fantasy football users could be genuinely empowered through the effective use of
data visualisation techniques, could it then compromise competitive advantage? While
fantasy football products may want to maintain an element of uncertainty in the data it
makes available, this offers many possibilities for other domains.
References
1. Cleveland, McGill: Graphical Perception: Theory, Experimentation, and Application to the
Development of Graphical Methods. Journal of the American Statistical Associa-
tion 79(387) (1984)
2. Bertin: The Semiology of Graphics: Diagrams, Networks, Maps. University of Wisconsin
Press, Madison (1983); Translated by Berg, W.J.: Original publisher, Mouton-Gautier. Esri
Press, Redlands
3. Ware: Information Visualisation: Perception for Design, 2nd edn. Morgan Kaufmann, San
Francisco (2004)
4. Tufte: The Visual Display of Quantitative Information. Graphics Press, Cheshire (1983)
5. Tufte: Envisioning Information. Graphics Press, Cheshire (1990)
6. Ahlberg, Shneiderman: Visual information seeking: tight coupling of dynamicquery filters
with starfield displays. In: Proceedings of the SIGCHI Conference on Human Factors in
Computing Systems: Celebrating Interdependence, Boston, Massachusetts, United States,
pp. 313–317 (1994)
7. Shneiderman: The eyes have it: A task by data type taxonomy for information visualisa-
tions. In: Proceedings of the IEEE Symposium on Visual Languages (1996)
8. Chen: Aesthetics versus functionality. Information Visualisation 1(2), 93–94 (2002)
9. Plaisant: The challenge of information visualisation evaluation. In: Proceedings of the
Working Conference on Advanced Visual Interfaces, Gallipoli, Italy, May 25-28 (2004)
10. Carpendale, S.: Evaluating information visualizations. In: Kerren, A., Stasko, J.T., Fekete,
J.-D., North, C. (eds.) Information Visualization. LNCS, vol. 4950, pp. 19–45. Springer,
Heidelberg (2008)
11. Lam, Bertini, Isenberg, Plaisant, Carpendale: Seven Guiding Scenarios for Information Vi-
sualisation Evaluation. Technical Report (2010)
12. Bruls, Huizing, Wijk: Squarified TreeMaps. In: Proceedings of Joint Eurographics and
IEEE TCVG Symp. on Visualisation. IEEE Press (2000)
13. Moere, Purchase: On the role of design in information visualisation. Information Visuali-
sation 10(4), 356–371 (2012)
14. Norman: Emotional design - Why we love (or hate) everyday things. Basic Books, New
York (2004)
A Pedestrian Navigation Method for User’s Safe
and Easy Wayfinding
Hiroshi Furukawa and Yutaka Nakamura
University of Tsukuba, Tsukuba, Japan

[email protected]
Abstract. In recent years, most of mobile phones have a function of pedestrian

navigation guidance. It was reported that users sometimes feel anxiety because
of low accuracy of the position estimation especially in urban area and delay of
information updating. In order to reduce the anxiety, a route planning algorithm
is proposed in this study, which weighs user’s difficulty (or easiness) of locat-
ing own current position as well as total physical distance of courses. The diffi-
culty is estimated by valuation functions based on the “recognizability” and
“visibility” of landmarks. An experimental study conducted in real situation us-
ing a prototype system to examine and refine the model for the optimal route
planning. As the result, a modified model is proposed as a promising method of
route planning for user’s easy wayfinding.
Keywords: navigation system, landmark, salience, cognitive model, anxiety,

recognizability, visibility.
1 Introduction
However pedestrian navigation guidance is one of the widespread services provided

by mobile devices [1-2], it was reported that users sometimes feel anxiety because of
low accuracy of the position estimation and delay of updating of the navigation dis-
play [3]. In areas with many high buildings, the accuracy of position estimation by
GPS can be low when a communication state with a satellite is harmed by the ob-
stacles. When the precise location is not given on a navigation display, geographical
information about the area is the key for users to locate their position.
The purpose of our work is to develop a pedestrian navigation aid to reduce the an-
xiety of users [4-5]. A route planning algorithm is proposed, which weighs user’s
difficulty (or easiness) of locating own current position as well as total physical dis-
tance of courses. Fig.1 depicted the basic idea of the method. We focused on the roles
of useful landmarks at the locating task. Two types of valuation functions are pro-
posed to estimate effectiveness of various landmarks. The one is a valuation function
for their “recognizability” and the other “visibility.”
The aim of this study is to examine and improve the models for the optimal route
planning. The validity of the model was confirmed by comparing the elapsed time of
user’s locating task and the subjective evaluation of the anxiety by participants with
A Pedestrian Navigation
N Method for User’s Safe and Easy Wayfinding 157
the estimated results by the proposed model. Based on the results of the validityy, a
modified model was propossed in this study.
Fig. 1. The basic idea of the proposed

p method considering the effectiveness of landmarks, and
the conventional method consiidering physical distance of a course
2 Related Researcches
Different types of route seelection methods are proposed for pedestrian navigattion
system to meet the needs of users. Miura and his colleagues focused on the condittion
about street illuminations [6]. They proposed a method evaluating the unsafe cost of
the illumination and the cosst of route distance. Akasaka and Onisawa proposed a m me-
thod reflecting individual preference
p for route selection [7]. User’s satisfaction deggree
of a road is evaluated by fu
uzzy measures, and used as one of the costs for route sellec-
tion. In the same way, the method
m proposed in this paper uses multi-objective evallua-
tion method, where the co ost is defined as the easiness of locating user’s currrent
position and the distance off routes.
Some studies have cond ducted about evaluation methods of landmark’s effectiive-
ness for pedestrian navigatiion systems. Nakazawa and his colleagues proposed a m me-
thod of evaluating landmark k’s effectiveness based on the frequency used for availaable
navigation maps [8]. They collected guide maps from web pages and magazines, and
counted frequency of the ap ppearance of landmarks in attribute classification, such aas a
restaurant and a conveniencce store. The frequency is defined as an index of the coggni-
tive ease of finding landmaarks. An issue is that the relationships between the efffec-
tiveness of landmarks and the frequency of appearance is not clearly explainedd or
confirmed. There is a possiibility that the correlation is weak. For example, the stuudy
reported that the frequency for business complexes or apartment buildings is high [8].
On the other hand, the efffectiveness of the landmarks may not be high enoughh to
158 H. Furukawa and Y. Nakamura
achieve user’s navigation tasks because of the low salience or low uniqueness. We are
trying to develop a rational and feasible method for quantitative evaluation of the
landmark effectiveness.
3 A Route Planning Method Considering Landmark’s

Effectiveness
Authors proposed a route planning algorithm which weighs user’s easiness of locating
own current position [4-5]. The easiness is quantitatively estimated using two types of
valuation functions: valuation functions for the recognizability and visibility.
3.1 Recognizability of Landmarks
The recognizability is an index that shows how easy for users to find out a landmark
which depicted on navigation maps. For example, it must be high when a store has a
shining signboard, or when the appearance of a store is well known for users. Also,
the index must be negatively correlated with distance between a landmark and user’s
position. The degree of influence on change of the distance must be different with
different kinds of landmarks. It becomes difficult to recognize a small store like a
convenience store being located in the position away from user’s current location.
The valuation model for recognizability was constructed through several cognitive
experiments [4]. For each type of landmarks, a regression line was determined as a
valuation function for the recognizability, which shows relationships between recog-
nizability and the distance.
3.2 Visibility of Landmarks

The recognizability index is argued about the easiness of finding a landmark assum-
ing that it is within user’s view. The visibility is defined as possibility that a landmark
is in sight of users. It must be low when there are many high buildings between the
landmark and the users. To consider a situation that a landmark is concealed by other
buildings from user’s view, two physical features are defined as factors in the visibili-
ty of landmarks: denseness of buildings around user’s current position and depth of a
landmark from its nearest road.
The first factor denseness is defined as a function of each base area of buildings
around user’s position and distance to the buildings from user’s position, as described
in Eq. (1).
denseness = ∑ (1)
where n: a building n which is within fixed distance from user’s current place,
An: the base area of a building n [m2], dn: the distance from user’s current place to a
building n [m].
A Pedestrian Navigation Method for User’s Safe and Easy Wayfinding 159
The second factor depth is defined to consider a situation that a user can see a
landmark on the road which extends from his/her current location.
The valuation function for visibility is modeled using a logistic function (Eq. (2)).
visibility = 1⁄ 1 (2)
where x: the depth of a landmark from its nearest street [m], y: denseness of buildings
around user’s current position, A: a coefficient for the depth, B: a coefficient for the
denseness, C: a bias, which represents the rapidly changing point of the function ac-
cording to the depth.
To acquire data for construction of the valuation function for visibility of each type
of landmarks, visual confirmation investigations of landmarks were conducted using
the Street View function of Google map [4].
3.3 Valuation Function of Landmark’s Effectiveness

It is thought that the easiness of locating current position at a place is equivalent to the
total effectiveness of landmarks at a place in this study [4-5]. The valuation model of
the total effectiveness is defined as Eq. (3), based on the valuation functions of recog-
nizability and visibility [5].
CV(n) = ∑ (3)
C’(l) = 1.514 C(l) - 0.028 dl – 2.174
where CV(n): the total effectiveness of landmarks at a place n, l: the lth nearest land-
mark which is within fixed distance from a place n, C(l): the recognizability of land-
mark l where C(x) is the valuation functions of recognizability for a landmark x, V(l):
the visibility of landmark l where V(x) is the valuation functions of visibility for a
landmark x, dl: the distance from user’s current place to a building l [m], wl: coeffi-
cients for the effectiveness of landmark l, δ: a bias. The product of the recognizability
and visibility functions, C(l) x V(l), synthetically evaluates the easiness of finding out
a landmark when it is actually within user’s view. The values of the coefficients wl
are 0.958, 0.004, -0.007, and the bias δ is 0.194.
3.4 Route Planning Algorithm for User’s Easy Wayfinding

In the general navigation systems, path planning is performed using the shortest path
planning algorithm by Dijkstra [9]. In this study, the cost function for evaluating
routes is expanded by adding elapsed time of user’s locating task at each crossing to
the original cost [4-5]. The elapsed time at each crossing is estimated based on the
valuation function of landmark’s effectiveness Eq. (3). When the cost between nodes
m and n is defined as d(m,n) in the Dijkstra method, the proposed cost, dn(m,n), can be
described as Eq. (4).
dn(m,n) = d(m,n) + dcv(n) (4)

where dn(m,n): a cost function defined in this study, d(m,n): the original cost function
based on route distance, dcv(n): the additional cost proposed in this study.
The dcv(n) is the additional cost which considers elapsed time of user’s locating
task at a place n. Eq. (5) shows the relationships between the elapsed time and the
total effectiveness of landmarks at a place n.
dcv(n) = 5.0 / {1 + CV(n)}, (5)
where CV(n): the total effectiveness of landmarks at a place n.
4 Cognitive Experiment
This section shows an experimental study conducted in real situation to acquire actual
data for the evaluation and refinement of the proposed method described in Section 3.
4.1 Method
The main tasks of participants were achievement of navigation tasks, i.e., following
an instructed route from a starting point to a destination by oneself, using a prototype
pedestrian navigation system via a smartphone, and evaluation of their own anxiety
during the navigation task. When they recognize that they made mistakes, they were
asked to go back to the point made mistakes and restart the task.
Routes. The experiment was conducted at three different areas in Tokyo, i.e., Taitou,
Setagaya, and Ikebukuro. These areas were chosen to satisfy a condition that the ex-
perimental routes include different types of circumstances. The sizes and shapes of
buildings and the number of available landmarks are different from each other.
The elapsed times of navigation tasks were estimated with routes by the conven-
tional method, considering only physical distance, and the proposed method for all
possible sets of starting point and destination among each area. The five sets were
selected among the three areas, where the difference in the estimated time between
the conventional and proposed method conditions is large.
Prototype system. Participants were asked to walk through the assigned route with a
digital map displayed on a smartphone (Sharp SO-04D). An example of the display is
shown in Fig. 2. The functions implemented to the system are listed below, which are
designed to simulate conditions of real navigational services (such as [1]).
• Indication of current position using information from the GPS

• Indication of a route from a starting point to destination
• Automatic periodic updating of the display
• Manual selection of the map scale
• Indication of the remaining distance
A Pedestrian Navigation
N Method for User’s Safe and Easy Wayfinding 161
The map data were supplied by ZENRIN DataCom Co., Ltd. Several landmaarks
are depicted on the digital map, including ones used in the proposed route plannning
method, i.e., convenience stores, restaurants, bank buildings, gasoline stations, schhool
buildings, and public buildiings.
Fig. 2. An example
e of the display of the prototype system
Evaluation of anxiety. At selected intersections, the participants were asked to eeva-

luate the anxiety using five--level index, where 5: “in great anxiety,” 4: “in little anxxie-
ty, 3: “middle,” 2: “with litttle assurance,” and 1: “with great assurance.” Because the
indications were shown on the smartphone display, it was easy for the participants to
achieve the evaluation task.. The selection of intersections was at random manner.
4.2 Procedures
Nine paid participants, maale university students, took part in this experiment. T The
participants were randomly y divided into two groups, five in Group A and fourr in
Group B. With each set of a starting point and destination, one group was assigneed a
route planned by the conven ntional method, and the other group assigned a route by the
proposed method. In a grou up, the method used for planning was assigned alternatelly.
Total time of the experiiment for a participant was between three and four houurs.
Total distance a participant walked was approximately 5.5 kilometers.
The data acquired in the experiments were listed below.
• the elapsed time of naviigation task for a route (time for waiting for the lighht to
change was deleted from m the measured data)
• the subjective evaluation n of participant’s anxiety at selected intersections (ussing
five-level index on the smmartphone display )
• the reasons of the evaluaation of anxiety (oral replies)
5 Evaluation and Refinement of the Model
This section shows the resu

ults of evaluation of the proposed models based on the ddata
acquired through the cognittive experiment, and refinement of the model for the totaal.
5.1 Evaluation of the Proposed

P Model
The first evaluation is done by using the objective data from the experiment. Figg. 3
shows the relationships bettween the total time for user’s locating task in a route eesti-
mated by the model describ bed in Section 3, and the actual total time acquired in the
experiment. The correlation n coefficient between the two variables is 0.075, where the
value indicated there is no dependency
d between them.
As the second evaluatio on, correlation coefficients were calculated between the
elapsed time of user’s locaating task and the subjective data about the user’s anxiiety
acquired in the experimentt at the same intersection. The first row, the title is “The
original model in Section 3”, of Table 1 shows the results of every participant. T The
values of five participants are
a lower than 0.3, which can be interpreted that the deppen-
dency between the two variiables is none or low.
These results show thaat the proposed model is not proper for estimating the
valuation of the total effectiiveness of landmarks or the user’s anxeity.
Fig. 3. The total time of useer's locating task Fig. 4. The total time for user's locating taask
in a route estimated by the proposed model in a route estimated by the model with E Eqs.
with Eqs. (4) & (5), and the actual total time (6) & (7), and the actual total time as tthe
as the result of the experimen
nt result of the experiment
5.2 Refinement of the Model

M
Two possible causes of thee issue were considered in this modification process of the
proposed model. The one iss that the assumption about timing of user’s locating task is
different from actual situattion. The other is possibility that the Eq. (5) is not prooper
model for estimating elapseed time of user’s locating task at a place.
Table 1. Correlation coefficients between the elapsed time of user's locating task estimated by
the proposed models and the subjective data about the anxiety at the same intersection
Correlation coefficients
Group Participants The original The model with The model with
model in Section 3 Eqs. (5) & (6) Eqs. (6) & (7)
a1 0.28 0.50 0.46
a2 0.44 0.59 0.61
A a3 0.30 0.37 0.33
a4 0.12 0.18 0.07
a5 0.36 0.41 0.40
b1 0.45 0.55 0.53
b2 0.53 0.56 0.59
B
b3 0.24 0.57 0.58
b4 0.08 0.45 0.44
The bold number indicates that the correlation is medium (>0.3) or strong (>0.5).
Modification Considering Timing of User’s Tasks. In the interview of the partici-

pants, some reported that they did not feel any anxiety when they were walking
through intersections guided to go straight. The reason is that locating tasks were not
necessary at those intersections. On the other hand, the proposed route planning algo-
rithm was developed on an assumption that users try to locate their current positions
at every intersection, as described in Section 3.4.
Based on the results of the interview, a modified model described as Eq. (6) is pro-
posed which consider the user’s rule of executing a locating task at each intersection.
, , Intersection n is guided to turn
, (6)
, , Intersection n is guied to go straight
Correlation coefficients were calculated between the elapsed time of user’s locating
task and the subjective data about the user’s anxiety at the same intersection acquired
in the experiment. The second row (“The model with Eqs. (5)&(6)”) of Table 1 shows
the results for every participants. The values of seven participants are higher than 0.4,
which suggests that the correlation between the two variables is medium or strong.
An independent two-sample t-test was conducted to compare correlation coeffi-
cients in conditions with the original model and the modified model with Eq. (6).
There was a significant difference in the correlation coefficients in the previous model
(M = 0.31, SD = 0.15) and the modified model (M = 0.46, SD = 0.13) conditions;
t(16) = 2.12, p = 0.035 (<0.05).
These results suggest that the modified model with Eq. (6) is more proper than the
previous model to estimate the elapsed time of user’s locating task and the anxiety.
Modification on the Model for Elapsed Time of User’s Task. The aim of study in
this section is to modify Eq. (5). The basic assumption of Eq. (5) is that the relation-
ships between the elapsed time and the effectiveness is monotone decreasing, i.e., as
the effectiveness is getting higher, the time getting shorter.
Several monotone decreasing functions were evaluated as the alternatives using

correlation coefficients with the measured data. The results of evaluation suggest that
Eq. (7) is a proper model for the estimation.
23.0, 1.0
(7)
23.0⁄ ,
Fig. 4 shows the relationships between the total times for user’s locating task in a
route estimated by the model (Eqs. (6) and (7)), and the actual total time measured in
the experiments. The correlation coefficient between the two variables is 0.83, where
the value indicated the correlation is very strong. On the other hand, the correlation
coefficient with the Eqs. (5) and (6) is 0.68, which is also strong but less. These re-
sults suggest that the modified model using Eqs. (6) and (7) is proper to estimate the
elapsed time of user’s locating task.
Correlation coefficients were calculated between the elapsed time of user’s locat-
ing task and the subjective data about the user’s anxiety at the same intersection
acquired in the experiment. The third row (“The model with Eqs. (6)&(7)”) of Table 1
shows the results for every participant. An independent two-sample t-test was con-
ducted to compare correlation coefficients in conditions with the modified model with
Eqs. (5) & (6) and Eqs. (6) & (7). There was no significant difference in the correla-
tion coefficients in the Eqs. (5) & (6) model (M = 0.46, SD = 0.13) and the Eqs. (6) &
(7) model (M = 0.45, SD = 0.17) conditions; t(15) = 2.13, p = 0.80.
Summary. Even though the superiority in estimating user’s anxiety is not confirmed,
these results suggest that the modified model with Eqs. (6) & (7) is a promising me-
thod of route planning in this study.
An issue should be concerned is that an individual difference in the estimation. The
low correlation coefficients of a participant (a4 in Table 1) indicate that there are
some other meaningful factors not concerned in the proposed model. Nevertheless, it
is expected that the proposed method is proper for supporting navigational tasks of
most of users.
6 Conclusions
In order to reduce anxiety of a pedestrian navigation user, a route planning algorithm

is proposed in this study, which weighs user’s difficulty (or easiness) of locating own
current position as well as total physical distance of courses.
Recognizability and visibility of landmarks are defined as basic factors which spe-
cify the effectiveness of landmarks in user’s position estimation. The cost function for
evaluating routes is expanded by adding elapsed time of locating intersections to the
original cost of route distance. The elapsed time at an intersection is estimated based
on the valuation functions of effectiveness of landmarks. When a user inputs his/her
destination to navigation system, a route with the lowest cost is selected to support
user to reach the goal without anxiety.
The aim of this study is to examine and refine the models for the optimal route
planning. The validity of the valuation functions was confirmed by comparing the
elapsed time of user’s locating task and the subjective evaluation of the anxiety with
the estimated results by the proposed model. As the result, this study proposes the
modified model as a promising method of route planning for user’s easy wayfinding.
It is expected that the proposed method is proper for supporting navigational tasks of
most of users.
Future tasks of this study are following:
1. The first task is a study on the individual difference in the relationships between
the elapsed time and use’s anxiety of locating tasks must be our next theme to im-
prove the usefulness of the route navigation method.
2. The second task is the validation test of the pedestrian navigation guidance with
participants of different conditions. The aim of the test is to confirm that the guid-
ance can reduce anxiety of users actually in real situations.
Acknowledgment. This work has been partially supported by Grants-Aid for

Science Research 23500109 of the Japanese Ministry of Education, Science, Sports
and Culture.
References
1. Navitime Japan (in English), https://www.navitime.co.jp/en
2. EZ Navi Walk (in Japanese), http://www.au.kddi.com/ez_naviwalk/
3. Tatenami, T., Matsushita, D., Munemoto, J.: Transition of level of anxiety in wayfinding
with pedestrian navigation system on mobile telephone. Journal of Architecture and
Planning 608, 59–64 (2006) (in Japanese)
4. Furukawa, H., Uto, H.: A quantitative evaluation method of landmark effectiveness for pe-
destrian navigation. In: UKSim 14th International Conference on Computer Modelling and
Simulation. IEEE Computer Society (2012)
5. Furukawa, H., Nakamura, Y.: Pedestrian navigation system based on quantitative estimation
of landmark’s effectiveness. In: The 4th International Conference on Applied Human
Factors and Ergonomics 2012, pp. 391–400. USA Publishing (2012)
6. Miura, H., Takeshima, S., Matsuda, N., Taki, H.: A study on navigation system for pede-
strians based on street illuminations. In: König, A., Dengel, A., Hinkelmann, K., Kise, K.,
Howlett, R.J., Jain, L.C. (eds.) KES 2011, Part III. LNCS, vol. 6883, pp. 49–55. Springer,
Heidelberg (2011)
7. Akasaka, Y., Onisawa, T.: Personalized pedestrian navigation system with subjective prefe-
rence based route selection. In: Ruan, D., Hardeman, F., Meer, K. (eds.) Intelligent Decision
and Policy Making Support Systems. SCI, vol. 117, pp. 73–91. Springer, Heidelberg (2008)
8. Nakazawa, K., Kita, N., Takagi, K., et al.: A Dynamic Map Based on Landmark’s Visi-
bility. Transactions of Information Processing Society of Japan 49, 233–241 (2008) (in
Japanese)
9. Dijkstra, E.W.: A note on two problems in connexion with graphs. Numerische Mathema-
tik 1, 269–271 (1959)
The Effect of Information Quantity on Cbp Interface
in the Advanced Nuclear Power Plant
Min-Chih Hsieh and Sheue-Ling Hwang
National Tsing Hua University, Hsinchu, Taiwan 300

[email protected]
Abstract. Computer-based procedures (CBP) are widely used in advanced

nuclear power plant (NPP). In order to improve operators’ performance and
decrease the probability of human errors, several different CBP interface
layouts have been provided by previous studies. However, there has been little
research discussing the information quantity that present on the CBP interface.
This study is to examine operator performance in using computerized
procedures from an information quantity point of view. A simulated CBP
system was developed to present three different information present styles, each
with different task complexity. The results indicated that medium information
quantity of CBP interface was better than other level of information quantity.
Keywords: Computer-based procedures; Nuclear power plant; interface;

information quantity.
1 Introduction
The procedures of system control that guide operators in performing tasks are widely
used in nuclear power plant (NPP) domain (Husseiny et al., 1989; Xu et al., 2008).
Computerized procedures (CP) systems in a nuclear power plant (NPP) provide
instructions to guide operators in monitoring, decision-making, and controlling the
plant. The CP systems have been accepted by various regulatory authorities and are in
use at several NPP around the world (Yang et al., 2012). The results of other research
studies of CP systems show that CP systems can provide some performance benefits,
such as tasks can be performed more quickly; mental workload can be reduced and
minimized; fewer errors may be made in transitioning through or between procedures
(NUREG/CR-6634; NUREG/CR-6749; Portmann and Lipner, 2002; O;Hara et al.,
2003; Yang et al., 2012).
Computer-based procedure (CBP) was a human-system interaction technology that
was found to be potentially safety significant (Yang et al., 2012). In order to improve
operators’ performance and decrease the probability of human errors, several different
CBP interface layout have been provided by previous studies (Jung et al., 2004; Xu et
al., 2008; Carvalho et al., 2008; Hong et al., 2009).
The Effect of Information Quantity on Cbp Interface in the Advanced NPP 167
Regarding to the control interface of CBP studies, Wourms & Rankin (1994)
pointed out that flowchart is better than text in the procedure format because of its
ability to relate procedure elements explicitly. Jun et al. (2004) published coherently
coupled frames that reduced the number of navigation, and flowchart and success tree
were used where total of the process information was shown on the interface. Xu et al.
(2008) compared two mainstreams graphical computerized procedures such as
flowchart procedures and success tree, the results showed that flowchart procedure
was better than success tree. Because the flowchart procedure resulted in a lower error
rate without introducing a longer task time and more workload. Carvalho et al. (2011)
developed a new CBP interface, and compared it with existing interfaces. The results
showed that the time spent by the operator to identify two accidents, through the new
interface, was faster than the existing interface. Hong et al. (2009) also developed a
computerized procedures system to assist operator executing the procedure tasks.
No matter what format that the procedure information presenting on CBP,
however, there has been little research discussing the information quantity that
present on the CBP interface. Operators who work in the NPP main control room
might be influenced by the information quantity from the computer screen (Wickens,
2000). The objective of this study, therefore, is to examine operator performance in
using computerized procedures from an information quantity point of view.
This paper is organized as follows: Section 2 describes the experiment design;
Section 3 presents the results of this experiment; Section 4 presents discussions and
conclusions drawn from the study.
2 Methodology
The information quantity issues of CBP interface were investigated by experiment

with subject participation. The details about independent and dependent variables, the
experiment process, and the information quantity of CBP interface will be introduced
in this section.
2.1 Independent Variables
Two independent factors were considered in this experiment: the information quantity
of CBP interface and task complexity. Information quantity of CBP interface
consisted of three levels and task complexity consisted two levels, as listed in Table 1.
A between- and within-subjects mixed experiment design was adopted where a
subject would participate in the experimental condition of one presentation style and
both task complexity levels.
Table 1. Factors and levels in the experiments
Factor Level 1 Level 2 Level 3

Information quantity High Medium Low
Task complexity Complex Simple -
168 M.-C. Hsieh and S.-L. Hwang
2.2 Information Quantity of CBP Interface
Three information quantity of CBP interface (high, medium, and low) were adopted
in this experiment. The measurement in information quantity used in this experiment
lay in the number of events on the CBP interface, which can be expressed by the
formula as follows (Wickens, 2000):
Hs = log2N. (1)
The information conveyed by an event Hs, in bits, and N is the number of alternative.
For example, high information quantity in this experiment with at least eight events’
information was shown on the CBP interface in each step, log28 = 3 bits. Medium
information quantity in this experiment with at most four events’ information was
shown in each step, log24 = 2. Low information quantity in this experiment only have
one event information was shown on the CBP interface, log21 = 0. Each level of
factors included flowcharts and brief instructions. The flowcharts located on the left
side of the computer screen displayed all steps of the procedure with the current step
highlighted. The flowcharts located in the center of the computer screen displayed the
detail structure of the current step. System states were located in the right side of the
display.
2.3 Task Complexity
In this study, the degree in task complexity lay in the numbers of steps and parameters
in the procedures (Xu et al., 2008). The procedure with high complexity had more
steps and parameters than the one with low complexity (table 2).
Table 2. Procedures with complex and simple task
Task complexity Number of steps Number of parameters

Complex 9 99
Simple 5 45
2.4 Dependent Variables
Both objective and subjective indexes were used to evaluate operation performance.
The objective indexes include operation time and secondary task performance; the
National Aeronautics and Space Administration task load index (NASA TLX) was
used to measure the subjective workload.
2.5 Participants and Experimental Task

Twenty-one college students from National Tsing Hua University were recruited as
participants in this experiment. The average age of the participants was 25.5 years and
ranged from 22 to 29 years. All participants were randomly assigned into one of three
different information presentation styles. Half of the participants in each presentation
style start the formal experiment from complex task, others participants start from
simple task.
The CBP interface was simulated using a computer program developed by Adobe
FLASH CS6, as shown in Fig. 1.
Fig. 1. Interface of the experiment platform (Information quantity: medium; Task: complex)
The participants were told to try their best to perform the task and avoid any mistakes
during the formal experiment. During the primary task, participants are required to
detect signal as a searching task. This signals which are randomly showed on the
board in front of the participant and the response time of participants were recorded.
If the participants miss it and do not detect the signal in 15s, it is recorded as a miss.
Before the formal experiment, the participants needed to listen to the experiment
explanation from the experimenter and practiced the CBP system 5 to 10 minutes. At
the start of the formal experiment, the participants’ operation time was recorded.
After the participants finished an experiment treatment, they were asked to take the
first NASA-TLX questionnaire and took a break and then continued to execute the
other experiment treatment. Finally, the participants took the second NASA-TLX
questionnaire and finished the experiment.
3 Results
In this experiment, all statistical analyses were carried out with MINITAB 16.
Multivariate analysis of variance (MANOVA) was computed for the operation time,
secondary task performance, and workload.
The results of MANOVA indicated that information quantity (Willks’ λ = 3.743, P
< 0.01), and task complexity (Willks’ λ = 10.289, P < 0.01) had significant effects on
the dependent measures, as shown in Table 3 and Table 4. There was no significant
interaction between information quantity and task complexity on operation time,
secondary task performance, nor workload (Willks’ λ = 3.484, P < 0.05).
Table 3. MANOVA for information quantity
Criterion Test Statistic F P

Wilks' 0.50943 3.743 0.003
Lawley-Hotelling 0.87084 3.919 0.003
Pillai's 0.53751 3.553 0.005
Roy's 0.74759
Table 4. MANOVA for task complexity
Criterion Test Statistic F P

Wilks' 0.47564 10.289 0.000
Lawley-Hotelling 1.10243 10.289 0.000
Pillai's 0.52436 10.289 0.000
Roy's 1.10243
With regard to the individual dependent variables, information quantity

significantly influenced the operation time (F = 9.66, P < 0.01), secondary task
performance (F = 5.37, P < 0.05), and workload (F = 5.41, P < 0.05).
Figure 2 shows the operation time under each information quantity. Comparing the
operation time between these three information quantities demonstrated that the
operation time in the high and medium quantity was faster than that in the low
quantity.
The secondary task performance under each information quantity is showed in
Figure 3. The secondary task performance in the medium quantity was significantly
better than that in the high and low information quantity.
Figure 4 shows the workload under each level of information quantity. Comparing
the workload between these three information quantities demonstrated that the
workload in the medium quantity were lower than that in the high and low quantity.
With regard to the individual dependent variables, task complexity significantly
influenced the operation time (F = 32.47, P < 0.01), but not the workload nor
secondary task performance.
Comparing the operation time between these two tasks complexity demonstrated
that the operation time in the simple task was faster than that in the complex task.
Fig. 2. Main effect plot for operation time

Fig. 3. Main effect plot for secondary task performance
Fig. 4. Main effect plot for workload
4 Discussion and Conclusion
In this study, an experiment was designed to explore how the information quantity of
CBP interface and task complexity influence the performance of operators from the
criteria of operation time, secondary task performance, and subjective workload.
The information quantity significantly influenced the operation time, secondary
task performance, and workload. When the CBP interface with high information
quantity, the participants need to spend more time and pay more attention on the task,
and may ignore the secondary task and increase mental workload. On the other hand,
when the CBP interface with low information quantity, the participants need more
actions to click on procedure steps producing a detailed system state of the procedure
step, and thus the participants spent more time on the low information quantity task.
The task complexity significantly influenced the operation time, in this experiment,
and complex task corresponded to longer task time.
Three levels of information quantity were compared in this experiment. Medium
information quantity could cause a significantly lower workload than high and low
information quantity. Furthermore, Medium information quantity resulted in the
highest secondary task performance. There are two reasons to explain the result of
this experiment. First reason is that it decreases the information quantity of detailed
system state from high information quantity of CBP interface. The other reason is that
medium information quantity of CBP interface decreases the number of click CBP
interface from low information quantity of CBP interface. Therefore, the participant
could identify system state more efficiently, and induce lower mental workload.
These results suggest that medium information quantity of CBP interface is better
than other level of information quantity.
The interface in an advanced NPP main control room are modernized fully
integrated digital design, where CBP have a range of capabilities that may support
operators in controlling the NPP (Yang et al., 2012). This study verified the
information quantity of CBP interface, including operation time, secondary task
performance, and mental workload. In conclusions, the results of this study may be
applied on the advanced NPP, as well as be generalized to the design of information
display of other computerized system, such as process control system, air traffic
control system and flight management system.
Acknowledgments. Partial of this study was supported by National Science Council,

Taiwan, under project No. NSC101-2221-E-007-013-MY2.
References
1. Husseiny, A.A., Sabri, Z.A., Packer, D., Holmes, J.W., Rodriguez, R.J.: Operating
procedure automation to enhance safety of nuclear power plants. Nucl. Eng. Des. 110(3),
277–297 (1989)
2. Xu, S., Song, F., Li, Z., Zhao, Q., Lou, W., He, X., Salvendy, G.: An ergonomics study of
computerized emergency operating procedures: Presentation style, task complexity, and
training level. Reliability Engineering and System Safety 93, 1500–1511 (2008)
3. Yang, C.W., Yang, L.C., Cheng, T.C., Jou, Y.T., Chiou, S.W.: Assessing mental workload
and situation awareness in the evaluation of computerized procedures in the main control
room. Nuclear Engineering and Design 205, 713–719 (2012)
4. NUREG/CR-6634, O’Hara, J., Higgins, J., Stubler, W., Kramer, J.: Computer-based
Procedure Systems: Technical Basis and Human Factors Review Guidance (NUREG/CR-
6634). U.S. Nuclear Regulatory Commission, Washington, DC. (2000)
5. NUREG/CR-6749, O’Hara, J.: Integrating digital and conventional human system
interface technology: Lessons learned from a control room modernization program
(NUREG/CR-6749). U.S. Nuclear Regulatory Commission, Washington, DC. (2002)
6. Portmann, F., Lipner, M.H.: An Operational Model for Using a Computerized Emergency
Operating Procedures System. Modern Power Systems (2002)
7. O’Hara, J., Pirus, D., Nilsen, S., Biso, R., Hulsund, J.E., Zhang, W.: Computerisation of
procedures lessons learned and future perspectives. OECD HALDEN REACTOR
PROJECT. HPR-355 (2003)
8. Jung, Y., Seong, P.H., Kim, M.C.: A model for computerized procedures based on
flowcharts and success logic trees. Reliability Engineering and System Safety 26, 351–362
(2004)
9. Carvalho, P.V.R., dos Santos, I.L., Gomes, J.O., Borges, M.R.S., Guerlain, S.: Human
factors approach for evaluation and redesign of human-system interfaces of a nuclear
power plant simulator 29, 273–284 (2008)
10. Hong, J.H., Lee, M.S., Hwang, D.H.: Computerized procedure system for the 1400
simulator. Nuclear Engineering and Design 239, 3092–3104 (2009)
11. Wourms, D., Rankin, W.: Computer based procedure (Tech Report CSERIADC-RA-94-
002). Wright-Patterson Air Force Base. Crew Systems Ergonomics Information Analysis
Center, OH (1994)
12. Wickens, C.D., Hollands, J.G.: Engineering psychology and human performance. Prentice-
Hall Inc., NJ (2000)
Development of Communication Support System
at Mealtimes Using Tabletop Interface
Junko Itou, Shizuka Takao, and Jun Munemori
Faculty of systems Engineering, Wakayama University, 930, Sakaedani,

Wakayama 640-8510, Japan
{itou,munemori}@sys.wakayama-u.ac.jp
Abstract. In this research, we aim to develop a system to support communica-

tion during meal times. It is reported that the communication skills of young
people have declined because of decrease of opportunity to speak face to face.
We focus on the meal that is essential for everyday life as a place for face-to-
face communication. The proposed system uses tabletop interface as a dining
table to provide intuitive control without additional burden. Users can decorate
the table and dishes with several digital items by only touching the surface of
the table. Furthermore the system displays various visual events randomly on
the table according to the condition of the items. By using visual effects, we
realize the support to provide improving communication and enhancing the en-
joyment of the meal. From the result of experiment, it was cleared that users
could take a cue of conversation through use of the system.
1 Introduction
We propose a system that people can enjoy communication during their meal using
table top interface. The research for approximately 400 college and high school
students by the Ministry of Health, Labor and Welfare[1] reported the decline in
communication skills of Japanese young people. One-third of the target of the re-
search felt that the ability to communicate was insufficient. As the main cause, the
report pointed out the spread of mobile phones and the advancement of information
networks.
In face-to-face communication such as daily conversations, people convey linguis-
tic information with non-verbal information including gestures and facial expressions.
On the other hand, the only verbal information is exchanged in communication via
networking tools. It is not necessary to meet other people and it will reduce the oppor-
tunity to express and to construe the nonverbal expressions. This situation is one of
factor of decline in communication skills. Therefore the support of face-to-face com-
munication is a very important issue. In order to support making an opportunity of
face-to-face communication and activating the communication, we focused on the
meal that everyone always do in daily life.
We adopt a table top interface as a dining table to support communication during
meal times. The table can provide intuitive control so users can use the system easily
Development of Communication Support System at Mealtimes 175
while eating. The existing researches[2][3] provide the fun of mealtime communica-
tion by sharing photographs or decoration items using cameras and a projector. How-
ever, it is limited to operate the table because the user’s head or his/her hand blocks
the camera’s view and only one person controls the table at a time. We realize an
interactive table which many users can control freely and easily at a time so that the
users enjoy their meal times by viewing visual effects. E-Table[4] is an interactive
ordering system that diners can order food and drinks directly from a menu projected
on to their table in their hands in a restaurant and bar. Diners also can search and
browse information of the dishes or the local neighborhood of the restaurant until a
wait staff serves, and call a taxi during their meal times using this table. Whereas
E-table focuses on rationalization in business and fun to pass the time until dishes are
served, our system support to make opportunity to speak face to face among diners by
displaying visual effects including accidental visual events on the table.
This paper is organized as follows: in section 2, we will describe the related servic-
es on support of mealtime communication and interactive systems for dinners using
table top interface. In section 3, we explain our system that supports to initiate the
conversation starter and enhancing the enjoyment of meal time communication. A
validation test of our system will be given in section 4. Finally, we discuss conclu-
sions and a future work in section 5.
2 Related Research on a Support System of Meal Time

Communication
There have been a number of communication support systems at mealtimes using a

camera, a projector, a meal plate and a table. We discuss an effective approach to
support meal time communication aiming to make an opportunity to speak face to
face in this section.
pHotOluck[2] is a mealtime communication support system that consists of dishes,
a projector and a camera. The projector mounted to the ceiling displays photographs
that users have taken previously, on a meal plate on a dining table. The camera recog-
nizes three markers on the plate and estimates the position to project an image even if
users change the plate’s position and its direction. However, this system has a prob-
lem with operability because users cannot act in such a way as to cover the markers
on the plate form the camera by their hands or heads. Furthermore, people hold
chopsticks, a knife, a fork and a spoon in their hands during meal time so it is neces-
sary to achieve more intuitive interface that users don’t have to pay mind to operate.
Mori et al. has proposed an augmented reality system in order to make meal time
fun[3]. This system takes a movie of a dish on a dining table with a video camera. The
system analyzes its colors and calculates colors to make the dishes look more deli-
cious based on the analysis. The result is projected onto the dish as a decoration of its
rim. In addition, the image that a cock designs with illustrations and messages to de-
corate the dining table, the meal plate and the dish in the kitchen is also projected.
The cock can change the contents which the projector shows according to remaining
176 J. Itou, S. Takao, and J. Munemori
amount of the dish. This research concluded that the visual effects improved palatabil-
ity and pleasure during meal time. On the other hand, the eater itself cannot change
the rim decorations and the illustrations. Furthermore it is limited that the meal plate
the image is projected on is only one. Therefore this system is unsuitable for the situa-
tion that several persons sit around the dinner table and communicate face to face.
Compurants Ltd. has released an interactive ordering system ‘E-table’ with a table
top interface[4][5]. Dinners can order foods and drinks directly touching the surface
of the table where an illustrated food and drinks menu is displayed on. Dinners also
can set their table ambience and watch information on the ordered dishes or the local
neighborhood of the restaurant until the ordered dishes are served. Anyone can oper-
ate the system intuitively by adopting table top interface. As just described, E-table
helps dinners and the restaurant staffs in the point of rationalization in business and
service. On the other hands, this system doesn’t play a role strongly in supporting
face-to-face communication rather E-table decreases the opportunity for conversation
among diners and staffs because the dinners can enjoy the meal time on his/her own.
Based on the researches and services on support of meal time communication, we
aim to develop an interactive system that makes users enjoy a meal time and commu-
nication. For support face-to-face communication during meal time, we employ table
top interface and several visual effects so that several users can operate at the same
time and the system trigger a conversation between them.
3 System
3.1 Goal
Our goal is development of communication support system that users can operate
easily and intuitively while they eat. The system also provides opportunities of con-
versations and makes users enjoy the mealtimes. In order to achieve this goal, we
design the system as follows. First we adopt a table top interface as a meal table to
reduce user’s burden in term of use during a meal. Secondly, the system projects vari-
ous kinds of visual effects on the table including in digital decorations with motions, a
background image and a topic presenting character. Users can move freely the digital
decoration items within the table even if on the dishes. Finally, several kinds of spe-
cial events occur randomly based on the state of the digital decoration items displayed
on the table so that the event livens up the table and it gives an opportunity of conver-
sation among users.
3.2 System Structure

In this system, we use Diamond Touch Table (hereinafter referred to as ‘DTT’) as a
table top interface[6]. DTT can identify up to four persons at a time. In addition, cam-
eras and markers are not required for recognition of users’ action on the table. Fig.1
shows the overview of our system. An image from the projector is displayed on the
screen that is the surface of DTT. Users treat DTT as a dining table and sit around
DTT as shown Fig.2 and Fig.3. Employing DTT which provides intuitive controls
reduces users’ burdens and makes it possible to project many types of visual effects
on the dishes and the table.
Fig. 1. Overview of proposed system
Fig. 2. Dishes and DTT as a meal table Fig. 3. The table and users during meal times
3.3 Contents of Visual Effects
Fig.4 illustrates the image on the table. Users can manipulate the DTT by touching the
surface of this table. The four white circles in Fig.4 mean the positions of meal plates.
This system is considered to be used by up to four persons at same time so the buttons
for operations are placed in each four corner of the table. Users can press the buttons
only in front of themselves. This system provides several types of operations.
Fig. 4. An example of a projection image Fig. 5. Decorations with flower items
Fig. 6. Change of a background image and projection of an image on a dish directly
Decoration Function
Users can decorate the table and dishes with flowers and circles. After touching the
corresponding button and any position on the table, then a flower item is displayed.
Any users can move the items freely by drag on the table. The color of these items is
fixed depending on the seat. When a user wants to use the other color item, the user
asks for the other users to produce the item and to move it. An example of a decorated
dish with various color flowers and circles are shown Fig.5.
Background Image
Users can select the background image of the table from black to 10 different patterns
as if they change the table cloth. We aim that the users will communicate when they
want to change the background, and that the background image enhances the mood of
their meal. The sunset image is used as a background image in the left image of Fig.6.
As shown in the right image of Fig.6, the image is directly projected on the dish. Us-
ers can switch the display if the background image has an overlap with the dish.
These operations are performed by only touching the surface of the table so users
can control even if with non-dominant hand.
Topic Presenting Character

The white cube object displayed in the center of Fig.4 is a character that provides
topics of conversation. The character shows today’s news and trivia in its balloon and
it has 25 patterns of the message. The character constantly moves and rotates within
the table as all users can read the message.
Visual Event Based on Items

Various events may occur at random according to the condition such as the item
number of each color circles and flowers that are displayed on the screen. The visual
event has three types. The first one is the event that special items are displayed. An
example of this event is shown in Fig.7. Four big flowers are blossoming at the center
of dishes in the left lower image of Fig.7.Users can move these items as with the other
normal items.
Fig. 7. Examples of visual events
Fig. 8. Projections of visual events on a dish and the table

The second type of the visual event is that some decorations with motions are dis-
played around the rim of the dishes. There are 27 kinds of patterns in this event. The
right lower image of the Fig.7 is an example of this visual event. The result of the
projection on a dish is shown in Fig.8. Some big hearts are projected in blinking and
rotating around the rim of the dish.
The third type of the visual event is shower of blossoms or snow crystals. This ef-
fect covers the whole table top and it has four patterns. The top image of Fig.7 shows
the shower of snow crystals on the table. The example image of projection of this
event is the right image in Fig.8. The all sizes snow crystals are displayed on the table
and the dishes.
By using these visual effects and decorations, we realize the support to provide
opportunities of communication and enhancing the enjoyment of the meal.
4 Experimental Result
We performed an evaluation experiment using proposed system to investigate how

our system effects on communication during meal times. The experimental subjects
were 24 persons in their twenties divided into six groups. The three groups are con-
sisted of people who had met for the first time. We instructed each subjects to sit at
the table and to eat spaghetti and fruits. The meal time is 30 minutes. The subjects
could talk freely during their meal. The conversation topic was not limited. We rec-
orded the scene of the meal by a camera.
The result of a questionnaire on uses of the system is described in Table 1. The
numbers of the Table 1 means the number of subjects who answered the each degree.
The number 1 means strongly disagree and the number 5 means strongly agree. By
comparing the result of item (i) and item (ii), we obtain higher value in proposed
system than the usual meal. According to (iii), it is cleared that most subjects more
enjoyed their meal using our system than usual meal even contains first meeting per-
sons. The five subjects who answered “I don’t have any idea” or “It is troublesome”
for the question “What do you think for your daily meal? Do you enjoy it?” selected
the number 5 on the questionnaire item (i).
Table 1. Questionnaire result on use of the system
Questionnaire item 1 2 3 4 5
(i) You enjoyed your meal using the system. 0 0 0 7 17
(ii) You enjoy your usual meal with your friends. 0 0 1 10 13
(iii) This meal was more enjoyable than usual meal 0 0 0 12 11
Table 2. Questionnaire result on the support of communication
Questionnaire item 1 2 3
(iv) This system provided conversation material. 1 9 14
(v) You enjoyed this meal with first meeting persons. 1 6 5
Table 3. Analysis result of video data
With friends With strangers

Group A B C D E F
During Time (minute) 16 10 28 28 23 18
eating Time without speaking
47 8 5 59 20 13
(second)
Number of conversation
4 8 18 22 23 16
triggered by the system
Number of operation of
73 48 121 252 128 63
the table
After Time (minute) 14 20 2 2 7 12
eating Time without speaking
32 0 0 0 0 5
(second)
Number of conversation
6 23 0 0 10 12
triggered by the system
Number of operation of
39 174 8 18 81 77
the table
Table 2 is the result of the questionnaire on the communication support. The num-
ber 1 means ‘not at all’, the number 2 corresponds to ‘sometimes’ and the number 3
means ‘many times’. As shown in Table 2, 23 subjects answered that they could get
the conversation material more or less. In addition, 92% of the subjects whose group
consisted of first meetings answered that the meal became enjoyable by using this
system as shown in Table 2 (v).
Table 3 shows the analysis result of video data. The group A, B and C consists of
friends and the group D, E and F consists of first meeting persons. The 86% of sub-
jects who answered they felt gap in a conversation during the mealtimes could resolve
the situation by using this system. Totally the time without speaking after eating is
shortens the time during eating in all groups. On the other hand, the first meeting
groups tended to operate the table more frequently after eating. Group C took time to
eat their meals as much as Group D did, however the time without speaking and the
number of operating the table totally differed. By comparing group C and group D,
the members of the first meeting groups often touched the table and tended to devote
themselves actions meanwhile declining speaking time. They also answered that this
system made the opportunity to speak by providing the topic materials such as ex-
changing flower items and displaying random events. So we can confirm this system
support the conversation among the subjects.
An important thing to note, by analysis of video images taken in several groups, it
revealed that users often moved flower items and decorated dishes of their partners
with conversations. Especially, the person who finished his/her meal decorated the
plate of a user who still ate. They had conversation about the action and the result of
the decoration.
5 Conclusions
In this article, we proposed a support system for communication at mealtimes aiming

to support the activation of face-to-face conversation. We applied DTT as a table for
meals and designed the system to provide an enjoyment of meals. From the experi-
ment, it was cleared that users could take a cue of conversation through user of the
system and responses from the system even contains first meeting.
References
1. The Ministry of Health, Labor and Welfare : Youth Employability Support Program (2009)
(Japanese site), http://www.mhlw.go.jp/bunya/nouryoku/yes/
2. Nishimoto, K., Amano, K., Usuki, M.: pHotOluck: A Homeuse Table-ware to Vitalize
Mealtime Communications by Projecting Photos onto Dishes. In: Horizontal Interactive
Human-Computer Systems (TABLETOP 2006), pp. 9–16 (2006)
3. Mori, M., Kurihara, K., Tsukada, K., Siio, I.: Dining Presenter: Augmented Reality system
for a dining tabletop. In: Supplemental Proceedings of the 11th Ubicomp 2009, pp. 168–169
(2009)
4. Compurants Ltd. : “E-Table”, http://www.e-table-interactive.com/
index.html (accessed February 19, 2013)
5. Compurants Ltd. : inamo restaurant, Soho, London - interactive oriental fusion restaurant
and bar, http://www.inamo-restaurant.com/ (accessed February 19, 2013)
6. Dietz, P., Leigh, D.: DiamondTouch:A Multi-User Touch Technology. In: Proceedings of
UIST 2001, pp. 219–226 (2001)
Improving User Performance in Conditional Probability
Problems with Computer-Generated Diagrams
Vince Kellen, Susy Chan, and Xiaowen Fang
DePaul University, College of Computing and Digital Media, Chicago, IL, USA
[email protected], {schan,xfang}@depaul.edu
Abstract. Many disciplines in everyday life depend on improved performance

in probability problems. Most adults struggle with conditional probability prob-
lems and prior studies have shown user accuracy is less than 50%. This study
examined user performance when aided with computer-generated Venn and Eu-
ler-type diagrams in a non-learning context. Following relational complexity,
working memory and mental model theories, this study manipulated problem
complexity in diagrams and text-only displays. Partially consistent with the
study hypotheses, complex visuals outperformed complex text-only displays
and simple text-only displays outperformed complex text only displays. How-
ever, a significant interaction between users’ spatial ability and the use of dia-
gram displays led to a reversal of performance for low-spatial users in one of
the diagram displays. Participants with less spatial ability were significantly
impaired in their ability to solve problems with less relational complexity when
aided by a diagram.
Keywords: Human-computer interaction, diagrams, Bayesian reasoning, rela-

tional complexity, spatial ability, working memory, individual differences,
mental models.
1 Objective
Many people struggle with conditional probability problems, which are common in
the medical field and in any discipline that needs to communicate and interpret diag-
nostic evidence [1,2,3]. The research regarding facilitating user performance with
computer visuals has yielded conflicting results. The objectives of this study were to
a) determine ways of improving user performance with conditional probability prob-
lems in non-learning contexts using computer-generated Venn and Euler type dia-
grams and b) provide a set of design guidelines for designers of systems that have to
convey conditional probability problems.
2 Background Literature
According to theories of working memory [4,5] and mental models research [6,7,8,9],
visual displays should aid in these tasks as they can make better use of working
184 V. Kellen, S. Chan, and X. Fang
memory, reduce cognitive load and facilitate mental model construction, thus generat-
ing better judgments. The conditional probability problem is a common and difficult
quantitative reasoning task. While a significant amount of research has examined why
people make errors in conditional probability reasoning problems [10,11,12] very
little of it has looked systematically at optimizing people’s performance using visuals.
Moreover, practically no research has applied theories of working memory to guide
facilitation of these kinds of problems.
Counter-intuitively, some research is finding that Venn and Euler type diagrams,
which try to clearly depict set inclusions and conditional relationships directly, may
actually hurt performance [12,13,14]. Mental models theory [6], relational complexity
theory [15] and explanations of working memory limits [16,17] can provide guidance
for designing visuals to help people solve these problems.
3 Conditional Probability Problems
Several studies show participants’ accuracy on conditional probability problems is

low to intermediate, ranging from 6% of the problems correct to 62% depending on
the textual format [1,18]. An example of a conditional probability problem, shown
below, has been used in several studies [1]:
The probability of breast cancer is 1% for women at age forty who participate in
routine screening. If a woman has breast cancer, the probability is 80% that she
will get a positive mammography. If a woman does not have breast cancer, the
probability is 9.6% that she will also get a positive mammography. A woman in
this age group had a positive mammography in a routine screening. What is the
probability that she actually has breast cancer?
The normative answer requires applying a form of Bayes’ theorem to calculate the
posterior probability that a woman with positive results has cancer. The symbols H
and –H denote the two hypotheses (cancer and no cancer, respectively) and D denotes
the data representing the positive test results. The formula for computing p(H | D) is
as follows.
p( H ) p( D | H ) (.01)(.80)
p( H | D) = = = .078
p ( H ) p ( D | H ) + p ( − H ) p ( D | − H ) (.01)(.80) + (.99)(.096)
Prior studies show that even physicians estimate probabilities incorrectly when pre-
sented in text form. Eddy [10] shows that 95% of the physicians queried estimated the
probability of p(H | D) in this problem between 70-80% not 7.8%. Other prior re-
search indicates people’s accuracy is typically poor for this text-only, probability
representation of the problem [1,2,12].
Gigerenzer and Hoffrage [1] have argued that expressing conditional probability
problems as natural frequencies rather than as probabilities improves peoples’ per-
formance. Natural sampling mimics the process of encountering instances in a popu-
lation sequentially. Participants should perform better when probability problems are
Improving User Performance in Conditional Probability Problems 185
expressed as natural frequencies. A natural frequency representation has a simpler

mathematical form of Bayes’ theorem:
d &h 8
p ( H | D) = = = 0.078
d & h + d & −h 8 + 95
3.1 Use of Visuals in Conditional Probability Problem Solving

A lesser but growing body of research has addressed how to teach people to solve
statistical problems [3,11]. Unfortunately, many everyday settings that require statis-
tical reasoning abilities involve people who haven’t been taught how to reason with
statistics or have failed to retain the associated skills. One common diagram, an Euler
diagram, failed to provide significant improvement in either a probability or frequen-
cy format [12] and performed worse than iconic displays [13] which depict each item
in the frequency of observations as an icon. Iconic displays are superior to Venn dia-
grams because they better approximate ‘actual ecological presentations’ [13]. In a
study examining deductive reasoning [13], researchers found a disadvantage for Euler
diagrams, which performed worse than similar text-only representations.
3.2 Mental Models, Working Memory and Reasoning

Both behavioral and neuropsychological mental models theory research [6,7,9] shows
that people reason in these problems not by following linguistic rules, but by means of
visuospatial models of the problem structure. While verbal processing and the phono-
logical loop would need to be employed to process the text of these problems and for
certain math operations solving the problems requires manipulation of the premises
and relations within the problem using visuospatial processing and central executive
cognitive resources. Several dual task studies show that the VSSP and CE are in-
volved in problem solving, consistent with mental models theory [8,19]. Mental mod-
els need not be visual, and can represent non-spatial relationships, such as kinship, or
non-visual precepts in deductive reasoning. Model based reasoning interferes with
other spatial tasks, not necessarily with concurrent visual tasks without a spatial com-
ponent [9]. Due to the overlapping use of the VSSP for both reasoning and visual
processing, it is possible that use of visual diagrams might be interfering with perfor-
mance on reasoning problems.
Cowan [20] defines short term memory as “faculties of the human mind that can
hold a limited amount of information in a very accessible state temporarily” and sur-
mises that working memory is used to retain partial results while working out a math
problem without paper or “to combine the premises in a lengthy rhetorical argument.”
Baddeley’s [4] influential model of working memory holds that verbal-phonological
and visuospatial representations are stored separately and managed and manipulated
by the central executive. In Baddeley’s model [4], working memory is generally
viewed as several components working together [20]. Although early research into
working memory capacity showed that people can recall about seven chunks of
information in short-term memory tasks [21] more recent and extensive research is
placing a lower limit of about four chunks on the capacity limits in short-term memo-
ry [16,17] across both visual and verbal tasks.
3.3 Relational Complexity
Relational complexity is defined as the number of relations that must be processed in

parallel, that is, at the same time, to perform a task [15,22] As relational complexity
increases, the processing complexity of a cognitive process increases. While a cogni-
tive process may be made up of several steps, the complexity of the process is the
measure of the complexity of its most complex task. Process complexity is affected
by how many interdependent elements have to be processed in parallel at one time,
not how many have to be processed over time. The maximum relational complexity
adults can normally process is the quaternary relationship which binds four interactive
elements. Planning a correct strategy to solve a problem depends first on representing
the complete structure of the problem. Relational theory predicts that people’s need to
process premises jointly will increase processing load [15]. Mental model construc-
tion is constrained by the complexity of the model itself and the working memory
capacity, especially the ability for the person to focus attention to the features of the
model. In a neurological study [23], researchers confirmed that increasing relational
complexity from 0 to 4 in a reasoning problem dramatically increased the percentage
of errors and the time in processing, as well as increasing activation in areas of the
brain associated with working memory.
4 How Visuals Can Improve Performance
Conditional probability problems typically contain three and can contain as many as
six or seven elements that might need to be processed in parallel. Based on relational
complexity theory, mental models theory, cognitive load theory and the theories of
working memory discussed here, for visuals to reliably help with these kinds of prob-
lems, designers need to consider the following:
1. Through perceptual cueing mechanisms such as highlighting and controlling what
is visible in the display, restrict the user’s focus of attention to the task of relating
two independent elements at a time.
2. Use visuals to depict relationships between entities rather than describing the rela-
tionship with words.
3. Repeat the process of cueing and processing fewer rather than more elements at a
time to help the user make a series of correct inferences.
Using this approach, designers ought to be able to improve user performance in mak-
ing correct inferences by minimize element interactivity in the display. This encou-
rages users to build partial models based on fewer interacting elements at a time and
substitute images for words to depict relationships.
5 Hypotheses
Based on relational complexity theory, working memory theory and mental models
research, diagrams perform better than text and displays with lower element interac-
tivity should perform better than those with high element interactivity. This leads to
two hypotheses.
• Hypothesis 1: Under both high and low levels of relational complexity, users will
demonstrate higher performance with diagrams than with text.
Diagrams should help by further relieving working memory by substituting visuals in

place of some text (reducing the amount of text processing needed) and by providing
visual cues and guidance than can’t be equivalently done in a simpler text-only repre-
sentation. While performance in the diagram or text-only condition is expected to be
strongly influenced by complexity, the role of the visual under these two conditions is
of interest.
• Hypothesis 2: Under both the text and diagram conditions, users will demonstrate
higher performance with problems of lower relational complexity than with prob-
lems of higher relational complexity.
Since working memory demands will be reduced with low relational complexity dis-
plays that have less element interactivity, users will generate more correct answers.
Some of the problems with the use of visuals in prior research may be a result of
researchers failing to consider the number of interacting elements in conditional prob-
ability problems, the overall working memory demands of these problems and the
limited working memory resources available.
6 Method
This study recruited 158 participants from a large research university. This study used
both students and staff participants with varying ages and backgrounds in order to
better approximate real-world populations. This study used a 2X2 factorial design
with the two factors being complexity and display type. Complexity has two levels:
low (three interacting elements at a time) and high (six interacting elements at a time).
Display type has two levels: text-only and diagrams. Because participants were ran-
domly assigned to one of the four groups in the factorial design, this was a between-
subjects design. The independent variables were relational complexity (low relational
complexity and high relational complexity) and display type (text-only and diagrams).
Four treatments were used. Text-only treatments with low and high relational com-
plexity and Venn-type diagram treatments of low and high relational complexity were
used and presented via a computer display. Participants expressed their answers in
whole numbers and were allowed to round up or down to the next integer. User
performance was measured by number of correct answers out of the ten in the
battery of problems. Participants had 60 minutes to complete the 10 problems and all
participants completed them within the allotted time. To test for individual differences
and their effect on problem performance, participants completed a pre-test survey that
collects background information (age, gender, educational level, probability reasoning
experience, skill, and whether English is a native language). Because spatial ability
may contribute to, if not be the primary source of people’s performance with regards
to relational complexity [24] and is used for processing spatial diagrams [25], this
study measured spatial ability [26] and used it as a covariate.
7 Results
Approximately one-third of the participants (57) were older than 22 and 28 of the
participants were not in college. None of the participant background variables were
found to be significantly or marginally associated with performance. Performance was
measured by the count of correct problems out of the ten problems presented. Table 1
shows the means, standard deviation and sample size for each condition.
Table 1. Performance means

Display Type
Complexity Diagram Text
M SD n M SD n
Complex 5.38 2.62 40 3.32 2.00 37
Simple 6.39 3.60 44 7.41 3.13 37
Since the hypotheses call for testing for main effect and for interaction with the co-
variate, a two-way ANCOVA analysis was conducted to test for the influence of
complexity and display type and to check for any influence of spatial ability (as
measured by the VZ-2 test score). The overall analysis of covariance was highly sig-
nificant (F(6,151) = 17.56, p<.0001, η2=.41). The analysis of the model effects are
shown in Table 2.
Table 2. ANCOVA for Performance
Source DF SS MS F Pr > F
Display type 1 11.47 11.47 1.76 0.1860
Complexity 1 236.97 236.97 36.47 <.0001
Spatial ability 1 276.10 276.10 42.50 <.0001
Complexity*Display type 1 62.53 62.53 9.62 0.0023
Display type*Spatial ability 1 75.39 75.39 11.60 0.0008
Complexity*Spatial ability 1 18.30 18.30 2.82 0.0953
This study showed a significant main effect for complexity but not for display type.
However, the main effects (complexity and display type) interacted and the covariate
(spatial ability) interacted with Display type. The ANCOVA analysis also showed
that the slopes of the regression lines for each treatment are not the same, also con-
firming an interaction between display type and spatial ability.
Pairwise comparisons to test the hypotheses for performance were conducted with
one-tailed t-tests to match the hypotheses, adjusted for the differing covariate means
and for multiple inferences. Looking at the first hypothesis, H1, participants in this
study demonstrated higher levels of performance with diagrams than with text for
only the high relational complexity (complex) treatments (5.38 versus 3.32 correct,
t(75)=3.29, p=.0034, Cohen’s d=.76). Participants did not demonstrated better
performance with diagrams than with text in the low relational complexity (simple)
condition (6.39 versus 7.41 correct, t(79)=-1.53, p=.7878). Thus H1 is partially
confirmed.
Looking at the second hypothesis (H2), participants demonstrated better perfor-
mance with low relational complexity (simple) displays over high relational complex-
ity (complex) displays for only the text treatments (7.41 versus 3.32 correct,
t(72)=6.57, p<.0001, Cohen’s d=1.55.) Participants did not demonstrate improved
performance with simple displays than with complex ones for the diagram treatments
(6.39 versus 5.38 correct, t(82)=1.46, p=.0840). Thus H2 is partially confirmed. Table
3 lists the comparisons for H1 and H2.
Table 3. Pairwise comparisons

M SD M SD t-test Cohen’s d
H1: Diagrams > Text
3D>3T 6.39 3.60 7.41 3.13 -1.53
6D>6T 5.38 2.62 3.32 2.00 3.29 ** .76
H2: Low RC > High RC
3D>6D 6.39 3.60 5.38 2.62 2.07
3T>6T 7.41 3.13 3.32 2.00 6.54 *** 1.55
**p ≤ .01 ***p ≤ .001
This study finds partial support for both hypotheses. Using diagrams helped partic-
ipants in the complex condition. Because diagrams enabled use of visual processing
and text processing that conserves working memory and provides cues and guidance
not possible in a text display, users performed better. Reducing relational complexity
also helps improve performance, but for the text-only condition
In looking at the ANCOVA results, diagram displays and spatial ability are affect-
ing user performance. To understand the nature of the main effects interaction it is
helpful to see performance by a split of
spatial ability (below). Low spatial partici-
pants performed 51% worse (4.14 versus
8.44, N=21 and 23 respectively) on the
simple diagram and 28% worse (4.44 ver-
sus 6.24, N=19 and 21 respectively) on the
complex diagram than the high spatial
participants. Additional two-tailed t-tests
confirmed this finding for both the
simple diagram (3D) displays (t(42)=4.79,
p<.0001, Cohen’s d=1.46) and the 6D
(t(38)=2.31, p<.0264, Cohen’s d=.74). Also of note, low-spatial and high spatial par-
ticipants did not differ in their performance on both the simple and complex text-only
versions of the problem (p=.5662 and p=.9267 respectively). The diagram displays
are impairing low-spatial participants, precipitously for the simple diagram (3D)
treatment and significantly for the complex diagram (6D) treatment. The text displays
performed equally well for high- and low-spatial participants.
8 Discussion
While overall, reducing relational complexity improves user performance and more
complex problems are aided by diagrams, spatial ability is interacting with display
type. Diagram displays hurt the performance of participants with less spatial working
memory. Spatial ability may be a shared working memory resource serving both non-
visual reasoning and perceptual processes. Perceptual processes, being bottom-up
processes, may be taking priority and interfering with the reasoning processes for
low-spatial users. In terms of Baddeley’s model [4], this may mean that the VSSP
may be required for non-visual reasoning processes like those used in conditional
probability problems as well as for processing an external representation, which in
this case is a diagram. Low-spatial users may have less capacity to process or inhibit
and restrict bottom-up processing, thus causing visual diagram processing to conflict
with spatial working memory needed to construct and validate a correct mental mod-
el. While these bottom-up perceptual processes may be the basis of improvement for
high-spatial users as it benefits them in offloading cognitive work to the visual
stream, these bottom up processes may be crowding out working memory needed for
reasoning in the low-spatial user.
Since spatial ability is also associated with other intelligence measures [27,28], one
might also expect that low-spatial participants would have reduced performance on
the text-only version of the problem due to presumed deficits in other non-visual cog-
nitive capacities. This study failed to identify that effect. Low and high spatial users
performed similarly on text-only versions of the problem. Since this study relied ex-
clusively on the computer presentation of each problem by not allowing any second-
ary tools such as notepaper to interact with, participants had no choice but to interact
with the computer display to solve the problem. This study suggests that capacity
limits in spatial working memory may be the culprit for the diminished diagram per-
formance for low-spatial users.
To help users with conditional probability problems in everyday, non-learning con-
texts, this study recommends the following guidelines:
1. Reduce relational complexity. With the exception of the simple diagram display for
low-spatial users, simpler problems helped users produce more correct answers.
2. For low-spatial users, use simpler, text-only displays. These displays will facilitate
performance better than diagrams will.
3. For high-spatial users, Venn and Euler-type diagrams can be safely used and can
improve performance.
4. Use natural frequencies. All of the problems used in this study used natural fre-
quencies and the best performing treatment, the simpler text-only displays, had a
mean of 7.51 out of 10 correct. If all components of the conditional probability
problem must be displayed at once, the diagrammatic display will provide the best
performance overall.
5. If spatial ability measures are available for the target audience, designers can per-
sonalize the display. Simpler diagram displays can be used with high-spatial users
and simpler text-only displays can be used with low-spatial users.
9 Conclusion
This study has further clarified how diagrams can (or cannot) aid in solving condi-
tional probability problems. Reductions of complexity can improve performance,
especially with text displays. Diagrams can help, but they can hurt too. Individual
differences in spatial ability matter when processing Venn and Euler-type diagrams.
The impairment with diagrams for people with lower spatial ability could be the
source of some of the conflicting findings on the use of diagrams to facilitate condi-
tional problem. The extensive collection of research into people’s performance in
solving conditional probabilities and Bayesian reasoning has not yet applied concepts
and frameworks from cognitive psychology like relational complexity and theories of
working memory. The recent research into mental models and the neural correlates of
working memory and reasoning can provide linkages between the functional and
physical descriptions of the mind. These frameworks can be applied to improve our
understanding of user performance problems with computer diagrams without having
to refer to less testable constructs such as ‘frequency coding in the mind’ posited by
frequentist interpretations of performance [18].
References
1. Gigerenzer, G., Hoffrage, U.: How to improve Bayesian reasoning without instruction:
frequency formats. Psychological Review 102, 684–701 (1995)
2. Kahnamen, D., Lovallo, D.: Timid choices and bold forecasts: a cognitive perspective on
risk taking. Management Science 39(1), 17–31 (1993)
3. Sedlmeier, P.: How to improve statistical thinking: Choose the task wisely and learn by
doing. Instructional Science 28, 227–262 (2000)
4. Baddeley, A.: Working memory. Clarendon Press, Oxford (1986)
5. Miyake, A., Shah, P.: An Introduction. In: Miyake, A., Shah, P. (eds.) Models of Working
Memory. Cambridge University Press (1999)
6. Johnson-Laird, P.N., Legrenzi, P., Girotto, V., Legrenzi, M.S.: Naive probability: A men-
tal model theory of extensional reasoning. Psychological Review 106, 62–88 (1999)
7. Fangmeier, T., Knauff, M., Ruff, C., Sloutsky, V.: fMRI evidence for a three-stage model
of deductive reasoning. Journal of Cognitive Neuroscience 18(3), 320–334 (2006)
8. Knauff, M., Jola, C., Strube, G.: Spatial reasoning: No need for visual information. In:
Montello, D.R. (ed.) COSIT 2001. LNCS, vol. 2205, pp. 447–457. Springer, Heidelberg
(2001)
9. Ruff, C.C., Knauff, M., Fangmeier, T., Spreer, J.: Reasoning and working memory: Com-
mon and distinct neuronal processes. Neuropsychologia 41, 1241–1253 (2003)
10. Eddy, D.M.: Probabilistic reasoning in clinical medicine: Problems and opportunities. In:
Kahneman, D., Slovic, P., Tversky, A. (eds.) Judgment under Uncertainty: Heuristics and
Biases, pp. 249–267. Cambridge University Press, Cambridge (1982)
11. Gigerenzer, G., Hoffrage, U.: Overcoming difficulties in Bayesian Reasoning: A reply to
Lewis & Keren and Mellers & McGraw. Psychological Review 106, 425–430 (1999)
12. Sloman, S.A., Over, D., Slovak, L., Stibel, J.M.: Frequency illusions and other fallacies.
Organizational Behavior and Human Decision Processes 91, 296–301 (2003)
13. Brase, G.L.: Pictorial representations in statistical reasoning. Applied Cognitive Psycholo-
gy (2008), doi:10.1002/acp.1460
14. Calvillo, D.P., DeLeeuw, K.E., Revlin, R.: Deduction with Euler Circles: Diagrams That
Hurt. In: Barker-Plummer, D., Cox, R., Swoboda, N. (eds.) Diagrams 2006. LNCS
(LNAI), vol. 4045, pp. 199–203. Springer, Heidelberg (2006)
15. Halford, G.S., Wilson, W.H., Phillips, S.: Processing capacity defined by relational com-
plexity: Implications for comparative, developmental, and cognitive psychology. Beha-
vioral and Brain Sciences (21), 803–865 (1998)
16. Cowan, N.: The magical number 4 in short-term memory: A reconsideration of mental sto-
rage capacity. Behavioral and Brain Sciences 24, 87–185 (2000)
17. Cowan, N., Elliott, E.M., Saults, J.S., Morey, C.C., Mattox, S., Hismjatullina, A., Conway,
A.R.: On the capacity of attention: Its estimation and its role in working memory and cog-
nitive aptitudes. Cognitive Psychology 51, 42–100 (2005)
18. Brase, G.L., Cosmides, L., Tooby, J.: Individuation, counting, and statistical inference: the
role of frequency and whole-object representations in judgment under uncertainty. Journal
of Experimental Psychology: General 127(1), 3–21 (1998)
19. Klauer, K.C., Stegmaier, R., Meiser, T.: Working memory involvement in propositional
and spatial reasoning. Thinking and Reasoning 3(1), 9–47 (1997)
20. Cowan, N.: What are the differences between long-term, short-term, and working memo-
ry? In: Sossin, W.S., Lacaille, J.-C., Castellucci, V.F., Belleville, S. (eds.) Progress in
Brain Research, vol. 169. Elsevier B.V. (2008)
21. Miller, G.A.: The magical number seven, plus or minus two: Some limits on our capacity
for processing information. Psychological Review 63, 81–97 (1956)
22. Halford, G.S., Wilson, W.H., Phillips, S.: Relational knowledge: The foundation of higher
cognition. Trends in Cognitive Sciences 14(11), 497–505 (2010)
23. Kroger, J.K., Sabb, F.W., Fales, C.I., Bookheimer, S.Y., Cohen, M.S., Holyoak, K.J.: Re-
cruitment of anterior dorsolateral prefrontal cortex in human reasoning: a parametric study
of relational complexity. Cerebral Cortex 12, 477–485 (2002)
24. Halford, G.S., Cowan, N., Andrews, G.: Separating cognitive capacity from knowledge: A
new hypothesis. Trends in Cognitive Science 11(6), 236–242 (2007)
25. Hegarty, M.: Individual differences in use of diagrams as memory in mechanical reason-
ing. Learning and Individual Differences 9(1), 19–42 (1997)
26. Ekstrom, R.B., French, J.W., Harman, H.H.: Manual for kit of factor-referenced cognitive
tests. Educational Testing Service, Princeton (1976)
27. Hegarty, M., Waller, D.: Individual differences in spatial ability. In: Shah, P., Miyake, A.
(eds.) The Cambridge Handbook of Visuospatial Thinking, pp. 121–169. Cambridge Uni-
versity Press, Cambridge (2005)
28. Miyake, A., Friedman, N.P., Rettinger, D.A., Shah, P., Hegarty, M.: How are visuospatial
working memory, executive functioning, and spatial abilities related? A latent-variable
analysis. Journal of Experimental Psychology: General 130(4), 621–640 (2001)
Establishing a Cognitive Map of Public Place
for Blind and Visual Impaired by Using IVEO Hands-On
Learning System
Qing-Wen Lin1, Sheue-Ling Hwang1, and Jan-Li Wang2

1
National Tsing Hua University, Hsinchu, Taiwan 300,
2
National Center for High-Performance Computing, NARL, Hsinchu, Taiwan 300
[email protected]
Abstract. In recent years, there has been an escalation of orientation and way-
finding technologies and systems for the blind visually impaired people. The
purpose of this study was to help visual impaired to access and explore spatial
information by establishing a cognitive map of public place using IVEO hands-
on learning system and tactile direction map. The IVEO hands-on learning
system combines touch, sound, and sight, using touch and sound learning mod-
alities to help the blind and visually impaired to learn faster and retain informa-
tion longer, meanwhile making learning more interesting and interactive.
Subjects were divided into four groups in terms of route familiarity and IVEO
Aids of exposure, with an equal number of subjects (N=4) in each group. Then,
the subject performed an actual Way-finding test to get back and forth between
Ximending and TDTB (Taiwan Digital Talking Books association). The depen-
dent variables were total time, number of miss route, and number of request.
The results revealed a significant effect of IVEO Aids, and the subjects could
perform better through IVEO Aids. This research provided a learning method of
orientation and mobility for the blind and visually impaired, and assisted them
to be familiar with the route safer and faster.
Keywords: Learning System, audio-tactile maps, cognitive map, Way-finding,

blind people.
1 Introduction
1.1 Background and Motivation
According to the statistics by the Ministry of Interior, Taiwan, the legal population of
the visually impaired increased from 44889 persons in 2002 to 56589 persons in 2012
in Taiwan. The population of the blind and the visually impaired has been increasing
sharply over the past 10years. (Taiwan Ministry of the Interior, 2012).
194 Q.-W. Lin, S.-L. Hwang, and J.-L. Wang
In an alien environment, the sighted adult usually need a map to find spatial infor-
mation while a tactile direction map is required for the blind and visually impaired
(Levine, 1982; O’Neill, 1999; Klippel, Freksa & Winter, 2006).
The blind and visually impaired can successfully use verbally presented route
information as well as survey information in constructing cognitive maps of environ-
ments, even if they are not as efficient as normally sighted people (Steyvers
and Kooijman, 2008).Thus audio-tactile maps, which allow a multi-modal co-
comprehension of tactilely perceived and verbally presented information, are a prom-
ising way to diminish some restrictions of tactile maps.
1.2 Objectives
The IVEO hands-on learning system combines touch, sound, and sight, using touch
and sound learning modalities to help the blind and visually impaired to learn faster
and retain information longer, meanwhile making learning more interesting and inter-
active (Gardner, 2006). We propose assistive technology system to aid in way-finding
based on IVEO hands-on learning system. The participants who learned the route by
using IVEO Aids and tactile direction map showed a more accurate level of confi-
dence in retrieving the spatial information necessary to complete the way-finding. The
IVEO hands-on learning system is shown in Figure 1.
1 3
Fig. 1. The IVEO hands-on learning system
Tactile maps can also be an extremely effective tool for representing spatial infor-
mation for the orientation and mobility (O&M) student (Bentzen, 1997). Blades et al.
(2002) found that the performance of visually impaired people can be enhanced by the
strategies used for learning a new route, e.g. by modeling a map of the route itself.
However, many orientation and mobility (O&M) counselors dedicate a great portion
of their lesson preparation time to the founding of detailed tactile maps to be used
Establishing a Cognitive Map of Public Place for Blind and Visual Impaired 195
only once or twice by the student. An IVEO creator software and View-Plus Emboss-
er for producing highly detailed, “one-off” maps would be of great advantage to these
counselors and their students, freeing the counselor's time for actual O&M counselor,
instead of the difficult task of tactile cartography.
2 Research Method
2.1 IVEO Hands-On Learning System Interface Design
Automatic tactile-audio map creation is the key component of our proposed interac-
tive solution. The step takes a single map image as input and generates output image
for easy tactile reproduction and an SVG file which combines graphical contents with
audio annotations. We applied the mainstream graphics software, such as Google
Sketch Up to design the map image, and interviews and questionnaire were designed
to identify the requirements that support the blind and visually impaired for Way-
finding activities. Moreover, use IVEO creator software to input the title and content
to customize tactile map. In addition, View-Plus Embosser was used to print the tac-
tile map and place it on the IVEO touchpad so that the users can press different areas
of the map to hear the titles and content spoken back to them. The Processing flow of
the IVEO System Interface Design in Figure 2.
2.2 Experimental Design
The primary criterion for subjects selection was that they are the blind visually im-
paired people. Moreover, they have experience to walk alone in the streets. The sub-
ject was asked to fill out a SDQS/sense of direction questionnaire-short form which
elicited information concerning his cognitive map, sense of direction, spatial memory,
familiarity and uniform into four groups accordance with scores. The independent
variables were route familiarity (low or high) and IVEO Aids of exposure (with or
without). The low familiarity was defined as one practice of walking from Ximending
to TDTB before proceeding Way-finding test, and the high familiarity was defined
as three times practices of walking from Ximending to TDTB.
A total of 16 visually impaired participated in this experiment, and the subject
was assigned to one of four groups in combination of experimental variables,
familiarity and IVEO Aids of exposure, with an equal number of subjects (N=4) in
each group. R1 group is low route familiarity and without IVEO Aids, R2 group is
high route familiarity and without IVEO Aids, R3 group is low route familiarity and
with IVEO Aids, R4 groups is high route familiarity and with IVEO Aids. The
dependent variables were total walking time, number of miss routes, and number of
requests.
Fig. 2. Processing flow of the IVEO System Interface Design
The experiment took around 3 hours, following steps:

1. The subject was asked to fill out a SDQS/sense of direction questionnaire-short
form.
2. The subject performed a Way-finding test from Ximending to TDTB (Taiwan Dig-
ital Talking Books association). Data collected primarily by two experimenters.
One was a cameraman who recorded all the scenes with a video camera along the
way. Another logger registered behaviors of the subjects and record his/her specific
activities.
3. Following the test, subjects were interviewed and asked to fill out NASA-TLX
questionnaire and interviewed for approximately half an hour about their mental
workload.
3 Experimental Results
3.1 Number of Missed Routes

Analyses of variance were used to detect significant difference among variables.
Table 1 showed two-way ANOVA results of number of missed route. There was a
significant effect of IVEO Aids on number of missed route (p<0.05), but the effect of
route familiarity was not significant. It means that with the IVEO Aids in actual Way-
finding test, the Number of Missed routes of subjects would decrease significantly.
Establishing a Cognitive Map of Public Place for Blind and Visual Impaired 197
Table 1. Two-way ANOVA on number of miss routes
Source DF SS MS F P
IVEO Aids 1 45.563 45.563 8.64 0.012*
Route
1 22.563 22.563 4.28 0.061
Familiarity
Interaction 1 5.063 5.063 0.96 0.346
Error 12 63.250 5.2708
Total 15 136.438
S = 2.296 R-Sq = 53.64% R-Sq(adj) = 42.05%
*p<.05.
3.2 Number of Requests

Likewise, number of requests were calculated during the experiment. Table 2 showed
two-way ANOVA results on number of request. There was a significant effect of
IVEO Aids on number of requests ( p<0.05), and the effect of route familiarity was
also significant (p<0.01).The route familiarity and IVEO Aids for the number of re-
quests is of decisive importance in actual Way-finding test.
Table 2. Two-way ANOVA on number of requests
Source DF SS MS F P
IVEO Aids 1 9 9 8.64 0.012*
Route
1 12.25 12.25 11.76 0.005**
Familiarity
Interaction 1 4 4 3.84 0.074
Error 12 12.5 1.0417
Total 15 37.75
S = 1.021 R-Sq = 66.89% R-Sq(adj) = 58.61%
*p<.05 **p<.01.
3.3 NASA-TLX
Subjects were asked to fill out a NASA-TLX questionnaire for measuring the mental
workload. According to the score of each item, the differences of mental workload
in these subjects were evaluated. Table 3 showed two-way ANOVA results of NASA-
TLX. The result could help us to evaluate the mental workload for cumulative learn-
ing effect. There was a significant effect of IVEO Aids on NASA-TLX score(p<0.01)
that mental workload of subjects would decrease significantly by IVEO Aids, and the
effect of route familiarity was also significant (p<0.05).
Table 3. Two-way ANOVA on score of NASA-TLX
Source DF SS MS F P
IVEO 1 2620.08 2620.08 46.69 0.000**
Route Familiar-
1 369.73 369.73 6.59 0.025*
ity
Interaction 1 240.04 240.04 4.28 0.061
Error 12 673.47 56.12
Total 15 3903.32
SS = 7.491 R-Sq = 82.75% R-Sq(adj) = 78.43%
*p<.05 **p<.01.
4 Conclusion
We can find significant effect of using IVEO Hands-on learning system for the blind
and visually impaired. Thus subjects could perform better through IVEO Aids in
terms of IVEO Aids and route familiarity. This research provided a learning method
of orientation and mobility for the blind and visually impaired with orientation and
mobility (O&M) counselors, and assisted them to be familiar with the route safer and
faster.
Acknowledgments. Partial of this work was sponsored by the National Science

Council(NSC), Taiwan, R.O.C. through project NSC 101-2511-S492-001 and the
work was enhanced by the gracious assistance of Ludy Lee who is Secretary General
of Taiwan Digital Talking Books association (TDTB ).The authors like to express
their appreciation to all of the members involved in this project.
References
1. Gardner, J.A., Bulatov, V.L.: Scientific Diagrams Made Easy with IVEOTM. In: Miesen-
berger, K., Klaus, J., Zagler, W.L., Karshmer, A.I. (eds.) ICCHP 2006. LNCS, vol. 4061,
2. Klippel, A., Freksa, C., Winter, S.: You-Are-Here Maps in Emergencies — The Danger of
Getting Lost. Journal of Spatial Science 51(15), 117–131 (2006)
3. Kobayashi, K., Lee, R., Sumiya, K.: Lost again on the way?: Measuring Human Map-
Reading Ability. DBSJ Journal 8(4) (2010)
4. Parente, P., Bishop, G.: BATS: the blind audio tactile mapping system. In: Proc. of the
ACM Southeast Regional Conference (2003)
5. Taiwan’s Ministry of the Interior, statistics the population of blind and the visually impaired
persons (2012), http://sowf.moi.gov.tw/stat/year/list.htm (retrieved)
An Application of the Ballistic Movement Method
for Evaluating Computer Mice
Ray F. Lin, Ching-Wen Chung, Yi-Chien Tsai, and Chi-Yu Huang
Department of Industrial Engineering & Management, Yuan Ze University,

135 Yuan-Tung Road, Chungli, Taoyuan, Taiwan 32003
[email protected]
Abstract. Lin and Drury [1] validated two ballistic movement models and sug-
gested these models could have certain benefits for evaluating input devices.
Hence, the study aimed at (1) validating the application of ballistic movement
models for evaluating computer mice and (2) demonstrating the superiority of
this method, compared to Fitts’ law. In a two-stage experiment, four partici-
pants used six types of computer mice to execute Fitts-type aiming movements
and ballistic movements, sequentially. The measured data were analyzed by
Fitts’ law and the ballistic movement models, respectively. The comparison of
the results obtained by the two methods showed that (1) ballistic movement
models can well fit the measured data and (2) the ballistic movement method
can provide independent performance information of “speed” and “accuracy”
that is not available by applying Fitts’ law. This study demonstrated an alterna-
tive method for evaluating computer mice.
Keywords: ballistic movement method, computer mouse, input device, human

movements, Fitts’ law.
1 Introduction
A computer mouse is one of the most effective and efficient input devices while inte-
racting with computers. To enhance the performance of this human-machine system,
researchers and designers make efforts to develop a variety of computer mice. In the
development procedures, it is necessary to evaluate developing mice according to
several criteria, such as production cost, durability, and usability. Most importantly,
task performance is an essential criterion for assessing computer mice.
To evaluate the task performance of computer mice, Fitts’ law [2] has been ac-
cepted as a useful approach, which is easy to apply and widely used. However, Fitts’
law has limitations of lacking theoretical support and confounding performance of
movement speed and accuracy. Lin and Drury [1] suggested a new evaluation me-
thod, called the “ballistic movement method”, for evaluating input devices. They
stated that the method could provide independent performance of movement speed
and accuracy. However, the method has not been tested in evaluating computer mice.
Hence, it is the goal of this study to test the application of the ballistic movement
method, compared to Fitts’ law.
200 R.F. Lin et al.
1.1 Fitts’ Law

As shown in Equations 1 and 2, Fitts’ law [2] predicts that the movement time (MT)
required to execute a Fitts-type aiming movement is linearly related to the index of
difficulty (ID) of that movement, defined as the dyadic logarithm of the quotient of
amplitude of the movement and target width (Equation 2).
(1)
(2)
where A is movement amplitude, W is target width, a and b are experimentally deter-

mined constants.
To compare two computer mice, for example, the practitioner asks participants to
perform Fitts-type aiming movements manipulated at several values of index of diffi-
culty. Fitt’s law, hence, is able to describe the tradeoff relationships between the
movement accuracy and speed by using two computer mice. By regressing measured
MT data on to manipulated ID values for two computer mice and then drawing two
regression lines on a two dimensional graph, researchers can easily compare the
performance of the two computer mice. Because Fitts-type aiming movements are
easily tested and the measured data can be well predict by Fitts’ law, Fitts’ law thus
become one of the most popular evaluation methods in the domains of Human Factors
and Human Computer Interaction.
Although Fitts’ law is easy to apply, as mentioned in Lin and Drury [1], Fitts’ law
has two limitations. First, Fitts’ law only allows practitioners obtain the performance
information that is confounded with the two motor properties: speed and accuracy. A
Fitts-type aiming movement that takes a longer movement time could result from
lower motor speed, lower motor accuracy, or a combination of both. However, Fitts’
law has difficulty discriminating the extent to which the two motor properties contri-
bute to the overall movement time. Second, the application of Fitts’ law has theoreti-
cal issues while reporting the throughput values and generalizing the measured results
to other conditions [see 3 for detials].
1.2 Ballistic Movement Method

In recent year, the general model proposed by Lin, Drury, Karwan and Paquet [4] and
Lin and Drury [5] indicates that a Fitts-type aiming movement is composed of ballis-
tic movements, which are basic movement unit. The movement time and the endpoint
variability of a ballistic movement are two essential factors that directly affect the
speed and accuracy of a Fitts-type aiming movement. Lin and Drury [1] further tested
two ballistic movement models for describing how these two properties are associated
to ballistic movement distance.
Ballistic movement time represents the required time for performing a ballistic
movement. In an experiment in which participants performed hand control move-
ments on a drawing tablet, Lin and Drury [1] verified that Equation 6, proposed by
Gan & Hoffmann [6], can effectively describe and predict the relationship between
An Application of the Ballistic Movement Method for Evaluating Computer Mice 201
ballistic movement time (tballistic) and the squared root of ballistic movement distance
( ).
(3)
where i and j are experimentally determined constants.

Ballistic movement variability describes the endpoint variability of a ballistic
movement. Because of certain noise existing in human motor control mechanism, the
ultimate endpoint of a ballistic movement may not exactly end at the anticipative
aimed point [7, 8]. No matter endpoint errors are measured in the movement direction
or perpendicular to the movement direction, Lin and Drury [1] found that the proba-
bility of endpoint location formed a normal distribution around the aimed point. In
order to predict two directions of endpoint variability, Lin and Drury [1] verified the
application of Equation 7, originally developed by Howarth, Beggs and Bowden [9].
(4)
where e and f are experimentally determined constants. As shown in the equation, the
endpoint variability is linearly related to the square of movement distance.
The two ballistic movement models (i.e., Equations 3 and 4) have been tested in
several conditions by Lin and his colleagues [1, 10, 11]. Lin and Drury [1] originally
verified the two models by asking participations to perform ballistic movements using
a drawing tablet. Further, the models were tested in executing three-dimensional hand
movements [10] and evaluating age effects on touchscreen usage [11]. With the bal-
listic movement method, Lin and Ho [10] were able to evaluate three-dimensional
hand movements in a detail manner in which the movement speed and accuracy could
be evaluated independently. Six right-handed graduated students participated in their
study. In terms of movement speed, some participants performed differently while
using left hand, compared to right hand, but some kept the same no matter which hand
was used. In terms of movement accuracy, however, all the participants had higher
accuracy in any of the three dimensions while using their right hands. Lin, et al. [11]
attempted to use the ballistic movement method for evaluating ageing difference
while using a touchscreen. Although no significant difference were found on ageing
effects by analysis of variance, the ballistic movement models showed that the older
participants performed slower movements, but had greater movement accuracy, com-
pared to the young participants.
1.3 Research Objective

Although Fitts’ law and Fitts-type aiming movements have been widely used as an
evaluation method, the method has theoretical issues and only provides confounded
results of movement speed and accuracy. Nonetheless, ballistic movements are essen-
tial movement unit to construct an aiming movement. The performance of executing
ballistic movements is associated to the performance of aiming movements. The two
ballistic movement models introduced above separate the movement performance into
“speed” and “accuracy”, independently. Hence, this study aimed at verifying the
202 R.F. Lin et al.
ballistic movement models as a new evaluation method for computer mice. This new
method was expected with superiority and more effectiveness, compared to Fitts’ law.
2 Method
2.1 Participants and Equipment

Two female and two male undergraduate students, aged from 19 to 20 years, partici-
pated in this study. They were all right-handed with normal or corrected-to-normal
vision.
Fig. 1. Six computer mice tested in this study
Experimental apparatus included a personal computer (PC), six types of computer

mice (see Fig. 1), and two self-developed programs, comprising the Fitts-type aiming
movement program and the ballistic movement program. The PC ran Visual Basic 6.0
using the developed programs that both displayed the experimental tasks and meas-
ured task performance.
2.2 Experimental Procedures

After informed consent procedures, the four participants utilized six computer mice to
performed both Fitts-type aiming movements and ballistic movements. To reduce
training effect, the participants had sufficient time (about 2-3 hours) to practice all the
computer mice and the experimental programs. Because Fitts-type aiming movements
were relatively easier to perform, all the participants executed Fitts-type aiming
movements in the first stage of experiment and then executed ballistic movements in
the second stage of experiment. To reduce the fatigue effect, the participants finished
only a measurement that tested one computer mouse in a half-day.
To perform Fitts-type aiming movements, as shown in Fig. 2(a), the participants
drew a line from the starting point to end within the target line. The movements were
all performed from the left to the right. The participants were asked to complete every
trial as fast as possible. If the cursor was moved over the target line, the participants
needed to immediately change the moving direction and continue the movement. By
clicking the cursor on any location of the screen, the participants could continue on
the next trail.
To perform ballistic movements, as shown in Fig. 2(b), the participants drew a line
from a starting point to the center of the cross target with a certain distance ( ). The
movements were all performed from left to right as well. The tasks started by pressing
down the stylus cursor on the starting point and then moving toward the cross target.
Once the cursor was moved away from the starting point, the cursor and the cross
target disappeared and the movement time started to record. When the movement
stopped, the cross target and the endpoint of that movement were immediately dis-
played on the screen. Similar to Fitts-type aiming movements, participants could con-
tinue on the next trial by clicking the cursor on the screen.
A du
W /2
(a) (b)
Fig. 2. The execution of Fitts-type aiming movements (a) and ballistic movements (b)
In order to reduce kinesthetic feedback caused by repetitive movements from

the same starting point, there were four different starting points decided by a horizon-
tal displacement (200 pixels) and a vertical displacements (200 pixels) in both the
experiments.
2.3 Experimental Variables
In the experiment of Fitts-type aiming movement, the independent variables were

Computer Mouse and ID and the only dependent variable were Movement Time. The
four ID values were 2, 3, 4, and 5 bits determined by the modified Fitts’ law [12, 13].
The experimental program of Fitts-type aiming movement replicated each ID move-
ments 24 times, resulting in a total of 96 trials. All the trials were randomly conducted
by each participant, taking about 15 minutes to finish.
In the experiment of ballistic movement, the independent variables were Computer
Mouse and Movement Distance and the dependent variables were Movement Time
and the movement errors measured in the movement direction (X error) and perpendi-
cular to movement direction (Y error). The errors consisted of constant error and vari-
able error. To analyze whether the independent variables had significant effects on
these two types of errors, five replications of each experimental combination were
utilized to calculate the constant error (measured by mean) and the variable error
(measured by variance). However, only the results of variable error are discussed in
204 R.F. Lin et al.
this article. The 14 values of movement distance (du) were 8, 17, 32, 53, 80, 113, 152,
197, 248, 305, 368, 437, 512, and 593 pixels. The experimental program of ballistic
movement replicated each movement distance 20 times, resulting in a total of 280
trials. All the trials were randomly conducted by each participant, taking about 20
minutes to finish.
3 Results
3.1 Analysis of Variance
Analysis of variance was first performed to test the effects of independent variables
on dependent variables in the two experiments. In the experiment of Fitts-type aiming
movement, the main effects and the interaction effect of Computer Mouse and ID had
highly significant effects on Movement Time (all the p values < 0.001). In the expe-
riment of ballistic movement, the main effects and the interaction effect of Mouse and
Movement Distance also had highly significant effects on Movement Time and two
directions of variable errors (all the p values < 0.001). Since all the main effects in the
two experiments had significant effects on the dependent variables, the applications
and comparison of Fitts’ law and ballistic movement models were performed.
3.2 Results Obtained by Applying Fitts’ Law
The means of aiming movement time (MT) of the six computer mouse were regressed
on to the index of difficulty (ID) to give six regression lines. Fitts’ law accounted for
98.8 % variance on average and at least 97.2 % variance of the data of movement
time. The regression lines of the six computer mice are shown in Fig. 3, which shows
1200
3M-Joystick
d)n Evoluent-Vertical
oc 1000 Humanscale-Switch
es Logitech-Gaming
llii Logitech-Standard
m
( 800 Logitech-Trackball
e
m
iT
tn 600
e
m
ev
o 400
M
gn
i
im 200
A
0
2.0 2.5 3.0 3.5 4.0 4.5 5.0
Index of Difficulty (bit)
Fig. 3. Fittings of Fitts’ law on aiming movement time by using six computer mice
good model fittings. In general, the movement times required to execute aiming
movements are close while using the computer mice of Logetich-Standard, Evoluent-
Vertical, and Logitecfh-Gaming, which had the best performance compared to
the rest. However, the computer mouse of 3M-Joystick required the longest move-
ment time for executing aiming movements, especially when the value of index of
difficulty were greater than three.
3.3 Results Obtained by Applying the Ballistic Movement Method
The means of ballistic movement time ( ) of the six computer mouse were
regressed on to the square root of ballistic movement distance ( ) to give the
slopes and intercepts. The model fitted the data very well. It accounted for 96.2 %
variance on average and at least 85.6 % variance of data of ballistic movement time.
The regression lines of the six computer mice are shown in Fig. 4, which also shows
good model fittings. As shown in the figure, the movement times required to execute
ballistic movement were close by using the computer mice of Evoluent-Vertical and
Logitech-Trackball, which resulted in the fastest movements. The ballistic movements
were also fast by using the computer mice of Humanscale-Switch and Logitech-
Gaming. However, the computer mouse of 3M-Joystick resulted in the slowest
ballistic movements.
700
)d 3M-Joystick
no 600 Evoluent-Vertical
ecis Humanscale-Switch
Logitech-Gaming
lli 500 Logitech-Standard
m
( Logitech-Trackball
e
m
iT 400
tn
e 300
m
ev
o
M200
cti
isll 100
aB
0
0 5 10 15 20 25
Square Root of Movemenet Distance (pixel^0.5)
Fig. 4. Fittings of the ballistic movement time model by using the six computer mice
The two error variances, calculated from the raw data for each distance, were re-
gressed on to to give the slopes, intercepts, and values. For X-variable error
( ), the ballistic movement time model accounted for 93.0 % variance on average
and at least 85.5 % variance of data. For Y-variable error ( ), the ballistic move-
ment time model accounted for 90.4 % variance on average and at least 76.1 % va-
riance of data. The regression lines of X-variable error and Y-variable error are shown
in Fig. 5, where X-variable error is about ten times larger than Y-variable error. In
terms of X-variable error (measured in the movement direction), the six computer
mice listed in descending order by endpoint variance were Logitech-Trackball,
206 R.F. Lin et al.
Humanscale-Switch, Evoluent-Vertical, 3M-Joystick, Logitech-Gamming, and Logi-

tech-Standard. The computer mouse of Logitech-Standard resulted in the smallest
X-endpoint variance, and the computer mouse of Logitech-Trackball resulted in the
greatest X-endpoint variance. In terms of Y-variable error (measured perpendicular to
the movement direction), the six computer mice listed in descending order by end-
point variance were 3M-Joystick, Logitech-Trackball, Humanscale-Switch, Logitech-
Gamming, Evoluent-Vertical, and Logitech-Standard. Again, the computer mouse of
Logitech-Standard resulted in the smallest Y-endpoint variance. The computer mouse
of Logitech-Trackball also had great Y-endpoint variance, but the computer mouse of
3M-Joystck had the greatest Y-endpoint variance.
0 100000 200000 300000
X-Variable Error (pix el^2) Y-Variable Error (pix el^2)
3M-Joystick
20000 2000
Evoluent-Vertical
Humanscale-Switch
Logitech-Gaming
15000 Logitech-Standard 1500
Logitech-Trackball
10000 1000
5000 500
0 0
0 100000 200000 300000
Ballistic Movement Distance Squard (pixel^2)
Fig. 5. Fittings of the ballistic movement variability model on the error variance by using the
six computer mice
4 Discussion
This study first showed that the ballistic movement model, as Fitts’ law, have good
predictions of the measured data. Fitts’ law, as expected, well predicts the relationship
between the movement time and the index of difficulty while using each computer
mouse to execute Fitts-type aiming movements. The two ballistic movement models
also well predict the relationships between the measured data of ballistic movement
time, X-variable error, Y-variable error and the movement distance. Although the
fittings of the models were not as good as that of Fitts’ law, the models accounted for
96.2 %, 93.0 %, and 90.4 % of three types of data variance, respectively.
While the evaluation results of the two methods were compared, this study then
showed the superiority of the ballistic movement method. By applying Fitts’ law, we
can easily obtain the movement times required by each computer mouse to finish
aiming movements set at a certain ID value by comparing the linear regression lines
obtained by Fitts’ law. However, as mentioned previously, the movement time ob-
tained by Fitts’ law is confounded by two motor properties of speed and accuracy.
Two computer mice may require same aiming movement times to operate, but they
may possess different combinations of speed and accuracy properties. By applying the
ballistic movement method, we can further obtain independent performance of speed
and accuracy properties of each computer mice by comparing the linear regression
lines obtained by the ballistic movement models. This superiority is fully supported
by our results. For example, by applying Fitts’ law, we only know that the computer
mice of Logitech-Standard and Evoluent-vertical required close movement times to
operate Fitts-type aiming movements (see Fig. 3). However, the application of ballis-
tic movement method showed these two computer mice possess different combination
of speed and accuracy properties. The computer mouse of Logitech-Standard had
better performance of accuracy (see Fig. 5), but had poor performance of speed than
the computer mouse of Evoluent-vertical (see Fig. 4). Moreover, by applying
Fitts’ law, we know that the computer mouse of Logitech-Standard required shorter
movement times compared to the computer mice of Evoluent-Vertical and Logitech-
Gaming. However, Fitts’ law provides no information whether the good performance
of Logitech-Standard is due to its speed property or accuracy property. By using the
ballistic movement method, as shown in Fig. 4 and Fig. 5, we realize that the reason
why the mouse of Logitech-Standard required the shortest movement time was not
due to its speed performance, but due to its highest accuracy performance. Finally, by
applying the ballistic movement method, we know the reason why the computer
mouse of Logitech-Trackball required the longest movement time was not due to its
speed property, but due to its accuracy property that was the worst among the six
computer mice.
Future research should validate the ballistic movement method in a comprehensive
way by recruiting adequate participants and controlling driver difference of computer
mice. Furthermore, the independent performance of speed and accuracy could help
designers focus on specific properties of computer mouse for improvement.
5 Conclusions
This study attempted to validate the two ballistic movement models proposed by Lin
and Drury [1] as an alternative method for evaluating the task performance of com-
puter mice. To this end, six types of computer mice were tested by simultaneously
applying both Fitts’ law and the ballistic movement method. By comparing the results
obtained by the two methods, we showed the superiority of the ballistic movement
method, compared to Fitts’ law. Fitts’ law only helps determine the time required
performing Fitts-type aiming movements for each computer mouse, whereas the bal-
listic movement method provides independent performance information about speed
and accuracy, which could help effectively improve the design of computer mouse in
the future research.
Acknowledgment. We would like to acknowledge the grant support from Taiwan

National Science Council (NSC101-2221-E-155-006) for funding the paper submis-
sion and presentation.
208 R.F. Lin et al.
References
1. Lin, J.-F., Drury, C.G.: Verification of two models of ballistic movements. In: Jacko, J.A.
(ed.) Human-Computer Interaction, Part II, HCII 2011. LNCS, vol. 6762, pp. 275–284.
2. Fitts, P.M.: The information capacity of the human motor system in controlling the ampli-
tude of movement. Journal of Experimental Psychology 47, 381–391 (1954)
3. Zhai, S.: Characterizing computer input with Fitts’ law parameters - the information and
non-information aspects of pointing. International Journal of Human-Computer Studies 61,
791–809 (2004)
4. Lin, J.-F., Drury, C., Karwan, M., Paquet, V.: A general model that accounts for Fitts’ law
and Drury’s model. In: Proceedings of the 17th Congress of the International Ergonomics
Association, Beijing, China, August 9-14 (2009)
5. Lin, J.-F., Drury, C.G.: Modeling Fitts’ law, in Ergonomics for all: celebrating PPCOE’s
20 years of Excellence. In: Lin, D.-Y.M., Chen, H.-C. (eds.) Selected Papers of the Pan-
Pacific Conference on Ergonomics, pp. 561–567. CRC Press, Taiwan (2010)
6. Gan, K.-C., Hoffmann, E.R.: Geometrical conditions for ballistic and visually controlled
movements. Ergonomics 31, 829–839 (1988)
7. Novak, K.E., Miller, L.E., Houk, J.C.: Kinematic properties of rapid hand movements in a
knob turning task. Experimental Brain Research 132, 419–433 (2000)
8. Milner, T.E.: A model for the generation of movements requiring endpoint precision. Neu-
roscience 49(2), 487–496 (1992)
9. Howarth, C.I., Beggs, W.D.A., Bowden, J.M.: The relationship between speed and accura-
cy of movement aimed at a target. Acta Psychologica 35, 207–218 (1971)
10. Lin, R.F., Ho, Y.-C.: Verification of ballistic movement models in a true 3D environment.
In: Proceedings of the 2nd East Asian Ergonomics Federation Symposium, October 5-8.
National Tsing Hua University, Hsinchu (2011)
11. Lin, R.F., Shih, S.-W., Jiang, B.C.: An application of ballistic movement models for com-
paring ageing differences while interacting with a touchscreen. In: Proceedings of the 4th
International Conference on Applied Human Factors and Ergonomics, San Francisco, Cali-
fornia, USA, July 21-25 (2012)
12. MacKenzie, I.S.: A note on the information-theoretic basis for Fitts’ law. Journal of Motor
Behavior 21, 323–330 (1989)
13. MacKenzie, I.S.: Fitts’ law as a research and design tool in human-computer interaction.
Human-Computer Interaction 7(1), 91–139 (1992)
AugmentedBacklight: Expansion of LCD Backlights
Using Lighting Methods in the Real World
Maho Oki1, Koji Tsukada1,2, and Itiro Siio3

1
Future University Hakodate, Hokkaido, Japan
{okimaho,tsuka,siio}@acm.org
2
JST PRESTO, Saitama, Japan
3
Ochanomizu University, Tokyo, Japan
Abstract. People experience various lights in daily life, such as sunlight

through leaves and candle light. However, digital contents shown in common
LCDs cannot express these impressive lights. In this paper, we propose a novel
display technique, AugmentedBacklight, which can express impressive lights in
the real world by expanding the LCD backlight. The system mainly consists of
a transparent LCD, a screen behind the LCD, and a projector behind the above
devices. The projector shows videos of impressive lights on the screen for the
backlight of the transparent LCD. Thus, users can see digital contents (e.g.,
e-books and pictures) under various impressive lights in the real world.
Keywords: Display, backlight, lighting.
1 Introduction
People experience various lights in daily life, such as sunlight through leaves, candle
light, and sun glare on water. These lights affect appearances of objects in the real
world. For example, people often feel different impressions by reading books under
sunlight through leaves compared to under fluorescent lamps. However, digital
contents shown in common LCDs cannot express these impressive lights since the
LCD backlight emits flat lights –same brightness in each pixel – to keep visibility. To
overcome this limitation, we propose a novel display technique, AugmentedBack-
light, which can express impressive lights in the real world by expanding the LCD
backlight.
2 Related Works
Seetzen et al. [5] and Chen et al. [1] proposed novel displays that have similar me-
chanisms with our system: that is, integrating a transparent LCD and a DLP projec-
tor/LED-based display to expand the LCD backlight. Suyama et al. [6] proposed a
Depth-fused 3D (DFD) display that enables a user to perceive an apparent 3D image
using overlaid two displays. Lee et al. [4] proposed a unique DFD display that
integrates fog screen and two projectors.
210 M. Oki, K. Tsukada, and I. Siio
While these systems focused on expanding dynamic ranges of pictures/videos

shown on the LCD or generating 3-D image, our system focuses on expressing ap-
pearances of contents under various impressive lights in the real world.
There were several art works [2, 3] that add visual effects to real time videos de-
pending on activities of people in the real world. Our system may effectively work in
such applications.
Fig. 1. Basic concept of AugmentedBacklight
3 AugmentedBacklight
The basic concept of AugmentedBacklight is to express various lights in the real

world by expanding the LCD backlight (Fig. 1). Many lights in the real world (e.g.,
sunlight and spot lights) have so high luminance that most people have difficulty to
look at them directly. These lights also have fluctuation depending on surrounding
environment (e.g., sunlight reflected on the water). In this research, we aim to express
these lights on the LCD using lighting video that includes impressive lights in the real
world. For example, our system may recreate an experience of reading a book under
sunlight through leaves by integrating texts and the video of such lighting.
The system mainly consists of a transparent LCD, a screen behind the LCD, and a
projector behind the above devices. As shown in Fig. 1, the projector shows videos of
impressive lights on the screen for the backlight of the transparent LCD. Thus, users
can see digital contents (e.g., e-books and pictures) under various impressive lights in
the real world.
3.1 Prototype
We developed a prototype as shown in Fig. 2. It consists of a projector (EPSON EB-
410W, 2000lm), a transparent LCD (TCR-3104M) and a screen (handmade acrylic
panel). The LCD and the projector are connected to a PC (Windows XP) via an USB-
VGA convertor (Buffalo, GXDVI/ U2AI) and a built-in VGA connector.
AugmentedBacklight: Expansion of LCD Backlights Using Lighting Methods 211
The main program running on the PC controls (1) videos of impressive lights shown
on the screen through the projector and (2) main contents (e.g., e- books, pictures or
web pages) on the LCD. In this paper, we defined the prior contents as “light con-
tents” and the later contents as “main contents”. Since impressive lights in the real
world (e.g., sunlight through leaves) often change their brightness and textures dy-
namically, we selected videos of such changing lights as the light contents.
Fig. 2. The AugmentedBacklight prototype
3.2 Design of Light Contents
We selected light contents in consideration that users can perceive impressive lights
on the display without bothered by uncomfortable feelings. In this paper, we focused
on mainly two categories: (A) impressive lights in the real world, and (B) impressive
visual effects created by computer graphics. The former category includes videos of
sunlight through leaves thrown on the wall and sunlight reflected on water. The
latter category includes visual effects of lightning and brilliant lights. These examples
of light contents are shown in Fig. 3. These contents are converted into gray scale to
avoid affecting colors of main contents.
3.3 Design of Main Contents

We selected still images as main contents to verify the basic performance of light
contents. For example, we prepared pictures of “green leaves”, “a flower”, “lake” and
“texts in an e-book” in corresponding to the impressive lights in the real world. We
also prepared pictures of “people” and “a glass besides the window” suited for the
visual effects. These examples of main contents are shown in Fig. 4.
4 Preliminary Study
We performed a preliminary study to verify the basic performance of the Augmen-

tedBacklight prototype. We have selected 5 subjects (3 females, aged between 23 and
33) from our laboratory.
4.1 Procedure
We explain basic procedure of the preliminary evaluation. First, we set each subject
down in front of the AugmentedBacklight prototype and explained the basic function
of the system. Next, the experimenter shows a main content along with the white
background instead of a light content. Then, the light contents are changed several
times. After the subject repeated these processes for all main contents, the experimen-
ter ask them to write the impression of the system.
Fig. 3. Examples of light contents
Fig. 4. Examples of main contents
4.2 Results
In this section we explain the results from the viewpoints of “effective examples” and
“findings”.
Effective Examples
First, we describe effective combinations of light contents and main contents. The
picture of each combination is taken with a digital camera (GR Digital 2) at a close
range.
Fig. 5 shows the combination of “green leaves” and “sunlight through leaves”. The
typical impressions written by the subjects were as follows:
• The green leaves look vivid by the movement of sunlight through leaves.
• Beautiful. The backlight looks actual sunlight through leaves.
Fig. 5. The combination of “green leaves” and “sunlight through leaves”
Fig. 6 shows the combination of “texts in the e-book” and “sunlight through leaves”.
The typical impressions written by the subjects were as follows:
• The gentle movement of the sunlight brought me comfort.

• I can read texts more easily than I imagined.
Fig. 7 shows the combination of “surprised person” and “lightning effect”. The im-
pressions of the subjects were as follows:
• Strange feeling. The person and other objects look floating on the air.
• Cool. The person seems to emit the aura.
• I felt retro impression like old films or stained glasses.
Fig. 6. The combination of “texts in e-book” and “sunlight through leaves”
Fig. 7. The combination of “surprised person” and “lightning effect”
Fig. 8 shows the combination of “a glass besides a window” and “brilliant lights”.
The impressions of the subjects were as follows:
• The picture looks like a 3D image. It seems to pop out from the display.
• I was interested in the effect since I have never seen such lighting in the real world.
Fig. 8. The combination of “a glass besides a window” and “brilliant lights”
Findings
All subjects reported that the pictures shown with the light contents look different
from ones with the white background. Some subjects felt the depth between light
contents and main contents. Others reported each main contents looks drastically
different depending on light contents.
Since most subjects felt the stereo effect in the picture of “a glass besides a win-
dow” in which the main object is simply located on the center, the light contents may
effectively works with pictures that have simple compositions.
As the combination of light/main contents, most subjects had comfort with the
combination of contents in similar situations. For example, “green leaves” and “sun-
light through leaves” as shown in Fig. 5 are one of the most popular examples in the
preliminary evaluation. On the other hand, some light contents (e.g., “sunlight
through leaves”) work rather effectively with variety of main contents. For example, a
subject mentioned “I would like to use the light content (sunlight through leaves) as
the background of my desktop”.
Meanwhile, visual effects (e.g., lightening and brilliant lights) were not suited for
simple main contents (e.g., texts in e-book) since these effects have too strong pres-
ence that most subjects have difficulty to read texts with them. However, when the
main contents are carefully selected, the visual effects may give great impacts on
users as shown in Fig. 7.
In this study, we confirmed some effective examples of contents for the Augmen-
tedBacklight. We would explore people’s impressions on various combinations of
light/main contents more accurately. Moreover, we also try to specify the cognitive
factor that affects unique experiences with our system.
5 Conclusion
In this paper, we proposed a novel display technique, “AugmentedBacklight”, which

can express impressive lights in the real world by expanding the LCD backlight. We
have developed a prototype system using a transparent LCD, a screen behind the
LCD, and a projector behind the above devices. We introduced the examples of light
contents shown on the screen through the projector and main contents shown on the
LCD. We also explained the effective examples through the preliminary evaluation.
Acknowledgements. This work was supported by JST PRESTO program.
References
1. Chen, H., Sung, J., Ha, T., Park, Y., Hong, C.: Backlight local dimming algorithm for high
contrast LCD-TV. In: Proc. of ASID 2006, pp. 168–171 (2006)
2. Fukuchi, K., Mertens, S., Tannenbaum, E.: EffecTV: a real-time software video effect pro-
cessor for entertainment. In: Entertainment Computing–ICEC 2004, pp. 341–406 (2004)
3. Iwai, T.: Piano–as image media. Leonardo 34(3), 183–183 (2001)
4. Lee, C., DiVerdi, S., Höllerer, T.: An immaterial depth-fused 3D display. In: Proceedings of
the 2007 ACM Symposium on Virtual Reality Software and Technology, pp. 191–198
(2007)
5. Seetzen, H., Heidrich, W., Stuerzlinger, W., Ward, G., et al.: High dynamic range display
systems. ACM Transactions on Graphics (SIGGRAPH 2004) 23(3), 760–768 (2004)
6. Suyama, S., Ohtsuka, S., Takada, H., Uehira, K., Sakai, S.: Apparent 3-D image perceived
from luminance-modulated two 2-D images displayed at different depths. Vision
Research 44(8), 785–793 (2004)
An Exploratory Study to Understand Knowledge-Sharing
in Data-Intensive Science
Jongsoon Park and Joseph L. Gabbard
Virginia Bioinformatics Institute, Industrial & Systems Engineering

Virginia Tech
{babirong,jgabbard}@vt.edu
Abstract. This paper describes a challenges associated with data-intensive

research processes, knowledge-sharing phenomena, and end-users’ expectations
in the field of bioinformatics. We developed a questionnaire to support deeper
understanding of user experiences with knowledge sharing activities. The
results reveal that there are several challenging issues biologists encounter
when using bioinformatics resources. A much smaller number of biologists
have engaged in passive knowledge sharing within their research fields than we
had expected. However, most biologists expressed their willingness to share
their own knowledge with others. This result reinforces the need for more user-
centered design approaches for supporting knowledge-sharing in rapidly
emerging fields of data-intensive science. At the same time, our results suggest
that more work is needed to examine how to best motivate users to further
engage and contribute knowledge in online scientific communities.
Keywords: User-centered Design Approach, Knowledge-Sharing, Data-

Intensive Science.
1 Introduction
In the past two decades, we have seen an exponential increase in the size and breadth
of available scientific data, demanding new integrated solutions to explore and elicit
valuable insights more efficiently (Kelling et al., 2009). As such, there is growing
interest in data-intensive scientific discovery through data integration, simulation,
visualization, and validation across distributed networks of heterogeneous resources.
A notable example is in biological sciences, which is currently undergoing a rapid
paradigm shift to data intensive science (Bell, Hey, & Szalay, 2009). Diverse
bioinformatics resources (e.g. online resources that integrate biological data and
analysis tools) have been developed, potentially enabling biologists to analyze huge
and complex data sets faster and more efficiently as compared to traditional methods
(Bull, Ward, & Goodfellow, 2000; Katoh, 2002; Yarfitz, 2000). However, most of
these resources have been developed by applied scientists (i.e. computer scientists and
bioinformaticians), and are not intuitive or rich enough to address most bench
biologists’ goals. At the same time, bench biologists are beginning to appreciate
the power and potential of bioinformatics resources, despite their poor usability.
218 J. Park and J.L. Gabbard
Generally speaking, biologists and other researchers in data-intensive fields are

grappling with how best to deal with “big data” while HCI researchers (collaborating
with bioinformaticians) are working to understand how best represent and interact
with such data. The question remains: how can we best close this gap to ensure that
these resources are both powerful and intuitive to varying user classes. There have
been various attempts to bridge gaps between technology-driven bioinformatics
resources and the broader spectrum of biologists’ work practices, such as creating
more goal-oriented solutions for data collection and storage, and conducting task
analysis and usability studies (Joan Bartlett, Ishimura, & Kloda, 2011; J. Bartlett &
Neugebauer, 2005; D. Bolchini, 2009; Davide Bolchini, 2009; Javahery, 2004; Mirel,
2009; Tran, Dubay, Gorman, & Hersh, 2004). In parallel, we see web-based
“knowledge-sharing platforms” as a growing trend to support data-intensive discovery
research by allowing scientists to exchange data, ideas, expertise, and scientific
literature online to improve the effectiveness of their processes and validity of their
outcomes (De Roure, Goble, & Stevens, 2009; Li, 2012; McIntosh et al., 2012;
Parnell, 2011). These knowledge-sharing platforms in scientific communities may
provide a timely mean to assist biologists working with large and consistently
growing diverse data sets. However, to date, little attention has been paid to a
comprehensive understanding of end-users’ characteristics and attitudes about
knowledge sharing based on the culture of these data-intensive research domains.
To support useful and seamless knowledge-sharing and reuse in data-intensive
research, we need to examine a series of higher level questions from the user’s
perspective, such as: What are the major shortcomings in current online
bioinformatics resources? To what extent do end-users have experience with
knowledge-sharing activities (e.g., knowledge-sharing and -reuse)? What are end-
users’ expectations of knowledge-sharing activities? What are users’ perceptions of
challenges and opportunities in knowledge-sharing environments? Examining these
questions will provide us with meaningful insights not only to identify unmet needs
and opportunities, but also to support cross-disciplinary scientific research in data-
intensive fields such as biology.
2 Objective and Research Question
This study has three objectives: 1) to understand end-users’ perspectives on

shortcomings of current online bioinformatics resources, 2) to identify to what extent
end-users have experience with knowledge-sharing activities to support their research
processes, and, 3) to elicit specific end-users’ concerns and expectations of
knowledge-sharing.
3 Methods
We developed a questionnaire to elicit users’ experience with, perceptions of, and

attitudes towards knowledge-sharing activities. The questionnaire is based on
previous studies (Bock, Zmud, Kim, & Lee, 2005; Preece, Nonnecke, & Andrews,
An Exploratory Study to Understand Knowledge-Sharing in Data-Intensive Science 219
2004) and feedback from domain experts who are familiar with biological “wet-lab”
experiments and have worked in fields of biology for five years or more.
The first set of questions elicits participant demographics (e.g., age, gender, current
work/academic role) and background (e.g., usage frequency of bioinformatics
resources). The second set of questions elicits information on various aspects of users’
experiences, challenges, and expectations with current bioinformatics resources and
knowledge-sharing related to their research processes.
After data were collected, we employed statistical analysis to describe
characteristics and behavior of users’ current knowledge-sharing activities. The
findings of this survey are intended to help identify and prioritize distinguishing
web resource features needed to support online knowledge-sharing in data-rich
scientific processes.
3.1 Participants
We collected responses from participants of workshops offered by the Virginia

Bioinformatics Institute and from PhD students in the Virginia Tech Genomics,
Bioinformatics, Computational Biology graduate program. However, the calculation
of a response rate was difficult because we do not know how many total PhD students
in the program were invited to participate. Eighty-one of eighty-four total responses
were usable (three were incomplete or incomprehensible).
More than half of the participants (55.6%) are male, and 72.9% are between 20 and
39 years of age. Almost all characterized their main research role as biologist (63),
with others self-reporting roles of bioinformatician (6), chemist (4), computer
scientist (4), mathematician (3), and other (11) such as microbiologists, biochemists,
and clinician (note: participants were asked to select all items that apply). Slightly
over 60% of participants (n = 49) have over five years of research experience in
biology, while about 40% (n = 32) have been conducting research for no more than
five years.
The frequency of use of bioinformatics resources ranges from every day to less
than once a month. A majority of participants (70.3%) reported using bioinformatics
resources more than once a week, 13.6% reported use as more than once a month, and
13.6% reported using bioinformatics resources around once a month.
3.2 Results
We performed data analysis using SPSS (version 18.0), defining statistical

significance at ρ < 0.05. In general, we observed similar response patterns among
participants with no more than 5 years of research experience. Those with over 5
years research experience also showed similar response tendencies. As such, in the
following discussion, we consider two broad classes of participants; those with no
more than 5 years of research experience and those with over 5 years of research
experience.
3.3 Limitations in Online Bioinformatics Resources

Our results show that most participants are currently challenged by the lack of
integration and inconsistent results across online bioinformatics resources (e.g.
different gene naming conventions, different annotations for the same gene). In the
same vein, they repeatedly highlighted limitations due to the poor quality of genomic
sequences and metadata. Some participants noted strengths of bioinformatics
resources such as multiple views on the same data and multiple comparisons across
different genomes. Inconsistency in user interfaces and general lack of usability were
cited as major difficulties for a number of participants, implying a steep learning
curve (i.e., long learning times) as a key usability issue. In addition, some participants
had trouble accessing data due to complex navigation structures typical of
bioinformatics resources. Lastly, data security was noted as an important issue, since
many researchers are leveraging these resources to support hypotheses generation,
publications or grants.
3.4 Important Factors of Bioinformatics Resources
To examine users’ expectations of bioinformatics resources, we asked participants

which resource features are the most important or valuable. Multiple responses
were categorized and tallied using the multiple-dichotomy frequency analysis. We
constructed a cross tabulation table to analyze the most dominant participants’ response.
Table 1 presents the cross-tabulation frequencies by years of research experience.
Table 1. Important factors of bioinformatics resources

Research Experience (yrs)
No more Total
than 5 over 5
Speed and responsiveness of resource 23 (71.9%) 36 (76.6%) 59 (74.7%)
Wealth of available data 22 (68.8%) 30 (63.8%) 52 (65.8%)
Breath of resource tools and functions 16 (50.0%) 29 (61.7%) 45 (57.0%)
Degree of data integration 17 (53.1%) 26 (55.3%) 43 (54.5%)
Ease of use 16 (50.0%) 26 (55.3%) 42 (53.2%)
Ability to upload my own data 13 (40.6%) 25 (53.2%) 38 (48.1%)
Ability to ask questions related to my research 12 (37.5%) 23 (48.9%) 35 (44.3%)
Ability to create publication quality images 10 (31.3%) 22 (46.8%) 32 (40.5%)
Advanced visualizations 13 (40.6%) 19 (40.4%) 32 (40.5%)
Ability to collect knowledge from others
9 (28.1%) 21 (44.7%) 30 (38.0%)
researchers
Ability to share knowledge with other
6 (18.8%) 9 (19.1%) 15 (18.9%)
researchers
Total # of participants 32 47 79
As expected, performance-related factors common to most web-based systems

ranked relatively high in “important factors of bioinformatics web resources”.
Namely, participants valued “high speed and responsiveness of resource” (74.3%),
followed by “wealth of available data” (65.8%), “breadth of resource tools and
functions” (57.0%), “degree of data integration” (54.5%), “ease of use” (53.2%), and
“ability to upload my own data” (48.1%).
Interestingly, nearly half of the participants with over 5 years of research
experience selected “ability to ask questions related to my research” (48.9%) and
“ability to collect knowledge from others” (46.8%) as an important resource features.
In contrast, a much smaller proportion of these participants (19.1%) appear interested
in sharing their knowledge with others. It can be inferred from these results that
experienced participants are more interested in enhancing the overall quality (and
performance) of their research by making use of others’ shared knowledge than
sharing their accumulated knowledge and skills. In comparison with the above
findings, participants with no more than 5 years of research experience showed little
interest in sharing and collecting knowledge as compared to other features.
3.5 Knowledge Sharing Experience
Our knowledge-sharing results suggest significant, but limited, online knowledge-

sharing activity among our sampled user population (See Fig. 1).
Fig. 1. Knowledge sharing experience
Specially, all participants reportedly engage in knowledge-sharing activities to some

degree (e.g., knowledge-sharing or -reuse), but nearly 80 percent of respondents
reported participating in knowledge-sharing activities in a passive manner. That is,
most users rarely share their knowledge, but instead tend to seek and use knowledge
shared by others. This is a typical example of a lurker; an individual who consumes
information and requests specific questions from others, but does not explicitly
contribute to the shared knowledge base (Preece et al., 2004).
To examine potential correlations among age, years of research experience, and
experience level in knowledge-sharing activities, we performed a Kendall’s tau
correlation analysis. As might be expected, we found a positive, statistically
significant correlation between age and research experience (Kendall's tau b = 0. 608,
ρ = 0.000). Contrary to our expectation, however, we found no significant correlation
between years of research experience and the experience level in knowledge-sharing

activities (Kendall's tau b = -0.154, ρ = 0.123). Moreover, there was no correlation
between age and knowledge-sharing experience (Kendall's tau b = -0.019; ρ = 0.847).
Even though one might expect younger researchers to be more attuned to online
knowledge-sharing opportunities.
We next examined the types of knowledge consumed in knowledge-sharing
activities; including specific examples of both implicit and explicit knowledge (Bock
et al., 2005; Choo, 2000). These results provide valuable insight into what types of
knowledge participants seek from others in support of their research (See Table 2).
Table 2. Types of shared knowledge employed
Research Experience (yrs)

Total
no more than 5 over 5
None 0 (0%) 6 (18.8%) 6 (11.1%)
Knowledge
General ideas 14 (63.6%) 15 (46.9%) 29 (53.7%)

Implicit
Accumulative research
experiences 9 (40.9%) 15 (46.9%) 24 (44.4%)
Unique opinions 6 (27.3%) 7 (21.9%) 13 (24.1%)
Articles published in books, websites,
13 (59.1%) 16 (50.0%) 29 (53.7%)
Knowledge
and documents
Explicit
Products, patents, databases, tools,

and prototypes 8 (36.4%) 7 (21.9%) 15 (27.8%)
Rules, routines, or operating
procedures 10 (45.5) 5 (15.6%) 15 (27.8%)
Total # of participants 22 32 54
The most frequently reported knowledge used is “general ideas” (53.7%), “articles
published in books, websites, and documents” (53.7%), followed by “accumulative
research experiences” (44.4%). The overall pattern of responses indicates that our
participants employ implicit and explicit knowledge in an evenly balanced way.
We also found that participants with no more than 5 years research experience
tended to rely on "general ideas" (63.6%) more than participants with over 5 years
research experience (46.9%), with open-ended responses suggesting that the relatively
inexperienced participants use shared knowledge extensively to generate research
questions and confirm hypotheses. In addition, these relatively inexperienced
researchers depend on others to confirm rules or procedures (45.5%) as compared
with more experienced participants (15.6%).
3.6 Knowledge Sharing Intention

Next, we assessed participants’ intention to share knowledge using questions adapted
from Bock et al. (2005). Participants indicated their agreement or disagreement with
statements using a seven-point Likert-type scale (where scores of 7 suggest a strong
willingness to share). Surprisingly, almost all participants (95%) reported a
willingness to share knowledge with others (Fig. 2). This result suggests there are
ample opportunities to promote and grow knowledge-sharing in data-intensive

sciences such as biology.
Fig. 2. We used a one-way MANOVA to determine whether there are any
differences between our two research experience groups on more than one kind of
knowledge. We found no significant group effects for the types of knowledge on their
intention to share knowledge, F (14, 117) = 0.738, ρ =0.743; Wilk's λ = 0.728, partial
ε2 = 0.54.
Fig. 2. Intention to share specific knowledge (7=strongly agree, 1=strongly disagree)
Next, we used an independent t-test to investigate differences in knowledge-

sharing intentions between the two research experience groups, and found no
significant differences between them across any of the five variables: articles
published in books, magazines, or website, t(75) = 1.539, ρ = 0.128; manuals, or
methodologies, t(73) = 1.316, ρ = 0.192; own opinion from the accumulative
experience, t(75) = -0.023, ρ = 0.982; know-how at the requests of others, t(75) =
0.124, ρ = 0.901; and expertise from education or training, t(75) = 0.314, ρ = 0.755.
3.7 Concerns about Knowledge Sharing

One of the significant results to emerge from our responses is that most participants
worry about source reliability (i.e. poster’s expertise, data integrity, experimental
verification, quality of information). Similarly, results indicate concerns about
encountering “incorrect information or annotations”. Other concerns include the need
for an easy way to contribute their own knowledge, suggesting that participants have
encountered usability issues when they have previously tried to share information.
Additionally, participants noted response time, frequency of knowledge update,
copyright, and privacy as potential barriers to engaging in knowledge-sharing
activities.
3.8 Expectations
We found the following three themes of expectations concerning knowledge-sharing
from the free text comments. First, there were many responses that referred to source
creditability. For example, participants expect to be able to authenticate contributors’
expertise as well as shared knowledge. Second, about 50% of participants said they
often need technical support to make use of bioinformatics tools (e.g. troubleshooting
advice, application tips, how-to guides). Participants noted the lack of well-organized
information repositories of shared knowledge by domain, technique, methodology,
etc. Third, there was some evidence that participants are interested in reusing
knowledge from others to improve their research processes and outcomes. Frequently
mentioned expectations include implicit knowledge related to protocols used for
conducting biological experiments such as, small “bench work tricks”, protocols of
best practices for cutting edge -omics research, and information about negative data.
4 Discussion
The results of this research indicate not only existing barriers to use of bioinformatics
resources but also opportunities to address users’ unmet needs in data-intensive
scientific communities. We found no significant differences across age groups and
years of research experience; suggesting that designs to support knowledge sharing
should consider other user class characteristics.
Our results indicate that biologists struggle to utilize bioinformatics resources
regardless of years of research experience mainly due to inconsistent results and poor
user interfaces. Moreover, users indicate that the greatest shortcomings of current
resources are often associated with the most important features. A possible
explanation may be that many online bioinformatics resources employ a system-
oriented development approach rather than user-centered design approach that aims to
better understand users’ unmet needs.
We found approximately 80% of participants do not actively engage in knowledge-
sharing than we expected, regardless of age or years of research experience. Only a
small portion of participants have actively engaged in research-related knowledge-
sharing. A possible explanation might be that current links exist between knowledge-
sharing platforms and biologists are limited or ad hoc despite the prevalence of online
knowledge-sharing resources in scientific communities. In other words, we can
suppose that many knowledge-sharing platforms are built with a focus on current
technological trends rather than user experience factors, which affect users’
motivation to engage in knowledge-sharing activities. Another possible explanation is
that characteristics of the scientific culture may influence an individual’s propensity
to engage in knowledge sharing and reuse. Thus, more study is needed to understand
how best to foster knowledge sharing and reuse in scientific communities.
Results from this study also show that participants with no more than 5 years
research experience rely more on implicit knowledge shared by other practitioners
than explicit knowledge. It seems possible that self-efficacy caused by accumulated
expertise may contribute to the differences found in usage patterns of implicit

knowledge.
The most significant finding is that nearly 95% of participants are willing to share
their knowledge, contradicting the very low levels of current involvement in
knowledge-sharing activities. The lack of quality user experiences to support seamless
and easy knowledge contribution suggests a need to develop knowledge-sharing
platforms that embrace user-centered design approaches.
Lastly, this work identifies several issues that must be ensured to facilitate
knowledge-sharing and reuse in data intensive settings aiming to support scientific
discovery. A majority of participants were concerned about the quality of knowledge
and the degree to which they can trust shared knowledge. These results are consistent
with those of previous studies (Golbeck, 2008; Levin, Cross, Abrams, & Lesser,
2002) and suggests that source credibility has a considerable impact on attitudes
towards knowledge-sharing. More study is therefore needed to better understand how
to cultivate trust in, and increase motivation to use, knowledge-sharing activities.
5 Conclusion
This research is one of the first studies to investigate knowledge-sharing in emerging

data-intensive sciences such as biology. Our results imply significant opportunities to
support knowledge-sharing in these communities, but that careful attention needs to
be taken to users’ perceived and actual needs. This initial study focuses on eliciting
basic user experiences with, and perceptions of, bioinformatics resources and online
knowledge-sharing activities. Results presented herein may inform future studies to
explore user experiences and knowledge-sharing activities in data-rich environments.
This study was mainly conducted among biologists. The outcomes of the study
could be strongly influenced by the culture of experimental science. Hence, to
determine whether these findings can be applied to a wide range of knowledge-
sharing platforms for rapidly emerging fields of data-intensive science, there needs to
be further study with additional participants from different background (e.g. applied
scientists) across other data-intensive fields (e.g., visual analytics, meteorology).
Acknowledgments. This work was supported in part with Federal funds from
the National Institute of Allergy and Infectious Diseases, National Institutes of
Health, Department of Health and Human Services, under Contract No.
HHSN272200900040C.
References
1. Bartlett, J., Ishimura, Y., Kloda, L.: Why choose this one?: Factors in scientists’ selection
of bioinformatics tools. Information Research 16(1), 15 (2011)
2. Bartlett, J., Neugebauer, T.: Supporting information tasks with user-centred system design:
The development of an interface supporting bioinformatics analysis. Canadian Journal of
Information and Library Science 29(4), 486–487 (2005)
3. Bell, G., Hey, T., Szalay, A.: Beyond the data deluge. Science 323(5919), 1297–1298
(2009)
4. Bock, G.W., Zmud, R.W., Kim, Y.G., Lee, J.N.: Behavioral intention formation in
knowledge sharing: Examining the roles of extrinsic motivators, social-psychological
forces, and organizational climate. MIS Quarterly, 87–111 (2005)
5. Bolchini, D.: Better bioinformatics through usability analysis. Bioinformatics (Oxford,
England) 25(3), 406–412 (2009), doi:10.1093/bioinformatics/btn633
6. Bolchini, D., Finkestein, A., Paolini, P.: Designing Usable Bio-information Architectures.
In: Jacko, J.A. (ed.) HCII 2009, Part IV. LNCS, vol. 5613, pp. 653–662. Springer,
Heidelberg (2009)
7. Bull, A.T., Ward, A.C., Goodfellow, M.: Search and discovery strategies for
biotechnology: the paradigm shift. Microbiology and Molecular Biology Reviews 64(3),
573–606 (2000)
8. Choo, C.W.: Working with knowledge: how information professionals help organisations
manage what they know. Library Management 21(8), 395–403 (2000)
9. De Roure, D., Goble, C., Stevens, R.: The design and realisation of the Virtual Research
Environment for social sharing of workflows. Future Generation Computer Systems 25(5),
561–567 (2009), doi:10.1016/j.future.2008.06.010
10. Golbeck, J.: Weaving a web of trust. Science 321(5896), 1640–1641 (2008)
11. Javahery, H.: Beyond power making bioinformatics tools user-centered. Communications
of the ACM 47(11), 58 (2004), doi:10.1145/1029496.1029527
12. Katoh, M.: Paradigm shift in gene-finding method: From bench-top approach to desk-top
approach (review). Int. J. Mol. Med. 10(6), 677–682 (2002)
13. Kelling, S., Hochachka, W.M., Fink, D., Riedewald, M., Caruana, R., Ballard, G., Hooker,
G.: Data-intensive science: a new paradigm for biodiversity studies. t BioScience 59(7),
613–620 (2009)
14. Levin, D.Z., Cross, R., Abrams, L.C., Lesser, E.L.: Trust and knowledge sharing: A
critical combination. IBM Institute for Knowledge-Based Organizations, 1–9 (2002)
15. Li, J.W.: SEQanswers: an open access community for collaboratively decoding genomes.
Bioinformatics (Oxford, England) 28(9), 1272–1273 (2012)
16. McIntosh, B.K., Renfro, D.P., Knapp, G.S., Lairikyengbam, C.R., Liles, N.M., Niu, L.,
Hu, J.C.: EcoliWiki: a wiki-based community resource for Escherichia coli. Nucleic Acids
Research 40(D1), D1270–D1277 (2012), doi:10.1093/nar/gkr880
17. Mirel, B.: Supporting cognition in systems biology analysis: findings on users’ processes
and design implications. BioMed Central Ltd. (2009)
18. Parnell, L.D.: BioStar: An Online Question & Answer Resource for the Bioinformatics
Community. PLoS Computational Biology 7(10), e1002216 (2011)
19. Preece, J., Nonnecke, B., Andrews, D.: The top five reasons for lurking: improving
community experiences for everyone. Computers in Human Behavior 20(2), 201–223
(2004), http://dx.doi.org/10.1016/j.chb.2003.10.015
20. Tran, D., Dubay, C., Gorman, P., Hersh, W.: Applying task analysis to describe and
facilitate bioinformatics tasks. Research Support, N I H, Extramural (2004)
21. Research Support, U S Gov’t, P H S. Stud Health Technol. Inform 107(Pt 2), 818–822
22. Yarfitz, S.: A library-based bioinformatics services program. Bulletin of the Medical
Library Association 88(1), 36 (2000)
Interaction of the Elderly Viewer with Additional
Multimedia Content to Support the Appreciation
of Television Programs
Kamila Rodrigues, Vânia Paula de Almeida Neris, and Cesar Augusto Camillo Teixeira
Department of Computer Science, Federal University of São Carlos/UFSCar,

São Carlos, Brazil
{kamila_rodrigues,vania,cesar}@dc.ufscar.br
Abstract. For many people, television is still the main form of entertainment.
The elderly population, in particular, spends much time at home and uses the
television for companionship and entertainment. However, in some countries,
due to low education level or the absence of it, to physical limitations and/or
cognitive difficulties, to the lack of prior knowledge, among other difficulties,
part of the elderly viewers do not enjoy enough the programs to have
satisfaction with them. With the interactivity provided by the Digital TV,
additional information may help the elderly viewer to better appreciate
television programs and have more fun. This paper presents data from an
observational case study conducted in the Brazilian scenario to analyze the
interaction of elderly viewers with TV programs enriched with additional
multimedia content in different formats. From the experience, some good
practices for the design of additional multimedia content to the elderly viewer
were formalized.
Keywords: Design, Interactive Digital TV, Elderly, Additional Multimedia

Content, Satisfaction.
1 Introduction
Technological advances and facilities for the dissemination and exchange of

information have prompted people to a more engaging socialization. However, part
of society faces obstacles to fully enjoy the benefits offered by Information and
Communication Technologies (ICTs). The reasons for this include economical,
educational, cultural and physical or cognitive problems. Part of the elderly
population also presents difficulties on handling of new technologies, including
ICTs.
In many countries, populations are aging. According to the United Nations (UN),
in its latest technical report "World Population Prospects", in the next 43 years the
number of people over 60 years old will be three times greater than the current one.
The elderly will represent a quarter of the estimated world population, i.e., about 2
billion people (in a total of 9.2 billion) [1].
228 K. Rodrigues, V.P. de Almeida Neris, and C.A.C. Teixeira
However, some researches indicate that in fact, elderly people have found barriers
to use ICTs [2]. Among the identified barriers are: the inability to read, due to the low
level of education of some populations, and psychological fears because of the cost of
some equipment [2]. The elderly user believes to be difficult to interact with new
technologies and fears to make mistakes which could damage the equipment [2].
However, the same technology that inhibits may be used to get the best of the benefits
it provides. The integration of efforts in different areas, including computing, can
promote a better integration of the elderly in the society.
In Brazil, no other segment of the population watches TV as much as the elderly
[3], [4]. With the interactivity provided by the interactive Digital TV (iDTV) and
applying the concept of Universal Usability and Integral Accessibility [5], it is
possible to provide specific services designed to the needs of specific population, as
elderly viewers, for example. These services, which generally require extra efforts
and investments, may not result in additional costs for TV broadcasters and viewers if
implemented appropriate business models.
Digital TV allows independent manipulation of different media that can be
properly synchronized for a multimedia presentation. This feature facilitates the
incorporation of content into television programs. The added content may be
referenced and non-intrusive. The possibility of representing the additional content in
different media is a key feature to meet different types of viewers’ needs and their
particularities. In this paper, the term “additional multimedia content” refers to extra
information (audio, image, text, etc.) added to the television media in order to
promote something extra to the viewer, aiming to clarify, inform or suggest,
synchronously with the television program in question.
Offering additional multimedia content to television programs is also a way to
provide accessibility to the elderly public, thus contributing to their involvement with
new technologies and promoting greater inclusion into the society. Such technological
solution based on additional information may help the elderly viewer to better
appreciate television programs and, hopefully, have more fun.
This paper presents data from an observational case study conducted to analyze the
interaction of elderly viewers with TV programs enriched with additional audio, text
and animation. Elderly of a community center were invited to watch a segment of a
documentary program and twelve of them volunteered to participate in the case study.
The main objective was to analyze the interaction of the elderly viewer with
additional multimedia content offered to television programs and to identify which
additional content is best suited for this audience. With the experience, some good
practices for the design of additional content to the elderly viewer were formalized.
This paper is divided as follows: section 2 reports surveys that provide information
about the interaction of elderly people with television programs and the difficulties
that they present due to the age; section 3 depicts some aspects of Digital TV, beyond
to emphasize the inclusion of additional content in television programs as a way to
provide additional information to the main content; section 4 describes the case study
conducted with the elderly audience; section 5 discusses the main obtained results
with the case study; section 6 presents some lessons learned and good practices for
the design of interactive additional content for the elderly viewers; the conclusions are
presented in section 7.
Interaction of the Elderly Viewer with Additional Multimedia Content 229
2 Television and the Elderly
Television plays several roles in the elderly lives. It is probably the main source to
keep them informed of news, besides being a way to be entertained, to look for
subjects to chat and share with friends, to relax, to relieve their loneliness, to prevent
boredom, among others [6].
Surveys show that news programs, comedy, drama, soap operas, sports and music,
are the preferred programs among elderly viewers [6], [7]. However, the pleasure and
satisfaction provided by TV can be reduced if the viewers acquire physical and
cognitive difficulties, typical of elderly people, such as the loss of hearing, vision or
understanding of the content broadcasted.
Aging process promotes reductions in the efficiency of the sense organs [8]. From
the cognitive point of view, some researches indicate that it is common that memory
failures occur with advancing age and it is generally easier for the elderly to
remember information when memory is tested by the recognition than when it is
tested by recall [8].
Carmichael [8] reports in the guide for the development of interactive TV services
for elderly viewers that, to design an appropriate solution for the elderly, it is
necessary to know this public and the difficulties acquired due to age. Important
issues should be considered when designing interactive television services for these
viewers. They are: a) Elderly suffer visual and auditory changes as they age and these
change decreases the accuracy of visual reception and also of auditory information
obtained; b) Speech should be presented clearly and accurately, to avoid a negative
impact on elderly viewers; c) Elderly have different changes during aging and, as
user, demand different requirements therefore, they cannot be considered a
homogeneous public. It should be careful with solutions that require dexterity and fine
motor coordination, such as those based on touch screens; d) Some important
information can be lost, because the nervous system of the elderly tends to process
data received by the sensor organs more slowly; e) Elderly forget acquired
information more quickly. Situations that require memorization should be replaced by
the presentation of situations familiar to the elderly; and, f) The fear of the elderly to
use new technologies can be minimized or eliminated if they understand the personal
benefits they can get. It is also important that the first experience in using the
technology be positive and successful.
In addition to potential commercial interests, the social responsibility requires
proposals of new ways to design interactive multimedia systems to meet the diversity
of the elderly public. From this perspective, it is a challenge to exploit the flexibility
offered by the new TV, digital and interactive, to provide accessibility services, using
different broadcasted media independently and simultaneously to the main program.
3 Interactive Digital TV and Additional Multimedia Content
Digital TV systems provide higher definition imaging and video, interactivity, and
new services [9]. Additional media content (audio, video etc), and applications that
synchronize, coordinate the presentation of various media and manage the interactions
promoted by the viewers can be sent with the main media (TV program). Examples of
these applications are Electronic Program Guide (EPGs), banking, educational
services, etc. Digital TV also enables the creation of nonlinear television programs
[10].
It is possible to exploit the flexibility offered by the digital TV, that allows other
media to be presented synchronized to the main program, and provide additional content
to the viewer with more information and explanations about the program. This solution
may offer, at the end of the program, more satisfaction to the viewers [11].
Current researches about the use of additional content as a form to enrich television
programs extend the paradigm of TV. These researches are based on the concept of
Social TV which explores the approach of consumption and sharing of television
programs with edition and reuse of the content by viewers [12]. In this kind of
research viewers may enrich parts of programs with information or indications and
share them with other groups or communities.
There are also researches that combine interactive Digital TV with hyper video. In
these cases the additional information are associated with the objects presented in
scenes. Hyperlinks are inserted into the video and users access the information
selecting a point on a single image area [13], [14].
Although these solutions are interesting, none of them considers the support for
different types of viewers and their particularities. Also, they do not consider the
difficulties and rejection of the elderly viewer, for example, when using new
technologies. Therefore, it is necessary to define proper contents to be added to the
TV program and also to define the best way to present them. The additional content
must meet the needs of different types of viewers and to follow the patterns of
production and exhibition of TV programs [15].
In the context of this work, it is necessary to think about a design process that
considers the elicitation, development, presentation and validation of additional
content, which is flexible and allows the offer of additional content for different types
of viewers, including the elderly. However, to provide services centered on users’
needs and with the application of the concepts of Universal Usability and Integral
Accessibility, it is necessary to get closer to the users and identify features that meet
their interaction requirements. It is also necessary to identify how to present, in the
user interface, these features preventing the elderly to reject the solution. The
interactive experience can bring more satisfaction if these issues are considered in the
design of process of television content.
In this scenario, this paper presents a case study conducted with elderly users. This
study observed the relationship of those users with an enriched television media. The
practice allowed the formalization of some good practices for the design of additional
content to the elderly user.
4 Observational Case Study
Aiming to investigate which format of additional content the elderly viewer feels
more satisfied during an interaction with digital TV and to evaluate solutions that
make this interaction more flexible to meet the needs of different types of viewers, an
observational case study was conducted with a group of elderly viewers. The main
objective was to analyze the interaction of the elderly viewer with additional
multimedia content offered to television programs and to identify which additional
media format are more appropriate for this audience. The researchers started from the
hypothesis that the additional content in formats of audio or animation, instead of text,
make more sense and are best enjoyed by the elderly public and by the public with
low education.
The case study was carried out in a center of social assistance (CRAS) in Brazil in
a city with 200.000 inhabitants. The various actions undertaken by that center attract
many people from various age groups. In particular, the center serves a group of
elderly people aged between 60 and 85 years old, who play games and attend to
physical and cultural activities. The goal of the partnership established between the
service center and university is, in addition to promote the activities mentioned, to
disseminate and promote the use of ICTs. In this sense, the case studies conducted
with this group provided them access to devices such as smartphones and last
generation television sets, including those with touch screen. The partnership was also
a way to promote digital inclusion for these viewers.
The environment set up for this observation was in the CRAS, exploring an area
that the target user attends. The environment was prepared similarly to a living room
with TV, remote control and seats. Elderly citizens, from the group mentioned above,
were invited to participate in an activity that consisted of watching a video enriched
with additional content and use the remote to interact with such content. Twelve of
them accepted to participate.
The elderly were separated into peers according to their age, education,
psychomotor skills and experience with the remote. The intention was to group people
with similar experience and profile and these data were collected from a profile
questionnaire applied in the first meeting with this group. The additional content
added to the video was in format of text, audio and animation. To evaluate the
interaction of the elderly viewer with additional multimedia content in television
programs, viewers were observed during watching sessions and interviewed in
follow-up meetings for clarification of issues relevant to the research.
The genre of the media used was also chosen from data collected with the profile
questionnaire. The data suggested that the target audience prefers to watch programs
like soap opera and newscast. From this observation, a media displayed on a newscast
was chosen to be enriched with additional content and to be used in the study. The
media had about seven minutes and was enriched using Nested Context Language -
NCL, a declarative language for authoring hypermedia. At each two minutes an
additional content was inserted. These contents were designed to add useful and
enlightening information for the audience and were inserted in intervals where there was
not character’s speech, in order to avoid damage to the main program. The first
additional content was offered in audio format, the second in text format (see Figure 1a)
and the third as an animation. The use of different formats allowed evaluating which of
them is most appropriate for the elderly audience interaction with the additional
multimedia content and which of them causes more satisfaction in this viewer.
During the ten seconds before the interval when additional content was inserted, an
icon indicating the presence of this content ahead was displayed in the upper right
hand corner of the TV screen (see Figure 1b). To trigger the additional content the
elderly should press a proper button in the remote control during these ten seconds.
The screen in which the media was displayed had 21.5 inches and the interactivity
icon was a static interface element that filled about 2% of the screen.
Besides the profile questionnaire, were elaborated for this study a form to support
the participants’ observation and a Term of Agreement. The observation form was
developed to analyze the interaction of the elderly viewer with additional content
during the session, using the remote control. From this form it was possible to observe
and report on whether the viewers observed the presence of the interactivity icon, if
they used the remote control to interact and which content has been triggered by the
remote control, if the viewers noted the presence of additional content after triggering
it and, moreover, if the viewer interacted with the peer during the session. The Term
of Agreement clarified to participants the objectives of the research, the risks and
benefits and their freedom to quit their participation at any time. The activity was
conducted following standards set by the Ethics in Research.
The second meeting with the elderly was divided into four parts. At the first
moment the peers (previously grouped) were invited to watch the enriched video. At
the beginning of the session, viewers were informed of how the session would be
conducted and the importance of their participation. In the second part, an
explanation about the elements involved in the session (mainstream media,
interactivity icon, additional content and remote control) and the procedures to be
followed to carry out the interaction with additional content through the remote
control were explained. In the third part, each peer watched and interacted with the
enriched video. The additional content was synchronized to the mainstream and there
was not possibility to pause the video to pay attention to the additional content.
There were 2 remotes, one for each viewer. When the interactivity icon was shown,
the viewer who wanted to see the added content could trigger it by pushing the green
button on the remote see (Figure 1c). The viewers could interact with each other during
the session.
Fig. 1. a) Additional content in text format. b) Interactivity icon. b) Viewers using remote control
to interact.
In the fourth part, after the end of the media presentation, each viewer participated
in an interview in which they were asked about their experience interacting with the
additional content, which content attracted more attention and which content they
considered more interesting and suitable to them.
5 Results
Among the six peers of elderly viewers who volunteered to participate in the case
study there was 1 man and 11 women. Data about this audience collected from profile
questionnaire show that most elderly live with some family member, but they have a
tendency to watch TV alone. The data also show that 10 out of 12 elderly viewers use
the remote to watch TV and they confess that the not understanding of parts of the
program may affect their satisfaction.
The first additional content was presented during the first 2 minutes of the video
and was a media in audio format. The second, appearing after 4 minutes of video, was
a textual media presented superimposed on the video scene in yellow fonts and in
accordance to guidelines and standards for Digital TV [16]. The third additional
content, an animation, was presented after 6 minutes.
The data collected from the observation form show that approximately 10 out of 12
elderly viewers evaluated recognized, at least once, the presence of the icon of
interactivity on the TV screen as being the interface element responsible for triggering
the additional content, but only 3 elderly noticed the additional content offered
thereafter. Many of them believed that the additional content was part of the main
content. To preserve the main content authoring issues and considering that the
awareness of the additional content is desired, this result suggests the need to provide
additional contents in a more highlighted format and in a way that draw more
attention. The 3 viewers, who noticed the presence of the icon of interactivity,
triggered the remote control to interact with the additional content.
The data also show that 8 of elderly viewers interacted with their partner during the
video presentation talking about the transmitted content. These findings reinforce the
fact that television has a characteristic inherent, the collective audience.
At the end of the session, during the discussion time, all viewers could see the
excerpts of the video with interactivity icons included and their related additional
content. Hence, they could visualize the proposed solution by the research that aims to
provide better appreciation of TV programs.
After watching these additional contents, viewers were asked about which of the
contents they most enjoyed and which of them were more useful as a complement of
the main program. The answers show the tendency of elderly to additional content in
audio format, confirming our initial hypothesis.
Seven viewers chose the audio content, 1 viewer opted for the text content, 1 chose
the content in the animation form and 3 said that all the contents were "cool". Among
the viewers who opted for content in the audio format, were 4 illiterate ones. Of these
4, 3 prefer the content in audio format, showing that the viewers with less education
prefer to interact with content in audio format. One viewer said: "... I can’t read, I
prefer to hear ...".
The viewer with higher education preferred the textual content. She stated: "…I
guess that the audio proposal does not disturb the video and it is cool, but I do prefer
the text...". This declaration indicates the need to offer flexible design solutions.
The viewer with higher education preferred the textual content. Observe the
following excerpt from his declaration: "…I guess that the audio proposal does not
disturb the video and it is cool, but I do prefer the text...". This declaration indicates
the need to offer flexible design solutions.
The following declarations reinforce the presented data about the trend of this
viewers group for the content in audio format: "...It is difficult to read..."," ...The
sound is easier to understand ...". For the content in animation format, one viewer
reveals "…it attracted attention because there is something moving on the screen ...".
6 Lessons Learned and Good Practices
Contributions of this study include lessons learned about the interaction of elderly
people with additional multimedia content. The elderly is a potential public to the
digital television audience and their interaction requirements should receive special
attention from interactive television designers.
From the questionnaire and observation results, it was possible to verify that a
static image filling only 2% of the TV screen does not appear to be sufficient to
attract the attention of this viewers’ group and, therefore, alert them about the
availability of an additional content.
Most of the time, they were so watchful to the main video that they did not notice
the presence of an icon for triggering additional content or they were so dispersed that
they did not pay attention to the program or to the presence of an additional content.
This situation can also occur with others profiles of viewers. It is necessary to
consider another strategy to attract the attention of audience. A possible solution
could be the use of a device for interaction that warns the viewer previously, for
example, vibrating or beeping [17].
It could also be noted during the discussions that the additional animated content
attracted the attention of some elderly viewers because, besides the audio, the moving
on the screen indicated them that something different of the content and the main
video was happening. This fact reinforces the hypothesis that a static interface
element does not attract enough attention to elderly viewers. It was also observed that
the content in textual format makes more sense to viewers with certain degree of
education, because the textual content requires more ability of reading and
comprehension.
From the data collected and observations made during the case study, and
considering the characteristics of the elderly viewer, as described in Carmichael [8] in
guide for development of interactive TV services, it was possible to formalize some
good practices for the design of additional content for the elderly: a) The interactivity
icon should be attractive, lively and preferably larger than 2% of the display screen;
b) The interactivity icon should remain on screen for a time greater than five seconds,
so that the user may become aware of its presence and trigger the remote control; c)
Give preference to additional content in audio format or animation/audio; d) If
possible, offer different options. Users with higher levels of education may prefer to
interact with text; e) The additional content should be highlighted or have some strong
evidence that it is not part of the original content. This would be useful for the elderly
viewer to identify the difference between them and; f) It is necessary use a resource to
alerting the elderly viewer, since there is a risk of them do not pay attention to the
icon of interactivity and miss the opportunity. One way to obtain attention and warn
of the presence of additional content is through the vibration or beeping of the device,
as suggested by Rodrigues et al. [17].
7 Conclusion
The case study allowed to evaluate the interaction of the elderly viewer with
additional content, and also to observe which media formats are most appropriate for
this group of viewers. By carrying out this study with diverse elderly viewers,
different ages and education levels, the initial hypothesis that participants in older age
and lower education level prefer the additional content in audio format was
confirmed.
The proposal to draw attention to the additional content with vibration or beeping
may be achieved by applications designed for mobile devices that already have such
features.
Moreover, to make the additional content a feasible solution, new content
professionals could be part of the production chain, complementing the content
produced by TV broadcasters. Service providers could make content available to
viewers via the Internet, for example.
This research also allowed that the elderly, who participate in the study, had the
first contact and interaction with technologies such as Digital TV, promoting thus an
initial step for the digital inclusion of those viewers.
Further studies may consider situations such as collective sessions, using different
genres and media sizes, as well as different profiles of viewers should be conducted in
order to identify other specificities and support the interaction design with additional
content. Finally, an architecture that allows meeting the viewer’s specificities,
including the elderly public, will be proposed.
Acknowledgements. We thank CAPES for financial support, CNPq, FAPESP, FINEP

and RNP that, through various projects of innovation and technological development,
provided the physical resources for this research. Thanks also to the physical educator
João Carlos, the Sao Carlos City Hall and the coordination of CRAS.
References
1. United Nations, http://www.un.org/esa/population/
2. Garcia, H.D.: The Third Age and the Internet: A Question for the New Millennium.
Master’s thesis submitted to Faculty of Sciences, Paulista State University, Marília – SP,
Brazil (2002) (in Portuguese)
3. Reis, D.S.M.: Specifics of Communication for the Elderly. Journal of the Faculty of
Communication FAAP 12, 65–71 (2004)
4. Colin, A.D., David, A.S., Wesley, K.T., Dilip, V.J.: Age, Affective Experience, and
Television Use. American Journal of Preventive Medicine 39(2), 173–178 (2010),
doi:10.1016/j.amepre.2010.03.020
5. Melo, M.A., Baranauskas, C.C.M.: Design for Inclusion: Challenges and Proposals. In:
Proceedings of VII Brazilian Symposium on Human factors in Computing Systems, IHC
2006, Natal - RN, Brazil, pp. 11–20 (2006), doi:10.1145/1298023.1298026 (in Portuguese)
6. Fouts, G.T.: Television Use by the Elderly. Canadian Psychology 30(3), 568–577 (1989),
doi:10.1037/h0079824
7. Davis, R.H., Edwards, A.E., Bartel, D.J., Martin, D.: Assessing Television Viewing
Behavior of Older Adults. Journal of Broadcasting 20(1), 69–76 (1976),
doi:10.1080/08838157609386376
8. Carmichael, A.: Style Guide for the Design of Interactive Television Services for Elderly
Viewers. Independent Television Commission (ITC), Kings Worthy Court, Winchester,
England (1999)
9. Brazilian Association of Technical Standards: NBR 15606-2: 2007, Digital Terrestrial
Television - Data Coding and Transmission Specification for Digital Broadcasting, Part 1:
Coding of data, corrected version - 07.04.2008, Brazil (2008)
10. Morris, S., Chaigneau, A.S.: Interactive TV Standards: A Guide to MHP, OCAP, and Java
TV. Focal Press, Burlington (2005)
11. Rodrigues, K.R.H., Melo, E.L., Nakagawa, P.I., Teixeira, C.A.C.: Interaction with
Additional Content to Support the Understanding of Television Programs. In: Proceedings
of the IX Brazilian Symposium on Human Factors in Computing Systems, IHC 2010, Belo
Horizonte -MG, Brazil, vol. 1, pp. 91–100 (2010) (in Portuguese)
12. Bulterman, D.C.A., Cesar, P., Jansen, A.J.: An Architecture for Viewer-side Enrichment of
TV Content. In: Proceedings of the 14th Annual ACM International Conference on
Multimedia, Santa Barbara, CA, USA, pp. 651–654. ACM (2006),
doi:10.1145/1180639.1180775
13. Gradvohl, A.L.S., Iano, Y.: Matching Interactive TV and Hypervideo. IEEE Latin America
Transactions 5(8), 579–584 (2007), doi:10.1109/T-LA.2007.4445709 (in Portuguese)
14. Teixeira, C.A.C., Linington, P.: Exploiting Interactive Video and Animation in Distributed
Environment for the Design of Hypermedia and Graphical User Interfaces. In: Proceedings
of VI Brazilian Symposium on Computer Graphics and Image Processing, SIBIGRAPI
1993, Recife - PE, Brazil, pp. 213–219 (1993)
15. DTV4All. Digital Television for All, http://www.psp-dtv4all.org
16. Instituto Nacional de Tecnologías de la Comunicación.: Digital Terrestrial Television
(DTT) Accessibility Recommendations Guide,
http://www.inteco.es/file/snb-6ZR2I2DaXKiMJlkT_g
17. Rodrigues, K.R.H., Pereira, S.S., Quinelato, L.G.G., Melo, E.L., Neris, V.P.A., Teixeira,
C.A.C.: Interaction with Additional Content Using Multiple Devices to Support the
Appreciation of Television Programs. In: Proceedings of Brazilian Symposium on
Multimedia and the Web, WebMedia 2011, Porto Alegre – SC, Brazil (2011) (in
Portuguese)
Exploring Children’s Attitudes towards Static
and Moving Humanoid Robots
Fang-Wu Tung1 and Tsen-Yao Chang2

1
Department of Industrial and Commercial Design, National Taiwan University of Science and
Technology, 43,Sec. 4, Keelung Rd., Taipei, 106, Taiwan
2
Department of Creative Design, National Yunlin University of Science and Technology,
Yunlin County, 64002, Taiwan
Abstract. This study investigates the uncanny valley for robots designed specifi-
cally for child users, and examines their attitudes toward humanoid robots with
different anthropomorphic appearances and behaviors. An uncanny valley was
identified in this study, indicating that the children were less attracted to images
they considered highly human-like, although they were distinguishable from hu-
mans. The results of this study support Mori’s uncanny valley hypothesis regard-
ing children’s perceptions of static robots. A significant finding is that moving
robots can moderate an uncanny valley plot, which contradicts Mori’s uncanny
valley theory that the movement of a robot amplifies a generated emotional re-
sponse. The moving robots exhibited various behaviors, such as facial expres-
sions, speech, gazing, and gestures, which can generally enhance children’s
perceptions of robots. These behaviors were derived from human-human interac-
tions, and can be considered social cues. The results of this study show that social
cues can be applied to child-robot interactions. Children perceive robots are more
socially and physically attractive when they exhibit sufficient social cues. Specifi-
cally, the display of social cues by robots that are less anthropomorphic can
significantly enhance children’s social perceptions of them. This has crucial im-
plications for the behavior a child anticipates from a machine-like robot compared
to a human-like robot. According to the study results, robots designed for children
do not require excessively human-like designs. Middle- to low-level anthropo-
morphic designs combined with appropriate social cues can enhance children
preferences and acceptance of robots. This also enables businesses to develop
educational, care, or entertainment robots for children at a reasonable cost.
Keywords: humanoid robot, child-robot interaction, behavior, social cue.
1 Introduction
The paradigm of robotics is moving from a specific industrial technology to the con-
sumer, home, and service markets. Service robots can be applied to people’s living
and working spaces as assistants, companions, or perform other social roles in
the future [1]. Because robots are designed and built for purposes such as service,
education, therapy, or entertainment, human-robot interactions become increasingly
238 F.-W. Tung and T.-Y. Chang
socially situated and multi-faceted. Social and emotional levels of interaction influ-
ence a person’s acceptance of a robot’s role [2,3]. Furthermore, considering a robot’s
functionality, previous studies have examined the social aspects of human-robot inte-
raction to evaluate how people respond to the appearance and behavior of robots in
various contexts [4,5,6]. To understand people’s views toward robot companions in
households, an investigation by Dautenhahn et al. [7] showed that the majority of
people were in favor of a robot companion and saw the potential role of a robot as
being an assistant, machine, or servant. People desired human-like communication
with potential robot companions. Attributing human-like characteristics to robots with
a human appearance is a design consideration because this could facilitate human-
robot interaction [8]. Robots designed with anthropomorphic characteristics are the
embodiment of a human-computer interface, and forms the basis for potential social
relationships. This argument is supported by research by Goetz et al. [9] which indi-
cates that people anticipate human-like robots would be best suited for interactive
tasks, whereas mechanical-looking robots would be best suited for routine jobs. Sco-
pelliti et al. [10] focused on people’s attitudes toward domestic robots across three
generations. Their findings showed that younger people scored higher on positive
feelings (e.g., amusing, dynamic, pleasant, or relaxing) toward a domestic robot com-
pared to adults and older adults people. Younger people who were born in the digital
era did not express any anxiety toward the idea of a domestic robot. They reported a
preference for robots with human-like attributes so they can interact with it in leisure
situations, rather than perceiving it as a useful device. Considering children as poten-
tial users, the use of human-like features such as behavior and appearance can engage
children to interact with robots. The large number of current humanoid robotics
projects exemplifies this tendency.
As discussed, various humanoid robots are designed specifically for children, for
educational, entertainment, and therapeutic purposes. Attributing human-like charac-
teristics to a robot can facilitate children’s understandings of its functionalities, and
establishes meaningful human-robot interactions. Such notions lead to an assumption
that increasing the realism of a robot has practical benefits. Although many research-
ers have pursued a highly human-like form of social robots, a potential danger is that
highly human-like robots might fall into Mori’s [11] uncanny valley. Uncanny val-
ley theory hypothesizes a positive correlation between the human-like appearance and
motion of robots with people’s positive emotional reactions towards them. However,
as likeness increases, there is a breaking point beyond which familiarity drops and
robots become “eerie”. When the emotional reaction is plotted against the robots’
level of anthropomorphism, a negative valley becomes visible, and is commonly re-
ferred to as the uncanny valley. Furthermore, Mori argued that a robot’s movement
amplifies the emotional response in comparison to static robots.
The uncanny valley hypothesis has received empirical support from several stu-
dies. These studies have focused on examining Mori’s hypothesis with adult partici-
pants, using static robot pictures as stimuli. However, whether the movement of a
robot amplifies the emotional response compared to that of static robots has seldom
been investigated. Robots are not developed to remain motionless, rather to be
responsive to and interactive with people. Therefore, understanding people’s percep-
tions of robots exhibiting human-like behavior provides insights that may be useful
when designing a robot’s appearance and behavior to facilitate better human-robot
Exploring Children’s Attitudes towards Static and Moving Humanoid Robots 239
interaction. Given that children are potential users of learning or entertainment robots,
numerous humanoid robots have been created to serve as social companions or learn-
ing partners for children. This study investigates the uncanny valley for robots de-
signed specifically for children, and examines their attitudes toward humanoid robots
with various anthropomorphic appearances and behaviors. Obtaining insight about
their perception toward robots that exhibit various degrees of anthropomorphic
appearance and behavior could provide designers with valuable a reference
2 Method
2.1 Participants
For this study, we recruited a large sample group of 578 (N) children. We assigned
267 children to the static condition group, comprising 87 fourth graders (42 girls and
45 boys), 86 sixth graders (42 girls and 44 boys), and 94 eighth graders (45 girls and
49 boys). The remaining 311 children were assigned to the moving condition group,
comprising 108 fourth graders (49 girls and 59 boys), 86 sixth graders (53 girls and
65 boys), and 94 eighth graders (43girls and 42 boys).
2.2 Experimental Stimuli

To conduct the investigation, we accessed numerous robot resources developed by
companies, institutes, research labs, and artists, and acquired 54 robots ranging from
“barely human” to “fully human”. The 54 robot images obtained primarily from the
Internet were edited by re-scaling and removing variables such as background color,
marks, and other objects to ensure that the images were presented in a standardized
format. Three professional designers subsequently examined the 54 images to remove
redundant or inappropriate images regarding the anthropomorphism scale, and identi-
fied 34 images for inclusion in the study.
Twenty-nine children aged between 10 and 11 years were recruited to sort the 34
images by means of hierarchical clustering individual assessment. Upon commence-
ment of a trial, the experimenters shuffled the cards in the set to ensure random pres-
entation. To facilitate children’s participation in this study, they were requested to sort
the cards into the following three categories of realism: 1) low; 2) middle; and 3)
high. Each subset was then divided into three groups based on identical criteria. Thus,
each participant sorted the 34 robot images into nine groups ranging from low to high
human likeness. Based on ranking data, we selected nine robots as experimental
stimuli to show how children rate robots on an anthropomorphic scale, from “little
resemblance to humans” to “highly human-like”, (Table 1).
The robot pictures shown in Table 1 are the stimuli used in the static condition,
whereas the robot videos are the stimuli employed for the moving condition. The
length of the edited video clip is approximately 50 s, and the content fully demon-
strated the behaviors of each robot, including locomotion, gesture, facial expression,
and speech. Among the robots, R5 and R7 were capable of speech, whereas the
remaining videos were accompanied by identical background music. Table 1 shows
the behaviors of each stimulus presents in the video.
Table 1. The experimetal stimuli
R1 R2 R3 R4 R5 R6 R7 R8 R9
Image
Behavior
Speech v v
Gesture v v v v v v v
Facial expres-
v v
sion
Locomotion v v v v v v
2.3 Measurement Tools
The dependent variables for social attraction and physical attraction were modified
from a version of McCroskey and McCain's [12] social and physical attraction scale,
as well as from relevant studies that adopted the same scale to measure users’ atti-
tudes toward computers, robots, or media [13,14]. Social and physical attraction are
two key dimensions of interpersonal attraction that have been found to facilitate inter-
personal communication that leads to the formation of friendships. This study investi-
gates whether various levels of anthropomorphic appearance and behavior influence
children’s social and physical attraction toward robots.
The social attraction scale comprises the following of five items: 1) I think this ro-
bot is friendly; 2) I like this robot; 3) I think this robot could be a friend of mine; 4) I
would like to have a friendly chat with this robot; and 5) This robot would be pleasant
to be with. The physical attraction scale comprises the following three items: 1) I
think this robot is good looking; 2) I find this robot very physically attractive; and 3) I
like the way this robot looks. We measured the two sets were measured using a set of
paper-and-pencil questionnaires that employed a 7-point Likert scale ranging from 1
(very strongly disagree) to 7 (very strongly agree). The wording used in the question-
naires was designed after discussions with teachers and children to prevent any
misunderstanding.
2.4 Procedure
Static condition group: the experiment was conducted using the discussed paper-and-
pencil method. Each robot image was presented as a high-quality color printed image
accompanying the aforementioned questionnaires. Participants were requested to
evaluate the images of nine robots by completing questionnaires. The order of the
stimuli was randomized for each child. The experiment was conducted at select
schools. Participants completed the questionnaire survey in their classrooms or in a
quiet place such as a school library.
Moving condition group: the experiment for the moving condition group was con-
ducted using computers in a computer class room. The robot videos were played in
random sequences. Each session contained 30 participants, and each participant was
provided earphones, and experimental personnel were present to assist the participants
while watching the nine robot videos and completing in the questionnaire. The total
duration of the experiment was approximately 40 min.
3 Results
We calculated the internal consistency (Cronbach’s α) to assess the reliability of these
scales. Cronbach’s α results for the social and physical attraction items were all more
than 0.7. According to Nunnally [15], Cronbach’s α values of 0.7 are adequate for
internal consistency and reliability. Therefore, the measures used in this study show
adequate reliability.
3.1 Social Attraction

Table 2 shows the mean social attraction scores rated by participants in the static and
moving condition groups. Robot R6 received the highest score for social attraction in
the static condition group, whereas Robot R5 received the highest score for the mov-
ing condition group. The social attraction that children felt toward the stimuli is plot-
ted in Fig. 1. Their attitude toward static robots supports Mori’s prediction, and an
uncanny valley emerged when children evaluated Robots R7 and R8. These robots
were highly human-like in appearance, although they could be distinguished as non-
human. However, this uncanny valley in the moving condition group was less appar-
ent, which does not support the hypothesis that moving robots could amplify people’s
emotional responses compared to static robots. Generally, participants felt higher
social attraction toward moving robots than those that were static.
Table 2. Mean Social attraction scores rated by Static and Moving condition groups
R1 R2 R3 R4 R5 R6 R7 R8 R9
Static condition 3.31 3.30 3.40 4.30 4.49 5.57 3.76 2.62 4.45
Moving condition 4.89 4.72 4.74 4.81 5.71 5.26 4.85 4.12 5.03
7
6
5
4 static condition
3
2 moving
1 condition
0
R1 R2 R3 R4 R5 R6 R7 R8 R9
Fig. 1. Each robot’s social attractiveness in the static and moving conditions
3.2 Physical Attraction

Table 3 shows the mean physical attraction scores rated by participants in the two
conditions, and Fig. 2 shows the physical attraction children felt toward the static and
moving robots. Similar findings were observed. Children reported that the moving
robots were more attractive in appearance than those that were static. The curve is
much flatter for the moving condition group than the static group.
Table 3. Mean Social attraction scores rated by Static and Moving condition groups
R1 R2 R3 R4 R5 R6 R7 R8 R9
Static condition 3.03 2.85 2.96 3.75 3.76 5.09 3.07 2.37 4.4
Moving condition 4.52 4.39 4.28 4.49 5.01 4.91 4.04 4.02 4.9
7
6
5
4 static condition
3
2 moving
condition
1
0
R1 R2 R3 R4 R5 R6 R7 R8 R9
Fig. 2. Each robot’s physical attractiveness in the static and moving conditions
4 Discussion
4.1 The Peak before the Uncanny Valley

The results from the static condition group show that the degrees of realism of robots
have a significant influence on children’s attitudes towards robots. Figures 1and 2
show a continuously positive change of realism versus children’s evaluation of social
and physical attraction until a point of realism beyond which children’s evaluation
decreases abruptly. As the appearance becomes less distinguishable from a human,
children’s responses become positive once again. Thus, an uncanny valley was ob-
served in this study, indicating children were less attracted to images considered highly
human-like, yet distinguishable from humans, which evoked a feeling of discomfort in
the observer.
In the static condition group, Robot R6 received the highest evaluation from child-
ren for social and physical attractiveness. Robot R5 received the highest evaluation
from the moving condition group. Robots R6 and R5 were at the peak before the
uncanny valley. This finding is identical to that obtained by Bartneck et. al [16],
which shows that toy robots and humanoids preferred over humans.
An objective of this study is to identify the threshold of “humanness” that is appro-
priate for robots designed for children. The implication of the observation is that de-
signers of robots intended for children might consider combining human and machine
features to put effort toward the first peak of the plot, rather than attempting to per-
fectly replicate human-like appearance. The finding that children preferred Robots R5
and R6 among all stimuli supports Woods’s [17] finding that children prefer robots
with cartoon-like appearances. People, including children, are sensitive to the particu-
lar pattern of features that form a face. Using mere representations of cartoon-like
faces can avoid the uncanny valley phenomenon and cover a large aesthetic range
[18].
4.2 Movement Can Moderate the Uncanny Valley
A significant finding of this study is that moving robots can moderate the uncanny-
valley plot. This contradicts Mori’s uncanny valley theory which states that move-
ment of a robot amplifies the emotional response more than a static robot. A compari-
son of children’s attitudes toward static and moving robots shows a relatively flat
curve for moving robots (Figures 1-2). Moving robots exhibiting behaviors such as
facial expressions, speech, gaze, locomotion, and gestures can generally enhance
children’s perceptions of robots. The behaviors other than locomotion derived from
human-human interaction can be considered social cues. Social cues (e.g., voice,
presence of a face, and facial expressions) have been adopted into user interface de-
sign to enhance human-computer interaction. Studies related to computers as social
actors (CASA) have empirically proven that people tend to treat a computer as a so-
cial entity when the computer adequately exhibits social cues to elicit social responses
from people [19, 20]. CASA principles argue that computers which exhibit social
cues can convey a sense of sociability and intimacy, thereby inducing social responses
from people. This consequently improves user attitudes toward computers and fosters
a more favorable relationship between users and computers. Social Agency Theory
[21] also argues that a greater frequency of social cues in an interaction improves the
quality of that interaction. The results of this study show that social agency theory can
also be applied to child-robot interactions. Children perceive robots more socially and
physically attractive when they exhibit more social cues.
Comparing the results of the static and moving condition groups, children gave
higher evaluations to Robot R5 than Robot R6, which received the highest scores
from the static condition group. In the videos provided in this study, Robot R5 could
speak and show simple facial expressions with its eyelids and lips, whereas Robot R6
could only walk with whole-body motions. The fact that Robot R6 did not exhibit
more social cues rather than locomotion could be the reason why it was the only robot
to receive lower scores from the moving condition group than from the static condi-
tion group. According to Piaget [22], self-moving is a crucial criterion used by child-
ren to judge the “aliveness” of an object. Results for Robot R2 receive from
both groups show that locomotion could enhance children’s attitudes toward robots,
especially for those with low-degrees of human-like appearance. The more human-
like robots appear, children might anticipate the robots to exhibit more human-like
behaviors. In this sense, Robot R6, a humanoid robot, might be expected to have more
social action capabilities rather than locomotion. This can also be observed in results
for Robot R9, which has the appearance of an attractive young humanoid robot with
influent whole-body motions. Robot R9 did not speak or show social cues such as
facial expressions and gestures in the video provided. Thus, the moving Robot R9 did
not significantly enhance children’s social attraction toward it. This has significant
implications for the accompanying behavior a person expects from a machine-like
robot compared to a human-like robot. For instance, people might expect a human-
like robot to have language capabilities, although they might not expect a machine-
like robot to have any language communication abilities at all. Results from the
moving condition group indicate that the human-like degree may not play an essential
role in influencing children’s emotional responses to robots. Appropriate action beha-
viors can enhance children’s perception of humanoid robots, especially for robots
with low- and mid-degree human-like appearances. This study recommends that de-
signers consider applying social cues to robots to improve their social and physical
attraction, in which child-friendly robot designs can be achieved at a reasonable cost.
5 Conclusion
In the innovation of robot service functions, developing educational or entertainment-

oriented robots for children is a focal point for robot development. In this study, we
investigated children’s attitudes toward humanoid robots. Ultimately, by adopting this
research as key reference, we hope to provide designers with design criteria for de-
veloping robots for children. According to the obtained results, robots designed for
children do not require excessively human-like features. Mid- to low-level anthropo-
morphic designs combined with appropriate social cues can enhance children’s prefe-
rences and acceptance of robots. The findings in this study also enable businesses to
develop cost-effective educational, care, or entertainment robots for children.
Acknowledgments. This material is based upon work supported by the National

Science Council of the Republic of China under grant NSC98-2221-E-239-012-MY2
and NSC100-2410-H-011-035-MY2.
References
1. Prats, M., Sanz, P.J., del Pobil, A.P., Martínez, E., Marín, R.: Towards multipurpose auto-
nomous manipulation with the UJI service robot. Robotica 25, 245–256 (2007)
2. Bartneck, C., Forlizzi, J.: A design-centred framework for social human–robot interaction.
In: Proceedings of the 13th IEEE International Workshop on Robot and Human Interactive
Communication, pp. 591–594. IEEE Press, Kurashiki (2004)
3. Kiesler, S., Hinds, P.: Introduction to This Special Issue on Human-Robot Interaction.
Human- Computer Interaction 19, 1–8 (2004)
4. Thrun, S., Beetz, M., Bennewitz, M., Burgard, W., Cremers, A.B., Dellaert, F.: Probablis-
tic algorithms and the interactive museum tour-guide robot minerva. Journal of Robotics
Research 19, 972–1000 (2000)
5. Falcone, E., Gockley, R., Porter, E., Nourbakhsh, I.: The personal rover project: the com-
prehensive design of a domestic personal robot. Robotics and Autonomous Systems 42,
245–258 (2003)
6. Kahn Jr., P.H., Friedman, B., Perez-Granados, D.R., Freier, N.G.: Robotic pets in the lives
of preschool children. In: Poster presented at the 70th Anniversary of the Society for Re-
search in Child Development-2003 Biennial Meeting, Tampa, FL, USA (2003)
7. Dutenhahn, K., Woods, S., Kaouri, C., Walters, M., Koay, K.L., Werry, I.: What is a robot
companion–friend, assistant, or butler? In: IEEE/Robotics Society of Japan International
Conference on Intelligent Robots and Systems (2005)
8. Friedman, B., Kahn Jr., P.H., Hagman, J.: Hardware companions? - What online AIBO
discussion forums reveal about the human-robotic relationship. In: Cockton, G., Korhonen,
P. (eds.) Proceedings of CHI 2003 Conference on Human Factors in Computing Systems,
pp. 273–280. ACM Press, NY (2003)
9. Goetz, J., Kiesler, S., Powers, A.: Matching robot appearance and behavior to tasks to im-
prove human-robot cooperation. In: Proceedings of the Twelfth IEEE International Work-
shop on Robot and Human Interactive Communication, Lisbon, Portugal (2003)
10. Scopelliti, M., Giuliani, M.V., D’Amico, A.M., Fornara, F.: If i had a robot at home
Peoples’representation of domestic robots. In: Keates, S., Clarkson, J., Langdon, P., Ro-
binson, P. (eds.) Designing a More Inclusive World, pp. 257–266. Springer, Cambridge
(2004)
11. Mori, M.: The uncanny valley. Energy 7(4), 33–35 (1970)
12. McCroskey, J.C., McCain, T.A.: The measurement of interpersonal attraction. Speech
Monographs 41, 261–266 (1974)
13. Moon, Y.: Similarity effects in human-computer interaction: Effects of user personality,
computer personality, and user control on attraction and attributions of responsibility, Un-
published doctorial dissertation, Stanford University, Stanford (1996)
14. Lee, K.M., Peng, W., Yan, C., Jin, S.: Can Robots Manifest Personality? An Empirical
Test of Personality Recognition, Social Responses, and Social Presence in Human-Robot
Interaction. Journal of Communication 56, 754–772 (2006)
15. Nunnally, J.C.: Psychometric Theory. McGraw Hill, New York (1967)
16. Bartneck, C., Kanda, T., Ishiguro, H., Hagita, N.: Is the Uncanny Valley an Uncanny
Cliff? In: Proceedings of the 16th IEEE International Symposium on Robot and Human In-
teractive Communication, pp. 368–373. IEEE Press, NJ (2007)
17. Woods, S.: Exploring the Design Space of Robots: Children’s Perspectives. Interacting
with Computers 18, 1390–1418 (2006)
18. Blow, M., Dautenhaun, K., Appleby, A., Nehaniv, C.L., Lee, D.: The art of designing robot
faces: Dimensions for human-robot interaction. In: Proceedings of HRI 2006, pp. 331–332.
ACM Press, New York (2006)
19. Nass, C., Steuer, J.: Anthropomorphism, Agency, and Ethopoeia: Computers as Social Ac-
tors. Human Communication Research 19(4), 504–527 (1993)
20. Reeves, B., Nass, C.: Media equation: how people treat computer, television, and new me-
dia like real people and places. Cambridge University Press, New York (1996)
21. Mayer, R.E., Sobko, K., Mautone, P.D.: Social cues in multimedia learning: role of speak-
er’s voice. J. Educ. Psychol. 95, 419–425 (2003)
22. Piaget, J.: The child’s conception of the world. Littlefield Adams, Savage (1929)
Part III
Interaction for Society and Community
Mobile Money Services in Uganda: Design Gaps
and Recommendations
Rehema Baguma
Makerere University, School of Computing & Informatics Technology, Kampala, Uganda

[email protected]
Abstract. Mobile money is a great service for developing countries where the
banking infrastructure is still severely limited and very few people can meet re-
quirements of banking institutions. While these services are being greatly
received due to the convenience that goes with electronic and mobile based ser-
vices, the reception and uptake is still limited to literate populations who are the
minority in these countries. This paper discusses the current design gaps of the
mobile money service in Uganda based on a survey carried out in Kampala and
Kayunga districts and provides recommendations on how these gaps can be ad-
dressed for the service to benefit more people especially the less literate poor
that have no access or cannot afford conventional banking services.
1 Background
According to Duga and Getachew [2], the number of mobile phone subscribers is
expected to hit 4.5 billion by 2012/2013 globally and this is attributed to emerging
markets in Asia, Latin America and Africa. Africa alone has more than 300 million
mobile phone subscribers according to the 2009 ITU report. In Uganda by June 2011,
the mobile phone subscription had reached 14 million [12]. As mobile phones are
multiplying in the developing world, new services are being innovated by Mobile
Network Operators (MNOs) in addition to the traditional voice and SMS to remain
competitive [9]. One of the prominent services is mobile money which loosely refers
to money stored using the SIM (subscriber identity module) as an identifier as op-
posed to an account number in the conventional banking sense [11]. A notational
equivalent in value is kept on the SIM within the mobile phone which is also used to
transmit payment instructions. The corresponding cash value is physically held by the
MNO, a bank or another third party depending on the business model [3]. MNOs and
their agents provide an interface through cash-out (issuing cash on demand) or cash-in
(convert cash into notational equivalent) functions providing convertibility between
mobile money and cash (Morawczynski, 2009).
In Uganda, all the 5 major MNOs currently offer mobile money service namely:
MTN Mobile Money from MTN, M-Sente from UTL, Airtel Money from Airtel, Warid
Pesa from Warid and Orange money from Orange Telecom. According to bank of
Uganda’s supervision report for 2011, mobile money services registered much more
significant growth compared to the previous year with the number of registered custom-
ers increasing from 1.7m in 2010 to 2.9m during 2011 [10]. MTN Mobile Money, the
250 R. Baguma
first of the five to launch and the biggest in the market, has reportedly registered more
than 1,000,000 customers, setup over 1,500 agents/outlets across the country and trans-
ferred more than UGX 590 billion (US$ 245 million) since its launch in March 2009.
Like the case of mobile telephony versus fixed wire, Africa is expected to benefit
more from mobile financial services because its financial services industry is not as
developed as in Europe, Asia and North America [4]. However, while these services
are being greatly received due to the convenience that goes with electronic and mobile
based services, the reception and uptake is still limited to the literate urban popula-
tions who are the minority in these countries [11]. UNESCO [14] reported that one of
the challenges of delivering mobile phone based services is that 41% of the popula-
tion in developing countries is non-literate and even the literate among the poor are
typically novice users of computer technologies. According to the 2011, Demographic
and Health Survey (UDHS), the literacy rate of 15-24 year olds is 76.1% [13].
UDHS defines literacy as the 15-24 year olds who have attended secondary school or
higher or who can read a whole sentence or part of a sentence. White [16] defines
literacy for mobile phone users to mean the ability to understand the text that is dis-
played to them on the screen in order to be able to navigate the menu system, and then
to understand the functionality that is subsequently presented to them. For technologi-
cal literacy, White [16] defines it as the ability to understand or familiarity with
common computer interaction paradigms such as the concept of menu systems, radio
buttons or data entry. Chipchase [1] observed that non-literate populations avoid
complex functions and primarily use phones for synchronous voice communication.
UNESCO’s Institute for Statistics defines the level of literacy as the percentage of
people aged 15 and above who can with understanding, read and write a short, simple
statement on their everyday life. Ndiwalana et al [11] noted that the bulk of mobile
money transactions in 2010 in Uganda (including receiving) were happening in Kam-
pala (the capital city of Uganda). In addition most respondents (72.6%) in Ndiwalana
et al. study [11] reported having access to other financial services through a personal
account in a formal financial institution and 72.7% reported being currently em-
ployed. This means that majority of the respondents in this study are not among the
conventional mould of users expected to gain most advantage from using mobile
money i.e. the unbanked rural population. White [16] noted that for businesses which
are targeting poorer populations, ensuring that the products and services can be un-
derstood and used by non-literate people will increase their potential to succeed.
Therefore there is need for research on how mobile money service can be designed
such that more low-literate and novice users of technology products who comprise the
bulk of the unbanked in Uganda and other developing countries can take it up and use
it. Medhi et al [8] revealed that non literate and novice users of technology face sev-
eral barriers with existing text-based mobile interfaces such as difficulties understand-
ing hierarchical structures, soft keys, scroll bars, non-numeric inputs, and specialized
terminology. This paper discusses the current design gaps of the mobile money ser-
vice in Uganda based on a survey carried out in Kampala and Kayunga districts and
provides recommendations on how these gaps can be addressed for the service to
benefit more people especially the less literate poor that have no access or cannot
afford conventional banking services. The rest of the paper is organized as follows:
methodology, design gaps for the mobile money service in Uganda, suggestions on
how the design gaps can be addressed, and conclusion.
Mobile Money Services in Uganda: Design Gaps and Recommendations 251
2 Methodology
To accomplish the objectives of this study, we conducted field interviews with current
mobile money users and Agents (business people who buy franchises from MNOs to
trade in the Mobile Money service) in peri-urban areas of Kampala city and rural
locations in Kayunga district all in Uganda. All the participants owned or had access
to a mobile phone. We interviewed 100 users (40 from Kayunga and 60 from Kampa-
la), and 25 agents (10 from Kayunga and 15 from the suburbs of Kampala City name-
ly; Najeera, Kiwatule, Natete and Kawempe. Kayunga District is a predominantly
rural settlement in central Uganda. Interviewed users were met at mobile money
shops and chosen randomly. The interview with users covered the languages spoken
and written, functions on a mobile phone used, whether they find the mobile money
interface easy to use, what is liked about mobile money, what they would change,
value of mobile money to their day today activities, challenges faced using mobile
money and possible solutions. The interview with mobile money Agents covered: the
languages spoken and written, whether they find mobile money interface easy to op-
erate, business viability of mobile money, what they like about mobile money, what
they would change, desired versus available services, challenges faced and possible
solutions. The obtained data was analysed using cross case/content analysis method to
examine themes, similarities, and differences across the feedback obtained. The next
section presents the findings from the study.
3 Findings from the Study
3.1 Social-Demographic Characteristics of the Participants

Of the 100 users interviewed, 60% were male and 40% female. 35 percent had a de-
gree or diploma, 25% had completed Advanced Level (High School), 20% had com-
pleted Ordinary Level (Middle School), 15% had completed primary school (junior
school) while 5% had no formal education. All the participants had ever used mobile
money. They were residents of the area of study and 40% percent had personal bank
accounts. Of the 25 agents, 80% were female and 20% male. The next section dis-
cusses the design gaps for the mobile money service in Uganda as per the feedback
obtained from interviewing users and agents.
3.2 Design Gaps for the Mobile Money Service in Uganda

English Interface: The interfaces of all the mobile money services in Uganda is in
English. This means the low literate masses cannot use it without intermediaries.
Hinman and Matovu [4] noted that one of the main causes of people not engaging
with mobile money services is due to illiteracy among other factors. From the inter-
views, users complained that customer care staff use English only and speak very fast.
On the other hand, an Agent noted that most people from villages do not understand
English hence they have to interpret for them but this keeps away some potential us-
ers who consider this an inconvenience or suspect that they can easily be cheated.
Therefore the popular premise that mobile money through an increasingly large
252 R. Baguma
mobile phone user base will provide a platform that could potentially be leveraged to
service the financial needs of the poor is yet to be a reality. Additionally, even the
literate but poor are not likely to use it or use it as much as they would due to associa-
tion of English language with the wealthy and prestigious members of society. In a
study conducted by Medhi et al [8], participants in the study on designing mobile
interfaces for novice and low literacy users strongly and positively associated the
English language with wealth and prestige. This was due to a combination of mindset
inherited from colonial history as well as the modern-day fact of greater economic
opportunities available to English speakers. The less educated that is those who can-
not read and write in English at all or those with difficulty doing the same find it hard
to learn and remember how to use mobile money. One user noted that “I stopped in
primary five therefore I can only read and write in my local language-Luganda but
mobile money is ‘written’ in English therefore I keep asking for help from Agents
whenever I want to use the service”. On the other hand, those who can read and write
English without difficulty, reported that they found learning and remembering how to
use mobile money services particularly the menu driven MTN mobile money very
easy because of the prompts like one said “I follow prompts which is easy for anyone
to follow if one can read English”.
Limited Understanding of How Some Mobile Money Services Work: Services
related to buying and selling goods or paying for utilities like water and electricity are
still least used. One of the reasons for this is limited understanding of the m(e)-
commerce concept as one respondent noted; “I am still not sure that mobile money
can reach when I use it to pay water bills otherwise it would be convenient”.
Limited Coverage of the Payment Function: The payment function of mobile mon-
ey service to-date covers a few aspects such as mobile phone credit, utility bills and
school fees. However users are also using mobile money to settle informally other
common payment obligations such as transport fares, professional fees, ho-
tel/restaurant bills, rent, etc. However, the problem faced with this informal arrange-
ment is that it depends on the willingness and flexibility of the recipient and in a
majority of cases, they decline this form of payment due to various reasons such as
having to bear the burden of converting mobile money into cash, limited understand-
ing and trust of mobile money due to lack of sensitization, non-membership to mobile
money services (transacting with a non-registered member attracts very high charges),
failure to agree on transaction charges which the payee has to add on the bill, etc.
Desire for Mobile Money Service to Provide Services Provided by Banks: Some
users view mobile money service as an alternative to conventional banks hence expect
it to offer most of the services provided by banks. During the interviews, users
requested MNOs to consider providing loans particularly airtime (mobile phone cre-
dit) and mobile money. They noted that this is particularly useful when one is in a
remote place far from home. They reason that payment can be redeemed as soon as
the user loads airtime/mobile money. At the moment, only Warid loans UGX 1000
airtime and none of the service providers loans mobile money. Viewing mobile
money service as an alternative to banks is not surprising given that 100% of the
respondents interviewed were all using sending, receiving and depositing money
functions of mobile money services that are traditionally provided by banks.
Lack of Feedback on System Status during Interaction: For some services of mo-
bile money such as paying payTV dues, users are not kept informed about what is
going on during interaction. One user noted that: “when you use a scratch card (an
alternative option for paying TV dues), there is a person at the other end during the
crediting process, so it is more reliable compared to mobile money where feedback is
not guaranteed and in case of failure one is left in suspense concerning how far the
transaction had gone”.
Long Menu: Some of the users interviewed complained that the menu for mobile
money services is very long hence difficult to remember how to use. MTN mobile
money has 7 items at the main menu and between 1-4 options at sub menu level. Air-
tel Money also has 7 items at the main menu and between 1-4 options at sub menu
level, M-sente has 8 items at the main menu and between 1-5 options at sub menu
level while Warid Pesa is command driven. When asked if mobile money was easy to
use, one interviewee responded: “someone just took me through once but the menu is
very long so it has taken me a lot of time to learn how to use it”.
Confusing Functions: Users noted that there are some confusing functions such as re-
dundant functions like non-mobile user function in MTN mobile money and the com-
bined sending and receiving functions in Warid pesa. One commented that “the non-
mobile user option under send money to in MTN mobile money confuses me for it serves
no purpose”. In Warid pesa, sending and depositing are not separated which confuses
users. Users would like them to be separated like is the case with MTN mobile money.
Weak Error Prevention and Correction: The design of the current mobile money
services has limited support for error prevention and correction in particular regarding
sending money to the right recipients. According to the users and agents interviewed,
MTN and Warid do not have adequate measures to protect against losses from errors
or mistakes e.g. reversing a transaction or freezing an account when an error is made
sending or depositing money on a phone. Many users interviewed reported having
sent money to wrong recipients or received money from unknown people which could
not be got back. Some of the responses given include:
“I have been receiving money from someone whose telephone number is almost
similar to mine except for one digit and every time, I have to send it back because
I am born again but I have heard of people who could not recover their huge sums
of money because the people they sent it to did not cooperate”.
“I Sent Money to a wrong Phone number and when I called the phone, the owner
did not pick it”
“I have never made a mistake. I am always careful when transacting but someone
sent me UGX 900,000 last month which I had to go through the inconvenience of
sending back”.
One Agent called Musa in Kiwatule trading centre, in the process of sending UGX
250,000 had the receiver credited 3 times due to poor network connection. When this
was reported to MTN’s customer care, there response was that they could not do any-
thing because the provision to block such numbers was removed since the process pro-
vides many levels of approval for the agent and customer to identify such anomalies
254 R. Baguma
before completion of the process. In another incident at another service centre in Najee-
ra run by Mariam Nansubuga, a customer ran away before paying a deposit of UGX
900,000 that had been processed. When MTN’s customer care was contacted, they gave
the same regret as in the first case. The other incident involved a fish monger called
Mzee Zakayo who paid UGX 700,000 by mobile money to his supplier but sent it erro-
neously to a number that was switched off and the number was still off at the time of the
interview. In desperation, the agents have resorted to informal negotiation with affected
customers and other third parties to avoid mob justice and costly law suites. On the side
of customers, they are losing confidence in the service particularly those that have been
affected or have heard about the security incidents. Hinman and Matovu [4] noted that
one of the reasons why some people are not yet engaging with mobile money services in
Uganda is due to fear of losing money in the transaction among other factors.
Operation Tied to Unnecessary Constraints: For Airtel money, all transactions are
limited to availability of airtime on one’s phone yet charges do not come from air-
time. Users suggest delinking transactions from availability of airtime on one’s phone
after all this is not where the charges are made.
System Interface Does Not Render Well on Computer: Agents of M-sente com-
plained that accessing the system on a computer which some Agents prefer is a prob-
lem. That is it does not render seamlessly as it does on a phone. So the current interface
of M-sente is device dependent which is a poor user interface design practice.
3.3 Other Gaps
Unstable Network. The network especially that of MTN is sometimes on and off like
in the months of November 2011 to January 2012 and when it is off, business comes
to a standstill hence loss of profits. This situation especially for MTN is attributed to a
larger customer base compared to other MNOs. To make matters worse, there was no
dedicated support line for mobile money Agents/customers for MTN, Airtel and Wa-
rid Pesa at the time of the interview, and the customer lines shared with voice services
are always busy. Some of the responses from respondents include:
“When the network is off, we cannot work”
“Airtel’s network is limited to towns- There is no network deep in the villages”
“You find people in villages wanting to use airtel money but network is a prob-
lem”.
“UTL needs to improve network performance so that more people can use it.
“I am breaking even because I am dealing in other things i.e. it is not the only
business I am involved in because availability of the network is limited to towns.
Beyond, there is no network so people cannot use the service so we have very
few customers”
Limited Sensitization/Education of the Public about the Service: Agents com-
plained that service providers have not sensitized customers about the service which
would create more demand hence more profitability.
Change of Terms without Consultation: Agents accused MNOs of regular change of

terms of engagement without consulting them e.g. initially commissions on each trans-
action would be deposited on an Agent’s account daily but at the time of the study, de-
posits were being made monthly which made it hard to track deposits. Agents and
users would like to be more involved in the introduction of changes as key stakeholders.
3.4 Recommendations for Addressing the Design Gaps

Provide an Alternative Luganda Interface: MNOs need to consider providing a Lu-
ganda interface for mobile money in addition to the English interface as one of the strat-
egies to increase usage among the less literate and poor sections of the population
whose proficiency in English is limited. One Agent estimated that this could increase
usage to more than 5%. Another Agent shared his opinion that a luganda version/option
would make it easier for many to understand how the service works and how to use it.
Some less educated users receive transactional messages much later after the transaction
and have to spend time and money going back to the agent for interpretation. An option
for the interface in a language they understand would save them money and time.
Consider Non Text Interfaces such as Spoken Input and Graphical Output:
Medhi et al [8] in their study on designing mobile interfaces for novice and low litera-
cy users established that text interfaces are unusable without literacy. They noted that
non-literate subjects need non-text user interfaces such as spoken input and graphical
output. Mobile money service providers should consider this option to cater for the
user segment in the country that cannot read and write.
Visual Representation of Functions: Most people considered non-literate have some
level of reading ability for example they can read numbers and can recognize at least
a handful of symbols and words. But the design of most mobile services has not ex-
ploited this opportunity to make them more usable to them. To address this gap,
White [16] recommends a visual icon based user interface that is in turn supported by
voice prompts. The voice prompts help people navigate and learn the user interface as
well as guide the users through the tasks supported by the service such as sending
money. The prompts can also serve to provide confirmation of successful accom-
plishment of the given tasks.
Provide Mechanisms for Error Prevention and Correction at the Front End: Ac-
cording to Nielsen (20110), even better than good error messages is a careful design
which prevents a problem from occurring in the first place. The mobile money design
should incorporate effective error prevention and correction mechanisms such as options
to suspend a transaction mid-way or immediately after (with in a given window period)
in case of an error or fraud, blocking withdrawals for accounts pending security issues
etc. Some MNOs already have some measures in place. For example, Airtel keeps mes-
sage history and this function has made it more secure compared to others. In addition,
during the sending process, the sender is given information about the name of the person
he/she intends to send money to using registration information therefore in case of a
mistake made one is able to know and correct it before proceeding. With M-sente during
transacting, one is asked to enter a phone number twice and has to confirm the amount.
256 R. Baguma
Encrypt Pin: Interfaces of all MNOs except Warid use encrypted pins. Warid should
also implement use of encrypted pins during user login to prevent pin tapping from
wrong elements.
Keep Users Aware of What Is Happening during Interaction: Nielsen (2012) in
his Nielsen’s 10 Usability Heuristics advises that a system should always keep users
informed about what is going on. This can be achieved through appropriate feedback
within reasonable time which Nielsen (2012) calls visibility of system status.
Use Fewer Menus and Dedicated Buttons: Dix et al (2009) noted that the short term
memory capacity of a human being is limited and can only hold effectively 7± 2
chunks at a time. Hence the menu design of mobile money needs to keep the number
of menu items at 7 maximum for a more usable interface. Jones et al. (2000) ques-
tioned suitability of menu-based navigation for novice users of computer technology
which is the case with mobile money users and recommends designs with fewer me-
nus and dedicated buttons for this target group. This view is supported by Lehrman
[7]. Related to this, there is need to remove redundant functions and separate com-
bined functions into distinct functions to make them easier to use.
Better Sensitization to the Public: MNOs should design and execute better sensiti-
zation campaigns for the public about the concept of mobile money to increase under-
standing of the service. This could partly be inbuilt within the service’s interface.
Hinman and Matovu [4] noted that the source of many people not engaging with mo-
bile money is confusion with the mental model of the service. They observed that
most promotional materials for mobile money services focus on awareness and bene-
fits of the service. Hinman and Matovu [4] suggest that in order to fill the conceptual
gap of how the service works, experience with creating and maintaining a formal
account such as a bank account or account with a mobile service provider is required.
The two noted that to the people in rural Uganda, the concept of money transfer is
foreign but they do understand and have a wealth of experience with trading in partic-
ular exchanging assets that are equal in value e.g. exchanging a bull for a cow or a
piece of land for a number of cows. Therefore it is such concepts that can be used in
sensitizing the public rather than the abstract money transfer terminology. In Hinman
and Matovu’s [4] study, people who seemed to conceptually grasp how mobile money
services work were the ones who related the service to buying and selling airtime and
they noted that unlike transferring money, selling airtime mapped to their existing
behavior of liquidating fixed assets when cash was needed or gifting fixed assets to
people such as in the case of remittances.
Avoid Unnecessary Constraints in the Interaction Process: Airtel ties transactions
to availability of airtime on one’s phone yet this is not where the charges are made.
This constraint and any others in this category that the study may not have found out
should be removed to make the service easier to use for users.
Device Independence: Users of mobile money use a variety of devices ranging from
high end to basic mobile phones as well as tablets, laptops and desktops. Therefore
it is important to make the mobile money interface device independent such that it
renders seamlessly on a variety of standard devices.
MNOs Should Share and Learn Each Other’s Best Practices: instead of reinvent-
ing the wheel, MNOs should collaborate, learn from each other and put in practice
each other’s good practices. For example, some MNOs already have some effective
error prevention and correction measures in place such as keeping message history
and giving the sender information about the name of the person he/she intends to send
money to using registration information such that in case of a mistake made one is
able to know and correct it before proceeding by Airtel,. With M-sente, during trans-
acting, one is asked to enter a phone number twice and has to confirm the amount. In
addition, Agents have special dedicated numbers to call in case of a problem for quick
help. These are measures MTN, Warid and Orange are lacking. Warid pesa and M-
Sente could also learn and implement MTN’s menu driven interface to replace their
command line interface that users and agents find hard to use. In addition, Warid can
learn from the rest the benefits of encrypting the pin during login and learn from them
if necessary on how to implement it to provide more security to the users.
Other Design Considerations Suggested by Users and Agents:

• Current statements are too short e.g. MTN’s mobile money statement covers the
last 4 transactions only. Users recommended coverage of the last 10 transactions at
minimum.
• Bonus: Users would like to be given a bonus for example when one sends/receives
500,000 Uganda shillings and above.
• Increase limit: Users would like an increment to the limit transferable in a day from
2,000,000 to 10,000,000 Uganda shillings.
• Provide for overdraft of a certain limit in times of emergencies: Users recommend-
ed an allowance for overdraft of at least Uganda shillings 50,000 to cater for
emergencies
• Interoperability across networks: Users noted that the ability to transfer money
between networks and banks would make the service more flexible to use for them.
At the time of the study, only Warid had this service.
• Connection to Bank Accounts: Users would like to have their mobile money ac-
counts connected to their bank accounts. This they said would help them link mo-
bile money transactions with their bank transactions which would make overall
management of their financial transactions easier.
4 Conclusion
Mobile money is a great service to developing countries where the banking infrastruc-
ture is still severely limited and very few people can meet requirements of banking
institutions. The short time it has been in operation in Uganda has seen many
achievements but also there are some challenges including design related that are
threatening to stifle its usage among the people so much in need of it. This work has
highlighted the design gaps and possible ways these can be addressed. We hope these
insights will contribute to the growing body of knowledge around mobile money ser-
vices and help to improve mobile financial services in Uganda as well as in other
emerging markets.
258 R. Baguma
5 Limitations
The work reported in this paper is based on factual findings from a group of respon-
dents in the out skirts of Kampala city and Kayunga district. It is largely qualitative
giving a description of the current design gaps of mobile money in the country ac-
cording to the user experience of users and agents rather than a quantitative picture of
users and Agent’s opinions about the service. This approach was by design because
the goal of the study did not require opinions but facts about the design gaps that
could be verified independent of the participants.
References
1. Chipchase, J.: Understanding non-literacy as a barrier to mobile phone communication
(2005), 789 hjyuv7v78hjyuv878yuhvjjjjjjj877vssxWebarekungarukamu juba (retrieved on
May 16, 2012)
2. Duga, N., Getachew, H.: Mobile Services and ICT4D: To the Network Economy - Bridg-
ing the Digital Divide, Ethiopia’s Case (UCC, 2010) (2012)
3. Esselaar, S., Stork, C., et al.: ICT Usage and its Impact on Profitability of SMEs in 13
African Countries. Info. Technologies & Inter. Development 4(1), 87–100 (2007)
4. Hinman, R., Matovu, J.: Opportunities and Challenges for Mobile-based Financial Servic-
es in Rural Uganda. In: CHI 2010, Atlanta, Georgia, USA, April 10-15 (2010)
5. Herzberg, A.: Payments and banking with mobile personal devices. Communications of
the ACM 46(5), 53–58 (2003)
6. Hu, X., Li, W., et al.: Are Mobile Payment and Banking the Killer Apps for Mobile Com-
merce? In: 1st Hawaii Inter. Conf. on System Sciences. IEEE, Hawaii (2008)
7. Lehrman, S.: Dialing in. Mobile phones target the world’s nonreading poor. Sci.
Amer. 296(5), 30–31 (2007)
8. Medhi, et al.: Designing Mobile Interfaces for Novice and Low-Literacy Users. Designing
mobile interfaces for novice and low-literacy users. ACM Trans. Comput.-Hum. Inte-
ract. 18(1), Article 2 (April 2011)
9. Mendes, S., Alampay, E., et al.: The innovative use of mobile applications in the Philip-
pines-lessons for Africa. Sida Publications, Sida (2007)
10. Mugenyi, A.: Wanted: A Mobile Money Regulator. The CEO Magazine (June 2012)
11. Ndiwalana, A., Morawczynsk, O., Popov, O.: Mobile Money Use in Uganda: A Prelimi-
nary Study. In: Mobile for Development (M4D) Conference 2009 (2010)
12. Tentena, Mobile phone subscribers hit 14 million in Uganda. Business week June 20 ,
2012 (2011), http://www.busiweek.com/index.php?option=
com_content&view=article&id=1207:mobile-phone-subscribers-
hit-14-million-in-uganda&catid=104:uganda& (retrieved on September 10)
13. Uganda Demographic Health Survey (UDHS) ( 2011), http://www.ubos.org/
onlinefiles/uploads/ubos/UDHS/UDHS2011.pdf (retrived on June 20, 2012)
14. UNESCO, UNESCO Institute of Statistics Data Centre, regional literacy rates (2007),
http://stats.uis.unesco.org/unesco/TableViewer/
tableView.aspx?ReportId=y201 (retrieved on June 12, 2012)
15. UNICEF, Info by Country Statistics (2012), http://www.unicef.org/
infobycountry/uganda_statistics.html (retrieved on June 16, 2012)
16. White, G.: Designing Mobile Services for non-literate communities. In: Proceedings of
2nd International Conference on M4D Mobile Communication Technology for Develop-
ment (M4D) 2010, Kampala, Uganda (2010)
A Static and Dynamic Recommendations System
for Best Practice Networks
Pierfrancesco Bellini, Ivan Bruno, Paolo Nesi, and Michela Paolucci
DISIT-DSI, Distributed Systems and Internet Technology Lab,

Dipartimento di Ingegneria dell’Informazione, DINFO,
Università degli Studi di Firenze, Firenze, Italy
{pbellini,ivanb,paolo.nesi,paolucci}@dsi.unifi.it
Abstract. Semantics computing technologies may be used to provide recom-

mendations and stimulate user engagement in many kinds of services, such as
social media, match making, best practice networks, technology transfer, etc.
The recommendation metrics used take into account both static information and
dynamical behaviors of users on a Social Network Platform. The recommenda-
tions provided include those realized taking into account also strategic and ran-
dom users. The set of recommendations have been assessed with respect to the
user’s acceptance, which allowed to validate the solution and to tune the para-
meters. The experience performed in creating and validating recommendation
systems adopted for ECLAP and APREToscana best practice networks is de-
scribed and results obtained are reported. The identified model has significantly
increased the acceptance rate for the recommendation on ECLAP.
Keywords: best practice network, semantic computing, recommendations,

social media, grid computing, validation model.
1 Introduction
Semantics computing technologies may be used to stimulate user engagement in many
kinds of services, such as social media, match making, best practice networks, technol-
ogy transfer, etc. The semantic computing is typically confined on the server side to
provide recommendations. Despite to the massive success of social media, most solu-
tions have limited semantic computing capabilities and provide simple recommenda-
tions about possible friends and on marginally similar content items. Among the possi-
ble combinations of suggestions related to users, content, ads, and groups only some of
them are viable [1]. Recommendations should be computed on the basis of relation-
ships UU, GC, CU, etc. where U means User, G: Group and C: Content/Item,
thus CU means proposing Content suggestions to Users. The earliest solutions for
guessing users’ intentions have been based on keyword-based queries (i.e., sponsored
search, or paid listing), which places ads and/or recommendations in the search results;
and content match, also called content-targeted advertising or contextual advertising,
which places ads on the basis of the web page content and content similarity [2], [3].
Contextual recommendations are widespread and many systems can extract keywords
260 P. Bellini et al.
from web pages to produce suggestions [4], sometimes using semantic approaches [5].
In order to predict which terms describing a product or service are more relevant, mod-
els based on clustering, collaborative filtering, logistic regression, etc., are used [6].
User’s ranking and reputation are connected to recommendations and trust and are
becoming essential elements of web-based collaborative systems [7]. Implicit trust
networks have been employed to incorporate trust and reputation [8], obtaining trust
relations from a record of results in previous recommendations, by semantic reasoning
and inference mechanisms upon recorded data.
In this paper, we reported the experience performed in creating and validating rec-
ommendation systems for services including complex descriptors: (i) ECLAP
http://www.eclap.eu a best practice network and service derived from research tools and
solution for providing services towards the community of performing art institutions;
ECLAP includes about 120000 contents, 1900 users and 35 groups; (ii) APREToscana
http://www.apretoscana.org a best practice service for supporting industries and re-
search institution to match demand and offer and accessing to European commission
founding; about 1800 users, 15 groups. They need recommendation systems for CU,
UU, GU, CC in order to facilitate contacts. Contacts are consolidated by estab-
lishing stable connections among colleagues. Initially, both the above solutions where
set up with a recommendation system developed for a medical best practice network,
namely, Mobile Medicine [1]. The results obtained were not satisfactory since the con-
text of ECLAP and APREToscana were quite different and thus a study phase has been
started to reshape a more focused and tuned recommendation system.
To this end, a new model to compute similarities and propose suggestions has been
developed. The proposed model to present suggestions has been validated. The vali-
dation aimed to verify if the modality and the parameters used to propose the recom-
mendations where acceptable for the users in the domain. This model considers both
the user profile and the information extracted by analyzing the actions that users per-
form on contents in the recent time. The list of recommended friends/colleagues it is
not comprised only by the users they can more probably accept but is realized taking
into account also strategic and random users, basing on the serendipity philosophy.
2 Requirements Overview
In this section, the main requirements for the recommendation system that can be
adopted in best practice networks are discussed. The requirements have taken into
account the lesson learnt from the management of a number of thematically different
best practice networks, such as: ECLAP on performing arts, APREToscana research
and technology transfer, Mobile Medicine medicine and emergency situations, IUF of
CSAVRI (http://iuf.csavri.org) e-learning and new companies, etc.
The following requirements are referred to user to user (UU) recommendation
systems that have to provide reasonable suggestions to users basing on:
• both static and dynamic aspects of user behaviour, and of the content descriptors.
User profile static aspects may include: age, languages, sex, city, job, education,
preferred content, joined groups, etc;
A Static and Dynamic Recommendations System for Best Practice Networks 261
• both new and regular users, avoiding the problem of the Cold Start. The new users
often risk to do not have any recommendation since their static data are frequently
not compiled and dynamic data are not yet collected;
• the last performed dynamic actions of the users and progressively forgetting the
older (less recent) activities;
• the experience of the users that may have similar interests, intentions and/or tem-
poral evolution in the portal;
• stimulating the connection to users in successive activities on the portal by creating
side effects in their home pages (e.g., content posted, new connections, new
groups). This action allows increasing the mean number of connection per user,
and helps peripheral users to get connected with those with greater centrality, high-
er number of connections, etc.
• the complementary suggestions that could be unexpected for the recipient, but that
can be accepted if well motivated (e.g., he likes this music genre, he visited Paris
recently, etc.). So that, the system could learn from the acceptance of those new
connections about the user preferences, despite the lack of related content and of
similar colleagues;
• progressively estimate of the recommendations and not constraining the system to
perform the systematic recomputation of all of them at each change of the dynamic
aspects of the user profile and behaviour;
• an identification and use of the minimum number of parameters as a compromise
from computation and effectiveness. This means that: (i) the computational com-
plexity (the costs) of recommender system may be strongly influenced by the num-
ber of parameters taken into account; SVD/PCA and other statistical techniques
can be used to reduce them as much as possible; (ii) the acceptance rate of a rec-
ommendation system may strongly depend on the presentation of the recommenda-
tions to the users.
3 Recommendations Model
As previously mentioned, recommendations can be computed through several differ-

ent techniques. In most cases, the elementary operation is the similarity distance
among descriptors. The estimation of distances among elements can be computation-
ally expensive in the presence of complex descriptors and/or millions of items,
depending on the complexity and on the high number of the descriptors. In the follow-
ing, the model adopted in ECLAP is presented.
The user static profile consists of the data that change slowly over time [1]. In ge-
neralist social networks, the static profile is usually not very detailed: users do not like
fill form online at the time of registration and sometimes tend to provide false infor-
mation. In small thematic networks, however, this kind of information is much more
reliable. The static profile takes into account: general information (name, surname,
gender, date of birth, personal description, place of origin, spoken and mother lan-
guages), contact information (email and instant messaging contacts), school and work
(school level, name of school or university, type of employment, name of the place of
work) and interest (a list of categories of interest). Some data are coded by using spe-
cific standard like: ISO 3166-1 alpha 2 and ISO 3166-2 for place of origin, ISO 639-1
for spoken languages. Type of employment and categories of interest depend on the
portal usage domain. Moreover, the static profile considers also subscribed groups,
friends list and user’s interested taxonomy topics.
Table 1. Example of dynamic profile taxonomy based

User last N Corresponding collection of
Interactions on multilingual Taxonomy descriptor documents
Promoted Arti figurative Danza Balletto, Arti figurative Danza Balletto Russi,
Performing Arts Russian Ballet, Lettres Danser Ballet Russe…
Downloaded Performing Arts Modern Dance Performance Utopy, Un altre Recerca
Gènere Tema, Performing Arts Modern Dance Performance Utopy,
Andre Forskning Genre Om, Andere Forschung Genre Gegenstand…
Played Video Musicale Rock, Video Musical Rock, Video Music Rock, Vid-
eo Musique Rock…
Favorites Drama Gènere, Coreografia Rendiment Contemporani Dansa Arts
Escèniques període històric Arts Escèniques, Performing Arts Modern
Dance Performance Utopy, Drama Genre….
The dynamic profile is established on the basis of the actions the users perform on
the portal. In ECLAP, the dynamic profile considers four types of user interaction:
content seen by the user (played), user's favorite content (favorites), promoted and
downloaded content. The user profile is built by providing a hierarchical taxonomical
classification for each of the these categories considering the N last recently used
content for each of them. Table 1 shows an example of a user that in the last period
has watched several videos of rock music and promoted Russian ballets.
3.1 Users Proximity Evaluation
On the basis of the above mentioned aspects, the calculation of the proximity (prox)
between two users A and B is defined by a linear combination of the values of
proximity calculated for the static ( ) and dynamic ( ) profile aspects and
defined as follows:
, , , 1
Where: , respectively weight the relevance of the static and dynamic distance.
is defined by the eq. (2) as a function of static similarity between users rela-
tive to: spoken languages ( , locality ( , interests ( , common friends
( , subscribed groups ( ), age ( ), interested taxonomy topics ( ) as pro-
posed in [1].
,
, , , , , , , , , , , , , 2
The dynamic proximity is calculated using the value of similarity (score) pro-
vided by the system of indexing in Lucene and given by the Lucene’s Pratical Scoring
Formula in the multilingual documents created from document as described in Table
1. The dynamic profile is based on the multilingual document ( ) which is in-
dexed. In this way, the dynamic proximity distance is defined as:
,
, 3
max ,
The score provided is normalized using the maximum score given to documents re-
sulting from the query.
3.2 Other Recommendation Criteria

The static and dynamic aspects satisfy a part of the requirements but not all. They are
unsuitable to provide recommendations: (i) in the case of cold start (new users), (ii)
that may stimulate new and unconnected users to get in contact to strong reputation
colleagues, (iii) to strong users about new users that may need help in entering into
the community. To this end, additional recommendation types have been added:
• Strategic recommendations are those that recommend to users who have a few
colleagues those with highest number of connections and vice versa solving point
(ii) and (iii).
• Random recommendations consist in suggesting to users a random selection of
other users (perhaps with completely different to their own interests), driven by cu-
riosity to new content, can create contacts with new friends, expanding the list of
his interests and thus changing his dynamic profile.
As described in the following, the early validation presented in this document aimed
at assessing the acceptance level of the ECLAP users about the proposed recommen-
dations: static, dynamic, strategic, and random. The questions was: provided that the
recommendations are performed by presenting some rationales about the similarity,
which of the above mentioned aspects and recommendations would get the highest
relevance from the user point of view, thus stimulating them to get in connection.
4 Computational Architecture
The architecture of recommendation system (Figure 1) consists of: the ECLAP portal, the
ECLAP Storage Area and ECLAP Back Office, implemented by using AXMEDIS
AXCP tools [9]. The ECLAP portal is responsible for: (i) collecting user static and dy-
namic information and storing all data respectively in the Static and Dynamic Data repo-
sitory; (ii) creating users connections and store them in the User Relationship repository;
(iii) providing suggestions to users and (iv) providing a survey to get a feedback by users
for assessing and tuning the system. Finally, the ECLAP back-office is responsible of
evaluation recommendations according to the above presented model and executes on a

distributed system multiprocessor algorithms:
• Potential Friends: it calculates the static proximity between users, performing the
computations related to the new users, and renovating the computations for the less
recently updated. This approach keeps the estimation as much light is possible to
update the static aspects that slowly change over time.
• S.L.I.M. (Suggest Lucene Index Manager): it deals with the dynamic data by build-
ing dynamic profiles and indexing them in the Lucene engine. Also this process is
periodic and estimates the new version of the documents (Table 1) indexing them
incrementally, thus updating only the dynamic profiles of the most active users and
those that have significantly changed their dynamic descriptors.
• US.TER (User Suggester): it calculates the vector of UU recommendations by
using the above mentioned method: (i) eq. (2) as described in the following, by
considering value produced by Potential Friends and SLIM, (ii) random, and (iii)
strategic.
ECLAP Portal ECLAP Storage ECLAP Back Office

Actions:
play,
preferred,
promoted,
donwload
Static Data Potential Friendes
Users
Dynamic Data Dynamic

S.L.I.M. Profiles
User
Connections
User Relationship
Suggestions
Suggestions
Us.Ter
Similar,
Survey
Strategic and
Validation Random
Survey Data Suggestions
Fig. 1. Computational Architecture of the ECLAP Recommendation System
Typically, suggestions are provided with the 50% of kind (i), and 25% for kinds
(ii) and (iii). The number of recommendations presented to the user have to be typi-
cally a small part of the whole recommendations computed, leaving at the users the
possibility of taking more recommendations on demand.
The architecture is implemented as grid processes to take advantage from the dis-
tributed computing and to calculate/update progressively the needed data for generat-
ing recommendations. To have fresh suggestions, grid processes run as periodic
processes according specific schedules defined by the administrator.
5 Validation and Results
Before moving to the real analysis of results achieved a system for validating the
method. This allows tuning the system to match the user’s preferences providing them
better suggestions. A survey has been defined and posted on the portal. It proposes to
each user the profile of 10 potential friends: each user is called to provide an answer
to the question “Are you interested in getting contact?” by giving a vote (from 1 to 5)
to indicate how much he is interested in the connection. To justify why the system
asks to each user for (possible) potential friend, some motivations have been provided
in the survey, as shown in Fig. 2. Such motivations have been built by considering:
a) similar users: similarities are shown according to the user profile as described;
b) strategic users: are motivated according to their activity or how much they are
connected (or are not connected) to other users, etc.
c) random users: are randomly selected and motivated providing user profile details
such as: the list of groups to which the users are registered, the profession, and the
taxonomic classification of the last content viewed.
Fig. 2. Examples of the online survey (names have been obscured)
This should avoid that a user is faced with a recommendation without any details
and thus considering it as not relevant. The range of the votes is 1-5. It is possible to
consider the following three categories: useful (‘Yes I want the user as my col-
league!’) if the vote is 4 or 5; not useful (‘No I do not want the user as my col-
league!’) if the vote is 1 or 2; not relevant if the vote is 3 (‘I do not know’). In order
to estimate the model weights, a number of users have been involved into a learning
phase in which we presented to them set of potential colleagues to be voted. Since the
recommendations proposed were of two kinds, (a), (b) and (c) above, the validation
votes have been taken into account in different manners, as reported in the following
subsections.
5.1 Validation of Static, Dynamic Recommendations

This analysis aimed at estimating a linear model to the vote of users considering the
values of metrics. More precisely, the model is defined by the following equation:
, , · , · , ·
, · , · , ·
, · 4
Where γi coefficient weights the relevance of the distinct proximity factors. The num-
ber of votes collected largely exceeded 10 times the number of γi weights of eq. (4).
This allowed us to perform Multilinear Regression to estimate the weights as reported
in Tables 2 and 3. Table 2 indicates that the estimated regression model has been
confident: Rsquare represents the percentage variation in vote given by users explained
by the model, and it is the square of Rm and it indicates the quality of the predictive
model. F is the ratio of the variation of the votes that are explained and those that are
not explained by the metrics: it is the variation of residuals. Frelevance represents the
probability that the vote can be random compared to the values of the metrics. In
Table 3, γi weights, Stat-t and P-values of each metric are present. Stat-t is the ratio
between the single metric coefficient inside the model and its standard deviation.
Table 2. Results on the data collected Table 3. Multilinear Regression Coefficients analysis
Results (MR, first time) Metrics Coefficients (γi) Stat-t P-value

Rm 0.9624 -0.0047 -0.7312 0.4673
Rsquare 0.9262 νtax 0.006 0.3906 0.6974
F 131.7795 νage 0.0328 4.7163 1.3753E-05
Frelevance 2.3389E-33 νlang 0.032 10.2136 5.4563E-15
νloc 0.0408 6.0958 7.2905E-08
νgroups 0.024 3.9658 0.0002
P-value indexes how the votes, compared to each metric, are relevant in the model:
the higher is the Stat-t value and lower is the P-value. The analysis of the relevance of
each metric used for the generation of suggestions (Table 3), reveals that the relev-
ance of νtax and are rather low. For this reason, a new Multilinear Regression
has been realized and the results are reported in Table 4 and Table 5. The results hig-
hlighted that users selected their friends mainly on static aspects and less on the tax-
onomical modeling of the content. The data shows that the motivations that drive
users to tighten social connections are: age, spoken language and location.
Table 4. Results on the data collected Table 5. Multilinear Regression Coefficients analysis
Results (MR, second time) Metrics Coefficients (γi) Stat-t P-value

Rm 0.9620 νage 0.0324 5.0306 4.1064E-06
Rsquare 0.9254 νlang 0.0154 11.6651 1.3683E-17
F 201.4967 νloc 0.0131 6.1930 4.4967E-08
Frelevance 1.6585E-35 νgroups 0.0027 4.4234 3.7840E-05
5.2 Validation of Other Kinds of Recommendations

On the basis of the votes collected, the strategic suggestions have been considered
very interesting and useful by users, that in the 73,81% of time admitted to have been
convinced of getting in connection with the proposed users. On the other hand, ran-
domly provided users did not give a real stimulus to get connected. In fact, in this
latter case, the percentage of votes in the three categories have been very similar:
useful 32,58%; not relevant 37,08%; not useful 30,04%.
5.3 Impact of the Produced Model on Acceptance Rate of Recommendations

Once discovered these issues, the derived model has been adopted in ECLAP, substi-
tuting the previous model that was in place since 28 months. Therefore, from Novem-
ber 2012 up to now, the new model has been adopted, and use data have been
collected, in order to assess the users’ appreciation. From the data analysis, it can be
noted that the increment of the average number of accepted recommendations is of the
42%: ECLAP registered an averaged increment of 42% of the accepted recommenda-
tions per week in the period Nov 2012 - Feb 2013. Moreover, by comparing the same
periods: 'Nov 2011-Mar 2012' (old model) against 'Nov 2012-Mar 2013' (new model),
an increment of connections of 241% has been registered.
6 Conclusions
In this paper a recommendation system integrated in a collaborative best practice

portal has been described. The system provides to users a list of potential colleagues
based on both static information and dynamical behaviors of users. These aspects
have been used to define metrics that combined together allow to estimate a proximity
assessment between two users. The model includes a set of weights to define the re-
levance of the different metrics. The acceptance rate of the recommendation is not
only an aspect related to user proximity but also the manner by which the recommen-
dations are proposed. In this work, we have validated the model to propose different
kinds of recommendations: static/dynamic based, random and strategic. The assess-
ment performed allowed us to tune the recommendation system increasing the pre-
vious solution. The analysis reveals that the system has been useful to increase the
number of acceptance rate of the new connections and suggested the actions to be
performed in order to improve the recommendation system efficiency.
Acknowledgement. The authors want to thank Dario Nesi, Lorenzo Calosci for their
collaborations in the implementation of some of the tools and all the single people and
partners involved in ECLAP, and the European Commission for funding in the Theme
CIP-ICT-PSP.2009.2.2, Grant Agreement No. 250481.
References
1. Bellini, P., Bruno, I., Cenni, D., Fuzier, A., Nesi, P., Paolucci, M.: Mobile Medicine:
Semantic Computing Management for Health Care Applications on Desktop and Mobile
Devices. International Journal Multimedia Tools and Applications 58(1), 41–79 (2012)
2. Ciaramita, M., Murdock, V., Plachouras, V.: Semantic Associations for Contextual Adver-
tising. Journal of Electronic Commerce Research 9(1) (2008)
3. Anagnostopoulos, A., Broder, A.Z., Gabrilovich, E., Josifovski, V., Riedel, L.: Just-in-Time
Contextual Advertising. In: CIKM 2007, Lisboa, Portugal, November 6-8 (2007)
4. Zhang, Y., Surendran, A.C., Platt, J.C., Narasimhan, M.: Learning from Multitopic Web
Documents for Contextual Advertisement. In: KDD 2008, Las Vegas, Nevada, USA, Au-
gust 24-27 (2008)
5. Broder, A., Fontoura, M., Josifovski, V., Riedel, L.: A Semantic Approach to Contextual
Advertising. In: Proc. of 30th Annual Int. ACM SIGIR Conf. on Research and Development
in Information Retrieval, SIGIR 2007 (2007)
6. Bartz, K., Murthi, V., Sebastian, S.: Logistic Regression and Collaborative Filtering for
Sponsored Search Term Recommendation. In: EC 2006, Ann Arbor, Michigan, June 11-15
(2006)
7. Anderson, E.F., McLoughlin, L., Liarokapis, F., Peters, C., Petridis, P., de Freitas, S.: De-
veloping Serious Games for Cultural Heritage: A State-of-the-Art Review. Virtual Reali-
ty 14(4) (2010)
8. Martín-Vicente, M.I., Gil-Solla, A., Ramos-Cabrer, M., Blanco-Fernández, Y., López-
Nores, M.: Semantic inference of user’s reputation and expertise to improve collaborative
recommendations. Journal of Expert Systems with Applications: An International Journal
Archive 39(9), 8248–8258 (2012)
9. Bellini, P., Bruno, I., Cenni, D., Nesi, P.: Micro grids for scalable media computing and in-
telligence on distributed scenarios. IEEE Multimedia 19(2), 69–79 (2012)
Connecting Electric Vehicles and Green Energy
Peter Dempster
BMW Group Technology Office USA

[email protected]
Abstract. This paper discusses the interrelationship between the purchase of

green energy (GE) and electric vehicles (EV) and the motivations for and val-
ues formed around the purchase of the combination of the two. The BMW
Group completed a two-year EV and GE user study, a joint project with the
PH&EV Research Center at UCDavis and multiple market research agencies.
Through a focus group study of US East and West coast all-electric MINI E
markets and a new car buyer online survey, the BMW Group assessed the value
of offering GE with EVs. A follow-on user experience survey, assessed the
success of two GE products in the US. 39% of MINI E and 29% of Active E
respondents have purchased home solar. A vehicle design game revealed that
adding GE options increased overall demand for EVs among conventional buy-
ers by 23%, with technology interest, environment and cost savings reported as
the motivation.
Keywords: Electric Vehicle, Green Energy, Solar, Environment.
1 Electric Vehicles Deliver a Positive User Experience
Vehicles with plug-in electric drive technologies deliver a positive user experience,
according to published customer satisfaction surveys. More than 90% of BMW Ac-
tive E, Nissan Leaf and GM Volt customers report being satisfied or very satisfied
with their electric vehicle purchase. After two years with the car, 88% of respondents
of the MINI E survey said that they would purchase an EV within the five years fol-
lowing the field trial. After one year with the BMW Active E, 91% of users responded
“yes” to the question; “Do you plan on buying or leasing an electric vehicle in the
next three years?” according to preliminary data from a recent survey.
EV customers report numerous advantages over vehicles with conventional drive
trains. These include a quiet interior space, smooth and quick driving experience,
compelling environmental benefits, the ability to refuel at home, increased energy
independence and decreased fuel and operating costs. EV drivers form new values
such as the intersection of clean and fun-to-drive, the ability to expand an electric
driving territory and to master energy use, in and out of the car [1]. EV users discover
that they can be independent from an energy system which is potentially unsustaina-
ble and are able use electricity from a portfolio of clean generation technologies,
including GE, to fuel their car.
270 P. Dempster
2 BMW Group Electric Vehicle History
BMW Group has introduced 7 electric vehicle models since 1972, but only two EV
field trials have been coupled with academic user research. The first, an all-electric
MINI E field trial occurred from 2009 to 2011 with 450 vehicles in regular customer
hands on U.S. East and West coast markets. The focus of the accompanying user
study - which included multiple in-person and phone interviews, online surveys, focus
groups and vehicle data logging - was driving and charging behavior, gauging the
interest of the market in electric drive technologies and renewable energy, range pre-
ference and demands on the electricity grid.
BMW of North America is currently leasing 700 BMW Active E vehicles to cus-
tomers in Boston, Hartford, New York City, New Jersey, Los Angeles, Sacramento,
San Diego and San Francisco for a period of 24 months. The BMW Active E is
unique compared to past BMW all-electric offerings, like the MINI E. First, the active
E is equipped with a standard charging port (J1772), so customers can use tens-of-
thousands of public charging stations which greatly enhances everyday usability.
Second, Active E customers can purchase home solar through a preferred partnership
with Real Goods Solar or a wind power renewable energy certificate package through
Green Mountain Energy. Third, the Active E has an advanced telematics system,
ConnectedDrive, and a smart phone application, MyBMWRemote, which allows the
customer to connect to the car to see its status, like battery state-of-charge and
remaining range, and set charging and preconditioning timers.
3 EV Charging and the Environment
The environmental benefits of driving an EV are linked to the timing, duration and
location of charging. For example researchers at the Institute for Transportation Stu-
dies at UCDavis identified through an electricity supply model that emissions asso-
ciated with generation vary greatly with hourly demand and power plant availability
[2]. Green house gas (GHG) emissions associated with electricity generation also vary
greatly across geographic regions. For example, the US South, Midwest, and Western
Rocky states have more than 3 times greater EV GHG emission rates than the West
Coast and Northeast [3]. An electric vehicle charged in California is likely to produce
60% less well-to-wheel GHG emissions than a conventional gasoline vehicle [4] be-
cause of the mix of electricity generation sources. In 45% of US regions, regardless of
when it’s charged, an EV produces less well-to-wheel GHG emissions than even the
cleanest hybrid vehicles [5]. The BMW i-3 is expected to achieve a 50% reduction in
well-to-wheel greenhouse gas emissions largely due to the integration of GE in both
production and use phases of the vehicle.
When charged intelligently, electric vehicles can be grid assets, lowering electricity
costs for all consumers and total emissions from the electricity sector. A fleet of plug-in
vehicles can act as: 1. a virtual power plant, eliminating the need to turn on expensive
natural gas power plants and reducing costly transmission and distribution system up-
grades, which will lower the cost of electricity for all consumers, and 2. a large battery,
Connecting Electric Vehicles and Green Energy 271
storing wind energy which may be produced in excess during the night when load is at
its minimum or other renewable energy which is intermittent so that it can be used later,
which will lower total grid emissions and enable the installation of more GE.
An ongoing BMW Group partnership with a major European energy service pro-
vider, Vattenfall, is demonstrating that intelligent charging of an EV can reduce costs
and GHG emissions associated with EV charging. In a Berlin field trial, a cloud com-
puting algorithm calculates and sends a charging profile to the MINI E based on a
wind energy forecast, a customer mobility calendar, and battery state-of-charge and
state-of-health information stored on BMW backend servers. Users are encouraged to
plug in according to their mobility calendar through gamification of the smart phone
application. Through intelligent charging of the MINI E, GHG emissions are reduced
by 3%. Further economic and environmental benefits can be realized through a true
vehicle-to-grid scenario; Excess wind electricity is purchased during off-peak hours
and fed back into the grid during peak periods when electricity is often more expen-
sive and polluting.
The BMW Technology Office USA, located in Mountain View CA, recently
launched eMobility Lab, a new research and demonstration platform for the evolving
ecosystem surrounding sustainable electric mobility. eMobility Lab is focused on five
Fig. 1. BMW Group Technology Office USA solar and EV energy visualization tool
272 P. Dempster
key research areas: 1. understanding the interaction between renewable sources of

electricity and EV charging (Fig. 1), 2. communication pathways that allow the con-
sumer and utility to visualize and interact with vehicle charging and home energy use,
the “smart home”, 3. second-life opportunities for EV batteries, 4. consumer studies
and 5. electric vehicle grid services. BMW Technology Office USA is keeping an eye
on emerging trends in technology innovation and learning from consumers about
emerging markets through focus group and survey based studies.
4 Green Energy and EV Users
A survey of customers who participated in a two-year all electric MINI E US field

trial revealed a strong user preference for charging EVs with solar (100%) and wind
(98%) energy. The same survey of former US MINI E owners found that 37% owned
solar by the end of the field trial. This result was re-confirmed by a recent survey of
California EV owners (96% Nissan Leaf owners) which found that 39% have invested
in home solar systems (n=1419), helping to “fuel” their vehicles with renewable solar
energy [6]. Most recently, 29% of Active E respondents said that they have home
solar according to preliminary results of a recent survey. An additional 25% respond-
ed that they intend to purchase solar within the next three years.
In order to gain a better understanding of the connections that EV drivers make
with GE and the motivations for combining the two, BMW Group and the Plug-in
Hybrid and Electric Vehicle Research Center at the Institute for Transportation Stu-
dies at UCDavis conducted eight focus groups in Beverly Hills, CA, Manhattan, NY,
and Montvale, NJ in June of 2011. Participants were drawn from three sets of house-
holds: all-electric MINI E customers, people who bought GE through their local
utility (ex. LADWP Green Power Program), and people who owned conventionally-
fueled MINI vehicles and who had no obvious connection to GE or electric mobility.
By creating a comfortable environment in a focus group, the moderator enabled
MINI E customers to tell a story about their common experience with owning, driving
and charging an EV. The discussion began with MINI E users describing how they
came to be an EV owner - how the MINI E made them feel at the start of the field
trial along with initial like and dislikes - and moved to how they currently see the
MINI E, electric mobility and fueling an EV. The moderator then asked if MINI E
users knew about GE options available to them and if talking about GE enhanced
their feelings about and experience with the MINI E. The group discussion was con-
ducted four times in three locations so that the research team could identify themes
and patterns. These themes were incorporated into the design of the survey compo-
nent of the mixed-method GE and EV buyers study.
During the focus groups MINI E customer response to the combined offering of
GE and EVs was truly mixed. Some participants had already made large investments
in home solar photovoltaic (PV) systems or were participating in voluntary GE
programs. Leasing the EV was an extension of a sustainable lifestyle which included
GE.
I put SunPower panels on my barn so I have a 20-kilowatt solar system and the MINI
E was kind of a segue to that.
Some customers for whom home solar was not accessible had thought about other
innovative pathways for coupling solar or wind power and EV charging, like a power
purchase agreement or intelligent charging.
I can buy a package of green power, renewable power, power generated from renew-
able sources that [makes me] feel good that a very healthy portion of my driving - if I
didn’t have solar - is covered by a program that’s really easy for me to sign up on or
with to drive basically in a non-emission-generating way.
A third group decided that because EVs were so clean, it was not necessary to
charge them with GE. They believed that regardless of how it’s produced, using coal
or other sources, electricity must have a lower environmental impact than gasoline
burned in an engine.
It would have to be a really compelling way in which you packaged whatever it is...I
already feel great about driving the electric car. I don’t need to feel great about pay-
ing Southern California Edison more money for the proper source, for me, just me.
The nationwide new car buyer online survey applied themes and language taken
from the focus group sessions to assess consumer interest in PEVs and green electrici-
ty programs. Deployed in July of 2011, the survey contained an EV and GE design
game. BMW Group and a market research agency recruited three U.S. samples to
complete a web-based survey, including recent buyers of new conventional vehicles
(CVB, n=1064), hybrid vehicles (HEVB, n=364) and plug-in electric vehicles, includ-
ing the MINI E, Chevrolet Volt and Nissan Leaf (PEVB, n=74).
Respondents were asked to design a vehicle, choosing the make, model and engine
type (conventional, hybrid, plug-in hybrid or electric). They then designed a home
energy plan, choosing between no green program (or current “green” program if al-
ready enrolled), a monthly green program applied to at least 20 percent of home elec-
tricity use with a user defined electricity source, a 2-year green lease supporting wind
or solar power through the voluntary purchase of renewable energy certificates (and
requiring a two-year commitment) and purchasing and installing residential solar,
financed as one monthly bill for 20 years (a power purchase agreement or PPA).
The final section of the survey asked them to choose a vehicle and electricity plan
side-by-side with an option to cover EV electricity use with GE. Respondents were
asked to identify motivation for choices of vehicle, energy plan and combinations of
the two.
Respondents that already have “green electricity” are more likely to design PEV in
the game. Most hybrid and plug-in buyers design solar (18-35%), or already own
solar (32-37%). Adding GE options increased overall demand for PEV designs among
CVBs by 23%, HEVBs by 20%, and among PEVBs by 5% [7]. For PEVBs the most
frequently designed combination (38%) is an EV with home solar; 86% combined an
EV and GE and none designed a conventional vehicle and rejected GE (Fig. 2).
274 P. Dempster
Fig. 2. EV buy
yers combine vehicle and energy plan (source: [7])
Motivations for combiniing GE and EVs vary within car buying groups. They in-
clude environment, total co
ost of ownership, energy independence, renewable eneergy
support and control over ellectricity sources. Conventional and hybrid vehicle buyyers
are more likely to be motiv
vated by cost savings. EV buyers are more strongly m moti-
vated by technology intereest and as well as environmental benefits. Customers’ re-
ported motivations providee guidance on policy and marketing strategies to advaance
GE and EV demand.
5 Green E Producct Offerings
In 2012 the BMW of North America offered Active E customers two GE optioons.
One, a wind-energy renew wable energy certificate (REC) package offered throuugh
Green Mountain Energy, prrovides a two-year green energy plan - approximately $448 -
that corresponds to the ammount of energy an Active E uses during the lease, 64400
kWh, with a mechanism forr “topping up” at the end (Fig. 3). The second is a preferrred
partnership with Real Goo ods for a residential solar PV installation which is apppro-
priately sized for home andd EV. A portfolio approach to green energy allows BM MW
Group to meet all US EV cu ustomer needs.
The most recent results from
f the Active E user experience survey show an encoour-
aging take-rate for BMW GE G products offerings. The largest customer of the Grreen
Mountain Energy REC paackage is DriveNow, a 70 EV car sharing program m in
Fig. 3. Active E customers can purchase RECs through Green Mountain Energy
San Francisco. GE gives DriveNow a competitive advantage over other EV car shar-
ing programs and is expected to attract environmentally minded consumers, which is
why the GE program is front-and-center in marketing materials for the company.
According to the preliminary results of the Active E user experience survey, 5% of
respondents have purchased home solar through Real Goods and 3% of respondents
have purchased wind energy through Green Mountain Energy.
6 Conclusions
In a multi-year project exploring the consumer demand for EVs and GE, BMW Group
and numerous research partners completed a novel study, in three parts. The first, a
literature review, aided the design of a unique multi-methodology study employing
both focus group and survey research. The second, focus groups with MINI E drivers,
conventional MINI drivers, and GE program participants in Los Angeles, New York
276 P. Dempster
and New Jersey generated themes and language used in a larger nationwide online
survey and design game. The third, a nationwide survey of new car buyers, which
included subsets of hybrid buyers and recent EV lessees and owners, was used to
justify the addition of two GE products for the BMW Active E.
EV customer preference for GE is very strong, with 29% of Active E customers
reporting having home solar. One question remains: Is the BMW Group observing an
early market phenomenon or a long term trend and opportunity? The vehicle and
energy plan design game results illuminate the benefits to an auto OEM of a com-
bined EV and GE product offering. Though conventional vehicle buyers are moti-
vated by different reasons, fuel cost savings, combining EVs and GE increases the
overall demand for EVs by 23% in that new car buyer segment.
EVs and GE are transformative and disruptive because the combination represents
a disconnection from current liquid transportation fuel systems along with the de-
coupling of environmental impacts of vehicle mobility from on-board fuel consump-
tion. GE for the home and EV has the potential to lower total household energy costs,
which is appealing for new car buyers across all segments.
References
1. Turrentine, T.S., Garas, D., Lentz, A., Woodjack, J.: The UC Davis MINI E Consumer
Study. Institute of Transportation Studies, University of California, Davis, Research Report
UCD-ITS-RR-11-05 (2011)
2. Yang, C., McCarthy, R.W.: Impacts of plug-in electric vehicle charging. EM Magazine, Air
& Waste Management Association (2009)
3. Lutsey, N.P., Sperling, D.: Regulatory Adaptation: Accommodating Electric Vehicles in a
Petroleum World. Energy Policy 45, 308–316 (2012)
4. Delucchi, M.A., Lipman, T.E.: Lifetime Cost of Battery, Fuel-Cell, and Plug-in Hybrid
Electric Vehicles. In: Pistoia, Gianfranco (eds.) Electric and Hybrid Vehicles: Power
Sources, Models, Sustainability, Infrastructure and the Market. Elsevier, Amsterdam (2010)
5. Anair, D.: State of Charge: Electric Vehicles’ Global Warming Emissions and Fuel Costs
Savings Across the United States. Union of Concerned Scientists, Cambridge (2012)
6. Ferry, M.: California Plug-in Electric Vehicle Owner Survey. California Center for Sustain-
able Energy (2012)
7. Kurani, K.N., Axsen, J., Caperello, N., Bedir, K., Tyree-Hageman, J.: Consumers, Plug-in
Electric Vehicles, and Green Electricity. Presented at Plug-in Electric Vehicles and Clean
Energy in California, Sacramento, CA (2012)
SOCIETY: A Social Reading Application to Join
Education and Social Network Experience
Elena Guercio1, Fabio Luciano Mondin1, Maurizio Belluati1, and Lucia Longo2
1
Telecom Italia S.p.A. Torino, Italy
{elena.guercio,fabioluciano,maurizio.belluati}@telecomitalia.it
2
Politecnico di Torino
[email protected]
Abstract. This paper describes a social reading application designed for educa-
tion and school communities based on electronic book reading experience. Both
design and evaluation followed a user centered approach in order to obtain a
custom made application called SOCIETY, to be used by community based
groups with similar reading interests, running both on IOS and Android System.
A special emphasis was given on personal and social aspects in writing and
sharing notes with the community groups and the “facebook or twitter” large
community to better understand how that can improve the learning process.
Beside showing a high interest on society@school app our trial evidenced
the need to separate school community from wider social networks, a need
which seems to be important for students. That stresses the importance of priva-
cy management in school applications design, because students seem to be wor-
ried about sharing information and mixing school and social networks contexts.
Keywords: Social reading, privacy/respecting, User Experience, School, User

Interface, Personal, Social.
1 Introduction
The number of social network users is expected to grow in the next few years. Ac-
cording to a research performed by eMarketer [1] in 2011 there were 1.20 billion
people using social networks, number which is expected to grow to 1,85 Billion
through 2014. Social networks are making multimedia content sharing very easy and
funny, also due to the possibility of exchanging opinions and thoughts together with
digital content. On the content side, expenses related to digital content are probably
going to grow as well as delivery and access to digital content becomes easier and
easier. The 2011 PwC report [2] forecasted consumer spending for entertainment and
media industry to surge from $1.4 trillion in 2010 to $1.9 trillion by 2015. Digital
content certainly include e-books, and in particular Children and Young Adult’s e-
books, which according to the American Association of Publisher showed a massive
+475.1% increase from 2011 to 2012 [3]. E-book adds a new dimension to traditional
books: interactivity. E-Book enables users interaction both with the media and be-
tween users, creating and sharing notes and content on top of the e-book content.
278 E. Guercio et al.
Market data and technical innovation aspects seems to open interesting opportunities
in the education field, and that is the main reason standing behind the choice of de-
veloping SOCIETY (SOCIal Ebook communiTY) application: giving students the
possibility to add a note reporting their opinion, summary, or link to external content,
making the e-book ‘live’ and dynamic. A note may contain multimedia content (i.e.
audio, video, image), extending the original plain text book as a real and new multi-
media dimension.
An application to be used by students should be designed in order to be suitable for
this scenario, finding a proper trade-off between a “natural” wider use of social
networks and privacy concerns. Sharing content, especially in the education environ-
ment, implies also the definition of instruments and rules for identifying ‘who is
sharing what and with whom’ and for moderating the shared extra contents in order to
avoid misleading or offensive notes.
Moreover the integration with general social networks (i.e. Twitter, Facebook),
should be addressed carefully. Giving right to all user profile information just for
posting a message and sharing it on user preferred social network, could be enough
for reject the authorization to proceed.
2 State of the Art
2.1 Technical View
As it was said previously the e-book is more than a different way to read: it is a new
interactive experience. Reading a book on a tablet gives the user the possibility to
perform actions on the book, spreading from basic operations such as changing font
style or size, to advanced operations such as adding notes and multimedia content to
the original book.
The most popular reading applications, Apple iBook and Amazon Kindle, some-
way give the user such kind of interactivity. The user can underline text or add textual
notes and save them on the cloud in order to find them among their different termin-
als, but it’s not possible to interact with people reading the same book except for the
possibility to post a note as a new status on the chosen social network (i.e. Twitter or
Facebook). Other applications such as Readmill, Inkling and Kobo extends the possi-
bility to share text annotations among users owning the same e-book and discuss by
using comments.
Adding text notes to e-books is anyway just a small part of a wider story. Content
could be extended by adding multimedia annotations by giving the users the possibili-
ty to add audio, photo and video to be enjoyed using the application itself.
On the other hand, sharing multimedia content requires keen user identification in
order to avoid privacy and moderation issues. Teachers should be able to delete inap-
propriate contents or promote the best contributions as new teaching material for the
entire class.
Moreover, a social reading solution should be able to manage content protected by
Digital Right Management system in order to make it work also with bestsellers and
editorial books.
A Social Reading Application to Join Education and Social Network Experience 279
2.2 User Experience View

As it was also explained previously, compared to traditional paper books, eBooks
have several advantages [4]. Digital books can give the possibility to overcome some
reading diseases by means of customizable displays, page and font layout changes or
by combining text reading with auditory stimuli. E-books are also easy to be trans-
ported and always available. According to Abram, these features can revitalize the
Past, with a sort of culture and thinking restoration, by re-inventing books in a crea-
tive way, through video, animations and sounds.
In the article “Mobile phones in the Classroom: If You Can’t Beat Them, Join
Them’’, Scornavacca , Huff And Marshall [5], by collecting data about 1200 students
in a large undergraduate class, showed “that both students and instructors can benefit
from an additional channel of communication – SMS messages via mobile phones - in
the classroom. The lecturer perceived a gain of quality and quantity of feedback from
the students. Students indicated that the system was useful - making classes more
interesting and interactive”.
It was 2002 when Takeda e Suthers [6] tried to create a social reading web tool
(Pink), while in 2008 Kim, Farzan and Brusilovsky [7] evaluated an interface in a
classroom study, designed to allow the users to identify spatial regions in textbook
pages and use them for bookmarking, commenting, and sharing opinions with oth-
ers. This work stated that “annotation is a natural and beneficial activity for users
because it assists them in reading, writing, sense-making, indexing and retrieving”.
Both works showed how users seem to need online tools to support discussions and
learning by making artifacts, references and annotations in a easily way .
Other researchers, such as Kiili et al. [8] tried to identify different profiles for col-
laborative reading in order to find out different styles in sharing content and found
five different profiles: co-constructers, collaborators, blenders, individually oriented
readers, and silent readers.
This is the context in which Society try to take its place. Our prototype is based on
the idea that reading can be a social and shared experience and annotations help the
reader in text understanding, storage and organization and the reading experience will
be made more and more interactive and significant on mobile devices such as tablet
PC.
3 Application Description
With SOCIETY (Fig.1) users can share their reading experience with other readers
or books’ authors. For instance in educational environment they can share their learn-
ing experience with classmates and teachers by creating and commenting annotations
on schoolbooks both from home and from school, experiencing a new and more dy-
namic way of collaborative teaching and learning. The design of SOCIETY kept the
focus on allowing the re-use of well-known interaction processes such as notifica-
tions, text selection, to build a new interaction process in App by “recycling” the pre-
vious know how and by using it in a simple and quick way. So that users can choose
Fig. 1. Application screenshots
the right level of visibility of

o their annotation, decide if they have to be private, shaared
only with the author, within n a group, or within the whole community and larger social
network (i.e. facebook or tw witter) too.
SOCIETY was then dev veloped as a prototype App, keeping the user at the cennter
of the design process, and once
o developed it was tested with final users.
In designing SOCIETY the attention was kept on the actions performed by ussers
using a paper book trying to rebuild this user experience in a simpler, digital w way:
then, using SOCIETY, the user does not only read e-books, he also underlines text,
put notes beside the text, he sometimes places bookmarks, keeps it, lends it, taalks
about it with his friends.
4 User Requiremeents and Field Trial
In general e-books make learning

l easier through collaboration. The students offten
prefer giving their commen nts to a text by writing them due to “social” implications of
speaking with their teacherss and classmates.
In designing Society@school app we had to take into account both data com ming
from trials and literature. We
W also involved some teachers and the Director of an hhigh
school in Trento, through so ome open face to face interviews.
The first challenge, was to combine basic requirements (text customization, annno-
tation sharing, access from m smartphone and web, etc) with some more specific re-
quirements, such as having separate views for private and public notes was highlighht.
The extension of school contents towards large communities (facebook and twiitter
overall) seemed to be a very promising opportunity as “everyday students usuaally
study and do their homewo ork with Facebook open”. Social Network sharing couldd be
A Social Reading Appliccation to Join Education and Social Network Experience 281
important in order to createe a link between the students’ school and private life, stim
mu-
lating them this way to writte and share contents about school texts or reading bookks.
The App was developed both for IOS and Android with the perspective to reachh as
many device as possible, keeeping anyway the same approach to logical design of G GUI
information, accessibility, navigation,
n number and types of functionalities and tryying
to give a common look and feel to both apps.
Once developed, the app p was tested by means of a six- months field in Trentoo, in
the same school in which we w collected most of the requirements.
The users’ sample involv ved 42 students in two different classes and 3 teachers. TThe
trial started with s short presentation of the App to the students and teachers bothh to
explain the App focal pointts and usages, and to support the users with the installattion
on their devices.
Afterwards, we constanttly monitored the app usage in different contexts with dif-
ferent teachers, trying to co
ollect as many information as possible from the anonym mous
log files, but also getting ex
xplicit and direct opinions or comments through a speccific
in-app feedback functionaliity. Finally we used a poll to get a general opinion abbout
the whole trial and the App p’s functionalities and performances. In order to stimulate
feedbacks a specific web portal
p section was developed for the description of the de-
velopment team actions relaated to users’ feedback.
The data collected duringg the field trial could be classified as:
• spontaneous input and su uggestions from the students during the App interactionn by
email and a specific App p software button
• students log files (accesss number, more clicked functionalities, etc…)
• group interviews (about 10 in 2 classes) with students (Fig. 2)
• 3 face to face interviews with teachers (Fig. 2)
• interviews with students’ parents (6)
• final questionnaire
Fig. 2. Face to face interrviews with students (on the left) and teachers (on the right)
We used this information too evaluate the usefulness of the App in the school conttext,
but also to optimize and re-design
r the App and to make it users-tailored. A lott of
interesting suggestions weree collected, the most important were:
• Notes were not enough: comments over notes were introduced

• explicit highlighting of thhe notes’ cited text.
• Number of notes and com mments shown in library section of the app.
• Visibility of notes on claass basis.
Unexpectedly “sharing actiivities to the large communities of Facebook and Twittter”

(i.e. not default sharing off notes towards community) from a student perspecttive,
should be kept apart from the friends community making them two worlds to bee at
least interconnected but nott tied.
Students asked for the same
s social approach and functionalities of their preferrred
social network, but they ad dopted a different behavior during the trial. They usedd to
share school data and expeerience just inside the class and use Facebook and Twiitter
for extra scholastic experiiences: independently from the fact that they receiveed a
positive mark from the teaacher or even if a note had a lot of comments from the
classmates.
We also offered the student a “sign up” functionality via Google, Facebook and
Twitter, in order to simplify
y the registration process, but in nearly every case studeents
weren’t interested in using it.
i
User were concerned abo out the number and nature of grants and permission to ggive
to the app on Facebook or Twitter.
T It's important to ask for the minimal and necesssary
grants in merging the personal and school “environments”. Asking access to the enntire
user profile information jusst for posting a single message (i.e. user status) seem too be
a strong concern for the ussers and should then be avoided. It’s necessary insteadd to
ask for the minimum possib ble number of grants and explain well why you are askking
for it.
To make clear this desig
gn choice for the user, we put the default to the direct ccon-
nection to Facebook and Tw witter “off”.
In general the survey showed a satisfactory experience (Fig.3) with Soccie-
ty@school: the “vote” of thhe app was 7.3 on 1 to 10 scale. App and the shared suggges-
tions and optimizations on App proposed both by teachers and pupils were taken iinto
account in the new App verrsion.
Fig. 3. Users’ comments on Society@school App

A Social Reading Appliccation to Join Education and Social Network Experience 283
5 Conclusions and
d Next Steps
Our trial results show us hoow high has been the interest on society@school app, as the
figure n. 4 shows (Fig. 4).
Fig. 4.
4 Final evaluation of society@school app
Over 85% of users have expressed a positive opinion on it, highlighting the addded
value to the traditional learrning. Trialists have also judged the app very user frienndly
and easy to use. In fact, in the interviewed sample, 82% of them have evaluated eeasy
to read notes, 71% easy to share
s them, and still 71% easy to insert new one or edit iit.
Very important is the “diiscovery” that, more for students than for teachers, a sepaara-
tion between school commu unity and the wider social networks community is necessary.
Thanks to the log files we traced, we have found that only a few percentage of ussers
(about 10-15%) has linked own society@school account to the Facebook or Twiitter
profiles. Moreover, even among those who have linked society@school profilee to
Facebook or Twitter, few users
u have posted their own society@school activities. T This
finding suggests how imporrtant is the privacy management in school applications de-
sign, because people, especiially teenagers, seem to be very cautious and worried abbout
sharing information and mix xing school and social networks contexts.
Our trial has been also very useful to optimize our app, adding new features re-
quested by users such as the opportunity to comment others’ notes or to highlight ttext
you are reading. Then abou ut privacy, the “shrewdness” we adopted:
• the new inserted note has to be checked by the student before be shared with othhers
(instead of that the note is
i just “private”)
• then, in a second step, student
s can decide if sharing or not his note on Faceboook
and Twitter
The next step of this experrimentation is to extend the trial to other primary and ssec-
ondary schools (Bolzano, Bologna
B and Caltanissetta are starting at the time of w
writ-
ing), to other classes in the same school in Trento and also to some university courrses
and community environmen nts.
The new trials should focus also on the Android App (which was designed follow-
ing the same device principles, but it has not been evaluated on the field yet).
It could also be useful and valuable to insert some specific didactical instruments
(cognitive maps, vocal recording or reading, test customized formats, etc…) in order
to extend the app with some tools helping dyslexic students both at school or home
with the final aim of minimizing the individual difference in reading and learning for
the student suffering diseases. In a context like the Italian school in which the num-
ber of students suffering DSA problems is increasing, social reading could be a way
to compensate some learning diseases. At the time of writing we are collecting re-
quirements in order to extend the app with tools for students suffering DSA.
References
1. http://www.workinghomeguide.com/10763/huge-social-network-
users-growth-worldwide-by-2014-facebook-reinforce-dominance
2. http://blog.highbeambusiness.com/2011/08/pwc-report-digital-
content-spurs-global-consumer-spending-growth/
3. http://www.publishers.org/press/62/
4. Stephen, A.: P-Books vs. e-Books: Are there education issues? Multimedia and Internet @
Schools (November/ December 2010)
5. Scornavacca, Huff, Marshall: Mobile phones in the Classroom: If You Can’t Beat Them,
Join Them. Communications of the ACM 52(4) (April 2009)
6. Suthers, T.E.: Online Workspaces for Annotation and Discussion of Documents. In: ICCE
International Conference on Computers in Education (2002)
7. Kim, Farzan, Brusilovsky: Social Navigation and Annotation for Electronic Books
8. Kiili, et al.: Working on Understanding During Collaborative Online Reaing. Journal of Li-
teracy Research 44(4), 448–483 (2012)
The Evaluation of a Voting Web Based Application
Linda Harley, Keith Kline, Jerry Ray, Carrie Bell, Andrew Baranak,
Chandler Price, Matthew Hung, and Brad Fain
Georgia Tech Research Institute, Atlanta, Georgia, USA

{linda.harley,keith.kline,jerry.ray,carrie.bell,andrew.baranak,
chandler.price,matthew.hung,brad.fain}@gtri.gatech.edu
Abstract. Ballot layout and the incorporation of assistive technologies into vot-
ing systems are plagued with inconsistencies across the United States. The
purpose of this study was to evaluate both ballot layout display configuration
(information density) and a variety of controllers (e.g., mouse, 2- or 5-button
controller) in order to assess performance and preference among voters. Par-
ticipants were presented with three mock ballots, each with different layouts
(scrollable pages, multiple columns or multiple pages per contest). Eye-tracking
data and selection time data were recorded and a usability questionnaire was
administered after each testing condition. The results of the study found that
participants preferred the multiple column display configuration and the use of
the mouse. The results from this study will be leveraged to design an iPad Vot-
ing Application with appropriate interfaces and controls. This will allow indi-
viduals with disabilities the opportunity to vote without requiring the dexterity
to use a paper and pencil ballot.
Keywords: vote, web based application, human factors engineering, interface.
1 Introduction
The goal of the Military Heroes Initiative and Military Voting project is to better un-
derstand and improve voting technology and voting processes that affect recently
injured military personnel who have returned from a combat zone with one or more
disability. According to statistics compiled by the U.S. Department of Defense, as of
February 15, 2013, 50,476 U.S. troops have been wounded during the conflicts in Iraq
and Afghanistan, many returning home with a range of disabilities including loss of
limbs, impaired vision, and traumatic brain injury [1]. These individuals may have
difficulty getting to the polls to vote or even casting their vote via absentee ballot due
to their functional and cognitive limitations.
To provide alternative solutions for absentee voting, GTRI developed a web-based
voting application test-bed (Voting App), for use with smart technologies such as a
smart phone or portable computer. This technology could facilitate obtaining an ab-
sentee ballot from an individual who is incapable of visiting the polling station. The
purpose of the Voting App is to provide an alternative way of completing the absentee
286 L. Harley et al.
ballot for those individuals who do not have the physical or mental capability to com-
plete a standard paper and pencil absentee ballot. In the Voting App test-bed two va-
riables were considered: (1) information density and (2) controllers.
1.1 Information Density
With an increase in digital reading, from reading on computers at work to leisurely

eBook reading, finding an optimal presentation of information is important. To date,
research investigating the optimal use of line length, multiple columns, and text justifi-
cation is inconclusive. Longer line lengths result in faster reading times [2-5], but re-
search suggests medium to short line length may result in greater comprehension [6-7].
Objective results support the use of both single columns of text [4], and multiple short
columns [8] but subjective preference seem to be multiple short columns [4], [9-10].
Alternatively information can be displayed on multiple pages or on a scrollable
page. Paging refers to navigating between pages where all text is replaced by a new
screen full of information. Scrolling involves one page where text is replaced when
new text appears on the screen. The objective is to ensure that individuals can move
from page to page as efficiently as possible. If designers are unable to decide between
paging and scrolling, it is recommended that they provide several short pages rather
than one or two long pages that involve significant scrolling. However, with pages
that have fast loading times, there may be no reliable difference between scrolling and
paging when people are reading for comprehension [11]. The results of this study
should help provide enough information to make confident decisions as to the appro-
priate layout of information.
1.2 Controllers
The most complete collections of hand anthropometry data have been collected by the
US Army [12-13]. Studies have shown no significant difference in anthropometric
hand measures between people with and without disabilities [14] or due to aging [15].
However, more longitudinal data are needed to fully confirm the aging effects. Where
the considerable differences are generally seen is in the areas of strength and motor
control, for example, with age there are considerable decreases in strength, dexterity,
precision, coordination, joint mobility and sensitivity [16-17].
Push buttons have been shown to have a small space claim and are easy to operate
[18]. According to recent occupation injury data, finger injuries are the most prevalent
of upper extremity injuries [19]. While performance with controls such as knobs and
dials are severely affected, push buttons show no performance difference between
disabled and able-bodied users [20]. To better determine differences in button confi-
gurations two separate push button configurations were created for this test. The first
of these configurations used two buttons, allowing for a tab and select control (Fig. 1.
(A) 2-Button and (B) 5-Button configuration). The other configuration was a five
button system, which allowed for tab, shift + tab, left (mapped to previous page
command of the Voting App), right (mapped to the next page command of the Voting
App), and select (Fig. 1. (A) 2-Button and (B) 5-Button configuration).
The Evaluation of a Voting Web Based Application 287
1.3 Hypotheses
The purpose of this study was to examine information density and controllers for the
Voting App.
Aim 1
Investigate the effect of different information density displays for the Voting App.
Hypothesis 1.1: The null hypothesis is that there is no significant difference between
scrollable (scroll), multiple column (columns) or multiple page (pages) display
configurations.
Aim 2
Investigate the effect of different controllers for the Voting App.
Hypothesis 2.1: The null hypothesis is that there is no significant difference between
the mouse, 2-button and 5-button controls.
A B
Fig. 1. (A) 2-Button and (B) 5-Button configurations
2 Method
2.1 Participants
Eighteen (18) participants (26±13 years of age, 12 males and 6 females) volunteered
for this study. Participants were 18 years of age or older and considered to be healthy.
The Georgia Institute of Technology Internal Review Board approved the study.
2.2 Experimental Setup
Participants were asked to sit comfortably behind a desktop computer (Fig. 2. Partici-
pant using the 2-button controller to navigate the Voting App at the eye-tracking sta-
tion.). The SmartEye system was used to track the participant’s eye movements in
order to determine areas of fixation on the screen. Three SmartEye cameras (Basler
acA640-100gm cameras with 8mm lenses), each with two IR flasher devices, were
used to track eye movement at a sampling rate of 60Hz. Three controllers were used
in this study: (1) mouse, (2) 2-button controller (Fig. 1. (A) 2-Button and (B) 5-Button
configuration) and (3) 5-button controller (Fig. 1. (A) 2-Button and (B) 5-Button con-
figuration). The buttons (Enabling Devices Compact Switch #745) were connected to
a switch interface device (X-keys XSI-38-US). The controllers interfaced the 3.5 mm
switch plugs to the PC via a USB port, and facilitated programming the switches with
the desired keyboard inputs. These buttons were mapped to standard keyboard keys,
which were then encoded in the Voting App.
Fig. 2. Participant using the 2-button controller to navigate the Voting App at the eye-tracking
station
2.3 Experimental Groups

Participants were pseudo-randomly assigned to one of three groups so that age and
gender were balanced between groups. This study used a 3x3 factorial design, with
the different controllers as a between-subjects group variable and the different display
configurations as a within-subject variable. Each participant, assigned to only one of
the three controllers, was asked to complete the mock ballot under each of the three
display configuration conditions. The order in which the display configurations were
tested was counterbalanced across participants by Latin square to eliminate order
effects.
2.4 Task Protocol
After consent had been obtained, participants were asked to sit at the SmartEye com-
puter while the cameras were calibrated to their individual anthropometric characte-
ristics. After calibration, the participants were instructed on how to use the controller
and presented with one of the ballot configurations. Participants were instructed to
experiment with the ballot and take their time. After completion of a ballot, partici-
pants completed the System Usability Scale questionnaire. Then participants were
given a few minutes to rest, while the SmartEye system saved the data. The process
was then repeated using a different ballot configuration. At the completion of the
study, participants were asked several interview questions to determine any additional
difficulties they may have encountered.
2.5 Data Analysis
Time Dependent and Userr Selection Variables

The Voting App contained a built-in function that logged each keystroke and selecttion
made by participants along with
w the time stamp. The following variables were calculatted:
• Ballot Duration: The tottal time it took the participant to complete the ballot. Tiime
started when the particip pant clicked on the “Start” button and stopped when the
participant clicked the “S
Submit” button.
• Mean Click Time: Click time is defined as the time between selections made by the
participant. Click time was averaged across the entire ballot to calculate M Mean
Click Time.
• Overvote: An overvote occurred when the participant attempted to vote for m more
participants than they were
w allowed to for that particular contest. The numberr of
overvote’s that occurred for the entire ballot was tabulated.
• Undervote: An undervotte occurred when the participant attempted to vote for lless
candidates than they werre allowed to for that particular contest. The number of un-
dervotes’ that occurred for
f the entire ballot was tabulated.
• Back Button: The numbeer of times the “Back” button was selected (Fig. 3. Screeen-
shot of ballot layout).
• Help Button: The numbeer of times the “Help” button was selected (Fig. 3. Screeen-
shot of ballot layout).
• Review Button: number of times the “Review” button was selected. This buttonn al-
lowed participants to reviiew their entire ballot (Fig. 3. Screenshot of ballot layout)).
• Next Button: The numb ber of times the “Next” button was selected. This buttton
allowed participants to go
g to the next page (Fig. 3. Screenshot of ballot layout).
Fig. 3. Screenshot of ballot layout

Eye Tracker Dependent Variables

The eye tracker system was used to determine how much time participants spent
viewing several regions of interest on the Voting App. The following variables were
recorded and analyzed using the SmartEye software. The look zones for the study
began recording data when the participant was presented with the first page and
stopped recording once the user submitted the ballot. The time spent viewing each of
the following regions of interest were monitored (Fig. 3. Screenshot of ballot layout):
(1) Candidates and referendums, (2) Title, (3) Voting Instructions, (4) Page, (5) Back,
(6) Help, (7) Review, (8) Next.
3 Results
3.1 Time Dependent and User Selection Variables
A two-way ANOVA was run with display configuration (pages, column and scroll)
and controls (5-button, 2-button and mouse) being the independent variables. The
statistical analysis found no significance between the independent variables for the
ballot duration, mean click time, number of undervotes, or number of times that the
“Help” button was selected (p>.05 for all comparisons). However, a significantly
greater number of overvotes occurred for the pages display configuration when com-
pared to both the column and scroll display configurations, F(2,30) = 8.73, p = 0.001.
This is supported by the fact that the number of times the “Back” and “Next” buttons
were selected was significantly greater for the pages display configuration when
compared to that of the column and scroll configurations, F(2,30) = 80.78, p<0.001.
For the controllers, the only significance among conditions was that the “Review”
button was selected more frequently for the 2-button controller than the 5-button and
mouse controllers, F(2,15) = 5.98, p = 0.012. There were no significant interactions
between display configurations and controls.
When considering the difference between display configurations, these results sug-
gest that many more errors occurred for the pages display configuration than the col-
umns and scroll configurations. This resulted in a greater amount of overvotes occur-
ring, meaning an increased navigation between pages as far as the number of times
that the “Next” and “Back” buttons were selected. One possible explanation may be
that cognitive load was high for the pages display configuration, since participants
were using working memory to recollect how many candidates they had selected for
the contest. When evaluating the difference between controllers, the only result of
note was that participants in the 2-button condition more frequently selected the “Re-
view” button. This strategy may have assisted participants in avoiding unnecessary
navigation between the pages, as the 2-button controller only allowed for forward
linear advancement and selection. Although not significant, participants using the
mouse had the shortest ballot duration and mean click times. These results indicate
that the most optimal controller was the mouse and that the pages display configura-
tion was not an optimal display configuration. There were no significant differences
between scroll and column display configuration.
3.2 Eye Tracker Dependent Variables

A two-way ANOVA was run with display configuration (pages, column and scroll) and
controls (5-button, 2-button and mouse) as the independent variables, and time spent
viewing regions of interest on the voting app as the dependent variable. The statistical
analysis found no significance difference between the independent variables for the time
spent viewing the “Candidates and referendum”, “Voting instructions”, “Page num-
bers”, “Back”, “Help”, and “Next” buttons. However, participants spent a significantly
greater amount of time viewing the “Title” for the pages display configuration when
compared to the column and scroll display configurations, F(2,20) = 10.99, p < 0.01. In
addition, results showed that participants spent more time viewing the “Title” for the 5-
button controls than the 2-button controls. In fact, the time participants viewed the
“Title” for the 2-button controls was small and negligible when compared to the other
conditions. This may be due to the linear nature of the navigation.
These results support that of the previous section, in that participants in the pages
display configuration had to pay much closer attention to the title of the page to
understand which contest they were on.
3.3 Subjective Rating Results
System Usability Scale

A 2-way analysis of variance was conducted on the System Usability Scale data, with
display configuration as a within-subjects variable and controls as a between-subjects
variable. The data are shown in Fig. 4. Subjective ease of use ratings obtained with
100
90
System Usability Scale
80
70
Ease of Use:
60
Column
50
Pages
40
30 Scroll
20
10
0
2-button 5-button Mouse
Fig. 4. Subjective ease of use ratings obtained with the System Usability Scale. Error bars
represent 95% confidence intervals.
the System Usability Scale. Error bars represent 95% confidence intervals The effect
of display configuration was significant, F(2,30) = 7.99, p = 0.002. The effect of con-
trols was not significant, F(2,15) = 3.15, p = 0.07. The interaction between controls
and display configuration was not significant F(4,30) = 1.82, p = 0.09.
Post-hoc paired t-tests were conducted on the display configuration conditions. The
critical p-value was adjusted to 0.0167 for multiple comparisons. Ratings for the col-
umn condition were higher than those in the pages condition, t(17) = 4.95, p<0.01, but
ratings for the column condition were not significantly higher than those in the scroll
condition, t(17) = 1.02, p = 0.32. Ratings for the scroll condition were not significant-
ly higher than for the pages condition, t(17) = 2.30, p = 0.03.
4 Discussion
The purpose of this research was to examine different information density display
configurations and different controllers for the Voting Application. The first null hy-
pothesis was that there would be no significant difference between the scroll, columns
or pages display configurations. The results of this study suggested that there was no
significant differences between display configurations when considering the length of
time it took participants to complete the ballot or the mean time between selections.
However, the number of overvotes that occurred during the pages display configura-
tion was significantly higher than the columns or scroll display configurations. Partic-
ipants also selected the “Back” and “Next” navigation buttons a significantly greater
amount of times for the pages display configuration when compared to the other con-
ditions. This implies that perhaps the cognitive load on recalling vote selections as
well as understanding what contest users were viewing was too complex. This is sup-
ported by the fact that participants spent an increased amount of time looking at the
“Title” of the contests for the pages display configuration than the other conditions.
Moreover, results from the System Usability Scale showed that participants had diffi-
culty with the pages display configuration. Participants reported that the pages dis-
play configuration was by far the most difficult, confusing and memory intensive.
There were no significant differences in performance of participants between the
scroll and column display configurations. Therefore, the research concludes that the
pages display configuration is the least optimal solution for presenting information on
a ballot, and that either scroll or column display configurations should be used.
The second null hypothesis was that there would be no significant difference be-
tween the mouse, 2-button, and 5-button controls. Participants utilized the “Review”
button more often in the 2-button condition when compared to the 5-button or mouse
conditions. This approach may have been useful in avoiding cycling through all the
contests to select only those contests in which they were interested. Recall that the 2-
button controls only allowed for one-directional (tab-forward) movement throughout
a page. Furthermore, participants spent little time looking at the “Title” for the 2-
button control (492±225 milliseconds). If participants navigated to the contest from
the review page, they would not need to read the title, as they would already know
what contest they were going to vote on. Participants reported that the 2-button
control was cumbersome. The mouse control was the easiest to use, which is likely
due to the fact that participants are already familiar with this technology. These re-
sults lead us to conclude that the 2-button control is the least optimal solution, while
the 5-button control is mediocre and the mouse is preferred. A more optimal button
solution may be 3-button controller, since it will allow forward and backward naviga-
tion as well as a selection button. This would be less demanding than the 5-button
controller and more flexible than 2-button controller.
One of the main goals of the Voting App test-bed is to utilize it to evaluate a varie-
ty of long-standing issues being discussed in the voting world, such as the use of plain
language in instructional content, font types, cognitive load, symbolism, and so forth.
In addition to the Voting App test-bed, research staff is also in the process of design-
ing an iPad hard case with a 3-button interface to be used in testing.
References
1. U.S. Casualty Status for OI, OND, and OEF,
http://www.defense.gov/news/casualty.pdf
2. Duchnicky, R.L., Kolers, P.A.: Readability of text scrolled on visual display terminals as a
function of window size. Human Factors 25, 683–692 (1983)
3. Dyson, M.C., Kipping, G.J.: The legibility of screen formats: Are three columns better
than one? Computers and Graphics 21(6), 703–712 (1997)
4. Dyson, M.C., Kipping, G.J.: The effects of line length and method of movement on pat-
terns of reading from screen. Visible Language 32, 150–181 (1998)
5. Dyson, M.C., Haselgrove, M.: The influence of reading speed and line length on the effec-
tiveness of reading from screen. Int. J. Human-Computer Studies 54, 585–612 (2001)
6. De Brujin, D., De Mul, S., Van oostendorp, H.: The influence of screen size and text
layout on the study of text. Behavior and Information Technology 11, 71–78 (1992)
7. Chaparro, B., Baker, J.R., Shaikh, A.D., Hull, S., Brady, L.: Reading online text: A com-
parison four whitespace layouts. Usability News 6(2) (2004)
8. Lam, K., Lam, Y., Liu, J., Shin, U.G.: Reading comprehension and rate: one column ver-
sus three columns. Dept of Computer Science Class Project, Univ. of Maryland, College
Park (2000)
9. Bouma, H.: Visual reading processes and the quality of text displays. In: Grandjean, E.,
Vigliani, E. (eds.) Ergonomic Aspects of Visual Display Terminals. Taylor & Francis,
London (1980)
10. Andrevey, V., Martynov, A.: Effects of splitting text into multiple columns. Dept of Com-
puter Science class Project, Unv. Of Maryland, College Park (2000)
11. U.S. Department of Health and Human Services: The research based web design and usa-
bility guidelines (2006)
12. White, R.M.: Comparative anthropometry of the hand. U.S. Army Natick Research & De-
velopment Laboratories (1980)
13. Grenier, T.M.: Hand anthropometry of U.S. Army Personnel. U.S. Army Natic Research,
Development, and Engineering Center (1991)
14. Feeny, R.: Specific anthropometric and strength data of people with dexterity disability. In:
Consumer and Competition Policy Directorate: Department of Trade and Industry (2002)
15. Molenbroek, J.F.M.: Anthropometry of elderly people in the Netherlands; research and
applications. Applied Ergonomics 18(3), 187–199 (1987)
16. Daams, B.J.: Human force exertion in user-product interaction. Doctoral Thesis, TU Delft,
Amsterdam (1994)
17. Haigh, R.: The ageing process: a challenge for design. Applied Ergonomics 24(1), 9–14
(1993)
18. Van Cott, H.P., Kinkade, R.G.: Human engineering guide to equipment design. United
States Dept. of Defense, Joint Service Steering Committee (1972)
19. Bureau of Labor Statistics: Nonfatal occupational injuries and illnesses requiring days
away from work. 2010 US Department of Labor (2011)
20. Pennathur, A., Mital, A., Contreras, L.R.: Performance reduction in finger amputees when
reaching and operating common control devices: A pilot experimental investigation using
a simulated finger disability. Journal of Occupational Rehabilitation 11(4), 281–290
(2001)
A Refuge Location Prediction System
for When a Tsunami Has Occurred
Akhiro Kawabe, Tomoko Izumi, and Yoshio Nakatani
The Graduate School of Science and Engineering, Ritsumeikan University,

Kusatsu city, Shiga, 525-8577 Japan
[email protected],
{izumi-t,nakatani}@is.ritsumei.ac.jp
http://www.sc.ics.ritsumei.ac.jp/
Abstract. During the 2011 Tohoku Earthquake and Tsunami, DMATs (Disaster
Medical Assistance Teams) could not rescue victims efficiently with accurate
location data, because the local governments had lost refuge location data and
resident registers due to damage caused by the tsunami. In this paper, to support
DMATs, a refuge prediction system based on the characteristics of disaster,
landscape, and victims’ psychology is proposed, which can function even if lo-
cal governments lose information about victims and refuge locations. As an ex-
ample, this system deals with tsunami. We demonstrate the effectiveness of this
system by comparing the data of the 2011 Tohoku Earthquake and Tsunami and
our prediction system.
Keywords: Disaster Medical Assistance Team, earthquake, local government.
1 Introduction
All over the world, suffering is caused by various kinds of natural disaster, including
typhoons, floods, volcanic eruptions, and earthquakes. Japan is particularly affected
by earthquakes, as the country is surrounded by four tectonic plates.According to
recent statistics, magnitude 2.0 earthquakes occur in Japan about 10,000 times a
year.For that reason, Japan is a country of frequent earthquakes and many earthquakes
happen every year[1].
The 2011 Tohoku Earthquake and Tsunami occurred on March 11, 2011, killing
more than 15 thousand people. This earthquake’smagnitude was9.0, which ranks
fourth highest in global statistics of the past 100 years.
In the 2011 Tohoku Earthquake and Tsunami, many hospitals and medical institu-
tionswere damaged by the tsunami, causing shortages of doctors. To solve this issue,
DMATs (Disaster Medical Assistance Teams) supported the disaster area. In spite of
the DMATs’ early dispatch to the disaster area, they were not able to rescue with
accurate location data, due to the fact thattown office administration stopped becau-
semany coastal town offices suffered from the tsunami. There were no means of
contacting victims, as the town offices had lost records of the sufferers’ addresses[2].
296 A. Kawabe, T. Izumi, and Y. Nakatani
It is predicted that the probability of a Nankai or Tonankai Earthquake occurring in

Japan within 30 years is over 60 percent [3].In future, when earthquakesoccur, it is
vital to know victims’ addresses, conditions and ages in order to make full use of
DMATs’ abilities.
In this study, we proposea refuge predictionsystem for supporting disaster medi-
cine from the viewpoint of disaster characteristics, regional characteristics and psy-
chological characteristics, which can be used even when local governments are not
able to acquire information on victims’ addresses and conditions.
2 Related Work
During a disaster, it is important to exchange disaster information. In the 2011 Tohoku

Earthquake and Tsunami, DMATswere not able to acquire sufferers’ addresses or refuge
location information, as the local governments that should have performed this task
were too damagedby the tsunami. In order to solve these problems, DMATs must estab-
lish who is in need of rescue, where they are, and acquire refuge location data. If infor-
mation cannot be obtained from sufferers, we need to predict refuge locations in some
way. In the 2011 Tohoku Earthquake and Tsunami, such support was not provided.
Over the past few decades, a considerable number of studies have been conducted
on predicting refuge locations. In this section, we briefly introduce such research.
In 2009, Asakawaet al. proposed a system that shares the location information of
the user by displaying it on a map [4]. This system attaches these data to Google
Maps using a mobile phone camera and GPS function. By sharing the data as map
information, the system links between Google Maps and the physical world.Thus,
Asakawa et al,’s system proposes an environment in which information can be ex-
changed as a reality.
In 2007, Tanida and Daito proposed a method for specifying victims’ locations us-
ing IC tags when the telephone network fails in the damaged area, and demonstrated
the system’s effectiveness by simulation [5]. In their system, a helicopter scatters
active type IC tags on the ground, and the system specifiesvictims’ locations by trila-
teration with three fixed IC tags and the location-unknown subject’s IC tag.
In 2011, Google constructed the “Google Crisis Response System”in response to
the Tohoku Earthquake and Tsunami [6]this system displays refuge locations with
markers in Google Maps, and displays the number of evacuees by marker color. In
addition, this system can output refuge location data for all forty-seven Japanese pre-
fectures in CSV andKML formats.
In 2011, Iizukaet al. proposed a systemthat collects disaster situation information
and showsit on a map when institutions such as universitiesareaffected by disaster[7].
It is thus clear from the above research that it is effective to show disaster informa-
tion on maps. However, sufferers were forced to take refuge in unexpected places by
the Tohoku Earthquake’s large tsunami, andrescue attempts by DMATs were hin-
dered because the tsunami destroyed many designated refuge places. Therefore, a
system is required that can predict refuge locations in advance, and give this data to
DMATs, thus supporting disaster medicine.
A Refuge Location Prediction System for When a tSunami Has Occurred 297
3 System Construction
We proposed “The refuge place forecast system for supporting disaster medicine “as
previous study stage.
In this study, we proposeda refuge predictionsystem for supporting disaster medi-
cine from the viewpoint of disaster characteristics, regional characteristics and psy-
chological characteristics, which can be used even when local governments are not
able to acquire information on victims’ addresses and conditions from huge tsunami
[8]. A previous study stage of system output is shown in Figure 1.
Fig. 1. Output example
In this previousstudy, the user inputs expected maximum tsunami height, refuge
place are not displayed if tsunami height is higher than the refuge location’s altitude,
sothis system shows refuge location names and the number of evacuees by compari-
son between regional population and refuge locationpredictionresultif a marker is
clicked, as shown in Figure 2.
Fig. 2. Output example: marker detail
According to the results of the system’s evaluation by Konan Fire Department and
Kyoto City Fire Department, it would be useful in future to predict the number of
victims, their locations and ages when DMATs are dispatched for early rescue in
disaster medicine, however refuge location prediction isn’t precise about absence of
sufferers’ addresses or refuge location information area because this system treats past
disaster sufferers’ addresses or refuge location information, and this system treats only
national census population and age ratio for refuge location prediction.Therefore, in
this paper try to accurate predict about the area hadn’t occurred disaster.
In this paper, we propose system based on previous study problem not only accu-
rate predict about the area hadn’t occurred disaster, but also assume aboutNan-
kai,Tonankaiearthquake will occur. After, we aim to build a platform in which local
residents can mutually register the information. For example, local residents could
register refuge locations and places where in the past ground liquefaction had oc-
curred, and we propose that system predict refuge place based on registered informa-
tion when earthquake occur. In Tohoku Earthquake, many evacuees took refuge un-
expected place. For example, a skating rink, a private house, a shrine, consequently,
DMATs needs get the information that local residents input unexpected place [9].
Also, local residents again realized each other to register unexpected place. For
example, local residents think “We didn’t know there places”, “We can take refuge
this place”. This system can add choice of refuge location based on refuge location
information if tsunami occurred.
A Refuge Location Prediction System for When a tSunami Has Occurred 299
4 System Proposal
4.1 About Prototype System

About prototype system register information, local residents input information by
smart phone, and prototype system manage database. Local residents register infor-
mation with text. Register information are time, name, capacity, also this prototype
system can upload photo of before disaster refuge place for rescuing if DMATs have
visited disaster area first time. This system shows stamp variety of attribute to register
refuge place, and can display refuge locations information in map by these stamp.An
image of prototype system is shown in Figure 3.
Fig.3.An image of prototype system
4.2 System Flow

In this research is composed of local residents, disaster countermeasures office to pass
date. The process of the system consists of the following web application tasks.
① Local residents inputs information about refuge location, and register information
for database.
② This system abstracts refuge location information (capacity, evacuees)from
database.
③ This system compares abstracted data and predictive refuge location of previous
study, and modifiesits data with real-world circumstances so that DMATs can get
accurate information.
④ This system sends modified data to disaster countermeasures office.
⑤ DMATs start rescuing based on modified data.
It is thus that this system can support DMAT because input data of local residents and
disaster countermeasures office cooperate. To support DMAT is Disaster medicines
make revitalize, finally be able to rescue.
5 Endnotes
In this paper, we proposed a system that predicts refuge locations in times of disaster
so that DMATs can rescue effectively based on this information. In large-scale natural
disasters such as the 2011 Tohoku Earthquake and Tsunami, DMATs were unable
toobtain victims’ locations and refuge location information because local governments
that held such information had suffered extensive damage from the tsunami. The pur-
pose of this research is to support disaster medicine.
According to problem of previous study,we propose system based on previous
study problem not only accurate predict about the area hadn’t occurred disaster, but
also assume about Nankai,Tonankai earthquake will occur. In future, we would like to
strengthen the alliance betweenlocal residents, disaster countermeasures office to
feedback based on information of local residents input.
References
1. Okada, Y.: Japanese earthquake map. Tokyo Shoseki, 7-9 (2004)
2. Shimbun, Y.: Special compact editionone month after the Tohoku-Pacific Ocean Earth-
quake. Yomiuri Shimbun, 64–65 (2011)
3. The Headquarters for Earthquake Research Promotion,
http://www.jishin.go.jp/main/p_hyoka02L.htm
(accessed August 27, 2012)
4. Asakawa, K., Hirano, K., Tsukada, T., Hayashi, Y., In, Y., Omiya, Y., Hamai, T., Muraka-
mi, H.: Collection and presentation of a system for safety information in disasters. ITE
Technical Report 33(11), 123–126 (2009)
5. Noriyuki, T., Masatora, D.: Agent Based Modeling for Finding Victims with IC tag
6. Google: Google Crisis Response:A Google.org project,
http://shelter-info.appspot.com/maps (accessed August 28, 2012)
7. Iizuka, K., Iizuka, Y., Yoshida, K.: A Real-time Disaster Situation Mapping System for
University Campuses. In: Ozok, A.A., Zaphiris, P. (eds.) OCSC 2011. LNCS, vol. 6778, pp.
40–49. Springer, Heidelberg (2011)
8. Kawabe, A., Nakatani, Y.: The refuge place forecast system for supporting disaster. The
74th National Convention of IPSJ 2ZG-3, 4-791–4-792 (2012)
9. Disaster information portal for Iwate,
http://www.pref.iwate.jp/~bousai/index.html (accessed December 6,
2012)
Further Benefit of a Kind of Inconvenience
for Social Information Systems
Hiroshi Kawakami
Kyoto University, Yoshida Honmachi, Sakyo, Kyoto, 606-8501, Japan

[email protected]
http://www.symlab.sys.i.kyoto-u.ac.jp/
Abstract. This paper discusses on social information systems from the

viewpoint of Fuben-eki that stands for FUrther BENEfit of a King of
Inconvenience. Analyzing several examples with benefit of inconvenience
elucidates the relation between inconvenience and their benefits that
in turn are organized into design guidelines. The guidelines applied to
several design fields including social information system design. A few
examples of such design are presented in this paper. One of the examples
accumulates and shares information on literature, and the others are for
establishing local communities.
Keywords: benefit of inconvenience, social information system, system

design theory.
1 Introduction
It is well-known that such convenient systems as automated ones do not always

guarantee the high performance of human-machine systems. On the contrary, the
ability of auto-adaptation of machine is sometimes harmful for human operators.
It spoils human ability of skill acquisition and of maintaining their motivation.
The result of the spoilage sometimes leads a seedbed of serious problems for
safety engineering. FUBEN-EKI [1] can be seen as one of the approach to the
solution of those problems.
Not only for safety engineering but also for social information systems, ben-
efit of inconvenience provides several notions. Generally, inconvenient systems
require users labor, skill and consumption of cognitive resources. Turning the
viewpoint from requirement to allowance, inconvenience allow users to interact
systems with time/effort, and to demonstrate/improve their skill. Such allowance
is one of the issue of benefits of inconvenience. Among wide range of the relevant
field of FUBEN-EKI, this paper focuses on social information systems. Social
information is, as defined in the research field of fundamental informatics [2],
exchanged and accumulated through organizing social life. To get leverage out
of social information, support systems for accessing such information are insuffi-
cient. They may be superficially convenient but spoil benefits of inconvenience.

302 H. Kawakami
bothering time difficult understand put

consuming system at ease
make feel foster self-
personalized affirmative feeling
require many constrained continuous get
operation operation I/O skillful
devise change multi-

become ways result form
degraded
have a failure lead big cognitive not
presence big risk load computerized unpleasure
physical
existence
Fig. 1. Relations among inconvenience and benefits
2 Brief Introduction of Benefit of Inconvenience

The procedure for incorporating benefits of inconvenience into systems design
consists of
1. defining relevant terms, e.g., benefit and convenience,
2. analyzing, in terms of the above definitions, existing examples that provide
users with benefits of inconvenience,
3. generalizing the relationships between inconveniences and benefits that are
derived from above analyses and are organized to design guidelines, and
4. applying the guidelines to target systems.
The same procedure is applied in the case of designing social information system.
Naive Definition of Relevant Terms: For common understanding of relevant

terms, e.g., convenience, inconvenience, and benefit, we searched for sentences
including the term “convenience” in more than 200 web pages. Naively following
the results of the web search, this paper defines the terms as follows:
convenient ≡ saving labor to attain a specific task
labor ⊃ time consuming
labor ⊃ special skills ⊃ mental load.
Relation between Inconveniences and Benefits: Following the above mentioned

definitions, we examined several examples with benefits of inconvenience, and
yielded relationships among inconveniences and benefits. Figure 1 overviews a
part of such relationships.
In the figure, rectangular nodes denote normal matter, circular nodes denote
inconvenient matter, each thick arcs denotes implicational relation, each dashed
arcs denotes contribution relation, and 3 is an operator of Modal Logic [3]. In
the alethic mode, 3p means that the proposition p is possible. In the deontic
mode, 3p means that p is permitted. In either mode, even if the value of p
varies subjectively depending on humans attitude, the value of 3p is objectively
determined.
Further Benefit of a Kind of Inconvenience 303
Utilizing Benefit of Inconvenience: Tracing nodes and arcs in Figure 1 suggests

several ways of getting benefit of inconvenience. For example, the hatching part
of the figure shows that there are inconvenient tools or methods that are super-
ficially difficult to operate but allow users to get skillful to operate by utilizing
continuous I/O.
The nodes in the figure are categorized into three types, i.e., benefits, guides,
and strategies. They are organized by means-ends relations. The benefits are
– fostering affirmative feeling,

– providing motivation to tasks,
– personalization [6],
– putting users at their ease.
As shown in Fig.1, we also get some abstract guide for realizing these benefits
as follows:
– allowing understand systems (visibility),

– allowing exploration (enhance awareness),
– allowing creative contribution to tasks and unsaturated proficiency.
Furthermore, the strategies for implementing such guidelines have been observed
in many examples, e.g.,
– utilize continuity and diversity of physical quantities,

– give a physical presence.
3 Psychological Aspect of Benefit of Inconvenience
Benefits of inconvenience are dependent on individuals. Therefore, psychological

notions are beneficial for discussing such benefits.
Living with Complexity: A cognitive psychologist D. A. Norman, known as the

presenter of POET [4], declares that adaptation of tools is nonsense and puts
emphasis on adaptation of users [5]. He shares the same standpoint with this
paper that values such characteristic of tools or methods as “do not disturb users
adaptation.” He also presented emotional design [6] with the notion of person-
alization that is one of the benefit observed in inconvenient tools and methods.
However, his recent book [7] claims that complexity of operation is inevitable
with the modern complex system, and if the system is simple to understand,
user should compromise with complexity. On the other hand, this paper regards
complexity as one of the key issues to implement benefits of inconvenient. Com-
plexity is not evil but the value of being hard to understand should be carefully
examined and feed off the process of systems design.
304 H. Kawakami
Ecological Psychology: Inconvenience does not always provide users with ben-
efits. The characteristic of tools or methods, which are generally said to be
inconvenient, enables users to aware the possibility of the behavior to get some
benefits. In the research field of ecological psychology, affordance is one of the
important idea. Affordance is the possibility of behavior, and tunability to affor-
dance differs from one individual to others. Only tuned individual can be afforded
to some behavior but not every tuned one behaves according to affordance. The
essential quality of benefit of inconvenience is same as affordance. The benefit p
is objectively declared 3p that means “p is possible” or “p is permitted.” While
affordance is directly tuned by animals and affords direct actions, benefit of in-
convenience attracts users to relatively long term behaviors, e.g., getting skillful,
and devising ways of usage.
Risk Homeostasis Theory: In the research field of human factors, risk home-
ostasis theory [8] argues that passive restraint devices do not always guarantee
the safety because users overly rely on the devices and tend to compensate risks.
In other words, convenient automatic safety systems encourage users risky be-
haviors. On the other hand, inconvenient systems foster users safety behaviors
and skills to operate the systems safely.
4 Applying Fuben-eki to Social Information Systems
Applying design guidelines derived from the relationships among benefits and
inconvenience, novel social information systems are yielded. Their performance
are now under examination in several research fields. This section introduces a
few examples of the systems.
Biblio Battle: For accumulating and sharing information on literature, book-

review is one of the efficient method and convenient information systems allow
us to submit and consult reviews anywhere, anytime and anybody. Biblio Bat-
tle [9] is an inconvenient review system that requires participants to physically
assemble on a certain time and place. Each participant presents his/her review
on his/her favorite book, then participants take a vote on reviews. The inconve-
nience, i.e., physical assembling provides high quality of review and generates a
community via review battles. Biblio battle, proposed by a researcher of Fuben-
eki, has grown to become a nationwide event hosted by the Tokyo metropolitan
government, and got the award “library of the year 2012” in Japan.
The simple rule of the battle provides high visibility of the battle system and
the inconvenience allows to explore strategies to win the battle and to aware the
other participants’ personalities.
Neko Media: Communities in cyberspace are convenient since getting in and out
are relatively easy than physical communities and interaction with community
members is not require so much responsibilities. Neko Media [10] is a incon-
venient system for establishing a community. Participants are required to be a
Further Benefit of a Kind of Inconvenience 305
friend of a certain stray cat (Neko in Japanese) for taking part of the community
because URL of the community site is only known by the QR code attached to
the collar of the cat. The difficulty to take part in the community enhances close
communications in the community.
Media Biotope: Mass media has enormous influence and is an accomplice of glob-
alization. Media Biotope [11] is an idea to structure media inspired by biotope of
ecological system. Based on this idea, Fuben-eki media systems are schemed [12]
to localize communication media in a appropriate size of communities. Mediators
are potted plants on a footbridge, sign boards on local bus, etc., that promote
communications in local societies.
5 Conclusion
It is well-known that not every inconvenience can provide benefits. Even more,
not every benefits of inconvenience is acceptable for Fuben-eki. It is compro-
mising that we live with laborious password for security. It is ignominious that
complex payment structure makes money. Fuben-eki eliminates such kind of
compromises and ignominious situations. Inconveniences should be introduced
for only getting benefits as shown in section 2. A promising way to get such
benefits is following guidelines as shown in the same section.
Some convenient social information systems are harmful. Job matching sites
accumulate recruiting information and facilitate application process. It is conve-
nient because either applicants and recruiters can save many steps. The trouble
is that either applicants and recruiters are with light hearts. Convenient facili-
ties multiply the amount of transactions and consequently enlarge the amount
of rejections because the amount of job remains unchanged. From psychological
and social perspectives, the infestation of rejections is not desirable.
On the other hand, even inconvenience requires users to make some efforts,
its benefits are, as shown in section 2, psychologically desirable. Any examples
shown in section 4 puts emphasize on subjective benefits that depend on indi-
viduals rather than on objective conveniences.
References
1. Kawakami, H.: Inconvenience utilizing Design. Kagaku-Dojin Publishing Company
Inc. (2011) (in Japanese)
2. Nishigaki, T.: Fundamental informatics from life to society. NTT Pub. Co. Ltd.
(2004) (in Japanese)
3. Kripke, S.A.: A Completeness Theorem in Modal Logic. The Journal of Symbolic
Logic 24(1), 1–14 (1959)
4. Norman, D.A.: The design of everyday things. MIT Press (1998)
5. Norman, D.A.: Human-Centered Design Considered Harmful. Interactions 12(4),
14–19 (2005)
6. Norman, D.A.: Emotional Design. Basic Books (2005)
7. Norman, D.A.: Living with Complexity. MIT Press (2011)
306 H. Kawakami
8. Wilde, G.J.S.: Target Risk 2 –A new psychology of safety and health. PDE Publi-
cations (2001)
9. Taniguchi, T., et al.: Bibliobattle: an informal social interaction design mediated
by book review. J. of Human Interface Society 12(4), 93–103 (2010) (in Japanese)
10. Suto, H., Okita, M.: Proposal of a Novel Communication Medium to Promote
Community Using Homeless Cats. In: Proc. the 6th Workshop on Social intelligence
Design, pp. 287–296 (2007)
11. Mizukoshi, S.: Media Biotope. Kinokuniya Publication Services (2005) (in
Japanese)
12. Suto, H.: Media Biotope: Media designing analogous with Biotope. In: Proc. Int.
Conf. on Computer Information Systems and Industrial Management Applications,
pp. 75–80 (2010)
The Present Condition and Problems for Elderly People
Participating in Communities
Noriyuki Kobayashi, Shohei Yamane, Kotaro Ohori, and Akihiko Obata
Fujitsu Laboratory Ltd.

{kobayashi.nryk,yamane.shohei,ohori.kotaro,
obata.akihiko}@jp.fujitsu.com
1 Introduction
The proportion of aged citizens in Japan has been growing and has reached the high-
est proportion in the world (23.1% in October, 2010 [1]). The time when we need to
tackle the social welfare problem of an aging society has arrived. It has been re-
ported that social welfare services are insufficient to reach all people who need
help, resulting in 32000 lonely deaths in a single year [2]. It is difficult to solve the
problem only by reforming formal care systems under conditions where the number
of senior citizens is increasing and the number of younger citizens is decreasing.
Community activity that supports elderly citizens is important when the proportion
of aged citizens is increasing to prevent overreliance on formal care. The idea
named PPK(pin-pin-korori [3]) thus emerged. This is the idea that the number
of people requiring long-term care can be decreased by increasing the number of
active senior citizens.
The following problems were reported in the annual report of the Cabinet
Office of Japan for community activities for supporting elderly citizens [4].
- Difficulty for continuing community activities due decreasing numbers of partici-

pants in senior citizen clubs and increase of their age.
- Increase in elderly citizens who are isolated due to weak human connections with
their regional community.
One of the approaches for solving these problems is promoting community activities
that support senior citizens. ICT support is also expected to support such community
activities. These mechanisms have been studied from the aspect of administration,
however, there has been little research focusing on the view point of the senior citizen
who actually participates in the community activities.
In this paper, we report an ethnographic study for finding out important design im-
plications for ICT support for the community activities. The participants of the study
had various problems and stories regarding senior citizens participating in community
activities. We then discuss the implications of design for ICT support.
308 N. Kobayashi et al.
2 Method
We conducted ethnographic interviews of fourteen senior citizens to find out about

problem situations in their daily activities. Fourteen active seniors were selected as
interviewees, since we could expect many tips from “lead users” [5].
In the ethnographic interviews, we asked interviewees to speak freely about their story
of daily activities and community activities instead of answering predefined questions.
Based on analysis of the transcriptions of the conversations, we collected various notes,
such as episodes of participation, problems of participating in the community, surround-
ing situations, etc., from the story of the varieties of their activities. The collected notes
were classified, and arranged into different types of problems and situations.
It is difficult to find out about episodes if you ask about the community activity di-
rectly. Therefore, we used the following interview sheets (Fig.1,2,3,4). Using the map
sheets in Fig.1, we asked about the places where the interviewees visited daily. We
then asked about the people who met at the places and their activities. Using the clock
sheets in Fig.3, we asked about their daily activities from morning to night, and then
clarified the daily activities that engaged in unconsciously. It is difficult to find out
about daily activities because unconscious behavior does not leave an impression.
However, it becomes possible to clarify these by looking back on the activities one at
a time by using the interview sheets.
Fig. 1. Map sheet

The Present Condition and Problems for Elderly People Participating in Communities 309
Fig. 2. Example of the entry in a map sheet
Fig. 3. Clock sheet

Fig. 4. Example of the entry in a clock sheet
3 Results
We got more than 100 episodes of problem situations for facilitating community ac-
tivities from the fourteen interview transcriptions. The followings are example of
episodes.
(1) A new community activity derives from an existing community activity.

We found that new communities were created triggered by community activities that
already existed.
At the interview, we observed the following discussion.
"I joined the softball club because I was invited by a friend of the neighborhood as-
sociation. The softball club is a very important community for me."
“I joined the sign language club because I have handicapped friends in another
community”. These episodes might suggest that community activity itself acts as a
trigger for a new community activity.
(2) Importance of participation by obligation

We found that social pressure was a one of the important factors for the participation
of elderly people in communities.
“I was reluctant to join the neighborhood association. However, I have to take the
role of the neighborhood council because it was my turn. But now, this community is
very important for me"
“After I joined the neighborhood association, I created a baseball team with friends
in the neighborhood association”
(3) Hesitation getting help from others

We found that a feeling of hesitation getting help from others is one of the reasons
why elderly citizens refuse participation in community activities. An elderly person
said "I feel uneasy about many people taking care of me free of charge in the commu-
nity activities.” “In a certain community, they reduce such hesitation by taking 100
yen as a registration fee.”
(4) Barrier between generations
We found that there were few opportunities for meeting other generations in com-
munities. Most regional communities are grouped by generation, such as boys and
girls clubs, women’s clubs, elderly citizens clubs, etc. There are few events for meet-
ing different communities. An elderly person said “I have nothing to talk about with
young men“. “Our community has no events with other communities”.
It is possible that a community in which most of the members are elderly will de-
cline because younger people will not join. An elderly said, “There are no successors
in our community. Who take cares of the community after we get older?” As a re-
sult, members who run the community were fixed, and there will be few chances for
expanding their community.
On the other hand, we found a regional traditional festival event “O-maturi” in
which different generations are participating fostered cross generational relationships.
For example, we got episodes such as “I talked with a father of my friends”, “I met
a person who lives close to my house”, etc.
4 Implications for ICT Design
As a result of the ethnographic interviews, we found that elderly people tended to not
join unknown communities where there were no acquaintances. Therefore, it is neces-
sary for the system to provide functions for encouraging them to join unknown com-
munities. We designed a function that supports community members inviting elderly
people. When an elderly person pushes the "NICE" button from the list of communi-
ties, members of the community get a message that tells them who showed interest in
their community and the profile of that elderly person. The members of community
can then decide who they should invite. We designed another function that supports
elderly people joining a community with their friends. When an elderly person pushes
the “NICE” button, the elderly person receives a message that shows their acquain-
tances registered as “Associate” who also showed interest in the same community.
As a result of interviews, after they joined a community out of obligation, they be-
came interested in the community activities or they found other interesting communi-
ties. They might leave the community if they could not find interesting communities
and activities during the duty. Therefore, we designed a function that supports finding
interesting communities. The system registers the elderly as an “Associate” when s/he
joins the community. The system then supports finding a friend who has the same
interest in the community by using the “NICE” button.
It was interesting for us that the interviewees felt guilty about getting kind help
from others without giving back in return. Therefore, we designed a function that
reduces the hesitation. The system supports selecting appropriate people for the task.
When a community member inputs a task such as taking photos of the community,
playing a guitar, etc., the system shows the appropriate persons based on the profiles
of elderly people. It becomes easy to join a community by not requesting “please join
us” but “please help us”.
It is difficult to resolve the barrier among different generations by only focusing on
supporting individual communities. We designed a new function supporting connec-
tions among different communities. A regional traditional festival could be a good
trigger for fostering communication among different communities. The owner of the
festival can call different communities to invite them to the festival using the new
function. Members of each community can join a new community with their friends
from their community. Invitation by community (not by individual) might reduce
their hesitation of participating in a new community, and also might foster relation-
ships among different generations of different communities.
Fig. 5. Image of ICT design
The new function supporting connections with different communities might also
help to activate and expand a community. Senior citizen communities tend to be run
by fixed members and no new events are created. Owners of communities have to
find people who help their events through their own network of acquaintances. The
new function might help to find new activities and events of other communities, and
easily get help for them using the new function.
5 Conclusion
We obtained design implications for community support based on interviews of

community activities of active elderly citizens. However, these implications are still
limited. We should investigate not only active elderly people but also non-active
elderly people to get further implications. We will continue our ethnographic study on
different types of citizens and other stakeholders and then conduct field experiments
using an ICT prototype in some regions based on the findings of the field study.
References
1. Annual Report on the Aging Society: Cabinet Office Japan (2011),
http://www8.cao.go.jp/kourei/whitepaper/index-w.html
2. NHK Muen Shakai Project, “Muen Shakai”, Bungei SyunJyu (2010) (in Japanese)
3. Akiyama, H.: Science of aging society and social structure, Kagaku Iwanami Shoten (2010)
(in Japanese)
4. A meeting for production of communities which elderly people etc. feel easy also alone and
can live: Ministry of Health, Labor and Welfare Japan (2008),
http://www.mhlw.go.jp/houdou/2008/03/h0328-8.html
5. Hippel, E.: Democratizing Innovation. MIT Press, Cambridge (2005)
Applying to Twitter Networks of a Community
Extraction Method Using Intersection Graph
and Semantic Analysis
Toshiya Kuramochi, Naoki Okada, Kyohei Tanikawa,

Yoshinori Hijikata, and Shogo Nishida
Graduate School of Engineering Science, Osaka University,

1-3 Machikaneyama, Toyonaka, Osaka, Japan
[email protected],
{hijikata,nishida}@sys.es.osaka-u.ac.jp
Abstract. Many researchers have studied about complex networks such

as the World Wide Web, social networks and the protein interaction net-
work. One hot topic in this area is community detection. For example,
in the WWW, the community shows a set of web pages about a certain
topic. The community structure is unquestionably a key characteristic
of complex networks. We have proposed the novel community extracting
method. The method considers the overlaps between communities using
the idea of the intersection graph. Additionally, we address the prob-
lem of edge inhomogeneity by weiting edges using content information.
Finally, we conduct clustering based on modularity. In this paper, we
evaluate our method through applying to real microblog networks.
Keywords: complex network, community extraction, intersection

graph, hierarchical clustering, text mining, microblog network.
1 Introduction
Many researchers, having studied complex networks such as the World Wide
Web, SNS networks, and the protein interaction network, have reported scale-
free characteristics, the small-world effect, the property of high-clustering coef-
ficient, and so on [1,2]. Recently, the community structure in complex networks
is gaining increased attention from many researchers. The community structure
shows the appearance of densely connected groups of nodes, with only sparse
connections among groups. Many community detection methods have been pro-
posed based on this definition [11]. They are applied to various complex networks.
Communities in SNS networks shows a set of people with the same background
or hobby. Communities in the WWW show sets of web pages related to a cer-
tain topic [4] and those in the protein interaction network show sets of proteins
having the same function [6].
For community detection, researchers have started to show interest in whether
overlaps between communities can be extracted [15,17,23]. The overlaps signify
that one node belongs to several communities. For example, the Apple Inc. page

Community Extraction Method 315
is categorizable among computer category pages and audio category pages. It is

important that a community detection method be able to assign a node not only
to one community but also to several communities. Weights of all edges in complex
networks are assumed to be the same in many community detection methods [19].
However, edges are rarely homogeneous in real networks. For example, various hu-
man connections such as those of businesses, hobbies, and organizations exist in
SNS networks. It is important that the weights of these edges are not be treated as
identical. They should be set individually. Many researchers conduct hierarchical
clustering methods to complex networks for community detection. Most hierarchi-
cal clustering methods require advance input [3]: the number of clusters. However,
the number of real communities is often unknown in real networks. Therefore, it
is important that the number of proper clusters be decided automatically.
We have proposed a novel community detection method that can solve these
three problems [8]. Our proposed method can extract overlaps between communi-
ties using the idea of the intersection graph. When there exist several subgraphs
individual members are connected densely, we can make a new graph where the
above subgraphs are converted to new nodes and edges are created when two
subgraphs have common elements. Graph created by the above process is called
intersection graph [10]. We also determine the weights of the edges in the in-
tersection graph using two types of information: the degree of overlaps of the
members and the similarity of content information such as text information and
attribute information which appear in the network. Moreover, we use the hierar-
chical clustering method based on modularity proposed by Newman [11,13,14].
This method does not necessitate manual input of the number of clusters. In
this paper, we evaluate general versatility of our method through applying to
microblog networks.
2 Related Works
The problem of community detection in complex networks has been examined

in various areas such as those of computer science and medical science [2].
Some researchers have attempted to extract communities in complex networks
including the overlaps between communities. The overlaps mean that one node
belongs to several communities. Everett et al. found them using the idea of the
intersection graph [3]. Palla et al. found them by detecting cliques whose size
was k and merging the cliques that shared k − 1 nodes [15]. Fuzzy clustering,
the method considers the notion of fuzziness and can assign one node to several
communities is often used to extract overlaps of communities [17,23].
Our study weights edges in complex networks for dealing with edge inhomo-
geneity. Weighting edges in a network (usually a document network or hyperlink
network) is popular in the area of the information retrieval. Some researchers
improved the effectiveness of link analysis using content information. Jiang and
Conrath measured the similarity between words using link information and
semantic information of words [7]. Hung et al. improved the HITS algorithm
by analyzing anchor text [5].
316 T. Kuramochi et al.
Many researchers use hierarchical clustering methods for detecting commu-

nities. The methods need input, which is the number of clusters preliminar-
ily. Newman and Borgatti reported modularity as an indicator of how well the
clusters are formed [12]. Newman proposed some clustering methods based on
modularity [11,13,14]. These methods do not obviates manual input.
Our proposed method [8] considers the overlaps between communities using
the idea of the intersection graph. Furthermore, we address the problem of edge’s
inhomogeneity by weighting edges using the degree of overlaps and the similarity
of content information between sets (nodes of the intersection graph). Finally,
we conduct a clustering method based on modularity, which does not necessi-
tate manual input of the number of clusters. No study deal with all the above
problems for detecting communities.
3 Proposed Method
In this section, we explain our proposed method [8]. The input of our proposed
method is a graph of G = (V, E), where V stands for the set of nodes and E
signifies the set of edges. Additionally, content information is given to the nodes.
We apply the following four steps to this graph.
Step 1. Enumeration of Dense Subgraphs: This method enumerates dense
subgraphs (generally, they are called cliques) from an input graph of G = (V, E).
Step 2. Conversion to the Intersection Graph: This method regards each
subgraph enumerated in Step 1 as one new node and converts the input graph
G to the intersection graph of G = (V , E ).
Step 3. Calculation of the Weights of Edges: This method calculates the
weights of edges E in the intersection graph G using the degree of overlaps and
the similarity of content information between nodes V (dense subgraphs) in G .
Step 4. Clustering Based on Modularity: This method divides nodes V
into clusters using a clustering method based on modularity.
We applied the method of Everett et al. [3] to Step 1 and Step 2. In Step 1,
their method enumerates maximal cliques as dense subgraphs in an input graph
of G = (V, E). A clique is a subgraph in which an edge exists between any two
nodes. Next, the method converts the input graph G into the intersection graph
G = (V , E ) in Step 2. Our method regards each dense subgraph enumerated
in Step 1 as one special node and makes the intersection graph G = (V , E )
from the input graph G = (V, E). When several sets (dense subgraphs) Si (i =
1, · · · , n) are enumerated, our method generates a special node vi for each set Si .
If a common element exists in two arbitrary nodes vi and vj , then a special edge
is put between them. The intersection graph is a new graph composed of special
nodes and special edges [10]. When the method puts a special edge between
special nodes, we can set the threshold of the number of common elements
between the subgraphs corresponding to these special nodes. Finally, the method
of Everett et al. conducts hierarchical clustering for the intersection graph. Our
method address the edge inhomogeneity (in Step 3) and automatically detection
of extracting communities number (in Step 4).
In Step 3, the proposed method calculate weights of edges in the intersection

graph generated in Step 2. We use the degree of overlaps and the content infor-
mation similarity between each subgraphs. There exist many types of measure-
ment for presenting the degree of overlaps between X and Y (d(X, Y )) such as
co-occurrence frequency, mutual information, Dice coefficient, Simpson coefficient
and Jaccard coefficient [9,16]. For example, Jaccard coefficient is defined as:
|X ∩ Y |
d(X, Y ) = . (1)
|X ∪ Y |
The proposed method uses vector space model [18] to calculate the similarity
of the content information between two arbitrary sets X and Y . The method
regards each set as one vector and calculates the tf ·idf score for the keyword
in the texts in the set. This tf ·idf score becomes the element of the vector.
Finally, the method calculates the similarity sim(X, Y ) between vectors x and
y corresponding to two sets X and Y using cosine similarity:
x·y
sim(X, Y ) = cos θ = . (2)
x y
Then, the proposed method calculates the weights w(i, j) for the special edge
between special nodes vi and vj (corresponding to set X and Y ) using the degree
of overlaps of sets d(X, Y ) and the similarity of content information sim(X, Y ).
We can use several types of calculation function. In this work, we use function
emphasizing the similarity of content information:
d(X, Y )
w(i, j) = w(X, Y ) = . (3)
1 + − sim(X, Y )
Here, (0 < < 1) is a constant used to keep the denominator from being 0.
Finally, in Step 4, the proposed method conducts clustering for community
detection in the intersection graph. When a method extracts several clusters in
a network, we must evaluate the currently detected clusters. Modularity is a
broadly accepted indicator for evaluation. The indicator is simple and intuitive.
Therefore, we adopt a clustering method based on the modularity that is pro-
posed by Newman et al.[11,13,14]. When k clusters are given and Pk is defined
as the sets of these clusters, the module function Q(Pk ) is the following.

Q(Pk ) = (eii − a2i ) = Tr(e) − |e2 | (4)
i
1

eij = 2m vw Avw δ(cv , i)δ(cw , j)
1
ai = 2m v kv δ(cv , i)
4 Applying to Microblog Networks

We apply our method to real social networks. We select Twitter for this study
which is the most popular microblog searvice. Twitter users post tweets (short
messages) and have conversations with other users through their tweets. If tweets
have the word “@username,” they are mentions—tweets for certain users. The
word “#hashtag” in tweets means these tweets concern certain topics. Therefore,
we can get the large amount of content information. In the Twitter network,
all people can follow anyone without approvals, then the network contains links
represent the unilateral interest. Links between users may represent relationships
in the real world (university friend etc.) or that of interest (hobby friend etc.).
The purpose of this evaluation is to verify three questions:
– Whether our method achieves better results than the conventional
method: We compare our method with the conventional method proposed
by Everett et al. [3]. The conventional method converts an input graph into
the intersection graph and conducts a simple hierarchical clustering.
– Whether it is efficient to use content information for weighting
edges: Our method uses not only information about the degree of overlaps
of sets but also the content information. We compare the method using both
kinds of information with the method using only the degree of overlaps of
sets. We confirm the effectiveness of the content information.
– Whether the kinds of content information affect the results: Twitter
networks have several types of content information. We examine whether the
results change according to the kind of content information.
4.1 Dataset
We make a dataset for the evaluation inviting test subjects who give true re-
lationships between them and each member in the extracted communities. We
followed users from a test subject to two in the radius (from the test subject
up to the friends of the test subject’s friends). Our experiment assume a sit-
uation extracting the communities of the real world or strong interest. There
are some celebrities followed by million users in the Twitter network. A test
subject may be not able to answer the proper relation between stranger users
connected via such celebrities. We think it is important that all test subjects
can answer the all relations between users in the dataset. Therefore, we set the
threshold of the number of follow users and that of followers. Then, we removed
such hub users (celebrities) in advance. Thresholds are set as 400 from prior
study. Additionally, we collected profile texts and tweets (contain @usernames
and #hashtags) as content information. The test subjects are 9 users who are
all university students.
4.2 Implementation
Parameter Settings of the Proposed Method. We adopt the maximal
clique as the dense subgraph in Step 1. We can use various sizes of the maximal
clique (the clique threshold). If the clique threshold is 5, then the method uses
only the maximal cliques that comprise more than four nodes. We set 3, 4 and 5
as the clique threshold. In Step 2, our method converts the original graph to the
Table 1. Statistical information of the dataset
method content information

Prf nouns in profile texts
Prf-H nouns and #hashtags in profile texts
Prf-N nouns and @usernames in profile texts
Prf-H+N nouns, #hashtags and @usernames in profile texts
Twt nouns in tweets
Twt-H nouns and #hashtags in tweets
Twt-N nouns and @usernames in tweets
Twt-H+N nouns, #hashtags and @usernames in tweets
intersection graph. Here, it creates a special edge between two dense subgraphs
(two special nodes) when they have common elements. We can set the threshold
of the number of common elements (the overlap threshold) in this step. We set
2, 3 and 4 as the overlap threshold (In our prior work [8], we have found that the
performance of the proposed method is very low when the overlap threshold is 1
under the influence of clustering method based on modularity). We conduct nine
threshold conditions (clique threshold, overlap threshold) = (3, 2), (4, 2), (4, 3),
(5, 2), (5, 3) and (5, 4).
In Step 3, we selected the Jaccard coefficient (eq. (1)) as the degree of overlaps
of sets. We also selected profile texts and tweets as the content information. Our
method extracts nouns, #hashtag and @username as keywords by conducting
morphological analysis of the content information. The method calculates tf ·idf
scores for all keywords within one maximal clique (corresponding to a special
node). The maximal clique can represent one vector. The similarity between
maximal cliques is calculated using eq. (2). Finally, the weights between maximal
cliques are calculated using eq. (3). We set = 0.1 in this experiment. We
use a greedy approach that repeatedly merges a pair of nodes to maximize the
increment of the modularity [13].
Implementation of Community Extraction Method. As described in sec-

tion 4.1, we use content information of two types: profile texts and tweets. We
respectively designate the cases using profile texts and tweets as Prf, Prf-H, Prf-
N, Prf-H+N, Twt, Twt-H, Twt-N and Twt-H+N (see Table 1). Hereinafter, we
regard Twt as a representative example of these cases using the content infor-
mation. We implemented the case using only the degree of the overlaps between
sets (Jaccard coefficient) as the weights of edges in Step 3 to examine the contri-
bution of content information analysis. We designate the case NonCA (without
content analysis).
We implemented the method Everett et al. proposed [3] as a baseline method.
The method comprises three steps. The first two steps of the method are the
same as the first two steps (Step 1 and Step 2) of our method. Unlike our
method, it uses a simple hierarchical clustering method in the third step. In
detail, the method searches for a pair of special nodes that have maximal Jaccard
coefficient and merges the pair, repeatedly. We set the number of output clusters
as the number of clusters including the test subject becomes the number of true
communities (provided by each test subjects). We designate this case as Everett’s
method.
4.3 Evaluation Method of Extracted Clusters

Clusters of two kinds are extracted by a community extraction method: a cluster
that includes the test subject and a cluster that does not contain the test subject.
It is difficult for test subjects to judge the connection of members in the latter
clusters. Therefore, we specifically addressed only those clusters containing the
test subject. To measure the accuracy of the extracted clusters, we adopt the
following evaluation process.
Step 1. A test subject enumerates all relation names for each person in the dataset
(number of relation names is regarded as the number of true communities). The
test subject can see the profile texts and latest 200 tweets of each person.
Step 2. We assume that the relation in the extracted cluster corresponds to
each relation name. We calculate the precision, recall andF -measure per relation
name for the cluster (For the calculation, we consider that the relation name is
the name of correct relation for the cluster). The precision, recall andF -measure
of a relation name N are calculated as follows.
# people whose relation name is N in the extracted cluster
Precision(N ) =
# people in the extracted cluster
# people whose relation name is N in the extracted cluster
Recall(N ) =
# people whose relation name is N in the dataset
2 · Precision(N ) · Recall(N )
F -measure(N ) =
Precision(N ) + Recall(N )
Step 3. We use the highest F -measure calculated in Step 2 among all relation
names as the F -measure of the extracted cluster. We regard the relation name N̄
that marked the highest F -measure as the relation of the cluster. We also use the
precision and recall of a relation name N̄ as the precision and recall of the clusters.
Step 4. We calculate the average values of the precision, recall and F -measure
of all clusters. These values are regarded as the evaluation value of one test
subject.
5 Evaluation of Extracted Clusters

We conducted an experiment of community extraction for clarifying three ques-
tions mentioned in Section 4. We show the average values of the precision, recall
and F -measure in all 9 test subjects in the conventional method and our method
in Figure 1-3.
We compare our method (NonCA and Twt ) with the conventional method (Ev-
erett’s method ). In precision (Figure 1), the results of our method tend to become
0.9
0.8
0.7
precision
0.6 Everett
0.5 NonCA
Twt
0.4
0.3
(3, 2) (4, 2) (4, 3) (5, 2) (5, 3) (5, 4)
(clique threshold, overlap threshold)
Fig. 1. Comparing the precision of Everett’s method and our method
0.9
0.8
0.7
recall
0.6 Everett
0.5 NonCA
Twt
0.4
0.3
(3, 2) (4, 2) (4, 3) (5, 2) (5, 3) (5, 4)
Fig. 2. Comparing the recall of Everett’s method and our method
0.9
0.8
0.7
F-measure
0.6 Everett
0.5 NonCA
Twt
0.4
0.3
(3, 2) (4, 2) (4, 3) (5, 2) (5, 3) (5, 4)
Fig. 3. Comparing the F -measure of Everett’s method and our method
better when both thresholds become are larger. This is because users connecting
strongly each other tend to survive while making the intersection graph. The pro-
posed method overcomes the conventional method in the case when the overlap
threshold become larger. In recall (Figure 2), the results of our method become
better when both thresholds become smaller. Our method shows much higher re-
call than the conventional method in all conditions. At last, in F -measure (Figure
3), our method overcomes the conventional method in all conditions. Overall, we
found our method brings a better result than the conventional method.
We compare NonCA and Twt to find out the effectiveness of content infor-
mation analysis. In many threshold conditions, the precisions of Twt are greater
than those of NonCA (Figure 1). Other hand, in recall, NonCA tends to over-
come Twt (Figure 2). We cannot determine which method is better for extracting
0.8
0.7
0.6
0.5 precision
0.4 recall
0.3 F-measure
NonCA
Prf-H
Prf
Prf-N
Twt-N
Twt-H+N
Everett
Prf-N+H
Twt
Twt-H
community detection method
Fig. 4. Evaluation in condition (4, 2) using content information of various kinds
communities because the difference in the F -measure is small (Figure 3). We think
the reason is that text information in Twitter dataset contains many colloquial
words and coined words. We conduct a simple morphological analysis that may
be not enough effective to accurate extracting nouns from tweets.
Finally, we examine whether the results change according to the type of content
information. We evalutate the cases using nouns, #hashtags, and @usernames in
profile text and tweets as the content information in the threshold condition (4, 2).
We present results of Everett’s method, NonCA, Prf, Prf-H, Prf-N, Prf-H+N, Twt,
Twt-H, Twt-N and Twt-H+N in Figure 4. The results of cases using content infor-
mation (all method without Everett’s method and NonCA) are mutually similar.
As we explained above, the influence to community detection of content informa-
tion is not enough strong in our method (because of a simply morphological anal-
ysis). Additionally, many users in this experiment do not post tweets containing
#hashtags or @usernames much. The volume of these information is not enough
to influence the performance of our method.
6 Conclusion and Future Work

As described in this paper, we evaluated the proposed method for community de-
tection [8] through applying to the Twitter dataset. We demonstrated the superi-
ority of the proposed method compared to the conventional method. Although
we compared the case using content information with the case not using that,
we cannot show the advantage of proposed method. We think that is because the
morphological analysis in our method is too simply to applying coined words and
colloquial words in tweets and our experimental dataset do not contain much
#hashtags and @usernames. As future workCwe will improve the morphological
analysis and apply our method to other test subjects.
References
1. Albert, R., Barabasi, A.-L.: Statistical mechanics of complex networks. Rev. Mod.
Phys. 74, 47–97 (2002)
2. Boccaletti, S., Latora, V., Moreno, Y., Chavez, M., Hwang, D.U.: Complex Net-
works: Structure and Dynamics. Phys. Rep. 424(4–5), 175–308 (2006)
3. Everett, M.G., Borgatti, S.P.: Analyzing Clique Overlap. Connections 21(1), 49–61
(1998)
4. Flake, G.W., Lawrence, S., Giles, C.L., Coetzee, F.: Self-Organization of the Web
and Identification of Communities. IEEE Computer 35(3), 66–71 (2002)
5. Hung, B.Q., Otsubo, M., Hijikata, Y., Nishida, S.: HITS Algorithm Improvement
using Semantic Text Portion. WIAS 8(2), 149–164 (2010)
6. Huss, M., Holme, P.: Currency and commodity metabolites: Their identification and
relation to the modularity of metabolic networks. IET Systems Biology 1(5), 280–285
(2006)
7. Jiang, J.J., Conrath, D.W.: Semantic Similarity Based on Corpus Statistics and
Lexical Taxonomy. In: Proc. ROCLING 1997, pp. 19–33 (1997)
8. Kuramochi, T., Okada, N., Tanikawa, K., Hijikata, Y., Nishida, S.: Community Ex-
tracting Using Intersection Graph and Content Analysis in Complex Network. In:
Proc. WI 2012, pp. 222–229 (2012)
9. Manning, C.D., Schütze, H.: Foundations of statistical natural language processing.
MIT Press (2002)
10. McKee, T.A., McMorris, F.R.: Topics in Intersection Graph Theory, vol. 2. SIAM,
Discrete Mathematics and Applications (1999)
11. Newman, M.E.J.: Detecting community structure in networks. Eur. Phys. J.
B 38(2), 321–330 (2004)
12. Newman, M.E.J., Girvan, M.: Finding and evaluating community structure in net-
works. Phys. Rev. E 69(2) (2004)
13. Newman, M.E.J.: Fast algorithm for detecting community structure in networks.
Phys. Rev. E 69(6) (2004)
14. Newman, M.E.J.: Finding community structure in networks using the eigenvectors
of matrices. Phys. Rev. E 74 (2006)
15. Palla, G., Derenyi, I., Farkas, I., Vicsek, T.: Uncovering the overlapping community
structure of complex networks in nature and society. Nature 435(7043), 814–818
(2005)
16. Rasmussen, E.: Clustering Algorithms, Information Retrieval: Data Structures and
Algorithms. In: Frakes, W.B., Baeza-Yates, R. (eds.), pp. 419–442. Prentice-Hall
(1992)
17. Reichardt, J., Bornholdt, S.: Statistical mechanics of community detection. Phys.
Rev. E 74(1), 16110 (2006)
18. Salton, G., Wong, A., Yang, C.S.: A vector space model for automatic indexing.
Communications ACM 18(11), 613–620 (1975)
19. Scott, J.: Social Network Analysis: A Handbook, 2nd edn. Sage Publications (2000)
20. Scripps, J., Tan, P.-N., Esfahanian, A.-H.: Node Roles and Community Structure
in Networks. In: WebKDD/SNA-KDD 2007, pp.26–35 (2007)
21. Tasgin, M., Bingol, H.: Community Detection in Complex Networks using Genetic
Algorithm. In: Proc. ECCS (2007)
22. Wasserman, S., Faust, K.: Social Network Analysis: Methods and Applications.
Cambridge University Press (1994)
23. Zhang, S., Wang, R., Zhang, X.: Identification of overlapping community structure
in complex networks using fuzzy c-means clustering. Physica A 374(1), 483–490
(2007)
Search Engine Accessibility for Low-Literate Users
Débora Maurmo Modesto, Simone Bacellar Leal Ferreira, and Aline Silva Alves
Departamento de Informática Aplicada, Universidade Federal do Estado do Rio de Janeiro,

Rio de Janeiro, Brazil
{debora.modesto,simone,aline.alves}@uniriotec.br
Abstract. Search engines are often used to retrieve content on the Web, but it is
not a simple activity for low-literate users since they have to know the technol-
ogy and create strategies to query and navigate. Their interaction with search
engines differ from high-literate users on strategies used, perception, communi-
cation and performance. In order to improve search engines and create solu-
tions, we need to understand these users' needs. This research aimed to identify
how search engine features influence the interaction of low-literate users. We
analyzed the interaction of ten users through user tests that were part of a case
study. Based on a limited set of features of a specific search engine, we identi-
fied what features were used, the perception about them and some barriers faced
by these users. This study led to a list of recommendations for the development
of search interfaces focused on low-literate users.
Keywords: Low-Literate Users, Search Engine Accessibility, Guidelines.
1 Introduction
The available tools on the Web are part of everyday life [1] and perform a social role
especially for users with disabilities [2]. Therefore, there should be no barriers to
access the Internet. There are users who have limitations related to literacy that can
jeopardize the interaction mainly because the available content on the Web is mostly
textual [3].
To help users to retrieve this content, search engines are often used: 80% of the
access to Web pages comes from these tools [4]. Search involves analyzing different
types of media, so it is a mentally exhausting activity that requires focus and attention
[5]. Low-literate users have some limitations, particularly related to the strategies they
use to conduct a search and the perception of interfaces [6-7]. Besides that, high-literate
users interact with such tools in a different manner than low-literate users [3], [8].
In order to improve low-literate users experience on the Web, search engines
should accomplish their mental models. Designers and developers should know how
they interact with this kind of tools, what features are used and how they influence on
user experience.
This research aimed to identify how search engine features influence the interac-
tion of low-literate users. We observed how the interaction takes place and identified
some barriers faced by these users. In order to do this, a case study was conducted
Search Engine Accessibility for Low-Literate Users 325
leading us to a list of recommendations for the development of search interfaces fo-

cused on low-literate users.
Besides this introduction, this article is structured as follows: section 2 presents lite-
racy concepts; section 3 presents related work; section 4 presents methodology; section
5 explains how the study was conducted; finally, section 6 presents conclusions.
2 Literacy in Brazil
According to United Nations Educational, Scientific and Cultural Organization

(UNESCO), literacy can be analyzed by different perspectives and conceptualized in
different ways. A common understanding about literacy involves oral, reading and
writing skills, besides abilities with logic, mathematics, symbolic analysis (images
and sounds) and text interpretation [9]. Nowadays it is a common approach to consid-
er that these skills must be contextualized and they are not developed equally among
different individuals. Besides, literacy concept also considers functional aspect that
means the ability to apply oral, reading and writing on different areas of daily life, as
in computing, ecology, health and other areas [9].
The concept of functional illiteracy varies from region to region. In Brazil, federal
agencies as Brazilian Institute of Geography and Statistics (IBGE) adopt the same
criteria as UNESCO to define function illiterates: people between 15 and 64 years
old, which lack mastery of skills in reading, writing, calculations and science, corres-
ponding to an education of less than four years of study [9-10]. In Brazil, it corres-
ponds to incomplete 4th grade of elementary education. A research performed by
IBGE showed that 29 million people in Brazil are functional illiterate that means
more than 20% of the total population [10].
In this work, we adopt this concept but we call the participants as “low literate us-
ers”, once one of the protocols used to guide the study recommended not to call them
functional illiterates [8].
There are other criteria that could be used on the study instead of considering only
age and years of study on formal education. There are institutions that developed
specific literacy and numeracy tests, with levels of difficulty and punctuation, to eva-
luate people’s skills in reading, writing, calculations and science. However, these tests
are private, so there are no public tools that allow us to classify users this way [8].
3 Related Work
3.1 Taxonomy of Web Search

Web search can be categorized in three groups: (a) informational, when users want to
find more information about a topic; (b) transactional, when users want to perform
operations after searching, as downloading a file; (c) navigational, when users want to
find a site from some organization they already know [4], [11-12].
326 D.M. Modesto, S.B.L. Ferreira, and A.S. Alves
3.2 General Users’ Behavior
Some studies address general users’ behavior when using search engines: On informa-
tional or transactional queries, users usually focus on the title and description of a
search result. Images are also an expected type of media, but videos can distract users
once it is not possible to comprehend its full meaning quickly [4]. On navigational
searches, users tend to ignore results from the fourth position on and they focus also
on the URL [4].
If users do not find what they want on first results, they tend to perform another
search [13]. Users are usually influenced by suggested results’ relevance, identified
through positioning [14-15].
Another study identified that experience influence users’ behavior. Novice users
adopt search strategies less flexible than expert ones, have difficulty to formulate a
query and do not know how search engines work [16]. Another study stated that only
a few users know how to use advanced features [17].
A study about terms suggestion showed that most users preferred to refine the
query manually [18]. On other study some people stated that this feature was helpful
but it was also a distraction. However, after a week of log analysis, users started to
use it in an iterative way [19].
Users usually do not know how the search engine works. It creates incorrect expec-
tations about outcomes. Explanations presented about how search engine works helps
users to understand outcomes [20].
Some studies showed that grouping results by categories is a good way to present
them [21-22]. Present a good description of the outcomes is also helpful. Some stu-
dies showed that summarization of the page and highlighting terms were good to im-
prove result analysis [23-24].
3.3 Low-Literate Users’ Behavior

Low literate users usually do not check correctness of information [6]. These users
usually cannot scan results as high literate users do [4], so they need to read every
word to understand information [6]. They tend to have more difficulty recovering
from errors or changing search strategy. They also become confused when navigating
in pages full of information and links. Their performance is usually worst when com-
pared to high literate users, since they take longer to finish tasks, spend more time on
the same page and visit more pages [6].
A study stated that they need features that support decision about results such as in
which link they should click or if information is relevant or not. The same study men-
tions that this kind of feature is more important than the ones to support them to de-
fine query [25].
No study was found about how search features influence low-literate users. This
paper shows how search features affect low-literate users’ experience, mainly Brazili-
ans whose first language is Portuguese.
4 Methodology
This was a qualitative and exploratory research. The research took place in Rio de Janei-
ro, Brazil, and consisted of a case study to analyze the interaction of ten low-literate
users with Google search engine through user tests. The participants were between 15
and 64 years old and had less than four years of study on formal education [9]. They
were all Brazilians from different regions of the country.
This approach can be considered limited because it does not consider extracurricu-
lar capabilities and years of study are related to a feasible educational goal by Brazili-
an government [4]. If other criteria were used to select participants, maybe other
results could be found.
Tests were operationalized through "Protocol for conducting usability testing with
a focus on accessibility" [26], that defined steps to accomplish planning, preparing,
conducting and reporting results, and "Protocols for Web accessibility evaluation
involving functional illiterates" [8], that details the approach and execution of the user
tests.
People were recruited in schools with youth and adults education classes, churches
and residential buildings. As recommended by the protocol [8], a portable usability
lab was set and used on the tests that were performed in various locations to facilitate
transportation for participants.
A questionnaire was applied in order to find more information about users’ profile.
Based on this, Google was chosen for observation once it was the main search engine used
for all participants. Besides that, search volume on this search engine corresponds about to
66% of all searches on the Web [27-29]. In 2011, Google Brazil (www.google.com.br)
reached more than 92% of searches performed by users in Brazil [30].
The search engine was explored on its default state and all features were available
with no customization. A limited set of features related to activities such as writing,
reading and formulating search, guidance, navigation and feedback were selected for
analysis. This set consisted of: auto complete, spell checker, related searches, ad-
vanced search, filters, layout (header, search bar, advertisements, pagination and foo-
ter), search results, keyboard navigation, "I'm Feeling Lucky" button, and Google
Instant features (as page and result preview). Some aspects such as simplicity of the
text, amount of terms used, how they elaborate a query, perception and orientation
were also analyzed.
The case study consisted of two units of analysis. On the first unit five users inte-
racted with Google search engine to perform five tasks that varied by difficulty level.
The first three tasks were considered easy and it was expected that all participants
concluded all of them to finish the test. The fourth task was considered a medium task
since it involved notions of magnitude, as maximum and minimum and advanced
vocabulary. The fifth one was considered a difficult task since it involved notions of
history, current events and interpretation skills. Tasks also varied by search goals,
since three of them were informational tasks, one was navigational and one transac-
tional. Test could finish after five tasks concluded or thirty minutes, what have
happened earlier. After the test, each user was interviewed and answered questions
about perception of the features´ utility.
On the second unit, five other users interacted with two resources that were not
used on the first unit of analysis by anyone, but were considered useful: filters and
advanced search. Before performing each task of this unit, a video was shown teach-
ing how to use each feature.
Textual material used in the study, including the questionnaire, tasks and informed
consent, were prepared with the assistance of a checklist for plain writing for Web
[31]. Data was analyzed following Four-Phase Framework for Search [32]. This
framework states that every search consists of four main phases that are formulation,
action, review of results and refinement. Features were grouped on these categories
and analyzed according to each phase´s goals.
5 Case Study
Three men and two women participated of each unit of analysis. Data collected on
questionnaires indicated that eight users were less than two years of experience with
internet and two were less than five years. Despite the wide range of age adopted as
criteria, we selected users who had similar education and experience with computers
in order to minimize a possible bias caused by age difference. Five participants said
they usually ask for help when using search engines. Eight participants indicated that
the main difficulty is to know whether a word is spelled correctly. The result analysis
was also considered a difficult task by six participants. On the other hand, seven par-
ticipants stated they were comfortable when they need to formulate a query. Regard-
ing to search topics, products were the most mentioned (cited by eight participants),
followed by music and videos (both mentioned by seven participants).
5.1 Recommendations for Search Engine Development Focusing on

Low-Literate Users
Below are presented the recommendations developed based on observation of low-

literate participants during user tests:
1. Provide features that help writing and problem formulation (like "spellchecker",
"autocomplete" and "related searches"). Show changes made by the "spellcheck-
er" as soon as the results are presented, positioned above the results, so they are
quickly visualized. Present results corrected by "spellchecker" and provide feed-
back, indicating that terms were changed. Show terms suggestion to complete the
query while user types next to the search field to enable quick viewing. Show the
feature "related searches" below the results for easier query refinement.
2. Provide "filters" not only for refinement, but also for formulating the query. In
both cases, filters should be formatted as categories or as a menu. Place "filters"
in areas with less emphasis, such as the header of the page. Present results related
to "filters" or categories in a different format from conventional results.
3. Provide a large text box to write the query. The terms that the user typed should
not be hidden so he does not forget the words he used and do not get confused
about the research problem.
4. Low-literate users cannot handle too much information at the same time and get
confused with lots of text, so display around seven main results at a time (at least
five and at most nine). In general, people feel more comfortable to handle this
amount of results [33]. Secondary outcomes that lead to internal pages of a web-
site can be displayed once they not hinder the understanding of users, but should
not be excessive.
5. Present the results divided into pages and display a paging feature to navigate be-
tween them, positioned at the end of the results page. Also use the expression
"see more" beyond the page numbers, because this term is more familiar to users.
6. Provide a visual indication of results' relevance, which is not only showed by the
page rank and positioning on the page. Positioning has not a clear meaning for
these users.
7. Do not present other media formats or filtered results among the conventional
results because it confuses users. Allocate a page area to present this kind of
outcome.
8. Display the title and description for a result emphasizing the first one. Show other
information as the URL on demand, only if user requests. Generally, low-literate
users do not visualize this information and do not use it to decide whether to click
on a result. Search terms should be highlighted on the description, in order to
keep user focused on the subject of task. Distinguish snippets extracted from dif-
ferent parts of a website through background colors, for example. Avoid using
suspension points for that.
9. Features that provide instant feedbacks are recommended, however, should be
prominently displayed so they can be readily seen.
10. Provide features to help users to decide whether or not to select a result. These
inputs should be showed only on demand. For example, "page preview" feature
provides inputs to the user to decide whether to select a result, but it's not used
by low-literate users once they cannot comprehend what is in the page without
reading carefully its content.
11. Advanced features such as "keyboard navigation", can be available if the inter-
face is also used by advanced users. However, low-literate users do not make use
of these resources.
12. The footer area of the results page is less visualized. Provide information that
does not need emphasis in this area.
13. Use tips and directions about the use of the interface so that it does not distract
users nor overload the page with lots of information.
6 Conclusion
During this research, a set of recommendations was developed, considering low-

literate users’ preferences, such as writing instead of reading and features that don´t
impact performance. They also addressed issues related to user’s needs such as: (1)
features to assist them in writing and formulating search problem, (2) minimize dis-
tractions, (3) provide clear and visible feedback, (4) provide instructions about search
engine behavior, (5) facilitate recovery from errors, (6) provide ways to stay focused
on one activity at a time, (7) decrease the amount of text and results, (8) organize
search results by categories.
These recommendations still need a validation since it was not addressed on this
study. Despite the similarity with other search engines, generalization should be carried
carefully once no tests on other tools were performed to validate the guidelines at this
time. These recommendations can assist developers in creating interfaces for search
engines or search features within websites. It is expected that interaction of low-literate
users is enhanced on this kind of tools and they find information more easily.
References
1. Nielsen, J., Loranger, H.: Usabilidade na Web: Projetando Websites com Qualidade. El-
sevier, Rio de Janeiro (2007)
2. Ferreira, S.B.L., Nunes, R.R.: e-Usabilidade. LTC, Rio de Janeiro (2008)
3. Kodagoda, N., Wong, B.: Effects of Low & High Literacy on User Performance in Infor-
mation Search and Retrieval. In: Proceedings of the 22nd British HCI Group Annual Con-
ference on People and Computers: Culture, Creativity, Interaction, pp. 173–181. The Brit-
ish Computer Society, Swinton (2008)
4. Thurow, S., Musica, N.: When Search Meets Web Usability. New Riders, Berkeley (2009)
5. Hearst, M.A.: Search User Interfaces. Cambridge University Press, New York (2009)
6. Kodagoda, N., Kahan, N., Wong, W.: Identifying Information Seeking Behaviors of Low
and High Literacy Users: Combined Cognitive Task Analysis. In: Proceedings of NDM9,
the 9th International Conference on Naturalistic Decision Making, pp. 347–354. The Brit-
ish Computer Society, London (2009)
7. Gupta, N.K., Rosé, C.P.: Understanding Instructional Support Needs of Emerging Internet
Users for Web-based Information Seeking. Journal of Educational Data Mining 2, 38–82
(2010)
8. Capra, E.P.: Protocolos para Avaliação da Acessibilidade Web com a Participação de
Analfabetos Funcionais. Master Dissertation. Universidade Federal do Estado do Rio de
Janeiro, Rio de Janeiro (2011)
9. United Nations Educational, Scientific and Cultural Organization.: Understandings of Literacy.
In: The Education For All Global Monitoring Report: Literacy for Life, pp. 149–159 (2006),
http://www.unesco.org/education/GMR2006/full/chapt6_eng.pdf
10. Instituto Brasileiro de Geografia e Estatística: Síntese de Indicadores Sociais: Uma Análise das
Condições de Vida da População Brasileira. (2010), http://www.ibge.gov.br/home/
estatistica/populacao/condicaodevida/indicadoresminimos/
sinteseindicsociais2010/SIS_2010.pdf
11. Broder, A.A.: Taxonomy of Web Search. ACM SIGIR Forum 36(2), 3–10 (2002)
12. Manning, C., Raghavan, P., Schütze, H.: Introduction to Information Retrieval. Cambridge
University Press, Stanford (2008)
13. Spink, A., Jansen, B.J.: A Study of Web Search Trends. Webology 1, (2004),
http://webology.ir/2004/v1n2/a4.html
14. Keane, M.T., O’Brien, M., Smyth, B.: Are People Biased in Their Use of Search Engines?
Communications of the ACM 51, 49–52 (2008)
15. Bar-Ilan, J., Keenoy, K., Levene, M., et al.: Presentation Bias is Significant in Determining
User Preference for Search Results: A User Study. Journal of the American Society for In-
formation Science and Technology 60, 135–149 (2009)
16. Hölscher, C., Gerhard, S.: Web Search Behavior of Internet Experts and Newbies. Com-
puter Networks 33(1-6), 337–346 (2000)
17. Machill, M., Neuberger, C., Schweiger, W., Wirth, W.: Navigating the Internet: A Study of
German Language Search Engines. European Journal of Communication 19, 321–347 (2004)
18. Anick, P.: Using Terminological Feedback for Web Search Refinement: A Log-based Study.
In: Proceedings of the 26th Annual International ACM SIGIR Conference on Research and
Development in Informaion Retrieval (SIGIR 2003), Toronto, Canada, pp. 88–95 (2003)
19. Anick, P., Kantamneni, R.G.: A Longitudinal Study of Real-time Search Assistance Adop-
tion. In: Proceedings of the 31st Annual International ACM SIGIR Conference on Research
and Development in Information Retrieval (SIGIR 2008), Singapore, pp. 701–702 (2008)
20. Muramatsu, J., Pratt, W.: Transparent Queries: Investigating Users’ Mental Models of
Search Engines. In: Proceedings of the Twenty-fourth International ACM Conference on
Research and Development in Information Retrieval, New Orleans, LA (2001)
21. Käki, M.: Enhancing Web Search Result Access with Automatic Categorization. Ph.D.
dissertation, University of Tampere, Finland (2005)
22. Käki, M.: Findex: Search Result Categories Help Users When Document Ranking Fails.
In: Proceedings of the SIGCHI Conference on Human Factors in Computing Sys-
tems (CHI 2005), Portland, Oregon, USA, pp. 131–140 (2005)
23. Tombros, A., Sanderson, M.: Advantages of Query Biased Summaries in Information Re-
trieval. In: Proceedings of the 21st Annual International ACM SIGIR Conference on Re-
search and Development in Information Retrieval (SIGIR 1998), New York, pp. 2–10
(1998)
24. White, R.W., Jose, J., Ruthven, I.: A Task-oriented Study on the Influencing Effects of
Query-biased Summarization in Web Searching. Information Processing and Manage-
ment 39(5), 707–733 (2003)
25. Gupta, N.K., Rosé, C.P.: A Foray into Understanding the Next Billion Search Users (2010b),
http://www.cs.cmu.edu/~nkgupta/papers/chi2010a_submitted.pdf
26. Henry, S.: Just Ask: Integrating Accessibility Throughout Design (2007),
http://www.uiaccess.com/accessucd/
27. Experian. Search Engine Analysis, http://www.hitwise.com/us/
datacenter/main/dashboard-23984.html
28. Comscore. comScore Releases U.S. Search Engine Rankings (January 2012),
http://www.comscore.com/Press_Events/Press_Releases/2012/2/
comScore_Releases_January_2012_U.S._Search_Engine_Rankings
29. Netmarketshare. Desktop Search Engine Market Share,
http://netmarketshare.com/search-engine-market-
share.aspx?qprid=4&qpcustomd=0
30. Serasa Experian. Share Mensal de Buscas do Google Atinge 92,15% em Janeiro,
http://www.serasaexperian.com.br/release/noticias/2011/
noticia_00381.htm
31. Barboza, E., Nunes, E.: A Inteligibilidade dos Websites Governamentais Brasileiros e o
Acesso para Usuários com Baixo Nível de Escolaridade. Inclusão Social 2(2), 19–33
(2007)
32. Shneiderman, B., Byrd, D., Croft, W.B.: Clarifying Search: a User-Interface Framework
for Text Searches. D-Lib Magazine, 1–15 (1997),
http://www.dlib.org/dlib/january97/retrieval/
01shneiderman.html#formulation
33. Miller, G.A.: The Magical Number Seven, Plus or Minus Two: Some Limits on our Ca-
pacity for Processing Information. Psychological Review 101(2), 343–352 (1955)
Identifying and Representing Elements of Local
Contexts in Namibia
Kasper Rodil1 , Kasper Løvborg Jensen2 ,

Matthias Rehm1 , and Heike Winschiers-Theophilus2
1
Dep. of Architecture, Design and Media Technology
Aalborg University
Denmark
{kr,matthias}@create.aau.dk
2
Mobile Future Lab and Dep. Software Eng.
Polytechnic of Namibia
Namibia
Abstract. In an attempt to represent local context in a 3D visualisa-

tionfor rural elders in Namibia we have foundmajor differences in thecon-
ceptualizationof this context between external and local partners in the
co-creation process. Through the evaluation of a mobile context capture
tool we found a clear disconnection of community members with both
abstract and absolute representations of points, paths and areas. From
this we discuss how the local concepts of space and time as frames of
reference can not be represented adequately with our current selection
of contextual data, and how we are engaging in participatory activities
to derive a common understanding of contextual representations.
Keywords: context, indigenous knowledge, Participatory Design,

context-aware, re-contextualization.
1 Introduction
Rural Herero communities in Eastern Namibia have for many years been self-
contained in terms of transfer of local knowledge across generations. Children
and youth have been listening to their elders, participating in chores and prac-
tical work in the villages, thereby sustaining an often tacit and uncodifiable
knowledge. In Wenger’s terms it could be considered a ‘repertoire’ of indigenous
knowledge in ‘communities of practice’ [1]. These rural communities perform
actions within their own cultural context. There are obvious local benefits in
transferring knowledge on husbandry, herbal lore etc. and the tacit knowledge
transferred through intra-personal interaction effectively adds to preserving local
culture, customs and traditions. Due to formal education the youths are increas-
ingly detached from their cultural traditions and context. The majority of the
youths attend schools often far away from the villages, where many only return
on holidays. The curriculum is compliant with international standards incon-
siderate of local conditions. Valuable indigenous knowledge is lost day by day,

Identifying and Representing Elements of Local Contexts in Namibia 333
and future generations could very well suffer from this absence of cultural roots
and traditions, such as lack of self-awareness, cultural adaptation and self-worth.
Moreover a major risk is the impact on the ecosystem in which the Herero tribe
has lived sustainably since their settlement in Namibia in the 17th century.
Since 2008 we have been in close collaboration with a group of Elders (knowl-
edgeable men respected in their community) in a village in Eastern Namibia.
The overall objective of the project is to preserve local knowledge, but also find
ways to transfer parts of that knowledge to de-situated youths from the region.
Early in the process we have come to terms with the different world views of
us, primarily Western trained designers, and local Herero elders. We have em-
phasized a dialogical approach with intensive collaboration and co-design. Our
methodological stand is within Participatory Action Research (PAR).
In the ethos of PAR lies the acknowledgment by designers that they are from
the onset limited in their understanding of users and the users’ context. As re-
ported by Nielsen et al.(2003) [2] one of the principles of PAR is mutual learning.
Co-designers (in this case Elders) acquire technological skills meanwhile design-
ers attempts to fine tune their sensitivity to the new setting and the skills and
knowledge within that domain. The differences are in particular revealed when
the participants (co-designers and designers alike) have their origin in different
epistemologies. As Participatory Designers we attempt to understand the con-
text through a shared perspective. Limitations will unavoidably occur, but we
must investigate were these cut-offs manifest in the design of context-aware sys-
tems. As we have reported earlier, the foundations on which we design interfaces
from might have little use in cross-cultural collaborations [3], [4]. Thus one of
our approaches has been to substitute traditional (Western) interface metaphors
with localized metaphors. Yet we hypothesize that it may still be desirable to
explore these differences in the interaction design to overcome HCI issues of
both textual and computer literacy and unnatural interface conventions such as
folders, menus and files born out of the earlier computing days.
2 What Is a Context?
The meaning and use of context in computing systems and HCI has been debated
for decades. Abowd et al. (1999) provided an operational definition of context
within the scope of context-aware computing as: “Context is any information
that can be used to characterize the situation of an entity. An entity is a person,
place, or object that is considered relevant to the interaction between a user and
an application, including the user and applications themselves.”[5].
When introducing context-awareness, Schilit et al. (1994) alluded that the
notion of context is dynamic and goes beyond static concepts such as location
[6], e.g. lighting, noise level, network connectivity, communication costs, com-
munication bandwidth and even the social situation is part of the context. Thus
a context is also defined by people and their actions, at least if we include peo-
ple interacting with artifacts. Dourish (2004) describes context as arising from
activity:
334 K. Rodil et al.
“Context isn’t just ‘there’, but is actively produced, maintained and enacted in
the course of the activity at hand”[7]. Even just from these three perspectives on
‘context’ we can establish that the characteristics and description of any context
is highly complex. We can consider it to be a temporal state of a changing space
influenced by those that occupy and interact with and within it.
Thus it is impossible to capture and represent an entire context or expect
to experience the exact same state once again. It is optimistic to think that a
system is able to capture the complexity of an ‘interwoven’ situation of actions
performed in space and time, as we always make a choice of which aspects of a
context we capture. As Grudin (2001) explains on capturing context digitally:
“The context that is captured is removed from its context, namely the context
that is not captured.”[8].
This implies varying degrees of implications for artifacts using ‘some’ of the
information available from the context. An application that address users based
on identity, their activities and their location might not require extra information
from the context they are in to achieve the set objectives. For instance, a mobile
application that identifies a user then pulls the user’s characteristics, activity
and location to then push a suggestion of the route to a preferred dining place.
As described by Baldauf et al. (2007) there are a variety of sensors available
to further interpret the context (e.g. biosensors, thermometers, cameras etc.)[9].
Though it is of high practical value in mapping and modeling efforts, but the
arguably “objective” snapshots of the perceivable space-time continuum often
offers a poor reflection of the messy, subjective world of cultural, historical and
spiritual meanings attributed to objects, places and situations by human beings.
The objective of capturing a context in order to represent it digitally while
actually producing a new context is a delicate matter. Especially when engaged in
cross-cultural design of artifacts and preservation of cultural knowledge. Instead
of taking the pessimistic stance of not being able to capture the whole context,
we could examine the affecting elements defining and capturing a context.
3 Context Discussed through a Case Study

Since 2010 we have investigated the potential of 3D visualizations as means for
preserving local knowledge. The main driver for the 3D visualization approach
is to capture the local physical context into a visible medium unrestricted from
requirements of local literacy and being transformable into digital experiences
as virtual worlds or serious games. A cornerstone in the project is to enable
rural elders to capture landmarks and objects of significant importance (houses,
fenced areas, water holes etc.) at the research site through tagging with external
sensors. This data set will represent the contextual backbone of recreating the
village as a 3D representation for de-situated youths to explore.
An example of use could be that the youths take an online virtual tour of
the village while listening to their grandfathers telling locally collected stories
at places represented in a virtual context yet separated in time and space from
the recording. A cardinal point is whether the realism of sensor-captured GPS
coordinates and compass bearings can translate into contextually “accurate”
experiences. Thus it is fundamental to the research of representing a context,
that we can pin-point some of the challenges to facilitate knowledge preservation
and knowledge transfer. In this article we refrain from discussing the limitations
of 3D visualization. As will be highlighted later, the action of representing a
context through an incomplete perspective into a different form is not without
implications.
4 CARACAL: A Tool to Capture Context
CARACAL is a context-aware tool designed for mapping and tagging objects

and places in the field using sensor-enabled mobile devices [4]. Initially the tool
was developed as an aid to the researchers of the project when engaged with
re-mapping the spatial layout of the village. It was iteratively designed and
implemented over a series of field trips and tuned to optimize the efficiency for
the researcher.
Fig.1 shows the flow of interaction from the prototype. As can be seen the
main screen allows quick access to capturing the core types of data points and
gives feedback to the user as data is entered. CARACAL logs spatial information
as either points(e.g. objects), paths(e.g. walking routes) or polygons (e.g. home-
steads), which can each be tagged with additional information. It also allows for
the capture of rich media through pictures, video or audio which is automatically
Geo-tagged using the GPS, electronic compass and further allows entry of meta
data by the user.
Fig. 1. The figure shows the CARACAL interface and flow of interaction
While a study showed that this approach was feasible for researchers and that
we could indeed gain a lot in terms of efficiency and better data sets, we also
found that there were several important shortcomings. Firstly, it would never
336 K. Rodil et al.
scale to the level we would want to use it due to the shear amount of informa-
tion to be captured and tagged in even a relatively small village. Secondly, we
need to capture the “right” data, i.e. the information that has meaning and is
important to the local users. It was thus evident that local participants needed
to be included and that the optimal solution would be a tool that would enable
them to capture and map out their own environment according to their local
perspective.
We conducted an exploratory field study of the application with a number of
participants from different groups in the community and got mixed results.
As it turned out, the best performers were the youngest participants as they
were more receptive to the instructions and fast to learn to interact with the
application. The elders experienced much more challenges as they had a harder
time with the touch screen based phone.
The biggest problems were however of more conceptual nature. It was evi-
dent that there were plenty of opportunities to enter ‘erroneous’ data because
the original application was developed from the researchers’ understanding and
mental model of what the data should be used for. For example, this resulted in
wrong Geo-tagging and or missing information about the captured pictures.
We also conducted an unsupervised field trial with a single user traveling
to his home village in the North of Namibia. He used the application to track
walks in the area while listening to an elder from his local community. The
user is an IT master student and fairly proficient with mobile technology, so
the fact that he did not report any serious problems with the application was
not a surprise. The most interesting result from this part was the data itself,
showing a spatio-temporal flow through the environment and context as they
walked around. The data was not comprehensive enough that we could actually
recreate the environment but simply overlaying it onto a satellite map in Google
Earth gave an insight into that space (see Fig.2). What we also found was that
the story behind the walk and the points tagged on the way were missing. Thus
a limitation of the data is the lack of narration and re-contextualization.
Fig. 2. The figure shows the student’s and elder’s walk

5 Discussion of Capturing and Representing the Physical

Domain
5.1 Spatial
The coordinates obtained from the CARACAL are supposed to be used for in-
stantiating objects at the ‘right’ places and with a Cartesian mapping in a 3D
world. But as reported in the evaluation of the interface:“The key findings were
however, that the conceptual idea of ‘points’ and underlying purpose of ‘tagging’
was not well understood by the participants.” [4]. The participants (elders) were
not: “distinguishing between taking a picture of a landmark from a distance and
tagging the very same from close proximity” [4].
Thus within the evaluation of the interface we began questioning if coordinates
perceived to be objective were indeed objective, or if local concepts of space
are different to the Western concept of mapping a space. Pursuing our concep-
tualization and representation carries the risk to override a tacit cultural way
of understanding space. This might not have severe implications in many other
application domains, but within our project the very objective is to represent
cultural and local knowledge. Research suggests that there indeed might be a
difference in cognition of space across cultures. Haun et al. (2011) describe stud-
ies investigating cultural differences of spatial cognition and spatial language
between Dutch children and children from the Akhoe Hai//om hunter-gatherer
group of Northern Namibia [10]. The Dutch children have a dominant ego-centric
cognitive frame of reference, where the participating Akhoe Hai//om children
have demonstrated a dominant Geo-centric (absolute) frame of reference. The
former preference is to imagine the coordinate system centered in the ego. The
latter is where the coordinate system is absolute with cardinal directions (N, E,
S, W). This has an impact on how people orient themselves in the world, thus
if a mapping into a new 3D geometric space is constructed, the mapping might
need re-alignment according to local concepts.
Thus the action of capturing absolute ‘points’ on a global grid presents the scaled
distance between objects and activities as a fixed and measurable length. A fun-
damental question is whether there are local differences in perception of distance
between points. While we have not conducted formal studies as of yet, we ask
ourselves if activities, paths or places within the village are locally perceived to
be of a different scale? This could relate to differences in considered significance
of paths over places, or vice versa. While it remains a speculation for now, we
believe that in order to capture a space we must make sure we also represent
the distances. Based on some conversations with community members we believe
that a local perception of distance is that objects within a homestead are con-
sidered being almost at the same place, whereas everything situated outside is
considered far away. Thus we attempt to acknowledge this in our representation.
5.2 Temporal
Another, yet unresolved, parameter to be represented is the concept of time.
While traditional context-aware systems are easily able to log time of events,
338 K. Rodil et al.
time between events etc. the fundamental question is: do Western academics
have a different perception of time than the co-designers and the end-user group?
Should time be used as a convenient separation of activities (as in meeting cal-
endars, time tables etc.) or is activity and time even connected and to what
extent? Research shows that local concepts of time can have different charac-
teristics. Janca and Bullen (2003) explain that: “the Aboriginal concept of time
differs from the Judeo- Christian perception of time in that Aboriginal people
do not perceive time as an exclusively ‘linear’ concept (i.e. past–present–future)
and often place events in a ‘circular’ pattern of time according to which an in-
dividual is in the centre of ‘time-circles’...”. [11]. We have experienced that time
is perceived differently in the village than e.g. any random day at our academic
institutions. Sometimes we wait hours for the participants in the village although
having agreed to meet at a certain place at a specific time. As we put ourselves
under pressure to collect data during the relatively short village stays, we often
struggle accepting delays relative to our own plans. However, we must remem-
ber that our research interferes with their daily routines and activities. It is also
important to note that while we can consider ending a meeting due to another
meeting pressing on in the calendar, they might not end theirs because there are
still important matters to talk about. While time can be a fix point for activities
we have experienced that social activities in the village are often defining time
in the village. From a methodologically perspective we try to adapt to the local
‘rhythm’ of doing things. It shows us how we perceive time as a separation of
activity. Practically it means that when we design databases, we should be cau-
tious to taxonomically appoint time as the separating factor of events/activities.
If we assume a re-contextualization to be truthful we should be careful not to
override local perspectives with Western conceptualizations.
The purpose of the examples here is not only to elucidate a potential difference
in concepts, but to stress the divergence of the underlying framework of reference.
Thus if we as designers presuppose something to be a fact or common ground
we might lose sensitivity to the fact that it might not be universal.
6 Participatory Design for Synthesizing the Context
When discussing the use of context in computer systems Dey (2001) provides a
description of how humans have some success in transferring concepts to each
other:
“Humans are quite successful at conveying ideas to each other and reacting
appropriately. This is due to many factors: the richness of the language they
share, the common understanding of how the world works, and an implicit un-
derstanding of everyday situations.” [12].
Most of us are familiar with the saying:“it must be understood in the context
that...”. This statement implies that individuals in dialogue indeed are talking
about the same state if sharing similar perspectives. Yet people having shared
experiences can argue about what has happened, or the meaning of something
although they were both part of the same event. This seemingly trivial example
on subjectivity is much more prevalent in a cross-cultural project due to the obvi-

ous background difference between parties. Differences in language, world-views,
value beliefs and a collaboration containing two distinct cultural traits as either
collective or individual oriented leads to substantial divergence of perception and
desirable representations.
In a classical example Quine (1960, Pp. 29) analyses the difficulties of es-
tablishing a common ground and a common understanding about the context
of the conversation, when you do not speak the same language [13]. His exam-
ple is called the Gavagai problem: Imagine walking in the bush with a Herero
guide and suddenly he points at a rabbit and says ”Gavagai”. Quinn’s argu-
ment about the indeterminacy of the situation (and possible translation of the
utterance) takes into account that it is possible to derive several conflicting rep-
resentations based on the observable data, e.g. Gavagai could mean rabbit, or
dinner, or actually the bush under which the rabbit is sitting. Additionally, he
states that there is no way to know which of the competing interpretations is
the correct one, which poses a slightly ontological problem, when you assume
that an objective representation (some kind of ground truth) exists (see [14] for
an in-depth discussion).
From these examples it might be too ambitious to expect parties to have a
similar notion of the context under dissection. Thus, there will consistently be a
mismatch on interpretation and representation. While seemingly impossible to
resolve we have tried to understand the differing perspectives through Mutema
(2003, Pp.5), who explains that the goal for interpretation is by fusing horizons:
“The inter-subjective nature of the research process allows for the researcher’s
interpretations to be checked, reinterpreted and evaluated by the actors.” [15].
This perspective on understanding difference is emphasized in our participa-
tory design approach. The arguably objective nature of the shared interpretation
is in the hands of the parties involved. In this example we could explain it like:
we attempt to increase our sensitivity to interpret an unfamiliar context through
dialogue. Meanwhile the participants acquire skills on technology and design in
order to be critical towards the shared representation. We have reached a stage
in the project where the elders are indeed critical towards some of the proto-
types. But we have experienced in many cases that at the point of critique new
knowledge about the context is brought forward. The following example is from
recently published work and highlight the value of the local dialogue [3]. We
considered the concept of the trash can desktop icon as not being universally
understandable. We re-constructed this icon into looking like a hole in the ground
which we saw a local researcher from the village used to dispose garbage in. In the
evaluation of this metaphor we were made aware by the elders that they do not
throw away, but keep those parts for later use. Thus without the local evaluation
of an interface action we would probably not have learned about that particu-
lar view point of not throwing away. Needless to say which consequence that
misinterpretation would have for a knowledge transfer tool preserving cultural
viewpoints. The misrepresentation is our context unaligned conceptualization.
They suggested us to implement that the objects targeted for deletion should
340 K. Rodil et al.
be moved back into the menu from where it came. We argue that the next de-
sign iteration being informed from this evaluation is not a representation from
either part, but a product of fusing horizons. Although we remain cautious to
any impact new concepts might have on their context.
7 Conclusion
We have argued, given the application domain of transferring indigenous knowl-
edge and local culture, that designers should be wary that transfer of partial
information from one context to another might have implications on the de-
velopment and research objectives. We have argued through experiences with
a Herero community in Namibia, that perceived value-free and objective mea-
surements might produce a distortion and that those measurements should be
investigated further. That interpretation as an activity is in the hands of all
actors –participants and designers.Our proposed solution is participation and
inclusion in the design process through dialogue.
Acknowledgements. We wish to thank the community members for their

commitment and friendships.
References
1. Wenger, E.: Communities of practice: Learning, meaning, and identity. Cambridge
University Press, Cambridge (1998)
2. Nielsen, J., Dirckinck-Holmfeld, L., Danielsen, O.: Dialogue Design-With Mutual
Learning as Guiding Principle. International Journal of Human-Computer Inter-
action 15(1), 21–40 (2003)
3. Rodil, K., Theophilus, H.W., Jensen, K.L., Rehm, M.: Homestead creator: a
tool for indigenous designers. In: Proceedings of the 7th Nordic Conference on
Human-Computer Interaction: Making Sense Through Design, NordiCHI 2012,
pp. 627–630. ACM, New York (2012)
4. Jensen, K.L., Theophilus, H.W., Rodil, K.: Tapping into local lore: toward scal-
able local mapping and tagging for rural Africa using mobile devices. In: Proceed-
ings of the 7th Nordic Conference on Human-Computer Interaction: Making Sense
Through Design, NordiCHI 2012, pp. 631–634. ACM, New York (2012)
5. Abowd, G.D., Dey, A.K.: Towards a better understanding of context and context-
awareness. In: Gellersen, H.-W. (ed.) HUC 1999. LNCS, vol. 1707, pp. 304–307.
6. Schilit, B., Adams, N., Want, R.: Context-aware computing applications. In: First
Workshop on Mobile Computing Systems and Applications, WMCSA 1994, pp.
85–90. IEEE (1994)
7. Dourish, P.: What we talk about when we talk about context. Personal and Ubiq-
uitous Computing 8(1) (2004)
8. Grudin, J.: Desituating action: Digital representation of context. Human–
Computer Interaction 16(2-4) (2001)
9. Baldauf, M., Dustdar, S., Rosenberg, F.: A survey on context-aware systems. In-
ternational Journal of Ad Hoc and Ubiquitous Computing 2(4) (2007)
10. Haun, D., Rapold, C.J., Janzen, G., Levinson, S.C.: Plasticity of human spatial
cognition: Spatial language and cognition covary across cultures. Cognition (2011)
11. Janca, A., Bullen, C.: The Aboriginal concept of time and its mental health impli-
cations. Australasian Psychiatry 11(s1), S40–S44 (2003)
12. Dey, A.K.: Understanding and using context. Personal and Ubiquitous Comput-
ing 5(1), 4–7 (2001)
13. Quine, W.V.O.: Word and Object. MIT Press, Cambridge (1960)
14. Raatikainen, P.: On how to avoid the indeterminacy of translation. The Southern
Journal of Philosophy 43(3), 395–413 (2005)
15. Mutema, G.: Phenomenology, hermeneutics and the study of indigenous knowledge
systems. Indilinga: African Journal of Indigenous Knowledge Systems 2(1) (2003)
A Framework for Community-Oriented Mobile
Interaction Design in Emerging Regions
Monica Sebillo1, Genoveffa Tortora1, Giuliana Vitiello1,

Pasquale Di Giovanni1,2, and Marco Romano3
1
Department of Management and Information Technology, University of Salerno, Italy
{msebillo,tortora,gvitiello}@unisa.it
2
School of Computer Science and Informatics, University College Dublin, Ireland
[email protected]
3
Universidad Carlos III de Madrid, Spain
[email protected]
Abstract. The interaction design of mobile applications for a specific commu-

nity of users, requires a deep knowledge of that community in order to establish
appropriate usability requirements and tune the subsequent development activi-
ties. This is especially true when the target community is situated in one of the
developing countries. There, mobile devices are increasingly playing the role
that personal computers play in ‘developed’ countries, in spite of quite different
cultural, social and technological constraints. In this paper we propose a ‘com-
munity-centered’ design approach, where different aspects of a community are
used to formulate usability goals and are taken into account throughout the de-
sign, prototyping and testing phases, so as to ensure the actual application dep-
loyment by users of that community. The case study of a project carried out
with a community of farmers in Sri Lanka is used to illustrate the approach.
Keywords: Analysis and design methods, Human Centered Design and User
Centered Design, Human Factors Engineering Approach, Interaction design
1 Introduction
In the last few years several design guidelines and hci patterns have been issued to
provide guidance in the design of mobile applications addressing common usability
challenges [10-12, 15-18]. For instance, knowledge of cultural norms of reading con-
ventions and how people process information is used to suggest how to design ele-
ments for an interface and place items on it so that users will understand. All such
guidelines and patterns share the basic assumption that mobile users are everywhere,
want to access information quickly and to be able to manipulate it easily. Mobile
users are in the design focus and the adoption of common best design practices is
recommended to guarantee usability with users who keep moving around with their
devices and whose focus on the screen can be frequently distracted by the surrounding
environment.
A Framework for Community-Oriented Mobile Interaction Design in Emerging Regions 343
However, the design of mobile applications for a specific community of users, re-
quires a deep knowledge of that community, that often goes beyond the single user
and introduces cultural and social constraints which may invalidate some of the gen-
eral guidelines. Knowledge of the community, in that case, helps establishing appro-
priate usability requirements and tuning the subsequent development activities on the
community itself. This is especially true when the target community is situated in one
of the developing countries [5-8]. There, mobile devices are increasingly playing the
role that personal computers play in the developed countries in spite of personal, so-
cial and technological constraints.
In this paper we introduce a ‘community-centered’ design approach, where the
social, cultural, technological and economic aspects of a community are used to for-
mulate adequate usability goals and are taken into account throughout the design,
prototyping and testing phases, so as to ensure the actual deployment of the applica-
tion by users of that community (Fig. 1). The paper is organized as follows. In Section
2 we describe the two-layered ethnographic observation method which characterizes
the proposed approach. We provide guidelines for community profiling in the mobile
world that can be used to separate the ‘once-for-all’ community observation activity
from the specific problem domain analysis. In Section 3 we illustrate the community-
oriented design approach on a concrete case study, i.e., the design of a mobile
application for a community of farmers in Sri Lanka.
2 Guidelines for Community Profiling in the Mobile World
The human/sociological/behavioral aspects which characterize a community of people

are usually derived from the observation of their Cultural, Social and Technological
contexts [9]. In the following, a set of guidelines is defined to be used for profiling a
community of people using mobile devices, independently of the specific problem
domain.
• Social Context (Social Organization, Ethical beliefs etc.)

− Consider the social organization of the community. It “describes the collec-
tion of values, norms, processes and behavior patterns within a community
that organize, facilitate and constrain the interactions among community
members”[1]
 Find out the social necessities and limitations.
 Find out social relationships between individual subjects belonging to the
same community.
 Find out possible participation in governative or voluntary organizations.
• Cultural Context (Language, Education)
− Consider the average cultural level of the target community. The designed in-
terface should be easily understood by any stakeholder.
− Consider the average education level of community members. Never make a
design choice based on incorrect knowledge assumptions about the user.
344 M. Sebillo et al.
− Different languagees may affect the use of text in the visual design of thee in-
terface. This is esp
pecially true for communities living in some Eastern cooun-
tries and where thee official spoken language uses an alphabet different frrom
the common Latin alphabet. Consider that:
 sentences could d be hard to represent on small screens,
 the (virtual) keeyboard could be missing some characters of the languuage
alphabet. This would
w again affect the choice of the mobile device.
− In some countries more than one language is spoken. So, consider the necces-
sity to design a muulti-language interface.
− Consider the sem miotics of the target community. Signs, colors, symbols,
metaphors can hav ve different meanings in different environments.
Fig. 1. Community-oriented design and development

• Technological Context (Available technology, Familiarity with mobile devices)

− Consider the technological means available in the geographic area of the
community. Also consider the average degree of familiarity with the mobile
technology you are planning to use and the attitude to learning new technol-
ogies.
− Mobile devices operations often depend on remote services. Therefore con-
nectivity issues are paramount in this context:
 consider the data quantity that the system needs to transmit on the wire-
less networks
 bad connections can cause loss of reliability and can make the application
progressively slower causing usability problems too.
− The application should run on the majority of the devices available in the
community. The world market trends can suggest some devices, however in
specific communities particular technological ecosystems could be found.
− Consider the device models available to that community in the specific
country. Advanced devices may be present in rich communities (but this is
not a rule), in some countries some models could not be available.
2.1 Problem Domain Analysis
Differently from the traditional approach, the analysis of current practices will aim at
observing people from the target community within scenarios related to the specific
problem domain [14]. Therefore, being aware of the common characteristics of the
community of users and of their general needs with respect to the mobile world will
allow designers to focus on the following points:
1. what is expected from the target mobile application

─ what tasks are currently performed?
─ what roles?
─ what artifacts?
─ what interactions?
2. why users need the application

─ motivations for adopting a mobile application supporting current activities
3. what the temporal context of interaction is

─ when is each task accomplished?
─ when should the system be deployed?
4. what the geographical context of interaction is
─ where do the observed activities take place?
Requirements elicitation will come out of the answers to the above questions. In the
following section we describe the community-oriented design methodology applied
on a concrete case study.
3 The Case Study of SLN4MOP
The Social Life Networks for the Middle of the Pyramid (SLN4MoP) is an Interna-
tional Collaborative research program started in 2011 that aims to provide real-time
information to support activities related to livelihood delivered by mobile phone ap-
plications targeted to meet the needs of people in developing countries [19].The goal
of SLN4MOP project is to leverage the enormous reach of mobile phones equipped
with myriads of sensors such as GPS and camera, to develop the next generation of
social networks that not only connect people to people, but also have the capability of
providing real-time, context-sensitive local information by aggregating data from a
variety of sources. In order to move towards that goal, Sri Lanka was chosen as the
country where a pilot research study could start.
In this section we show how our approach can be adopted for the community of Sri
Lankan people who are willing to use mobile phones to improve their daily working
activities. In particular, the case study deals with the design of a mobile application
meant to support farmers in the initial phases of crop selection and planning.
3.1 Profiling the Community of Sri Lankan People

The Sri Lankan society presents opposing aspects: people reflect both some of the
typical facets of the modern advanced western societies and some aspects strongly
connected to a multiethnic culture, full of ancient traditions. As a result of the initial
ethnographic study, we were able to depict a general community profile for people
living in Sri Lanka, who may benefit from the use of mobile devices for their daily
activities [2, 3]. This represents the common knowledge about the social, cultural and
technological contexts, which could be exploited for several application domains.
In the following we present some relevant aspects of the community profile, which
directly impacted some of the design choices made for the farming application.
Cultural Context - Sri Lanka is a real multicultural nation. The community is made
of two main ethnic groups, namely the Sinhalese and the Tamil. 83% of the members
speak Sinhalese and the remaining 17% speak Tamil. However, English is the third
official language, mainly spoken in the cities.
Social Context - Young people have not been extremely influenced by modernity
and the effects of westernization. 80% of them have religious beliefs, the majority
being Buddhists. Mobile devices are quite widespread but there are still barriers to
their adoption, especially among parents who are often concerned with security and
reputation issues. Moreover, people’s general attitude to ICT is often influenced by
the opinion of prestigious members of the community, such as local temple priests.
Technological context - Like in many developing societies, a significant gap in ICT
(digital divide) can be observed in Sri Lanka, especially among people living in
villages, sometimes even missing electric power supply. In order to encourage the
adoption of technology and support the diffusion of culture, the local government has
recently created the so called “Nenasala telecenters” supply [4]. Computer desktops
and laptops are absent for the majority of individuals, while mobile devices are wide-
spread enough. Statistics say that 86.5% of the population owns at least one SIM
(Subscriber Identity Module). Nevertheless, even if the Sri Lankan mobile network
currently covers almost the totality of the urban areas, temporary lacks of connectivi-
ty, may affect services which heavily rely on network uptime.
In this context, the use of ICT and especially mobiles, have the potential to reduce
the information and knowledge gaps in the affected areas.
3.2 Problem Domain Analysis: The Cultivation in Sri Lanka

Agriculture in this developing economy employs the largest share of the workforce;
yet it contributes the least to GDP (Gross Domestic Product) when compared to the
Industry and Services sectors. The cultivation pattern of Sri Lanka is bi-modal. This is
mainly due to the two monsoons, which bring rainfall to the island in two distinctively
different periods resulting in two distinct cultivation seasons [13].
From the survey we conducted with local farmers, it was noticed that due to this
distinction there are specific crops to be cultivated in each season. Other factors that
contribute to the crop selection are type of soil, weather, water, financial status, pest
and diseases. Although several factors may affect the selection of crops for cultiva-
tion, farmers are used to choose crops primarily based on the profit. As a result they
are biased by the market price of the previous year and they tend to select crops that
produce a high yield within a short time period.
None of the interviewed farmers had a specific land extend per crop in mind at the
time of cultivation, though they have a better understanding about the different types
of crops that can be grown on a particular season. This knowledge is gained mainly
based on the practical experience that came via ancestors. Moreover, during the study
it was further identified that farmer awareness with respect to what others grow is
limited to the neighboring farmers. This is mainly gained by watching what others are
growing. They were reluctant to get this information by formal communication due to
social phobia and the competition among farmers.
Different selling mechanisms are adopted by farmers in Sri Lanka. Some bring the
harvest directly to the market, while around 90% depend on a middle person, namely
the transport agent or the shop keeper. However, none of them get help from the gov-
ernment to sell the harvest. Another interesting fact is the behavior with respect to the
selling prices. The selling price is a dynamic value which changes very frequently at a
particular market. The farmers reported that they were often unable to predict the
price as it changed vigorously within few hours. However, it was revealed that, ac-
cording to their experience, they were unable to gain a good price for their harvest at
the market, when all farmers tend to grow the same crop at the same time. As a result,
in some cases farmers gave up farming and joined other industries to find their living.
If this trend would continue in the future, it might have a huge impact for the econom-
ic growth of the country.
In the following, we summarize the most important claims about the domain
specific issues observed from the field study:
− users are disposed to use some technological instruments provided that they are
not invasive;
− governmental centers aimed at supporting agricultural activities are located all

around the farms;
− there is a very low level of trust among the members of the same community,
insomuch as not sharing basic information about their crop production.
3.3 The Community-Oriented Design

The user interface and application functionalities have been designed on the basis of
the social, cultural and technical contexts characterizing the Community Profile of Sri
Lankan people. Based on that, we understood that the adoption of smartphones was
paramount in this work, since users are distrustful of technology with the exception of
mobile devices, which are widespread. Moreover, farmers may need training that
could be provided by experts of governmental centers.
The application was also expected to provide support to a critical task on the basis
of data updated directly by users. Therefore, it was paramount to augment the percep-
tion of the data reliability.
The social and cultural contexts characterizing the profile of Sri Lankan communi-
ty, besides the existing technological constraints, led us to choose the Android operat-
ing system for our mobile application, due to the high variety of device models
running that system that are sold in that country. The technical and cultural issues,
along with the small display size of the devices on which it will be executed also
deeply influenced the choices regarding the design of the application user interface.
Major technical limitations are represented by the low resolution of the screen and
the aforementioned limited sizes, the low computing power and the limited amount of
main memory available. Such limitations, as well as the necessity to preserve as much
as possible the battery life, were a determining factor in our decision to limit the
amount of data processed directly on the smartphone. However, the need to commu-
nicate, via the mobile network, with a remote server has not been a trivial challenge
and a lot of efforts have been made to handle with the payload of our application on
the network. Even if the Sri Lankan mobile network currently covers almost the to-
tality of the urban areas, we had the necessity to find the right tradeoff with regard to
the data exchange modality and the quantity of data transmitted, while limiting as
much as possible the impact that a temporary lack of connectivity, typical of a GSM /
UMTS network could have on the general usability of our system.
More specifically, in order to decrease the total payload of exchanged network data
we adopted two main solutions. First of all, we chose JSON (JavaScript Object Nota-
tion) as a lightweight data-interchange format. JSON essentially allowed us to present
the data choosing the desired level of detail, with less verbosity than XML.
The second solution we adopted is based on a geographical information filter. Data
sent from the remote server are preventively filtered on the basis of the farmers' loca-
tion. The server collects information about products adequate for the soil type and
climatic conditions of the region of interest and about the products which have been
selected by neighbors so far. Such information is then aggregated, combined and
computed in order to send only relevant data through the network. It is worth to point
out that the given solutions addressed a general implementation problem characterizing
the target community and may therefore be adopted in different domains without
further efforts. The same consideration applies to the visual interaction paradigm we
decided to adopt, based on colors and images with little textual descriptions.
From the social context we knew that even if English is spoken among the mem-
bers of our target community, users feel more comfortable using their own language
so we decided to design our interface by allowing multiple languages. At the very
beginning we designed the interface for the English language and then we translated
the text in Sinhalese. We noticed that the graphical representation of this language is
often significantly longer than the English text representation. This caused a visuali-
zation problem and a resulting redesign of some screens of the application that made
the interface auto-adapting to the text dimensions.
At a later stage, the knowledge of the social phobia among farmers, who are reluc-
tant to share with others the information on what crops they are growing, led us to
design the interface so that the needed information could be provided anonymously
and visualized using a ‘traffic light’ metaphor. Figure 2 shows the crops catalog fil-
tered by the farmer's geographic position. For a better readability, a zoomed portion of
the screen is also displayed. The colored background of each crop indicates the ap-
proximate quantity already in production. Figure 3 shows the comparison screen.
A farmer can use this feature if he wishes to get more information about a specific
subset of crops.
Fig. 2. The Crops Catalog

Fig. 3. Crops Comparison
4 Conclusion
The development of mobile applications is paramount to support users living in de-

veloping countries to improve their lives. However, a deep understanding of the target
community is crucial for the realization of an application that is effectively accepted
by those users. In this paper we propose a community-centered mobile design metho-
dology, based on the idea that relevant aspects of the community can constitute a bulk
of knowledge upon which mobile interaction designers can build appropriate solu-
tions to a specific problem. The case study of Sri Lankan farmers presented in the
paper is part of a wider international project, which is meant to provide ICT support
to people from developing countries, with the goal to improve their quality of life. In
the near future we plan to validate the proposed methodology designing other mobile
applications for users who belong to the same socio/cultural community. This could
be the case, for instance, of an application meant to support fishermen of Sri Lanka in
their trading activities.
References
1. Mancini, J.A., Martin, J.A., Bowen, G.L.: Community capacity. In: Gullotta, T.P., Bloom,
M. (eds.) Encyclopedia of Primary Prevention and Health Promotion, pp. 319–330. Kluw-
er Academic, Plenum, New York (2003)
2. Di Giovanni, P., Romano, M., Sebillo, M., Tortora, G., Vitiello, G., Ginige, T., De Silva,
L., Goonethilaka, J., Wikramanayake, G., Ginige, A.: User Centered Scenario Based Ap-
proach for Developing Mobile Interfaces for Social Life Networks. In: Proceedings of the
2012 First International Workshop on Usability and Accessibility Focused Requirements
Engineering (UsARE 2012), Zurich, Switzerland, June 4, pp. 18–24 (2012) ISBN 978-1-
4673-1846-4 © 2012 IEEE
3. Di Giovanni, P., Romano, M., Sebillo, M., Tortora, G., Vitiello, G., De Silva, L., Goone-
thilaka, J., Wikramanayake, G., Ginige, T., Ginige, A.: Building Social Life Networks
through Mobile Interfaces–the Case Study of Sri Lanka Farmers. In: Proceedings of the IX
Conference of the Italian Chapter of AIS, Rome, September 28-29 (2012)
4. Telecenters in Sri Lanka: The Nenasala project. Technical Report. Skill International Pri-
vate Limited, Lakshman Building, Colombo
5. Medhi, I., Patnaik, S., Brunskill, E., Gautama, S.N.N., Thies, W., Toyama, K.: Designing
mobile interfaces for novice and low-literacy users. ACM Trans. Comput. Hum. Interac-
tion 18(1), Article 2 (2011)
6. Danis, C., Ellis, J., Kellog, W., Hoefman, B., van Beijima, H., Daniels, S., Loggers, J.:
Mobile phones for health education in the developing world: SMS as a user interface.
ACM DEV, London (2010)
7. Putnam, C., Rose, E., Walton, R., Kolko, B.: Mobile phone users in Kyrgyzstan: a case
study of identifying user requirements for diverse users. In: Professional Communication
Conference (IPCC). IEEE (2009)
8. Jayaweera, P., Senaratne, R.: Mobile service portal for rural fisher community develop-
ment. In: Proceedings of the International Conference on Advances in ICT for Emerging
Regions (ICTer), pp. 66–70. IEEE, Colombo (2011)
9. Dey, A.K.: Understanding and Using Context. Personal and Ubiquitous Computing 5(1),
4–7 (2001)
10. Preece, J., Rogers, Y., Sharp, H.: Interaction Design, 3rd edn. John Wiley and Sons (2011)
11. Mayhew, D.J.: The usability engineering lifecycle. Morgan Kaufmann Publishers (1999)
12. Longoria, R.: Designing Software for the Mobile Context, 1st edn., vol. XVII, 152 p. 72.
Springer, Longoria (2004)
13. Central Bank of Sri Lanka Annual Report (2010)
14. Rosson, M.B., Carrol, J.M.: Scenario-based design, The human-computer interaction
handbook: fundamentals, evolving technologies and emerging applications. Lawrence Erl-
baum Associates Inc., Mahwah (2002)
15. Hoober, S., Berkman, E.: Designing Mobile Interfaces. O’Reilly Media (2011)
16. Neil, T.: Mobile Design Pattern Gallery. O’Reilly Media (2011)
17. Di Chiara, G., Paolino, L., Romano, M., Sebillo, M., Tortora, G., Vitiello, G., Ginige, A.:
The Framy User Interface for Visually-Impaired Users. In: Proc.s of Int. Conf. on Digital
Information Management, ICDIM 2011, pp. 36–41. IEEE Press, Melbourne (2011)
18. Ginige, A., Romano, M., Sebillo, M., Vitiello, G., Di Giovanni, P.: Spatial data and mobile
applications - general solutions for interface design. In: Tortora, G., Levialdi, S., Tucci, M.
(eds.) Proceedings of the International Working Conference on Advanced Visual Interfac-
es, Capri Island, Naples, Maggio 21-25, pp. 189–196 (2012)
19. Social Life Networks for the Middle of the Pyramid, International project,
http://www.sln4mop.org/sln/home/index/home
The Effects of (Social) Media on Revolutions –
Perspectives from Egypt and the Arab Spring
Christian Sturm1 and Hossam Amer2

1
Hamm-Lippstadt University of Applied Sciences
[email protected]
2
The German University in Cairo, Egypt
[email protected]
Abstract. The Arab Spring has been titled as “social media revolution”. While
there is no doubt that Facebook and twitter contributed significantly to the
course of events, its role can only be understood when put into a broader tech-
nological and historical context. Therefore, we looked at the role of technology
during prior revolutions such as the French Revolution in 1789. It turns out that
media technology has played an important supportive role in social and political
movements throughout history. The changing characteristic, however, has al-
ways been the speed of information diffusion. Therefore, it is concluded that
social media should be seen as supportive but not exclusively responsible for
the events of the Arab Spring.
Keywords: Social media, society, revolution, evolution, twitter, Facebook,

democratization, media technology, Arab Spring, French Revolution.
1 Introduction
It is indisputable that the recent developments in the Middle East are of global magni-
tude. The so-called “Arab Spring” got worldwide social, political and media attention
due to its characteristics and connectedness with local, regional and international
interests. From the starting point in Tunisia and spreading over to Egypt and other
countries, people are revolting against suppressive forms of government and leader-
ship. One of the biggest achievements consists in an emerging public dialogue within
the societies and different social groups of the Middle East about the governing prin-
ciples of everyday life. This has been possible through the ousting of long-term presi-
dents in the region. Social media such as Facebook and twitter was mentioned by
researchers and the press as an enabler for these happenings. Therefore, what
happened in the Middle East is often referred to as “social media revolution” [1].
Nowadays, the use of social media is an established component for public and pri-
vate communication in the region. The Egyptian use of these services increased
through the January 25th revolution tremendously. It moved mainstream for a lot of
people. The majority, however, is still not connected due to a high level of illiteracy
and a lack of accessibility to information technology.
The Effects of (Social) Media on Revolutions 353
Nevertheless, the importance of social media cannot be denied. Therefore, a lot of

groups are debating its effects on revolutions. Some argue that the Internet is a double-
edged sword that could be a potential tool for suppression [2]. Others clearly state that
social media did have a significant role beyond being a source of news: “We use Face-
book to schedule the protests, Twitter to coordinate, and YouTube to tell the world” [3].
The actual role of social media, however, whether supporting or causing the develop-
ments during the Arab Spring, can only be determined by putting the events into a broad-
er context. Thus, the comparison with historic social developments on the one hand and
technological advances on the other hand will be the focal point for the following
chapters in order to see if the Arab Spring can be named a “social media revolution”.
2 (Media) Revolutions
2.1 Revolutions
The concept of “revolution” is extremely diverse without a unified definition. Its Lat-
in origin “revolution” means “turn around”. This refers to “change” in the broadest
sense. This change can be social, political, technical, cultural, psychological, personal
or any other domain that would change fundamental principles. In the context of this
paper, we look at revolutions of political and social systems. Tanter and Midlarsky
[14] define four different types of revolutions based on characteristics such as mass
participation, duration, domestic violence and intensions of the insurgents. Table 1
provides an overview of the types and their respective characteristics.
Table 1. Characteristics of revolutions according to Tanter and Midlarsky [14]
Type of Mass Duration Domestic Intentions of

revolution participation violence insurgents
Mass revolution High Long High Fundamental
changes in the
structure of politi-
cal authority and
social system
Revolutionary Low Short to Low to Fundamental
coup moderate moderate changes in the
structure of politi-
cal authority and
possibly some
change in the so-
cial system
Reform coup Very low Short, Low Moderate changes
sometimes in the structure of
moderate political authority
Palace revolution None Very short Virtually Virtually no
none change
354 C. Sturm and H. Amer
According to Tanter and Midlarsky [14], the French Revolution in 1789 as well as
the German Revolution in 1848 could be categorized as mass revolutions. Revolutio-
nary coups would be found in Turkey (1919), Germany (1933) and Egypt (1952).
Argentina (1955), Syria (1956) and Pakistan (1958) experienced a reform coup while
Venezuela and Brazil had a palace revolution in 1948 and 1955 respectively.
2.2 Media in Revolutions

Based on the schema for the categorization of revolutions mentioned above, the Arab
Spring can be defined as a mass revolution. As such, it is clearly not an individual but
a group phenomenon. Like every mass movement, communication between the dif-
ferent groups and actors plays a central role for the course of events. Due to its very
nature, this mass communication is only possible if it is mediated in order to over-
come the limitations of time and space. Therefore, it can be concluded that media
technology tend to play an important role for mass revolutions in order to address
these challenges.
Already the French Revolution “can be considered an epochal media event” [4].
The pre-industrialized printing press allowed the distribution of information in an
amount and speed not known before. Tens of thousands of pamphlets were printed
and distributed. The number of newspapers grew from just a couple “to over three
hundred weekly and daily newspapers between 1789 and 1790” [4], while most of
them were produced by individuals. This was possible only due to the low production
costs. It is estimated that the total number of newspapers created during the revolution
adds up to around 1,600 [4]. A lot of content was transmitted as well with printed
pictures due to an illiteracy rate of more than 50% for men and more than 70% for
women at the time [5].
In conclusion, media technology can be seen as an important component for the
information and mobilization of the masses already in the 18th century. Further exam-
ples for this supporting role from recent history include the so-called twitter revolu-
tion in Iran [6] and the Multimedia or SMS revolt in the Philippines [7]. They show
clearly that access as well as low cost of production and distribution account for poss-
ible easy dissemination of new ideas by individuals. Normally, television would not
be considered as having these advantages. The protesters in Eastern Germany,
however, were able to use the Western television and other media for informing the
population of the former GDR on the protests that emerged increasingly in 1989 [8].
In summary, it can be stated that media technology, easy access and low entry bar-
riers played always an important role in mass revolutions throughout history. It is not
a unique consequence of the latest developments in information and communication
technology. Therefore, the recent events in the Middle East might have happened
sooner or later with a different course of events even without the use of social media.
2.3 Revolutionary Timelines
Looking at mass revolutions in the last centuries, it turns out that they are composed of
single events and actions of different magnitudes. Although the Egyptian revolution is
referred to as the “January 25th Revolution”, there is a set of previous events leading up
to it [9]. The same can be observed for the time after the presidents of the region were
ousted. The process of negotiating the new principle foundations of everyday life is
contains violent events with a smaller magnitude than the main happening. This can be
seen in both successful revolutions such as the French Revolution that peaked in 1789
and not successful ones such as the German Revolution in 1848.
Figure 1 shows a smoothed graphical representation of such a typical course of
events. The horizontal axis is considered the timeline from left to right while the ver-
tical axis refers to actions of opposing parties with different magnitudes. In the case of
the Egyptian or Tunisian revolution, the central peak would relate to the ousting of the
president. The waves on the left are the events and counter-events preceding the oust-
ing. The waves after the peak on the right refer to events that represent the struggle of
power between different parties. Interestingly, this course of events can be
represented mathematically by the formula:
f(x) = sin(x)/x
The following chapter is going to look at possibilities to describe different mass

revolutions throughout the history by this formula.
Fig. 1. Graphical representation of the course of revolutions [as displayed by google.com

searching for f(x)=(sin(x)/x)]
3 The Changing Characteristic of Media Technology
The previous chapter concludes that media technology was always an important com-
ponent of mass revolutions. So what has changed since the French Revolution? What
are the differences between the technology available then and now?
It has been argued that mediated communication is essential for dissimilating in-
formation and ideas to the masses. Looking at the historical developments of other
domains helps to understand the most important changing characteristic.
The sector of transportation saw tremendous changes in the last centuries. The dai-
ly distance possible to travel in the 19th century with a carriage drawn by horses was
about forty miles maximum [10]. Today’s high speed trains allow traveling this dis-
tance in 12min while it takes a plane about 4 minutes only. Besides the increased
comfort, the speed of transportation is 165 times faster than centuries ago.
The changes that occurred in the business and financial sector are even more dras-
tic. Before the invention of the telegraph, it took traders at the stock exchange be-
tween days and weeks in order to finish a transaction. The New York Stock Exchange
today is able to close a transaction in 0.5ms [11]. The factor for the increase in speed
is of hundreds of millions.
The distribution of news did undergo a similar development. According to [12] it
took from 21 October 1805 to 6 November 1805 to publish the Battle of Trafalgar in
the London Gazette. Similarly “the French invasion of Russia on 24 June 1812” was
“published in The Times on 13 July 1812”. Nowadays, a tweet or post on Facebook is
published in a couple of seconds. The combination of influencers with a lot of con-
nections and average users that “quickly convey the information from one neighbor to
another” [13] results in the phenomenon of spreading the information faster than
established media would do.
In summary, speed of distribution can be seen as the changing characteristic of
technology. Taking both the changes in speed and the mathematical representation of
revolutions presented in the previous chapter, it even might be possible to come up
with a formal way of describing the course of revolutionary events by taking into
account various factors such as speed of communication provided by type of media
technology, number of participants and intentions of insurgents.
4 What’s Next?
The diffusion of activities, statements and information increased tremendously

through the usage of social networks such as Facebook and twitter. Compared to the
high-speed trading executed at the stock markets, however, it is still in its infancy.
The digitalization of the message with cameras and keyboards together with the dis-
tribution is still in the realm of seconds or even minutes. Projecting the latest technol-
ogical advances in the fields of ubiquitous computing and augmented reality,
however, could lead to a boost in the speed of diffusion that would enable types of
revolutions not yet known.
A simple scenario combining the Google Glasses with an iWatch measuring and
combining a huge set of data could lead to a true distribution of perceptions, emotions
and happenings in real-time. Thus, if a huge amount of people share and synchronize
such information, the network effect might be bigger and more influential than all the
mass revolutions seen before.
5 Conclusion
This paper looked at the role of media technology in the recent social and political
developments in the Middle East. The so-called “social media revolution” has been
put into a broader technological and historical context in order to gain a deeper under-
standing of its impact. It has been shown that principal characteristics such as easy
access and low entry barriers in terms of costs were already one of the main factors
that support the developments in the French Revolution in 1789. The only basic ele-
ment that changed constantly over the centuries was the speed of creation and trans-
mission of information. Therefore, it is argued that the Arab Spring should rather be
called a movement supported by media technology rather than a “social media revolu-
tion”. The events would not have happened without the social and political reality on
the ground. This reality, however, could have led to the same movement even without
social media involved.
References
1. Tayler, K.: Arab Spring really was social media revolution. TG Daily (2011),
http://www.tgdaily.com/software-features/58426-arab-spring-
really-was-social-media-revolution (last access: March 1, 2013)
2. Shanthi, K., Taylor, B.: Open Networks, Closed Regimes: The Impact of the Internet on
Authoritarian Rule. Carnegie Endowment (2003)
3. Chebib, N., Sohail, R.: The Reasons Social Media Contributed to the 2011 Egyptian Revo-
lution. International Journal of Business Research and Management (2011)
4. Reichardt, R.: The French Revolution as a European Media Event (2012),
http://www.ieg-ego.eu/en/threads/european-media/
european-media-events/rolf-reichardt-the-french-revolution-
as-a-european-media-event (last access: March 1, 2013)
5. Histoire de la scolarisation en France, http://fr.vikidia.org/wiki/
Histoire_de_la_scolarisation_en_France (last access: March 1, 2013)
6. Burns, A., Eltham, B.: Twitter Free Iran: an Evaluation of Twitter’s Role in Public Diplo-
macy and Information Operations in Iran’s 2009 Election Crisis. In: Communications Pol-
icy & Research Forum 2009, November 19-20. University of Technology, Sydney (2009)
7. Coronel, S.: The media, the market and democracy: The case of the Philipines. Jav-
nost=The Public 8(2), 109–124 (2001)
8. Friedensgebete und Montagsdmonstrationen Bundeszentrale für politische Bildung und
Robert-Havemann-Gesellschaft e.V. (2008),
http://www.jugendopposition.de/index.php?id=638
(last access: March 1, 2013)
9. Wolman, D.: The Digital Road to Egypt’s Revolution. The New York Times (2012),
http://www.nytimes.com/interactive/2012/02/12/opinion/
sunday/20120212-tahir-timeline.html (last access: March 1, 2013)
10. Ritter, W.: Fremdenverkehr in Europa. Eine wirtschafts- und sozialgepgraphische Untersu-
chung über Reisen und Urlaubsaufenthalte der Beweohner Europas. Leiden (1966)
11. Schewe, P.: Relativistic trading: The speed of light isn’t fast enough for some market
transactions (2010), http://phys.org/news/2010-11-relativistic-isnt-
fast-transactions.html (last access: March 1, 2013)
12. How fast did news travel? http://www.historyhouse.co.uk/articles/
speed_of_news.html (last access: March 1, 2013)
13. Doerr, B., Fouz, M., Friedrich, T.: Why Rumors Spread Fast in Social Networks. Commu-
nications of the ACM 55, 70–75 (2012)
14. Tanter, R., Midlarsky, M.: A theory of revolution. The Journal of Conflict Resolu-
tion 11(3), 264–280 (1967)
Memotree: Using Online Social Networking
to Strengthen Family Communication
Tsai-Hsuan Tsai1, Yi-Lun Ho1, Hsien-Tsung Chang2, and Yu-Wen Li2

1
Department of Industrial Design, Chang Gung University, Taoyuan, Taiwan
2
Department of Computer Science and Information Engineering,
Chang Gung University, Taoyuan, Taiwan
[email protected]
Abstract. This study describes the design and development of Memotree - a

basic service platform for family social interaction. Memotree features a Family
Tree which displays the family relationships and links among family members,
with features to encourage family interaction based on the Family
Communications Scale. The system and its associated hardware are designed to
be optimally accessible to family members of different generations. Finally,
Heuristic Evaluation was used to evaluate Memotree usability. Test subjects
affirmed the Memotree concept and provided positive feedback.
Keywords: social networking sites, family tree, family communication scale,

heuristic evaluation.
1 Introduction
In the context of filial piety, Asian societies place particular emphasis on kinship
relations and rely on family trees to record the family's history and clan associations.
Family trees are based on kinship ties, using the family surname to establish a family
genealogy, recording the history of lineages and family factions descending from a
common ancestor in a presentation integrating vertical and horizontal descent lines and
spherical descent tables. The eldest in the family holds a place of particular respect
within the family network[1, 2]. Traditional Chinese families typically consist of the
elderly parents, adult children and non-adult grandchildren, commonly referred to as
"three generations under one roof". In such arrangements, the grandparents are typically
cast as the family historians, maintaining the family's history, recording the family's life
experiences, and holding the family close together[3]. Generational interaction and
mutual support not only contributes to general life satisfaction and well-being for the
elderly, but also influences the family values and attitudes of the adult children and non-
adult grandchildren. However, as society has become increasingly urban, industrial and
commercial, an increasing number of families consist of conjugal or nuclear families, at
the expense of extended families. These changes to residence style and family type have
reduced the frequency of contact between generations which not only affects the
psychological well-being of the elderly and the fulfillment of their social needs, but also
raises critical issues for the family relations and support of modern families[4, 5].
360 T.-H. Tsai et al.
The application of internet technology today allows non-cohabitating family

members stay in touch[6, 7]. The rapid development of modern online social
networking sites (SNSs) has not only changed traditional social models, but have
emerged as a critical social tool for today’s youth[8]. Elderly people, despite their
relatively low rates of use of computers or the internet, have been found to express a
high degree of interest in learning to use such tools as a means of increasing their
interaction with the younger generation[9]. Research has found that SNS-based
communication do not replace face-to-face or telephone communication, but rather
creates a new domain of family communication, providing family members with rich
opportunities to maintain contact[10]. Unfortunately, currently most research into
SNSs focus on personal networks rather than maintaining family relationships or the
promotion of intergenerational exchange. Therefore, this study describes the design
and development of Memotree - an only social platform suitable for promoting
familial social interaction. Memotree not only imports the family genealogy concept
from online SNSs, but also focuses on social interaction and communication between
family members, thus promoting the maintenance of family relationships and
intergenerational communication.
2 Memotree Design and Development
The product of collaboration between designers and engineers, Memotree offers

social functionality similar to today’s SNSs, but presented from the perspective of the
user’s family tree in the context of the user’s family history and kinship structure,
creating the elements of a Family Communication Scale to promote communication
among family members. In addition, Memotree is designed to take into account the
difference between older and younger people in terms of usage habits and experience
to provide appropriate social platforms and hardware, ensuring that users of different
generations are able to get the most out of the experience.
2.1 Importing the Family Tree Design Concept
In this research, the Family Tree feature is used to diagram the user’s family
relationships, family’s lineage and kinship organization. In Memotree, the user can
manually manage the Family Tree by adding or deleting family members, or using the
Calendar feature to record family holidays, birthdays, anniversaries, etc. It’s worth
mentioning that, in the Family Tree network, the relationships are closest among the
three above-mentioned generations of family members, thus Memotree prominently
features these intergenerational relationships in its main display, allowing the user to
locate themselves within the context of their preceding or succeeding generations
(Fig. 1). Once the Family Tree is populated, it can automatically determine the
kinship between various users. If a relationship meets the criteria for kinship, users
can establish the relationship, eventually recreating their extended family. In addition,
the system automatically recognizes the identity of the user within the family and the
specific relationship with other members.
Memotree: Using Online Social Networking to Strengthen Family Communication 361
Fig. 1. Family tree Fig. 2. MemoTree main page
2.2 Functional Design Based on the Family Communication Scale
To promote interfamily communication and interaction, this study based the design of
Memotree functionality on Olson’s Family Communication Scale (FCS), which
identifies five general types of family communication skills by which the degree of
family communication can be assessed: listening skill, speaking skill, self-disclosure,
tracking, and respect and regard[11, 12]. This study first analyzed the characteristics
of the social functions of current SNSs. Ten pairs of grandparents and grandchildren
were then recruited as respondents, including 7 grandmothers and 3 grandmothers
(with an average age of 72.4 years), along with 2 grandsons and 8 granddaughters
(with an average age of 25.5 years). By evaluating the degree of family
communication among these pairs and interviews regarding their family interaction
needs, this study determined the gap between these needs and the functionality of
current SNSs. Finally, based on the research results, we designed interactive functions
to match the FCS results and users’ family interaction needs (see Fig. 2). The
interactive functions are described as follows:
• Listening skill: Listening skill is defined as empathy and attentive listening.

Memotree’s Listening functions include video conferencing and chat rooms which
not only allow users to hear each other’s voices but also help family members
share opinions and viewpoints. However, video conferencing is generally easier for
elderly users than text-based chat rooms.
• Speaking skill: Speaking skill is defined as talking to another person. Memotree’s
Speaking features include video conferencing and chat rooms which not only allow
users to converse with others about personal issues but also help people express
their feelings. In addition, the video feature allows users to sense each other’s
presence while speaking, thus mimicking face-to-face interaction.
• Self-disclosure: Self-disclosure is defined as sharing one’s feelings with others.
Memotree’s Self-disclosure features center around status updates and photo
sharing. The status update feature allows users to share their current mood or
recent events with family members. In addition to allow test subjects to freely
share personal information with privacy, Memotree’s Self-disclosure space is

divided into public and private areas, providing users with the ability to share
feelings and thoughts so that family members might better understand one another.
• Tracking: Tracking is defined as pausing on a topic. Memotree offers tracking
through its comment feature. Family members can post text-based or icon-based
replies on a given topic, thus providing family members emotional support and
encouragement. It’s worth pointing out that the Memotree system collects
messages for each topic and, based on the message properties and the topic header,
saves them to the Family Calendar system, allowing all family members view the
daily, weekly and monthly activity of other family members, thus giving them
insight into their recent situation.
• Respect and Regard: Respect and Regard is defined as the emotional side of family
communication. Memotree includes a Family Calendar system which records
important family events such as birthdays, memorial days and anniversaries.
Memotree users can specify certain days as Family Days, setting the Family
Calendar to send out appropriate messages to other family members to show
respect and concern.
• Family Communication Measure: Taking Olson’s Family Communication Scale as
the basis for evaluation, the Memotree system collects statistics on monthly
activity by family members to calculate the level of family communication. In
addition, the system will also provides statistics on average number of
communication instances between family members which not only helps users
better understand the level of interfamily communication, but also provides timely
recommendations and reminders to help users and their families maintain better
contact (Fig. 3).
Fig. 3. Family communication measure Fig. 4. System architecture of Memotree Service
2.3 System Interface and Hardware Designed to Accommodate User Habits
The Memotree user interface provides a Family Notepad to emphasize interaction

between family members and to record family events (Fig. 2). The Memotree system
also links to popular SNSs such as Facebook and Google+, allowing younger users to
simultaneously interact with the Memotree system and other social platforms. In
terms of hardware, the Memotree system supports most browsers and smartphone
devices, including tablets, thus allowing elderly users who are potentially unfamiliar
with PC operation to easily operate the Memotree family communication platform.
Figure 4 illustrate the overall Memotree system architecture. The system is divided
into a server side and a client side. The server side includes an intermediary server
responsible for managing all APIs for community service software and
communications. The intermediary server is designed so that updating the server
software obviates the need for API updates and updates to the client side software. In
addition, the intermediate server also conducts data Fusion, synchronously a user’s
data throughout different community services. The Memotree server is responsible for
all system operations and controls. Most data is obtained through the server, which
synchronously updates user information with community services. The Memotree
system only records some statistical data and account data to an Internal
Supplementary DB. Users can access their Memotree services through a tablet-based
application or through their PC browser. Overall, the Memotree system uses Fusion to
integrate the band-end community services and uses the Re-Render concept to re-
interpret community services.
3 Validation
Heuristic Evaluation was proposed by Nielson as a means of evaluating usability, and

is now widely used to assess a variety of systems[13, 14]. For example, Hart et al.
applied the approach to evaluate interface usability problems on Facebook[15],
Pinelle et al. applied it to assess interactive games[16], and Lee & Kozar used it to
evaluate the relationship of usability to purchase intention on e-commerce sites[17].
Accordingly, this study uses Heuristic Evaluation to assess usability of the Memotree
family social interaction platform.
3.1 Test Subjects and Experimental Design
We invited three experts in user interface design or ergonomic engineering to

participate in the usability assessment of the Memotree system, including two experts
with rich experience in information engineering system design, and one expert with a
strong background in interface design and usability. The experts were asked to
assume the role of a general user for system evaluation. All interaction with the
system was observed and recorded to identify problems, solicit opinions, determine
usability design flaws and analyze their causes. In addition, to obtain the point of
view of actual users, the solicited 42 subjects between the ages of 20 to 30 to test the
Memotree system. After testing, subjects were asked to provide recommendations
regarding the system interface and functionality.
3.2 Results
We invited three experts in user interface design or ergonomic engineering to

participate in the usability assessment of the Memotree system, including two experts
with rich experience in information engineering system design, and one expert with a
strong background in interface design and usability. The experts were asked to
assume the role of a general user for system evaluation. All interaction with the
system was observed and recorded to identify problems, solicit opinions, determine
usability design flaws and analyze their causes. In addition, to obtain the point of
view of actual users, the solicited 42 subjects between the ages of 20 to 30 to test the
Memotree system. After testing, subjects were asked to provide recommendations
regarding the system interface and functionality.
(a) Expert evaluation results
Heuristic Evaluation uses ten usability evaluation principles to identify potential
interface usability issues[13, 14]. The three experts were asked to assess the
Memotree system according to the ten criteria in Neilsen’s Heuristic Evaluation. In
the inspection process, the researchers recorded whether the experts considered the
system to present usability issues for each criteria, with an increasing number of
reports indicating an increased need for remedial action. The expert evaluation results
for the ten Heuristic Evaluation criteria are as follows:
• Visibility of system status: In terms of system status presentation, the experts

indicated that the user could easily recognize his or her location within the system,
and that system information displays were clear and useful. However, they felt that
the display text size may be too small for older users, and thus cause them
difficulty in reading.
• Match between system and the real world: The experts felt that the Memotree
system corresponds well to user operation, and the user can be made immediately
aware of the changes caused by his/her actions within the system.
• User control and freedom: In terms of user control and freedom, the Memotree
system is fairly uncomplicated. Action selection are clearly differentiated, and the
experts felt that further user control mechanisms were unnecessary.
• Consistency and standards: The experts felt that the Memotree system’s color
scheme and type face selections fulfilled design principles for consistency, but
some minor adjustments should be made to ensure the consistency of some icons.
• Error prevention: Memotree system does not provide an error-handling
mechanism. The experts felt that the system’s relatively uncomplicated design and
ease of use should reduce the user error rate.
• Recognition rather than recall: The experts felt that the Memotree system’s menu
system was highly navigable. This reduces the amount of information the user
needs to remember, thus reducing the user’s cognitive loading. However, the
expert evaluators also mentioned that the clickable buttons should be adjusted to
clearly differentiate clickable links and general information.
• Flexibility and efficiency of use: The three experts noted that all Memotree system
functions are shown on the system interface, such that users do not need to search
for the location of function. Therefore, although the system does not provide any
shortcuts, the expert evaluators all affirmed the current approach in terms of
flexibility and efficiency of use, and did not raise any further recommendations.
• Aesthetic and minimalist design: The expert evaluators felt that the Memotree
system interface messaging was concise, jargon-free, and easy to understand. In
addition, the system page only displays information needed to help users easily
recognize and operate the desired function.
• Help and documentation: Memotree system does not provide user manuals and
guides. The three experts recommended including simple instructions to help new
users quickly learn to operate the Memotree system.
(b) User viewpoint

Forty-two young test subjects, all familiar with the operation and function of online
SNSs (e.g., Facebook, etc.) were invited to help test the Memotree system. To solicit
their viewpoint, the testers were first asked to use the system and then provide
comments and suggestions for the system interface and functionality. Feedback from
the test subjects generally held that the interface design was attractive, the functions
are well-defined, and the system provides meaningful and effective information.
Three functions were singled out for specific recognition and appreciation: (1) the
Family Tree specifies the kinship relationship between various family members as
they interact in the system, (2) the interface design of the Family Calendar is not
found in other current SNSs, and (3) the Family Day function allows family members
to express emotion and concern, thus promoting cohesion among family members.
Selected tester responses are organized below:
• Family Tree: Unlike other current SNSs, the Memotree system only provides
communication between family members. Test subjects felt that the Memotree
system’s Family Tree feature depicts the identity of individual family members and
their specific relationships with other members, while also providing the basic
information and current status of each member (e.g., birthday, recent photos and
current status). This allows users to get to know their family members better and
also helps promote the continuation and maintenance of the traditional Chinese
sense of family and familial culture. In addition, the Family Tree function
automatically searches for family relationships and contacts between various
members of an extended family, thus creating an elaborate kinship diagram for an
extended family.
User quote: “… This family tree feature is amazing – it organized my entire

family’s information…”
User quote: “… I think this automatic editing feature in the family tree is
very good – it can let me clearly see who’s in my family…”
• Family Calendar interface design: Designed on the concept of conventional

calendars, the Family Calendar interface is easy for novice users (especially elderly
users) to intuit and learn, thus reducing the amount of time and effort needed to
learn to use the system. The 42 young test subjects indicated that the interface
design was attractive, providing a simple and intuitive mode of operation which
meets the cultural expectations and habits of elderly users, such as the inclusion of
the lunar calendar. In addition, the calendar offers links to day-to-day family events
and records the feelings of family members.
User quote: “… I’d show my grandfather the lunar calendar feature – he’ll
like that.”
User quote: “… the calendar is organized on a daily basis, allowing me to
clearly see how everyone in my family is doing…”
User quote: “… it’s like reading a diary. It’s very convenient and provides a
way for family members to interact.”
• Promotion of emotional expression and concern through Family Day: The Family
Calendar not only allows users to set specific Family Days (e.g., birthdays,
anniversaries, family activity days, etc.), but the system also reminds users of
upcoming Family Days to they can express their concern for the family. From an
emotional point of view, this kind of interaction is a type of rich emotional transfer
which can increase the intimacy between family members. In addition, the test
subjects indicated that the inclusion of icons and emoticons along with text gave
them a direct and convenient way to express their emotions and concern.
User quote: “On birthdays, I want to send my family members flowers or a

birthday cake, and these cute icons help me express my best wishes…”
User quote: “… these little gift [icons] are really great. I hope we can have
a bigger selection of gifts. It would make me very happy to received such
gifts….”
4 Conclusion
To extend traditional Chinese family values and ancestral culture, and to adapt these
concepts to modern popular digital social interaction modes, this study developed
Memotree – a basic service platform for family social interaction. Memotree features
a Family Tree which displays the family relationships and links among family
members, with features to encourage family interaction based on the Family
Communications Scale. The system and its associated hardware are designed to be
optimally accessible to family members of different generations. Finally, Heuristic
Evaluation was used to evaluate Memotree usability. Test subjects affirmed the
Memotree concept and provided positive feedback. Specifically, the Family Tree,
Family Calendar and Family Day features were seen as being conducive to promoting
strong family bonds and rich communication and interaction among family members,
thus satisfying the usage needs of different generations of the same family.
Acknowledgments. The authors would like to thank the National Science Council,
Taiwan, R.O.C., for supporting this research under grant NSC-99-2410-H-182-032.
References
1. FamilySearch (December 25, 2011), https://familysearch.org/
2. Ancestry.com. (December 25, 2011), http://www.ancestry.com/
3. Hagestad, G.O.: Transfers between grandparents and grandchildren: The importance of
taking a three-generation perspective. Zeitschrift für Familienforschung, pp. 315–332
(2006)
4. Silverstein, M., Cong, Z., Li, S.: Intergenerational transfers and living arrangements of
older people in rural China: Consequences for psychological well-being. The Journals of
Gerontology Series B: Psychological Sciences and Social Sciences 61, S256–S266 (2006)
5. Domingo, L.J., Asis, M.M.B.: Living arrangements and the flow of support between
generations in the Philippines. Journal of Cross-Cultural Gerontology 10, 21–51 (1995)
6. Selwyn, N.: The information aged: A qualitative study of older adults’ use of information
and communications technology. Journal of Aging Studies 18, 369–384 (2004)
7. Lehtinen, V., Näsänen, J., Sarvas, R.: A little silly and empty-headed: older adults’
understandings of social networking sites. Presented at the Proceedings of the 23rd British
HCI Group Annual Conference on People and Computers: Celebrating People and
Technology, Cambridge, United Kingdom (2009)
8. Boyd, D.: Why Youth Social Network Sites: The Role of Networked Publics in Teenage
Social Life. In: John, D., Catherine, T. (eds.) Foundation Series on Digital Media and
Learning (2007)
9. Tsai, T.H., Ho, Y.L., Tseng, K.: An investigation into the social network between three
generations in a household: bridging the interrogational gaps between the senior and the
youth. Online Communities and Social Computing, pp. 277–286 (2011)
10. Chen, Y., Wen, J., Xie, B.: I communicate with my children in the game: Mediated
Intergenerational Family Relationships through a Social Networking Game. The Journal of
Community Informatics 8 (2012)
11. Barnes, H.L., Olson, D.H.: Parent-adolescent communication and the circumplex model.
Child Development 56, 438–447 (1985)
12. Olson, D.H.: Circumplex model of marital and family sytems. Journal of Family
Therapy 22, 144–167 (2000)
13. Nielsen, J., Molich, R.: Heuristic evaluation of user interfaces, 249–256 (1990)
14. Nielsen, J.: Enhancing the explanatory power of usability heuristics. In: Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems: Celebrating
Interdependence, pp. 152–158 (1994)
15. Hart, J., Ridley, C., Taher, F., Sas, C., Dix, A.: Exploring the facebook experience: a new
approach to usability. In: Proceedings of the 5th Nordic Conference on Human-Computer
Interaction: Building Bridges, pp. 471–474 (2008)
16. Pinelle, D., Wong, N., Stach, T.: Heuristic evaluation for games: usability principles for
video game design. In: Proceeding of the Twenty-Sixth Annual SIGCHI Conference on
Human Factors in Computing Systems, pp. 1453–1462 (2008)
17. Lee, Y., Kozar, K.A.: Understanding of website usability: Specifying and measuring
constructs and their relationships. Decision Support Systems (2011)
Part IV
HCI for Business and Innovation
It Was Nice with the Brick So Now I’ll Click: The Effects
of Offline and Online Experience, Perceived Benefits,
and Trust on Dutch Consumers’ Online Repeat
Purchase Intention
Ardion Beldad and Mariël Segers
Department of Corporate and Marketing Communication,

University of Twente, The Netherlands
a.d.beldad@utwente, [email protected]
Abstract. The present study investigated the factors influencing Dutch con-
sumers’ intention to continue purchasing from a brick-and-click clothes shop’s
online channel after an initial commercial exchange. Results on the online
survey with 513 respondents reveal that their repeat online purchase intention is
influenced by the quality of their previous online purchase experience with the
retailer, their perception of the benefits associated with online purchasing, and
their trust in the retailer. Analysis also shows that respondents’ previous offline
purchase experience and online purchase experience contribute to their trust in
the retailer. Additionally, respondents’ previous online purchase experience po-
sitively influences their perception of the benefits associated with purchasing
online. The modified model also shows that respondents’ trust in the retailer
positively influences their perceptions of online shopping benefits.
Keywords: offline purchase experience, online purchase experience, online

trust, repeat online purchase intention, blick-and-click.
1 Introduction
The increasing popularity of online shops has prompted an increase in the number of
companies, which operate exclusively offline, to offer their products and services
online. Apparently, shops need to sell their products and services through both offline
and online channels not only to survive a stiff competition but also to better serve
customers whose buying behaviors have become increasingly mediated by electronic
technology, eventually resulting in the metamorphosis of online transactions from
pure clicks to brick-and-clicks [19]. With this phenomenon comes significant atten-
tion to the relationship between a customer’s interaction with a retailer in an offline
environment and the customer’s propensity to transact with that same retailer in the
online environment.
One may assume that a company that operates purely online is at a disadvantaged
position compared to a company with both offline and online channels, consequently
referred to as brick-and-click companies, since companies with an offline presence
372 A. Beldad and M. Segers
are more likely to be trusted in the online environment than those without offline
presence [16]. Online trust, as several studies have shown, is a strong determinant of
people’s intention to engage in exchanges with organizations online. Although trust in
a company could positively influence customers’ intention to purchase something
from that company’s online channel, other factors that could potentially increase such
an intention deserved to be identified. What is clear is that the factors influencing the
decision to buy online for the first time could be partly different from those that might
increase customers’ propensity to continue buying from a shop that operates exclu-
sively online and from an online shop with an offline presence.
While empirical studies that have identified the determinants of first-time online
purchase and repeat online purchase abound, specifically in the context of companies
that sell exclusively through their websites, there is still a dearth of research into the
factors influencing customers’ willingness to purchase again from an online shop with
an offline presence. This research primarily aims at addressing this gap. An online
survey with a brick-and-click clothing shop in the eastern region of the Netherlands
was conducted to test the research hypotheses.
2 Determinants of Online Repeat Purchase Intention

Companies continuously strive not only to attract customers to buy something from
them but also to maintain customers’ loyalty, primarily expressed in their propensity
to engage in a repeat purchase. One question deserving research attention, however,
pertains to the determinants of people’s online repeat purchase intention, referring to
the likelihood that people will buy something from the same company online after a
first-time purchase.
2.1 Experience with the Shop both Offline and Online

People’s decision to engage in an initial online transaction with a company is often
predicated on factors such as trust, organizational reputation, and the expected bene-
fits that can be derived from such form of transaction. However, upon the acquisition
of transactional experience and experience with the transaction partner, people can
already base their repeat purchase intention on first-hand information. While custom-
ers of shops that operate exclusively online could base their decision to buy again
from the shop solely on their online purchase intention, those who have purchased
something from the shop’s offline and online channels can predicate their online re-
peat purchase intention on their transactional experience with both channels.
An important point to consider, however, is that experience does not entirely suf-
fice to increase repeat purchase intention. As experience can be in a pole with satis-
faction and enjoyment at one side and dissatisfaction to disappointment at the other
side [2], a satisfying experience could logically increase repeat purchase intention.
Indeed, studies have shown that a satisfying online transaction experience is a strong
indicator of the aforementioned intention [22, 25].
It is also argued that people’s experience with a shop offline is often translated into
their buying behavior online, as the service obtained from a shop through its physical
It Was Nice with the Brick So Now I’ll Click 373
outlet could positively impact customers’ willingness to purchase from the shop
through its online channel [24]. The following hypotheses, therefore, are advanced.
H1a : Customers’ positive experience with a shop in the offline environment posi-
tively influences their intention to buy again from the shop’s online channel.
H1b : Customers’ positive experience with a shop in the online environment posi-
tively influences their intention to buy again from the shop’s online channel.
A positive relationship between a positive experience and trust in an online shop
exists as customers who are satisfied with their initial online exchange experience
tend to be more trusting of the company in the online environment [5, 10, 18]. Based
on this assertion, the next set of hypotheses is advanced.
H1c : Customers’ positive experience with a shop in the offline environment posi-
tively influences their trust in the shop’s online channel.
H1d : Customers’ positive experience with a shop in the online environment posi-
tively influences their trust in the shop’s online channel.
2.2 Trust in the Shop’s Online Channel

Referred to as one party’s reliance on the voluntarily assumed responsibility of an-
other party to safeguard the rights and interest of the former [11], trust has been re-
garded crucial in prompting individuals and organizations to engage in various forms
of exchanges with others. The success of online exchanges (e.g. e-commerce, e-
government) is predicated on trust [7] considering the uncertainties and risks associ-
ated with online transactions characterized as distant, anonymous, and impersonal
[18]. Certainly, trust does not eliminate risks although it is highly instrumental in
reducing the degree and the type of risks (e.g. financial losses, information privacy
violation) [3] people perceived from engaging in transactions with the trust (or dis-
trust) target [3, 13, 23]
Initial encounters, when both parties have no prior knowledge about each other’s
performance, are partly predicated on initial trust, which is based on the knowledge
about the trustee a trustor obtains from third parties [9] and on the expectation that the
trustee will not defect for fear of negative outcomes [17]. After an initial encounter,
the two parties’ decision to engage in succeeding exchanges could already be
grounded on trust resulting from the transacting parties’ knowledge about each other
[17]. Trust, therefore, matters not only in initial encounters but also in succeeding
encounters. This prompts the next hypothesis.
H2 : Customers’ trust in the (brick-and-click) shop’s online channel positively in-
fluences their online repeat purchase intention.
2.3 Benefits of Online Shopping

The acceptance of novel technologies and newer ways of doings things is partly de-
pendent on their usefulness, as the Technology Acceptance Model postulates [8]. The
usefulness of online shopping can be measured in terms of the benefits it extends to
customers such as convenience and comfort [6, 14]. The expected benefits people
associate with online shopping have been found to strongly influence their intention
to buy something online for the first time [14].
As people gain more online transaction experience, nonetheless, the extent to
which they were satisfied or displeased with their previous transactions could even
serve as a relatively objective indicator of their evaluation of the benefits associated
with an online transaction. In fact, empirical studies have shown that beliefs in the
usefulness of computer-mediated commercial exchanges are an important determinant
of online repurchase intention [25]. This prompts the hypothesis below.
H3 : Customers’ perception of the benefits of purchasing online positively influ-
ences their intention to continue purchasing from a (brick-and-click) shop’s online
channel.
3 Method
3.1 Sampling
An online survey was conducted with consumers of one brick-and-click clothes shop
in the eastern part of the Netherlands. The retailer, which sells primarily women and
children apparel, also assisted the researchers in distributing the questionnaires to the
study’s target respondents since the retailer maintains a database of contact informa-
tion of consumers who have purchased a product from both the offline and online
channels of the retailer. With the assistance of the retailer, a link to the online ques-
tionnaire was sent to 8,147 customers whose e-mail addresses were stored in the re-
tailer’s electronic database at the time of the study. The focus of the study required
that only respondents who have purchased from the retailer through both its offline
and online channels should be included in the study. After a four-week data collec-
tion, 706 filled out online questionnaires were returned. However, 193 questionnaires
had to be removed since they were not completed, resulting in 513 questionnaires
used for analysis or a final response rate of 6.29%.
3.2 Respondents
As the retailer used for this study sells women and children apparel, almost all res-
pondents (N=499, 97%) were women. Respondents’ age ranged between 21 and 60
(M = 39.06, SD =7.99). In terms of level of education, 57% (N = 291) of the respon-
dents have obtained higher education (university or bachelor degrees). Most respon-
dents indicated to have high levels of Internet experience, as those who have been
using the Internet for more than 11 years accounted for 45% (N = 233) of the study’s
sample.
3.3 Research Instrument

New items were formulated for the constructs ‘offline experience with the shop’ and
‘online experience with the shop’ – with three items for each construct. Examples of
items for ‘offline experience’ included ‘My experience with the shop offline was
positive’ and ‘My previous purchase with the shop offline proceeded without prob-
lems’. For ‘online experience’, typical items included ‘My experience with the shop
online was positive’ and ‘I never had negative experiences with the shop’s online
channel’. Cronbach’s alpha scores for ‘offline experience’ and ‘online experience’
were pegged at .90 and .93, respectively.
The three items to measure ‘trust’ (α = .85) were based on the scales of Carter and
Belanger [4] and Jarvenpaa et al. [13]. Examples of items included ‘I trust the shop’s
online channel’ and ‘I believe the shop’s online channel is trustworthy’. ‘Perceived
benefits of online shopping’ (α = .85) were measured with four items by Kim, Ferrin,
and Rao [14]. Examples of items for the construct included ‘Buying from the shop’s
online channel saves me time’ and ‘I can always buy something from the shop’s
online channel anytime I want to’. The dependent variable ‘repeat online purchase
intention’ (α = .90) was measured with three newly formulated items such as ‘I would
most likely buy an item from the shop’s online channel again’ and ‘I am intending to
purchase an item from the shop’s online channel anytime soon’.
Presented on Table 1 are the Cronbach’s alpha, mean, and standard deviation
scores for the different constructs of the study. All items used for the different con-
structs were measured on a five-point Likert scale (with 1 corresponding to ‘fully
agree’, 2 ‘agree’, 3 ‘neither agree nor disagree, 4 ‘agree’, and 5 ‘fully disagree’.
Table 1. Cronbach’s α, mean, and standard deviation scores for the different constructs
Constructs Cronbach’s Mean Std.
α Deviation
Offline experience with the shop .90 1.55 .71
Online experience with the shop .93 1.92 .84
Trust in the shop’s online channel .85 1.80 .56
Perceived benefits of online shopping .85 2.05 .66
Repeat purchase intention .90 2.34 .71
4 Results
To test the different research hypotheses and to determine whether the proposed re-
search model fits the data, structural equation modeling (SEM) using AMOS 18.0 was
performed. The use of SEM subscribed to the two-step approach proposed by Ander-
son and Gerbing [1], which indicates that the measurement model should be assessed
first prior to testing of the structural model. Model fit was assessed using the follow-
ing indices: RMSEA (root mean square error of approximation), CFI (comparative fit
index), TLI (Tucker-Lewis index), and SRMR (standardized root mean square resi-
dual) [15, 20]. CFI and TLI values should be greater than .95 [12], the RMSEA value
lower than 0.07 [21], and the SRMR value below 0.08 to reach a conclusion that a
relatively good fit between the proposed model and the observed data exists [12, 15].
Results of Confirmatory Factor Analysis (CFA) show that the measurement model
excellently fits with the data [X2(80) = 265.37, X2/df = 3.32, p = .00, CFI = .97, TLI =
.96, SRMR = .03, RMSEA = .07 (CI: .06, .08)]. Test of the structural model resulted
in a relatively acceptable fit [X2(83) = 313.76, X2/df = 3.78, p = .00, CFI = .96,
TLI = .95, SRMR = .05, RMSEA = .07 (CI: .07, .08)]. However, inspection of the
modification indices suggested that model fit could still be improved by establishing a
causal relationship between two independent variables. The original model, therefore,
was modified by adding a path from ‘online experience’ to ‘perceived benefits’ and
from ‘trust’ to ‘perceived benefits’. The test of the modified model yielded a slightly
improved fit [X2(82) = 278.28, X2/df = 3.39, p = .00, CFI = .96, TLI = .95, SRMR =
.04, RMSEA = .07 (CI: .06, .08)].
Analysis shows that respondents’ repeat online purchase intention is influenced by
the quality of their previous online purchase experience with the retailer (β = .32),
their trust in the (brick-and-click) shop’s online channel (β = .13), and their perception
of the benefits associated with purchasing online (β = .39). This results to the accep-
tance of hypotheses 1b, 2, and 3. However, the quality of respondents’ previous of-
fline purchase experience with the retailer has been found to have no influence on
repeat online purchase intention, leading to the rejection of hypothesis 1a.
Moreover, analysis indicates that respondents’ previous offline purchase expe-
rience (β = .34) and online purchase experience (β = .46) contribute to their trust in
the shop’s online channel. Thus, hypotheses 1c and 1d are accepted. The modified
model also shows that respondents’ previous online purchase experience positively
influences (β = .36) influences the perceived benefits of purchasing online. Further-
more, respondents’ trust in the (brick-and-click) shop’s online channel positively
influences perceptions of the benefits of online shopping (β = .31)
5 Discussion
For retailers that used to operate exclusively offline, the expansion of product and
service delivery to the online environment entails a thorough understanding of the
ways to ensure that customers who purchase offline would also exchange in a com-
mercial exchange with the retailer through its online channel. There certainly are co-
pious studies into the factors influencing initial online purchase intention and repeat
online purchase intention, especially in relation to shops with online channels only.
However, adequate academic attention has not been extended to the mechanism be-
hind people’s willingness to purchase through the online channel of a retailer with an
offline presence.
Results of this study show that people’s intention to purchase something again
from a brick-and-click retailer’s online channel is predicated on three important fac-
tors, namely, the quality of their previous online commercial exchange, their level of
trust in the retailer’s online channel, and the perceived benefits of online shopping.
These findings strongly support the results of previous empirical studies into the
impact of these three factors on purchase and repeat purchase intentions.
It should be noted that of the three, however, the perceived benefits of online shop-
ping strongly determine the extent to which customers would opt to purchase an item
from an offline retailer’s online channel after an initial exchange. While this variable
lies outside a specific retailer’s influence, the retailer can still ensure the continuance
of an online purchase among its customers by providing them with a satisfactory
online transaction experience.
Although it was initially surmised that the quality of customers’ experience with a
retailer offline would influence their repeat online purchase intention, the study’s
results indicated that the former has no impact on the latter. This is surprising since
one would expect that the extent to which customers were pleased with their purchase
through the retailer’s offline channel would determine their willingness to shop
through that retailer’s online channel. A possible explanation for this is that when
deciding whether or not to shop online, customers may not really pay attention to their
offline exchange experience with the retailer behind the online channel.
Even if the evaluation target – for quality of previous experience and trust – is the
same (the retailer), the channels used for the commercial exchange differ in several
aspects. Purchasing an item offline provides a customer with the possibilities to inte-
ract with a flesh-and-blood salesperson and to feel or try a product, which would
totally be impossible if one opts to buy a similar item through the retailer’s online
channel. Thus, when deciding whether or not to continue buying from a retailer’s
online shop, customers might pay more attention to the quality of their previous
online transaction than to their purchase experience offline.
The fact that trust in the retailer’s online channel positively influences customers’
repeat purchase intention amplifies the need to understand how trust could be devel-
oped. Trust has been known to evolve depending on the depth of the relationship be-
tween a trustee (in this case, the retailer) and a trustor (the customer). While initial
encounters could be predicated on trust that emerges from the knowledge a trustor
obtains from others, succeeding encounters could already be based on trust that de-
velops out of the quality of the interaction between the two parties. Studies have
shown that customers who were satisfied with their previous transactions with a com-
pany are more likely to continue trusting the company after an initial encounter.
Results of this research strongly supports the relationship between quality of pre-
vious transaction experience and trust online. The quality of respondents’ previous
online exchange experience with a blick-and-click is a strong determinant of their
trust in the retailer’s online channel. Although the quality of respondents’ offline ex-
change experience does not impact repeat online purchase intention, results further
reveal that it is an important determinant of trust in the retailer’s online channel.
While trust in the shop, in general, is not identified as a construct in this study, this
may have been captured by the quality of the respondents’ offline exchange. Custom-
ers who were satisfied with their transaction with the retailer in the offline environ-
ment have a strong reason to trust the retailer, in general, and such trust could
influence the extent to which they would trust the retailer’s online channel.
Modification of the structural model indicates that factors such as ‘quality of
previous online experience’ and ‘trust in the retailer’s online channel’ positively in-
fluence respondents’ perception of the benefits of online shopping. An important
implication of this finding is that online customers’ estimation of the benefits of on-
line shopping is partly predicated on the quality of their previous online transaction
experience. And this further suggests that when the previous online transaction with
the retailer was not gratifying, customers would be most likely inclined to undermine
the merits of online shopping.
Additionally, as the analysis shows, the benefits of online shopping would hardly
be appreciated if the channel used for the online transaction is deemed untrustworthy.
Trust, therefore, is a critical factor as it substantially impacts not only online custom-
ers’ behavioral intention but also their attitude towards the object of the intention –
online shopping.
6 Implications
Results of this study have important implications not only for electronic commerce
researchers but also for brick-and-click retailers. While it is known that customers’
experience with a retailer offline does not prompt repeat online purchase intention,
such an experience, which should primarily be satisfactory, is an important determi-
nant of their trust in the channel the retailer uses for commercial transactions online.
Brick-and-click retailers, therefore, should continuously strive to provide their cus-
tomers with a gratifying experience offline as it can influence customers’ trust in the
retailer in the online environment, which partly determines their repeat online pur-
chase intention and the extent to which online shopping is seen as beneficial.
What is more important is that brick-and-click retailers should incessantly ensure
that online shoppers are satisfied with their online shopping experience, as it is a criti-
cal determinant of repeat online purchase intention. Furthermore, such an experience,
as this study’s results reveal, shapes people’s evaluation of the benefits of online
shopping, which positively influences repeat online purchase intention. Although it is
beyond a brick-and-click retailer’s ability to influence people’s belief in the benefits
of online shopping, retailers could still have an impact on it by maintaining custom-
ers’ trust in the channel for an online exchange and by guaranteeing that customers
would be satisfied with such transaction – the two pivotal determinants of the extent
to which online shopping is viewed as advantageous.
The model tested in this study is remotely comprehensive. The effect of customers’
experience with a brick-and-click retailer on trust in the retailers’ online channel is
interpreted as a consequence of the trust people have in the retailer, which results
from customers’ positive exchange experience with the retailer. However, trust in the
retailer, in general, proves to be an important variable that needs to be included in a
more comprehensive model for repeat online purchase intention.
Results of the current study might have been skewed by the unusually high number
of female survey respondents, considering the type of brick-and-click retailer used for
the research. One can, therefore, assume that the findings could be a reflection of the
mechanism behind female shoppers’ repeat online purchase intention. The relatively
high impact of perceived benefits of online shopping on the behavioral intention of
interest might be a consequence of having a large proportion of women in the sample,
as it can be assumed that female shoppers might consider online shopping benefits
such as convenience and time-saving as decisive factors for their repeat online pur-
chase intention. The model, therefore, could also be tested with male respondents to
see whether or not the factors influencing repeat online purchase intention vary
according to the gender of shoppers.
References
1. Anderson, J.C., Gerbing, D.W.: Structural equation modeling in practice: A review and
recommended two-step approach. Psychological Bulletin 103, 411–423 (1988)
2. Beldad, A., De Jong, M., Steehouder, M.: How shall I trust the faceless and the intangible?
A literature review on the antecedents of online trust. Computers in Human Behavior 26,
857–869 (2010)
3. Beldad, A., Van Der Geest, T., De Jong, M., Steehouder, M.: Shall I tell you where I live
and who I am? Factors influencing the behavioral intention to disclose personal data for
online government transactions. International Journal of Human-Computer Interac-
tion 28(3), 163–177 (2012)
4. Carter, L., Belanger, F.: The utilization of e-government services: Citizen trust, innovation,
acceptance factors. Information Systems Journal 15(1), 5–25 (2005)
5. Casalo, L.V., Flavian, C., Guinaliu, M.: The influence of satisfaction, perceived reputation
and trust on a consumer’s commitment to a website. Journal of Marketing Communica-
tions 13(1), 1–17 (2007)
6. Chiu, C.M., Wang, E.T.G., Fang, Y.H., Huang, H.Y.: Understanding customers’ repeat
purchase intentions in B2C e-commerce: the roles of utilitarian value, hedonic value and
perceived risk. Information Systems Journal (2012), doi:10.1111/j.1365-
2575.2012.00407.j_407
7. Corritore, C.L., Kracher, B., Wiedenbeck, S.: Online trust: Concepts,evolving themes, a
model. International Journal of Human-Computer Studies 58, 737–758 (2003)
8. Davis, F.D.: Perceived usefulness, perceived ease of use, and user acceptance of informa-
tion technology. MIS Quarterly 13(3), 319–340 (1989)
9. Doney, P.M., Cannon, J.P., Mullen, M.R.: Understanding the influence of national culture
on the development of trust. Academy of Management Review 23(3), 601–620 (1998)
10. Flavian, C., Guinaliu, M., Gurrea, R.: The role played by perceived usability, satisfaction,
and consumer trust on website loyalty. Information & Management 43, 1–14 (2006)
11. Hosmer, L.T.: Trust: The connecting link between organizational theory and philosophical
ethics. Academy of Management Review 20(2), 379–403 (1995)
12. Hu, L., Bentler, P.M.: Cutoff criteria for fit indexes in covariance structure analysis: Con-
ventional criteria versus new alternatives. Structural Equation Modeling 6, 1–55 (1999)
13. Jarvenpaa, S.L., Tractinsky, N.: Consumer trust in an Internet store: A cross-cultural vali-
dation. Journal of Computer-Mediated Communication 5(2) (1999),
http://jcmc.indiana.edu/vol5/issue2/jarvenpaa.html (retrieved)
14. Kim, D.J., Ferrin, D.L., Rao, H.R.: A trust-based consumer decision-making model in
electronic commerce: The role of trust, perceived risk, and their antecedents. Decision
Support Systems 44, 544–564 (2008)
15. Kline, R.B.: Principles and practice of structural equation modeling, 2nd edn. Guilford,
New York (2011)
16. Kuan, H.H., Bock, G.W.: Trust transference in brick and click retailers: An investigation
of the before-online-visit phase. Information & Management 44, 175–187 (2007)
17. Lewicki, R.J., Bunker, B.B.: Developing and maintaining trust in work relationships. In:
Kramer, R.M., Tyler, T.R. (eds.) Trust in Organizations: Frontiers of Theory and Research,
pp. 114–139. Sage Publications, Inc., Thousand Oaks (1996)
18. Pavlou, P.: Consumer acceptance of electronic commerce: Integrating trust and risk with
the Technology Acceptance Model. International Journal of Electronic Commerce 17(3),
101–134 (2003)
19. Ranganathan, C., Goode, V., Ramaprasad, A.: Managing the transition to bricks and clicks.
Communications of the ACM 46(12), 308–316 (2003)
20. Schreiber, J.B., Stage, F.K., King, J., Nora, A., Barlow, E.A.: Reporting structural equation
modelling and confirmatory factor analysis results: A review. The Journal of Educational
Research 99, 323–337 (2006)
21. Steiger, J.H.: Understanding the limitations of global fit assessment in structural equation
modeling. Personality and Individual Differences 42, 893–898 (2007)
22. Tsai, H.T., Huang, H.C.: Determinants of e-repurchase intentions: An integrative model of
quadruple retention drivers. Information & Management 44, 231–239 (2007)
23. Van der Heijden, H., Verhagen, T., Creemers, M.: Understanding online purchase inten-
tions: Contributions from technology and trust perspectives. European Journal of Informa-
tion Systems 12, 41–48 (2003)
24. Verhagen, T., van Dolen, W.: Online purchase intentions: A multi-channel store image
perspective. Information & Management 46, 77–82 (2009)
25. Wen, C., Prybutok, V.R., Xu, C.: An integrated model for customer repurchase intention.
Journal of Computer Information Systems 52(1), 14–23 (2010)
Interaction between Enterprise Resource Planning
Systems and Organizational Culture
Carlotta Herberhold
University of Wuerzburg, Chair of Business Information Systems,

Wuerzburg, Germany
[email protected]
Abstract. Building up on findings of IT culture research and ERP research, the

assignment is to investigate organizational cultural variables impacting on
changes through enterprise resource planning (ERP) systems. The study sug-
gests seven cultural factors that are influence by ERP usage. The results are
grounded in both a review of the literature on ERP, organizational culture, and
an analysis of qualitative data collected from ERP experts. The findings will
provide explanations that cultural factors should be considered when using an
ERP system.
Keywords: Computer-augmented environment, organizational culture,

Enterprise Resource Planning (ERP), user behavior, organizational behavior.
1 Cultural Battle with Enterprise Resource Planning
Enterprise Resource Planning (ERP) packages are implemented in enterprises with

different organizational and national cultures. Culture refers to human practices, in
contrast to conditions that are dictated by nature. Human values and behavior or
communication patterns result from a cultural character. The latter decides whether
certain behavior is accepted in a country or not. Thus, different cultures lead to
diverse organizational structures in firms. They influence the culture of debate, the
hierarchical understanding, organizational behavior, structures and information
processing [3, p. 192; 4, p. 371].
A particular challenge is the introduction of new information systems and their
changes since the users are not ready to give up old habits. Therefore, in order to main-
tain the old workflows they do not rely on the standard components. Ideally, however,
the existing processes are not used as a basis for new target processes. Instead, the
structure of the organization should base on the established processes of the standard
software (reference processes) [12, p. 147,148; 3, p. 185,186].
The technical implementation of ERP systems is accompanied by a behavioral
change in work processes, for example, that orders are processed accurately, material
is dispatched on time or products are produced as needed [6, p. 231, 12, pp. 45, 55].
For management, it is important to realize that any change in the company also brings
a change in corporate culture and sometimes requires it. Since the introduction of an
382 C. Herberhold
ERP system has some significant impact on the way people work, the organizational
culture is affected. Without a deep understanding of its impact on organizational cul-
ture, a company is unlikely to be able to develop with regard to the ERP system.
The maxim of each introduction or adaptation of an ERP system should not be that
any resistance by the user is suppressed, punished and therefore useless—in other
words a cultural battle—but that the consistent use of ERP systems improves
processes and thus leads to a new company and organizational culture in the interest
of the company and also its employees.
2 Research Gap and Research Questions
The objective of this study is to reveal the clash of cultural forces between the (cultur-
al) practices embedded in ERP software and organizational culture and structure. The
aim is to identify organizational culture values related to the meaning of practices
embedded in ERP software (information, integration etc.). In detail, what characteris-
tics, behaviors and attitudes counteract in regard to different organizational cultures
with the features of an ERP system or what organizational cultures work easier with
an ERP system.
Scholars looked at following aspects: Information system development, inter-
organizational relationships, IT adoption and diffusion, IT management and strategy,
IT use and outcomes, IT culture, IT impact on culture, organizational impact of IT.
LEIDNER observes in her literature review that “a strong interest in understanding
the relationship of IT and culture and determining how social groups interact with and
apply IT in organizational settings“ exists [4, p. 371]. Although culture has been cited
widely as a challenge in ERP initiatives, and although many studies have considered
the implications of organizational culture on ERP, no study addresses the change of
organizational culture through an ERP system in a concrete and observable manner
(see figure 1). Studies on ERP and culture generally include the adoption of ERP in
different countries in order to explain the role of culture in implementing ERP sys-
tems. This article adds to the literature by identifying organizational culture variables
that change through ERP systems in an organization.
Variables
? ? Organizational
ERP ? culture
?
?
UI
UI = User Interface
Fig. 1. Interaction between ERP and Organizational Culture

Interaction between Enterprise Resource Planning Systems and Organizational Culture 383
Specifically, the following research questions (RQ) will be answered: What orga-
nizational culture variables change by ERP?
This aspect will contribute to the understanding of ERP systems and the associated
organizational changes in companies.
3 Research Design
The author pursues to contribute to IT culture research by identifying organizational

culture variables that change through ERP systems in organizational settings. An
explicit study of variables in order to identify requirements for organizational culture
through ERP has not been made yet in any work. For this reason, it is first necessary
to develop an understanding of this topic. A qualitative approach, as opposed to a
quantitative method, is a suitable research approach, if a topic has not been unders-
tood sufficiently. They offer the variety and wealth of data to get an understanding of
the context to be examined [5, pp. 19-22].
In the first part, the author strives to provide components and requirements of the
ERP system that is necessary for the analysis. The important point is that only com-
ponents related to the organization are looked at.
The second part has the objective to identify organizational culture variables that
change through ERP systems in an organization. In order to understand the interaction
between culture values and ERP, this paper reports on an exploratory interview study
with ERP experts that identify these values [13, pp. 76; 9] (see table 1). The inter-
view guide provides questions about organizational structure, processes and data as
well as information usage and usage in general. The survey of these people allows to
get a better understanding of the ERP-cultural relationship and thus to detect causal
relationships than it would be possible through a survey. The selected group of ex-
perts has years of experience and knowledge in dealing with the implementation and
use of ERP software in companies [13, 14].
The third part has the objective to measure the variables before and after ERP in-
troduction. The case study approach is chosen as research method. The case study
approach enables to explore and point out those questions with the concurrent lack of
knowledge about identify organizational culture variables that change through ERP
systems (see table 1). The case studies are based on observing different units of analy-
sis in its natural setting. The filtered data of numerous single observations (each
building one case study) is then combined and examined against the study objectives.
Propositions are deduced from the results and can, at a later stage, be verified by a
large-scale quantitative survey.
The overall objective of the empirical part is to identify requirements for organiza-
tional culture through ERP systems based on organizational culture theories (e.g.
Schein) and provide recommendations to support the ERP change management
process through adaptation of organizational culture (ERP culture alignment).
384 C. Herberhold
Table 1. Research process
Empirical study Objective(s) Method

I (present paper) Provide ERP components and Literature review
requirements (unit of analysis)
II (present paper) Identify organizational culture variables Interview study (ERP
that change through ERP systems in an experts)
organization (RQ 1).
III Measurement of the variables before and Embedded case study
after ERP introduction (e.g. MM-Module, (different units of
OTC process, PTP process). analysis in an
organization and
different experts)
IV Identify requirements for organizational Argumentative-
culture through ERP based on deductive
organizational culture theories (e.g.
Schein) (RQ 2).
V Recommendations to support ERP change Argumentative-
management process through adaptation deductive
of organizational culture (ERP culture
alignment)
4 Literature Review
4.1 Enterprise Resource Planning
Business software or enterprise systems are different names for Enterprise Resource
Planning (ERP) systems. An ERP system is a special type of information systems,
which are defined as socio-technical-systems. It is used for the quantitative and time-
oriented scheduling of equipment, personnel and capital. The system controls the
process depending on the material types, document types, groups of accounts or pay-
ment but independent of the people who work in the various areas of the company.
Compared to other information systems, ERP is characterized by best-of-breed busi-
ness processes. These standard software systems were designed and developed for the
interaction between different task fields from the very beginning [12, pp. 147, 148].
For the whole research process, the following four components and requirements
of ERP system are relevant [3,11, 12]:
• (Organizational-) structure. When using an ERP system it is necessary to illu-

strate the organizational structure of the enterprise in the structure of the standard
business software. That means it is necessary to transfer the real organization into
system elements like e.g. client, company code, plant and storage location. Be-
sides this the company is divided into functional areas, like materials management,
production planning, sales and distribution, financial accounting, controlling etc.
Also, in order to capture process information and the logical structuring of the
workflow various functional areas within the company have documents as control
and steering elements.
• Processes. The usage of ERP systems is meant to improve business processes.
These improvements are related to e.g. activities, division of labor, responsibili-
ties and deliverables. Besides, ERP means automation of functionality. For ex-
ample, transformation rules calculate and derive something in the pricing process.
• Common data. ERP systems also share a common database, which enables con-
sistent organizational data collection and processing. A standardized way of
communication and information distribution is meant by ERP.
• User Interface (UI). The UI is the interface between system and user. Thus, a
user interface for humans must be adapted to their needs and abilities in order to
be useful to them. Usability criteria analogous to websites are colors, grouping of
content and spatial distribution. The culture takes an affect on how the user is
dealing with the product and why people from different cultures have different
ideas of usability.
Since this paper is about dealing with ERP in the context of different organizational
culture dimensions, this rather technical aspect of the UI is not further included in the
analysis. It also rather concerns national-specific differences than organizational cul-
ture. The organizational culture influences how users work with the ERP system and
why people have different understandings of using the system.
4.2 Organizational Culture

Companies differ not only by location, size, products and their success; they are also
characterized by different objectives, principles, manners and behaviors. Not only the
country in which they reside shapes corporate cultures. In turn, corporate culture is
the basis for rules, procedures, working conditions, habits, decision-making, know-
ledge sharing, and information usage [2].
In his organizational culture framework Edgar Schein (1988) describes culture as
follows: „Culture is a pattern of basic assumptions, invented, discovered, or devel-
oped by a given group, as it learns to cope with its problems of external adaptation
and internal integration, that has worked well enough to be considered valid and,
therefore is to be taught to new members as the correct way to perceive, think, and
feel in relation to those problems.” His model illustrates culture on three different
levels, which influences each other: Visible organizational structures and processes,
values and beliefs [10, p.9].
Other frameworks include dimensions or values to detect organizational culture.
With information technology, in particular the following content areas of corporate
culture, denominated as cultural dimensions are put in contact: target orientation,
result orientation, cost orientation, open communication, process orientation, ability
to learn and adapt. These dimensions are partially interrelated or interdependent or
mutually dependent [1, 4, 8, 9].
386 C. Herberhold
5 Preliminary Results
The interview study was organized on the basis of the factors identified in the literature
review. The review leverages these components of ERP: structure, processes and data.
The results of the literature review were then discussed in semi-structured interviews
with ERP experts who have experienced a high number of ERP implementations and
change processes. The interview results revealed the following organizational culture
variables that change through ERP systems:
• Decisions. In order to make a decision managers need information about all the
different enterprise divisions. Through ERP this process has changed. Also the
users are confronted with a new situation. For instance the system has automated
some process steps changing the work environment of the user. This may influ-
ence the way of decision making.
• Transparence. An ERP system establishes the principle that information is
centrally available. It enables the transfer of information that was previously
available only in one place. However, this requires open communication and the
willingness to share information. The interview results showed that on the one
hand users were sometimes exchanging more information among each other and
on the other hand higher transparence were frightening them in the beginning.
• New skills of user. With the introduction of an ERP system the user has to learn
how to use the system technically, but he also needs to learn new things like
thinking in a structured and logical way, accuracy, unambiguity and the ability to
abstract.
• Evaluations. ERP enables to evaluate transactional data i.e. business transac-
tions. A customer-oriented, objective and logically correct process control takes
center stage.
• Responsibilities and accountabilities. Working with ERP systems is goal-
oriented, structured and systematic according to schedule. Organizational units
and employees are assigned clear roles and responsibilities. An organizational
culture that includes this already works—a positive effect on the ERP applica-
tion.
• Process Owner with a clear mandate and responsibility. ERP systems provide
the basis to overcome divisional thinking and going over to enterprise-wide
process thinking. The central information management and consultative informa-
tion processing enables process integration and collaboration among departments
with cross-functional teams. This will only work with a process owner provided
with power of decision.
• Avoidance of central data storage. ERP systems are based on the idea of data
integration. Important information is collected, stored and processed at the origin.
This is important for the smooth running, but also for a high level of information
security (compliance). For example, every movement of goods triggers changes
in accounting and cost accounting. The scheduling and procurement costs, inven-
tory levels and the cost of capital are controlled. Causalities become clear. This
new way of data and information usage is new to the user.
Interaction between ERP and Organizational Culture. The overall goal of this
research was to show how organizational culture values are reflected in relation to the
ERP systems. The challenge to use a business information system in the correct way
is not only a technical matter but, like the interview results show, a matter of interac-
tion of system requirements and organizational culture like decision making, know-
ledge sharing, information usage, structures and processes, values as well as beliefs
(variables).
Further research should answer the question, how organizational culture influences
the understanding for changes induced by ERP. The objective is to design this interac-
tion into practical innovation for a better adoption of ERP systems. Also a organiza-
tional culture should be described which fits best the intended usage of ERP
systems and how an iterative IT-based change leads to a new ERP culture with its
own information and IT values.
References
1. Hofstede, G., et al.: Measuring Organizational Cultures: A Qualitative and Quantitative
Study Across Twenty Cases. Administrative Science Quarterly 35(2), 286–316 (1990)
2. Kobi, J.-M., Wüthrich, H.A.: Unternehmenskultur verstehen, erfassen und gestalten. Ver-
lag Moderne Industrie, Landsberg/Lech (1986)
3. Krumbholz, M.: The Implementation of Enterprise Resource Planning Packages in Differ-
ent Organisational and National Cultures. Information Systems (26), 185–204 (2001)
4. Leidner, D., Kayworth, T.: A Review of Culture In Informations Systems – Towards a
Theory of Information Technology Culture Conflict. MIS Quarterly (30), 357–399 (2006)
5. Mayring, P.: Qualitative Inhaltsanalyse: Grundlagen und Techniken. Beltz Verlag. Weinheim
(2010)
6. Merkens, H.: Interkulturelles Management: Die Herausforderung des 21. Jahrhunderts. In:
Interkulturell denken und handeln: Theoretische Grundlagen und Orientierung für die ge-
sellschaftliche Praxis, pp. S.229–S.239 (2006)
7. Richter, T., Adelsberger, H.: Kulturspezifische Untersuchungen in der gestaltungsorien-
tierten Wirtschaftsinformatik: Bestandsaufnahme und Analyse des derzeitigen Umgangs
mit “Kultur”. Universität Duisburg-Essen (2011), http://duepublico.uni-
duisburg-essen.de/servlets/DerivateServlet/Derivate-28380/
017_FinalDuePublico_WI_Umgang_Kultur_Richter_Adelsberger.pdf
8. Sackmann, S.A.: Welche kulturellen Faktoren beenflussen den Unternehmenserfolg?
Universität der Bundeswehr München, Neubiberg (2006)
9. Sagiv, L., Schwatz, S.H.: Cultural values in organisations: insight for Europe. European J.
International Management 1(3), 176–190 (2007)
10. Schein, E.H.: Organizational Culture and Leadership. John Wiley & Sons, New York (2004)
11. Thome, R., Hufgard, A.: Continuous System Engineering. Oxygon Publisher, Munich
(2006)
12. Thome, R.: Grundzüge der Wirtschaftsinformatik – Integration der Informationsverarbei-
tung in die Organisation von Unternhemen. Pearson Studium, Munich (2006)
13. Walsham, G.: Interpretive case studies in IS research: nature and method. European Jour-
nal of Information Systems 4, 75–81 (1995)
14. Yin, R.K.: Case Study Research: Design and Methods. SAGE Publications, Thousand
Oaks (2003)
Reconfiguring the Corporate and Commons: Mobile
and Online Activism as a Form of Socio-technical Design
Constance Kampf
Business & Social Sciences, Aarhus University, Aarhus, Denmark

[email protected]
Abstract. This paper describes a model for understanding how digital activists
contribute to (re)defining the ways in which socio-technical interactions are
(re)designed at the Macro-level. This macro level is explained in three parts—
1) the cultural production of a new form of practice, 2) an active, contested
knowledge process, and 3) as producing knowledge about the interface between
society and business. Two kinds of digital activists are examined—critical
digital activists and activists from within the system.
Keywords: Socio-Technical Design, Culture, Knowledge.
1 Introduction
What is digital activism and why does it matter? Since the Lotus Marketplace and
Clipper Chip online protests [1], technology has begun to shift the balance of power
towards consumer activists who are able to use it effectively to interrupt product de-
velopment and suggest changes for corporate strategy. With critically focused activ-
ists such as Wikileaks, the hacker’s group Anonymous, and digital artist activists such
as Ubermorgen [2], technology is being used to engage with power structures and
shift the balance of power between the corporate and the commons.
Both critical Digital activism and “activism from within the system” can be seen as
constituting a force for (re)design of social technical interactions. [3] calls for two
levels in our understanding of socio-technical design, a micro level focused on user
level interaction, and a macro-level focused on society level interactions. These socie-
ty level interactions between organized groups of activists and corporations work to
shift the attention of business from itself and its own goals to surrounding issues.
This paper focuses on discussing these two approaches to Internet activism as
emerging forces working to shape and shift the balance of power and the face of tech-
nology based interactions between corporations and the commons as a form of socio-
technical design. Under-standing these two approaches to Internet activism helps
demonstrate ways in which activists towards business engage in socio-technical
(re)design at the macro level.
Reconfiguring the Corporate and Commons: Mobile and Online Activism 389
2 Method
This paper uses argument and example to lay out a theoretical frame-work for under-
standing how different forms of online activism provide forces shaping socio-technical
(re)design at the macro level. Both primary and secondary data are used to point out
patterns among cases of Internet activism, and reflect on their implications for explain-
ing how different kinds of activism function as a force shaping the (re)design of
socio-technical interactions. Primary data consists of social media postings in Blogs,
YouTube, Facebook as well as websites collected from 2008-present. Primary data is
focused on an NGO—CarrotMob, and a B corporation, the Good Guide.
This primary and secondary data is approached through a multidisciplinary lens
combing sociotechnical design, organizational theory and sociological concepts.
These concepts draw from two distinct levels of understanding the interaction be-
tween people and technology. First, from a workplace level, applying socio-technical
design and organizational theory we can find a perspective on knowledge that is both
socio-technical and process oriented in nature. Adding a more general level, that of
society, we can expand that perspective to include an epistemic or knowledge produc-
ing space that affects and interacts with culture at the societal level.
This lens draws on the following concepts. From a traditional workplace focused
perspective on socio-technical design, Suchman et al [4] describe systems develop-
ment as cultural production of new forms of practice. From the field of Management,
Blacker argues for under-standing knowledge as an active process in a culturally lo-
cated system [5]. According to Blackler, this process is contested, provisional, and
pragmatic. In the case of a Macro approach to Socio-Technical design, this cultural-
ly located system can be understood as an epistemic, or knowledge producing. Knorr
Cetina puts forward the notion of a general knowledge culture in which specific
knowledge processes are embedded [6], from the field of sociology. She uses the term
“epistemic” or knowledge-producing to describe a knowledge culture, focusing on the
ways in which knowledge is produced and reproduced through what she terms “epis-
temic circuits” connecting actors and technologies. We can link these three perspec-
tives together for understanding a macro level of socio-technical design. This macro
level can be constructed through examining Blackler’s active, contested knowledge
processes as part of a global epistemic culture following Knorr Centina’s sociological
frame of reference, in which activists are engaging with social and mobile media to
produce contested, provisional and pragmatic knowledge about the interface between
society and business. This use of social and mobile media can be informed by Such-
man et al’s connection between technology and cultural production to a global, ma-
cro-level for socio-technical design [3].
To operationalize this multidisciplinary lens, Critical digital activism and “Activ-
ism from within the system” will be examined through 1) the cultural production of a
new form of practice [4]; 2) as an active con-tested knowledge process [5]; and 3) as
producing knowledge about the interface between society and business [3,6]. These 3
aspects work together to demonstrate how each type of activism can be understood as
a force for (re)design of socio-technical interactions through linking together practice,
process and the larger picture of the interface between society and business.
390 C. Kampf
3 Critical Digital Activism as a Force for (re)design of

Socio-technical Interactions
Critical digital activism reconfigures the commons and the corporate in unexpected
ways. The technologies of email, websites, and social and mobile media open up spaces
for critics of business strategy to be heard and gather followers with global reach.
3.1 The Cultural Production of Reach in the Public Critique of Business

Strategy
In the late 1980s, the movement of email, bulletin boards, and discussion lists from
the workplace to society enabled a shift in the ability for large groups of protesters to
communicate with each other and companies. An important type of protester was the
consumer activist. Lotus Marketplace is an example of a product that was widely
protested through these new channels [1]. In the end, consumer actions led to change
in company strategy and the product was cancelled before it reached the market.
Another type of critical digital activist is the artist-activist. The artist-activists do
not always begin as activists, but can react to the pressures of companies when they
feel the company is inappropriate. An example of this can be seen in the etoy war,
arguable the most expensive art performance in history. The digital artist activist
group etoy.com was legally pressured by a U.S. new technology startup, eToys.com
to stop using their domain name because it was too similar to etoys.com and
etoys.com customers who mistyped the URL were offended by etoy.com content.
First the company sent them a request for their domain name, and when it was re-
fused, the company eToys.com went to court and won an injunction preventing the
digital artist group etoy from using their domain. The group was forced by US legal
action to comply, and they felt it was unjust because they had owned the domain etoy
and built their brand in Europe before the company eToys.com was started. In reac-
tion, they opened a new domain, Toywar.com, and started “involving the enemy in an
insane ToyNam situation.” [2]. The new domain, Toywar.com, supported a game that
awarded points for actions that lowered Etoys.com stock. The artists recruited 1799
hackers to play the game. The result of this art performance was bankruptcy for
eToys.com, which was sold at a loss of billions of dollars [2]. Here, the artists not
only affected the business strategy, but they felt attacked by it, and so they coordi-
nated a responding attack. The reach of these artists enabled them to affect eToys.com
in an unprecedented manner.
3.2 Critical Digital Activism as an Active, Contested Knowledge Process

In the cases of Lotus Marketplace and the eToy war, the activists engaged in commu-
nicating ideas which addressed a contested boundary between business and society. In
the case of Lotus Marketplace, the product itself was selling personal information that
was already available publicly, but packaged in a CD-ROM format that (for that point
in time) would make it easier for marketers to target consumers. In the late 1980s,
consumers were upset about a product that they understood would affect their privacy.
Thus, protesting Lotus Marketplace as a product could be understood having roots in
an active discussion. Gurak describes the discussion as taking place over a period of
several months, during which computer specialists and consumers engaged in tech-
nical arguments about the product. She describes postings as having “a range of
functions: to inform, to debate the protest, to debate the product, to mobilize, to di-
vulge “secret information” and to discuss other privacy related issues.” Thus, the
activism had roots in a contested discussion about the nature of privacy at the inter-
face of business and society [1, p.27].
On the other hand, the digital activists focused on issues of ownership and legal
rights, contesting the perspective of the corporation and the US courts by statements
on their website telling the story, as well as by actions of creating a space where the
corporation was not prioritized because of its’ resources, but rather reconfigured to
become the enemy of free art and free speech. Here, the contested understandings of
the situation were not discussed as much as enacted in the art performance, described
as a virtual “ToyNam” by the artists [2]. Because the force of courts had set up the
knowledge that a company has the right to take away a domain that had been held by
another group before the company was created, the artists chose to act out their disa-
greement rather than discuss it, and within the framework of the game offered differ-
ent options for supporting “toy soldier” hackers to lower the stock of Etoy.
Both of these cases demonstrate that activism occurs within an active and contested
space, and the duration of the Lotus Marketplace discussions over a period of several
months, as well as the duration of the ToyNam performance over about a year indi-
cate that the activism was not a single effort, but rather an ongoing, engaged process.
3.3 Critical Digital Activism as Producing Knowledge about the Interface

between Society and Business
Both Lotus Marketplace and the ToyNam art performance by the etoy artist collective
offer examples of activists focusing on the interface of business and society as it is
shaped and (re)shaped by technology. Each case focuses on a different key issue. For
the Lotus Marketplace product, privacy and consumer perception of privacy became a
key issue. At the time, the knowledge produced around privacy was communicated
through then current technology: usenet groups, bulletin boards, MOOs and MUDs,
and email. This knowledge addressed but did not resolve issues related to privacy, and
we are still dealing with them in different forms today. For the ToyNam art perfor-
mance, the issue was virtual domain ownership—do companies have the right to
interfere with previously owned domains because it can affect their business? The
response of the artist activists set a clearer boundary by creating the knowledge that a
group of unhappy digital artists and their supporters had the ability to affect the
success of a large and well funded business which had, in effect, attacked the digital
artists’ brand.
Both of these issues—issues of privacy and domain ownership, can be seen as
cultural in nature. The ways in which people choose to question or accept privacy
standards become cultural norms over time. The question is to what extend are we
392 C. Kampf
drawn into adapting our cultural practices to fit the needs of the business? For exam-
ple, recent Google and Facebook practices have taken the notion of transparency and
turned it over towards the consumer, setting up consumer transparency as a key issue
rather than consumer privacy.
4 “Activism from within the System” as a Force for (re)design

of Socio-technical Interactions
“Activism from within the system” reconfigures the commons and the corporate
around a principal focus or single topic approach, and is designed to push an agenda
in way that creates a “win-win” situation for both business and society. These digital
“activists from within the system” use technology to facilitate changes within the
current corporate-commons balance. We will examine two examples: CarrotMob, an
NGO led by Brent Schulkin, and Good Guide, a B-Corporation led by Dara
O’Rourke. These organizations use technology and social media to build awareness in
the commons around the issues of Corporate Social Responsibility (CSR) and sustai-
nability, and put pressure on corporations to change by engaging them in win-win
scenarios.
4.1 The Cultural Production of Corporate Social Responsibility in the Public

Critique of Business Strategy
Like the critical activists, activists from within are also engaged in using or appro-
priating technology for cultural production. In this case though, they go beyond the
cultural production of reach already established by earlier digital activist movements
to the cultural production of CSR. The framework for CSR is contested, and corpora-
tions work at sensegiving [7] through their communications as much as NGOs, and
government bodies also work to define it. In producing cultural norms for CSR, Car-
rotMob and Goodguide each focus on a different definition. For CarrotMob, the cul-
tural production of CSR is done using video campaigns in YouTube to create a dialog
with local business owners and activists around sustainability issues. The key video of
the organization is a cartoon that shows business as polluting the environment, and
not responding to government, but rather to people who have pooled their resources
enough to gain corporate interests [8]. The goal of the organization is to create a new
cultural norm or resource through the activity of organizing purchases and negotiating
with businesses to engage in CSR such as reducing their environmental footprint.
These issues engage with different cultural norms in the different locations of the
CarrotMob events. For example, the CarrotMob event in Jönköping, Sweden, focuses
on purchasing fair trade goods in local grocery stores [9]. In contrast, the CarrotMob
event in Eugene, Oregon has a number of videos that focus on reducing waste, energy
use, and engaging a local supply chain in local coffee shops [10]. There are currently
328 videos from around the world that show aspects of CarrotMob events. Each of
these videos is localized to create a cultural production of the practice of coordinated
purchasing appropriate to the local cultural context.
The Good Guide also engages in a different kind of cultural production of CSR. It
focuses on the coordination and interpretation of scientific data about products to
rate them with respect to three categories of CSR—health, environment and commu-
nity. The Good Guide is unique in its’ focus on health as a component of CSR for
products. To create this type of activism from within the system, the Good Guide is
a for profit company with a NGO style mission. Thus, companies hire Good Guide to
feature their products in Good Guide’s information system. This system is available
as a database from their website, as well as an iPhone app which scans in product bar
codes at the point of purchase. The Good Guide app then shows their rating system
for each of the their three key categories, allowing the user to find more information
about the product’s ingredients, as well as keep track of how socially responsible
their purchases are. The rating system is a 10-point scale, with 4 shades of color to
remind consumers about their choices. These shades are green for products that are
good, and shades of brown, orange and red for products that have scientific data that
demonstrate issues in one of the three areas. Thus, by using the Good Guide smart
phone app or database to evaluate products before purchase, consumers enact re-
sponsible purchasing, which then ostensibly puts pressure on companies to produce
more responsible products.
4.2 “Activism from within the System” as an Active, Contested Knowledge

Process
Both CarrotMob and GoodGuide offer different approaches to opening up conversa-

tions around CSR. For CarrotMob, these different approaches to CSR knowledge
can be seen in the video content, with different sponsors emphasizing various aspects
of CSR from fair trade products, to waste reduction and local supply chains, to work-
ers rights. Although the CarrotMob web presence does not specifically offer a space
for discussing issues related to sustainability, the format for Carrotmob events is open
enough to allow organizers to focus in on issues, which they find important and rele-
vant in their context. Thus, the collection of CarrotMob videos operates as a contested
knowledge process in which consumers from around the globe are engaging in build-
ing definitions of CSR appropriate to their local contexts. For GoodGuide, these ap-
proaches focus on coordinating and organizing scientific research in their database, as
well as encouraging consumers to interact with government regulators with issues
related to health. In the website database, consumers are offered the ability to recom-
mend, suggest avoiding, and/or comment on their experience with each product. This
opens up a formal space for dialogue and allows for an active, contested knowledge
process to build knowledge about each product while including multiple perspectives.
4.3 “Activism from within the System” as Producing Knowledge about the
Interface between Society and Business
CarrotMob and GoodGuide both work to produce knowledge about how consumer
action and dialog affects business strategy in win-win activist scenarios. In addition,
they work to produce knowledge about ways in which consumers can engage in
394 C. Kampf
consuming with responsibility in order to encourage companies to offer opportunities

for responsible/sustainable consumption. Thus, the interface between society and
business in these cases focuses on sharing knowledge about products, and brings a
critical perspective to consumption. The idea behind this type of activism is then, that
this communication loop between consumers explaining their strategy for sustainabili-
ty, or their needs for a product that protects their health as well as the community and
environment and corporations will lead to (re)definition of CSR as well as a
(re)definition of the roles for consumers and business to engage in co-producing CSR
through their actions, which in turn affects the possibilities for knowing how consum-
ers and businesses should interact at the interface between society and business.
5 Conclusion
In conclusion, this model for understanding the effects of digital activism on culture
can be seen through the overall idea of a Macro-level for socio-technical design. This
Macro-level focuses on cultural changes in society. It can be operationalized through
the three aspects used to examine digital activists—1) cultural production, 2) an ac-
tive, contested knowledge process, and 3) producing knowledge about the interface
between society and business. These three aspects of a Macro-level for socio-
technical design were examined using examples from two kinds of digital activists—
critical activists and “activists from within”. The activists in both categories examined
here followed similar processes, which can be understood as a macro level for
socio-technical design. Each group of activists worked towards engaging in cultural
production, an active contested knowledge process and producing knowledge about
the interface between society and business. By following the Macro-level for socio-
technical design, one can see a way of explaining how technology affects our possi-
bilities for action, activism, and engaging in both processes of creating knowledge
and producing culture. In order to develop a more nuanced understanding of a
macro-level for socio-technical design, connecting these three aspects to definitions of
culture in global contexts would be a useful next step.
References
1. Gurak, L.: Persuasion and Privacy in Cyberspace: The Online Protests over Lotus Market-
Place and the Clipper Chip. Yale University Press (1997)
2. Kampf, C., Cox, G.: Using Digital Art to make the tensions between Capital and Com-
mons Transparent: Innovation in shaping knowledge of Internet business practices as a
form of cultural knowledge. In: Cultural Attitudes Towards Communication and Technol-
ogy Proceedings, Aarhus 2012. Murdoch University Press (2012),
http://blogs.ubc.ca/catac/proceedings/proceedings-2012/
3. Kampf, C.: Revealing the Socio-technical Design of Global e-Businesses: a case of digital
artists engaging in radical transparency. International Journal of Sociotechnology and
Knowledge Development 4(4) (2012)
4. Suchman, L., Bloomberg, J., Orr, J.E., Trigg, R.: Reconstructing Technologies as Social
Practice. American Behavioral Scientist 43, 392 (1999)
5. Knorr Cetina, K.: Culture in global knowledge societies: knowledge cultures and epistemic
cultures. Interdisciplinary Science Reviews 32(4) (2007)
6. Morsing, M., Schultz, L.: Corporate social responsibility communication: stakeholder in-
formation, response and involvement strategies. Business Ethics, A European Re-
view 15(4), 323–328 (2006)
7. Gioia, D.A., Chittipeddi, K.: Sensemaking and Sensegiving in Strategic Change Initiation.
Strat. Mgmt. J. 12, 433–448 (1991)
8. Shulkin, B.: How organized purchasing can change business. YouTube Video (2008),
http://www.youtube.com/watch?v=AkGMVQl8_jI (accessed March 1, 2013)
9. Rättvis handel Jönköping. Carrotmob premiers in Jönköping (2010),
http://www.youtube.com/watch?v=pZ_Xlrn2PQU (accessed March 1, 2013)
10. Eugene Carrot Mob (2990). Eugene CarrotMob Teaser,
http://www.youtube.com/watch?v=pZ_Xlrn2PQU (accessed March 1, 2013)
Social Media in the Product Development Process
of the Automotive Industry: A New Approach
Andreas Klein and Götz Spiegel
Spiegel Institut Mannheim GmbH & Co. KG, Schwetzinger Str. 32, D-68782 Brühl, Germany
[email protected]
Abstract. This paper introduces a new methodology for implementing social

media monitoring into an important stage of the innovation process within the
automotive industry — the prototype stage. The information gathered on social
media channels was used for project management, quality management and
idea-generation purposes in order to support the parallel development of future
generations of electric vehicles. As the project is still ongoing the first general
results which show the effectiveness of the methodology can be presented in
this paper.
Keywords: Social media monitoring, electric mobility, electric vehicles,

product development, social networking sites.
1 Introduction and Theoretical Background
In the project this paper describes social media sources are used in an automotive
context in order to support a real-world prototype stage. As the project and the moni-
toring process are still ongoing and final results will probably be available in autumn
2013, the focus of this paper is on the innovative methodology and a few preliminary
results.
The importance of social media as a communication channel for companies is ob-
vious, especially after the spectacular IPO of Facebook at the beginning of 2012. But
social media is not only about social networking sites as Kaplan et al. [1] already
pointed out. It also includes blogs, virtual worlds, games as well as collaborative
projects like Wikipedia and most importantly content communities. Commonly ac-
knowledged by the scientific literature is the usage of social media tools for marketing
purposes, e.g., the introduction of the new BMW X1 [2]. The emphasis of such activi-
ties lies more or less on an expected increase of public attention and on targeting new
customer groups. But often it is also used as an instrument to gain information about
the requirements and wishes of the customers. This scientific analysis of internet
communities in order to learn more about the members, their mode of communication
and the subject of their discussion was very much influenced by the work of Robert
V. Kozinets who used ethnographic methods for online communities [3], [4]. These
methods are widely applied, e.g., to learn about brand communities of the automotive
industry [5]. Social media sources seem to be particularly suited to give input for the
innovation process — from the idea creation to virtual product tests [6].
Social Media in the Product Development Process of the Automotive Industry 397
2 Project Description
In 2007 BMW Group´s task force “project i” decided to include electric vehicles at a
larger scale into the product portfolio [7]. According to the communicated timeline,
BMW is expecting its first series-production full electric vehicle with the BMW i3 at
the end of 2013. To achieve this goal, electric prototypes on the basis of existing cars
with internal combustion engine were produced and tested in order to gain more expe-
rience with electric vehicles. The first converted concept was the MINI E. Several
hundred units were produced and internationally tested by private and fleet users be-
tween 2009 and 2012 accompanied by an all-embracing scientific research approach
in order to learn about the acceptance, requirements and barriers for electric mobility
[8]. In this context the BMW Group was closely supported by the German research
institute “Spiegel Institut Mannheim”.
The second converted vehicle concept introduced by the BMW Group is the BMW
ActiveE, derived from the 1-series. In this vehicle, series components for the follow-
ing series-production electric vehicles like the BMW i3 have already been integrated.
After the usual intensive internal testing by BMW regarding technological and safety
issues, the car was ready for being shipped to the customers. About 700 units have
been shipped to the United States since January 2012 and handed over to customers
for 24 months. These pay a monthly leasing fee of 499$ [9].
3 Research Methodology
Because of the nature of the BMW ActiveE trial — the test of a near-to-market tech-
nology while the mass-market product is at its final stages of development — two
aspects are of major importance for scientific research:
─ quality management aspects in order to identify potential product, service or usa-

bility problems
─ idea generation from the driver community about potential product improvements
for enhancing the user experience.
Therefore the research methodology has to meet some requirements. It needs to be

flexible and any critical findings need to be communicated to the development de-
partments instantly. Furthermore, it has to be considered that the time of occurrence
of quality management aspects cannot be predicted as an electric vehicle is dependent
on outer conditions like charging infrastructure, climate conditions etc. Additionally
even single problems might have a big impact on the mass-market product and need
to be reported to the development teams.
Considering these requirements, in addition to a “classic” market research ap-
proach, an additional social media research strategy was developed for the BMW
ActiveE. A primary research revealed four adequate social media sources for the later
analysis (see table 1). Besides the official BMW ActiveE web forum three additional
discussion groups have been founded by BMW ActiveE drivers on social networking
398 A. Klein and G. Spiegel
sites. Most members either drive a BMW ActiveE or are very much interested in the
product. Many drivers are members of the forum as well of the discussion groups at
the same time. All discussion groups are open and can be fully read by the public.
Table 1. Sources for Social Media Analysis1
Number of
Name Members Founder
posts
The BMW ActiveE Forum 1.806 9.709 BMW Group
BMW ActiveE 384 24.367 Private Users
BMW ActiveE San Diego 17 91 Private Users
BMW i3 – ActiveE – 75 139 Private Users
Megacity EV
The core instrument is a data base with all posts and comments assigned to 55 differ-
ent categories. On a weekly basis all new posts are transferred manually into the data
base. The categories include a simple sentiment analysis (positive/neutral/negative) and
a more detailed view on all vehicle related aspects. A special focus lies on new aspects
that bring electric vehicles to the customers, e.g., regenerative braking or battery
Social Social Social Social

Dealer Dealer
media media media media
reports reports
source 1 source 2 source 3 source 4
Social media database After sales database
Weekly
Indepth reports
reports
„Classic“ Product
Project After Quality
market develop
management sales management
research ment
Fig. 1. Social Media in the research process
1
In February 2013.
lifetime. The database provides the possibility to run in-depth data analysis for certain
topics and to identify trends. In order to track the project progress, to early identify criti-
cal project developments, and to generally get direct insight into the topics of current
users’ discussions, a weekly report including a summary of the most important discus-
sions and reported problems on the monitored social media channels is generated.
While the most important stakeholders for the results within BMW are the devel-
opment and quality management departments, the weekly reports also provide for the
project management and after sales departments the possibility to gain nearly instantly
direct insight into the customers’ most important discussed issues and thus to react if
necessary (see Fig.1.). And finally, the social media insights also provide a potentially
valuable source of qualitative data for the development of the accompanying “classic”
market research.
4 Results
While comparing the user activity on the official BMW ActiveE forum and the most
important discussion group, it is obvious that although more people have subscribed to
the forum the more active communication is on the social networking group.2 The dis-
cussion on the four social media channels can roughly be divided into three categories:
─ social communication about events, e.g., motor shows/meetings with other users
─ discussion about problems/product-related questions/help from other drivers.
─ statistics about driven miles/used electricity for BMW ActiveE.
While each of the categories applies to the forum and the social networking groups
equally, the direction of the discussion on the social networking groups is more within
the users, while the posts on the forum mostly aim at BMW, e.g., desired improve-
ments for future vehicles. From the sentiment analysis it is difficult to draw a “ba-
lanced” picture of the advantages and disadvantages of the product as positive
attributes are mostly expected while negative attributes contain a certain news value
and are therefore reported more often. So a proper assessment of the vehicle features
has to be carried out by the accompanying “classic” market research.
Nevertheless, it is a good tool to identify potential product problems. To give an
idea of the effectiveness of the methodology for quality management, the findings
from the social media sources have been compared to the information collected by the
conventional quality management process in a BMW internal after sales database.
The after sales database is based on the service data from the dealers. So if BMW
ActiveE customers encounter a vehicle problem and ask the dealers for help, the deal-
ers would use the after sales database for support.
Within a monitored period of the first nine months the BMW quality management
team thereby identified 39 product-related problems which occurred at least twice.
The social media analysis meanwhile revealed 35 different product-related problems.
These problems did not all demand a stopover at the garage but influenced the
2
Nardi et al. give a good description of the motivation for activity on social media sites [10].
400 A. Klein and G. Spiegel
product experience in a negative way. About 32 problems were already covered by

the dealer reports. One reported problem with the vehicle door could not be validated
by the BMW engineers. Two problems had only been identified with the support of
the social media analysis and could be tackled by quality management measures. One
serious problem was already known but the frequency of occurrence was a lot more
than previously expected by the experts.
In order to gain new product ideas, an in-depth analysis of the four social media
sources identified 23 new ideas or suggestions for the BMW ActiveE or electric ve-
hicles in general. Thereof 12 ideas have already been part of the product strategy for
the successor, the i3. About eight ideas are currently in a review process by the devel-
opment departments. Only three ideas had been declined because of not fitting to the
product strategy.
Finally, the social media monitoring also turned out to be a valuable source for the
project management. So far the weekly report is currently subscribed by more than
130 involved experts from different departments throughout the entire BMW Group,
involving project management, development, communications and sales.
5 Review of the Social Media Monitoring Research

Methodology
To summarize, the described methodology has a number of advantages and disadvantages.
Advantages:
─ nearly real-time feedback on the product from the users’ perspective over the trial
period
─ non-intrusive methodology; users explain problems in their own language; topics
are set by the users themselves
─ cost-effective compared to other near real-time “classic” market research metho-
dologies
─ useful way to gather qualitative feedback as input for a later broader quantitative
market research design.
Disadvantages:
─ no exact quantitative results about the occurrence of problems or about the opinion
on ideas possible, because only a subset of the users post on social media channels
─ no “balanced” picture of the product; the communication is rather more problem-
driven.
6 Conclusion
Although social media monitoring during the product prototype stage contributed a
significant part to the collected information, it did not provide the full picture. It is
therefore rather an important element than a full replacement of a broader “classic”

research strategy. Furthermore, it was very helpful to do the research with a high-
involvement product such as a new electric vehicle which also attracts rather tech-
savvy customers who are more likely to use social media channels. Finally, the
limited number of participants in the trial also limited the amount of data, which
facilitates the analysis process.
Acknowledgments. The authors would like to acknowledge the BMW Group for
allowing us to listen to the customers’ voice, especially Roman Vilimek and Andreas
Keinath for their guidance and support throughout the international MINI E and Acti-
veE projects. And, of course, we would like to thank all the BMW Electronauts who
supported the entire development process of the i3 and its successors with their valua-
ble feedback. Finally many thanks also to the Spiegel team especially Alexander
Stech for summarizing the weekly reports and designing the database, and Julia Mes-
serschmidt for her valuable in-depth analyses.
References
1. Kaplan, A.M., Haenlein, M.: Users of the world, unite! The challenges and opportunities
of Social Media. Business Horizons 53(1), 59–68 (2010)
2. Douglas, T., Lorenz, M., Oheimb, C.: Pre-Kommunikation im Internet — Der BMW X1
trifft Kunden im Social Web. Marketing Review St. Gallen 26(4), 11–15 (2009)
3. Kozinets, R.V.: E-tribalized marketing? The strategic implications of virtual communities
of consumption. European Management Journal 17(3), 252–264 (1999)
4. Kozinets, R.V.: The Field Behind the Screen: Using Netnography for Marketing Research
in Online Communities. Journal of Marketing Research 39(1), 61–72 (2002)
5. Füller, J., Matzler, K., Hoppe, M.: Brand Community Members as a Source of Innovation.
Journal of Product Innovation Management 25(6), 608–619 (2008)
6. Füller, J., Bartl, M., Ernst, H., Mühlbacher, H.: Community based innovation: How to in-
tegrate members of virtual communities into new product development. Electronic Com-
merce Research 6(1), 57–73 (2006)
7. BMW Group: BMW i. Born electric, http://www.bmw-i-usa.com/en_us/,
(checked on October 21, 2012)
8. Vilimek, R., Keinath, A., Schwalm, M.: The MINI E field study. Similarities and differ-
ences in international everyday EV driving. In: Advances in Human Aspects of Road and
Rail Transportation. Advances in Human Factors and Ergonomics Series, pp. 363–372.
CRC Press (2012)
9. BMW Group: BMW Delivers First All-Electric ActiveE in the US. (2012),
https://www.press.bmwgroup.com/ (checked on October 28, 2012)
10. Nardi, B.A., Schiano, D.J., Gumbrecht, M., Swartz, L.: Why we blog. Communications of
the ACM 47(12), 41–46 (2004)
Do All People Enjoy the Benefits from Technology
Innovation?
Yoon Jeon Koh1, Jae Heon Park1, and Byung Do Chung2

1
Economics & Management Research Lab., kt, Sungnam, Korea
{yoonjeon.koh,jaeheon.park}@kt.com
2
Department of Systems Management Engineering, Sungkyunkwan University, Suwon, Korea
[email protected]
Abstract. The spread of emerging devices such as smartphone and Tablet PC

enables people to experience an unprecedented technology innovation. Howev-
er, some smartphone users show their intentions to reuse the feature phone due
to the increased stress and pressure caused by a lot of functions of the smart-
phone. This study explores that the technology innovation doesn’t provide all
the people with the same level of values and satisfactions. A paper and pencil
survey is employed with 3,798 mobile users randomly selected from 10’s to
60’s in Korea and asks them to answer the devices, ICT services, and applica-
tions in use, and the degree of their satisfactions as well as their perceived val-
ues and reuse intentions. Results show that the effects of technology innovation
are different depending on mobile users’ perceived value of technology innova-
tion and price-value scheme, satisfaction of mobile service in use, and ages.
Keywords: technology innovation, smartphone, perceived value, satisfaction.
1 Introduction
With the development of technology, the users of the emerging devices such as
smartphone and tablet PC are rapidly growing. For example, the penetration rate of
mobile phones in Korea has already surpassed 100% and about 60% of mobile phone
users switched feature phones to smartphones in 3 years only after kt first introduced
the iPhone into the Korean market in 2009. The tablet PC market is also growing in
Korea, even though the usage rate is less than 20% out of the mobile phone users
now.
The spread of emerging devices mentioned above enables people to experience an
unprecedented technology innovation in User Interface (UI) such as voice search and
touch-sensor screen. In addition, the technology innovation enables people to use
diverse functions of the new devices in a simple and easy way. As a result, people
become more enthusiastic for the UI which is regarded as a key factor to influence the
competitiveness of the emerging devices, driving the direction of the technology de-
velopment such as motion and voice sensing.
However, some smartphone users show their intentions to reuse the feature phone
due to the increased stress and pressure caused by the smartphone such as frequent
Do All People Enjoy the Benefits from Technology Innovation? 403
email checks and notification sounds. On the other hand, a newest smartphone with
high-specification is a just useless device to those who are not familiar with applica-
tions and functions which are based on the new UI, or those who use only voice and
text services with their mobile phones. Thus, this study attempts to explore that the
technology innovation doesn’t provide all the people with the same level of benefits,
values, and satisfactions, which are influenced by their life style and attitudes toward
the technology.
2 Literature Review
Research has been shown that customer satisfaction is an important factor positively
influencing customer loyalty and post-purchase intention in telecommunication mar-
ket (Kim et al., 2004; Kuo et al., 2009). However, customer satisfaction is a broad and
complex concept, which can be viewed from various perspectives including price,
product design and service quality. Product design and demographic factors have
been investigated as influencing the level of customer satisfaction of a mobile phone.
For example, Han et al. (2004) and Ling et al. (2007) investigated critical design fea-
tures of a mobile phone affecting overall user satisfaction. Yuen et al. (2011) ex-
amined Malaysian citizens to indentify the demographic factors influencing customer
satisfaction.
More attentions have been given to new characteristics and functionalities due to
an increasing number of value-added features on current mobile phones, especially
smartphones. A recent study by J.D. Power and Associates (2011) found that the key
factors of overall satisfaction with a smartphone were not only design features such as
a specific size and weight but latest technological advancement. Among the studies on
the relationship between the new features and customer satisfaction, Park and Lee
(2011) analyzed how user experience and device characteristics of smartphone affect
customer satisfaction. In their pilot study, they showed that instant connectivity is the
most important factor but smartphone stress and enjoyment aren’t. Ling et al. (2006)
showed that color screen, mobile Internet and voice dialing strongly impacted satis-
faction level. Also, Haverila (2011) found that the most important feature is bat-
tery/talk time among male users. Dobrota et al. (2012) analyzed how customer satis-
faction is related to the customers’ foreknowledge, awareness and information about
mobile operation system. Bayraktar et al. (2012) compared customer satisfaction and
loyalty efficiency of six mobile phone brands in Turkish telecommunication market.
While a lot of research has focused on the relationships between new features of
smartphone and customer satisfaction, some have shown the relationships between
smartphone price and customer satisfaction. In Tan et al. (2012), the most important
feature of a smartphone is price plans by telecom operators rather than other features
including technology-related features, design-related features, application-related
features and image-related features. Also, smartphone price is ranked 5th among
11 smartphone features. Lee (2011) claimed that the price of a smartphone has no
relationship with customer loyalty to the mobile service provider.
404 Y.J. Koh, J.H. Park, and B.D. Chung
According to Kuo et al. (2009), customer’s perceived value is also closely related
to customer satisfaction and can be considered from different perspectives including
money, quality, benefit and social psychology ((Bishop (1984), Zeithaml (1988),
Sheth (1991)). From the perspective of quality in mobile services, studies have inves-
tigated the perceived value of short message service (Lai (2004)), mobile commerce
(Lin and Wang (2006)), mobile value-added services (Kuo et al. (2009)). In addition,
self-disclosure, flow and social presence have been considered as affecting user satis-
faction of instant mobile messenger in smartphone (Park et al. (2011)). However, few
have explored the relationship between the satisfaction of smartphone and the per-
ceived value of technology innovation of smartphone. Therefore, this study will ex-
plore how users’ perceived value of technology innovation of smartphone affects their
satisfaction of the smartphone, and include personal characteristics like price-value
perception and the satisfaction of their mobile service provider as important factors.
3 Research Model and Hypotheses
The research model includes demographic and psychological variables determining

users’ satisfaction of their smartphones as shown in Fig. 1. We assumed that Per-
ceived Value of Technology Innovation (PVTI) and Price-Value Perception (PVP)
would determine Satisfaction of Smartphone (SS). We also hypothesized that demo-
graphic characteristics (Gender and Age) and users’ satisfaction of their mobile ser-
vices (SMS) would be related to Satisfaction of Smartphone (SS).
Demographic
Characteristics
(Gender)
H4a
Demographic
Characteristics
(Age) H4b
Perceived Value of
Technology Innovation H1
(PVTI) Satisfaction of
Smartphone (SS)
Price-Value Perception H2
(PVP)
Satisfaction of H3
Mobile Service
(SMS)
Fig. 1. Research Model

Perceived Value of Technology Innovation (PVTI) of a mobile phone represents how

much a smartphone user recognizes and utilizes the innovative value of a smartphone in
contrast to a feature phone. The innovative value of a smartphone can be explained by
the function of data communication. Although a feature phone also provides data com-
munication services, they are very constrained compared to a smartphone. PVTI was
measured in terms of the perceived ratio of the value of data communication in the
whole value of smartphone and the perceived ratio of the value of application messenger
in users’ communication usages. Considering that customer’s perceived value is related
to customer satisfaction, we propose the first hypothesis:
Hypothesis 1. PVTI is positively related to SS.
In general, smartphone users pay higher telephone rates as well as use more expensive
devices, compared to feature phone users. All smartphone users do not think that
higher price provides higher value or quality, affecting the satisfaction of a smart-
phone. We measured Price-Value Perception (PVP) in terms of user's price-quality
schema (Bearden et al., 2011). Therefore, the second hypothesis is:
Hypothesis 2. PVP is positively related to SS.
Mobile phone users use mobile voice and data communication services. Because
functions of a mobile phone are realized based on a mobile telecommunication ser-
vice, we assumed that SMS affects SS. Thus, we hypothesize the following:
Hypothesis 3. SMS is positively related to SS.
Utilities of technology are various according to users’ demographic characteristics. In

general, individual's gender and age are basic considerations in segmentation for mar-
keting. It is important to examine the effect of users’ gender and age on the satisfac-
tion of a smartphone. Thus we propose the following hypotheses:
Hypothesis 4a. Gender is related to SS.
Hypothesis 4b. Age is related to SS.
4 Methods
This study employed a face-to-face survey to obtain 3,798 questionnaires. The res-
pondents were randomly selected from all of major cities and provinces of Korea,
according to the national demographic statistics. Thus, the survey data represents a
summarization of nationwide characteristics of Koreans. Smart PLS was used in the
analysis of this study. First, a measurement model was tested using confirmatory
factor analysis. For the test, individual item reliability, internal consistency and dis-
criminate validity were analyzed. After the measurement model test, a structural
equation model was analyzed and the hypotheses were tested.
Table 1. Characteristics of Respondents

Category Frequency Percentage (%)
Male 2003 52.7
Gender
Female 1795 47.3
13~19 458 12.1
20~29 783 20.6
30~39 905 23.8
Age
40~49 876 23.1
50~59 646 16.7
60~69 140 3.7
M1 2099 55.3
M2 410 10.8
Smartphone
M3 448 11.8
Manufacturer
M4 489 12.9
Etc. 352 9.3
T1 1898 50.0
Mobile
T2 1191 31.4
Service
T3 677 17.8
Provider
Etc. 32 0.8
5 Results
5.1 Reliability and Validity of Measures
Individual item reliability was tested using factor loading of measured variables for
each latent variable. If the factor loading of an item are greater than 0.5, the item
reliability is demonstrated. Composite reliability indexes were used to test internal
consistency of latent variables. The acceptable level of composite reliability index is
0.7. To test the discriminant validity, the square root of AVE (Average Variance Ex-
tracted) was used. It should be greater than 0.5 to verify the discriminant validity.
Table 2 shows the results of the reliability and validity test. Factor loadings are great-
er than 0.5 and the square root of AVE and composite reliability are greater than 0.5
and 0.7, respectively. Thus, reliability of measured variables and discriminant validity
of latent variables are justified.
5.2 Analysis of the Structural Model

Bootstrapping of Smart PLS was used to analyze the research model. Table 3 and Fig.
2 show the results. PVTI and PVP were positively related to users’ satisfaction of
smartphone (SS). The path coefficients of them are 0.100 and 0.091, respectively, and
p-values are less than 0.001. Thus, H1 and H2 are supported. SMS is found to be
significantly related to SS. The path coefficient is 0.324 and p-value is less than
0.001. H3 is supported. The path between Gender and SS was found to be insignifi-
cant. The p-value is greater than 0.05 and thus H4a is not supported. On the other
hand, Age is related to SS. The path coefficient is 0.038 and p-value is less than 0.05.
Table 2. Results of Validity and Reliability Test
Factor SQRT Composite

Loading (AVE) Reliability
Perceived Value of Technology Innovation 0.86 0.85

What do you think is the ratio of the value of mobile internet 0.87
service among all of the services provided by mobile com-
munication?
What do you think is the ratio of the usage of communication 0.86
though messenger application among all of your mobile
communication?
Price-Value Perception 0.94 0.96
Generally speaking, the higher the price of the product, the 0.94
higher the quality
The price of a product is a good indicator of its quality 0.94
You always have to pay a bit more for the best 0.93
Satisfaction of Mobile Service 1.00 1.00
How much are you satisfied with your mobile communication 1.00
service provider?
Gender 1.00 1.00
What is your gender? 1.00
Age 1.00 1.00
What is your age? 1.00
Satisfaction of Smartphone 1.00 1.00
How much are you satisfied with your smartphone? 1.00
Table 3. Results of Hypotheses Test
Hypothesis Path Path Coef- t-value p-value

ficient
H1 Perceived Value of Technology Innovation 0.100 5.305 0.000
→ Satisfaction of Smartphone
H2 Price Value Perception → Satisfaction of 0.091 4.966 0.000
Smartphone
H3 Satisfaction of Mobile Service → Satisfaction 0.324 18.508 0.000
of Smartphone
H4a Gender → Satisfaction of Smartphone -0.018 1.200 0.237
H4b Age → Satisfaction of Smartphone 0.038 2.097 0.042
Demographic
Characteristics
(Gender) -0.018
(1.200)
Demographic
Characteristics
0.038 *
(Age)
(2.097)
Perceived Value of
Technology Innovation 0.100 * * *
(PVTI) (5.305) Satisfaction of
Smartphone (SS)
Price-Value Perception 0.091 * * *
(4.966) R2 = 0.127
(PVP)
0.324 * * *
Satisfaction of (18.508)
Mobile Service
(SMS) * : p<0.05, ** : p<0.01, *** : p<0.001
Fig. 2. Results of Research Model and Hypotheses Test
6 Discussions
6.1 Summary of Findings
This study provides a strong evidence that the satisfaction of smartphone is affected
by the perceived value of technology innovation of a smartphone. Two measured
variables, the value of mobile internet service and the ratio of the usage of communi-
cation through messenger application, were used and showed to be reliable to
represent the perceived value. This means that although people accept new technolo-
gy, they have different acceptance level of value provided by the technology and thus
the satisfaction of a new technology is influenced by not only its features but also
users’ acceptance level.
Another finding is that the relationship between the price-value perception and the
satisfaction of smartphone is significant. This means that some people may be dissa-
tisfied by the high price although they get great value from technology innovation. A
user’s satisfaction with the mobile service provider he/she subscribes to is a signifi-
cantly associated with the satisfaction of the smartphone. This shows that satisfaction
of a technology which is enabled by another one is influenced by the other one. As
expected, a user's age has a significant relation with the satisfaction of smartphone.
On the other hand, the gender is not significantly related with the satisfaction of a
smartphone. This shows that there are some other factors which are affecting on the
satisfaction of a smartphone and have in relation with age but not with gender.
6.2 Contributions and Limitations

This study expands the concept of perceived value by showing that people show the
perceived value from the perspective of technology, in addition to that of money,
quality, and benefit. The results show that the higher the perceive value of a smart-
phone is, the higher the satisfaction of smartphone is. Additionally, the findings sug-
gest that individuals’ price-value perception affects the satisfaction of smartphone.
This study also provides that the satisfaction with a telecom service provider af-
fects the satisfaction of a smartphone, meaning that the satisfaction of a smartphone
should be considered by the combination of the device and the service of a smart-
phone. Finally, findings suppose that all people do not fully enjoy the benefits from
the technology innovation, showing that users have different acceptance levels of the
smartphone value. Thus, different approaches and services should be prepared de-
pending on the level of acceptance, in order to help people who do not easily follow
and accept the technology innovation.
In spite of those contributions, this study has some limitations. Although the study
supposes that the users' personal characteristics have a relation on the satisfaction of a
smartphone, other factors were not considered except the price-value perception. Such
as involvement and technology readiness can also be used as affecting factors.
The model can be expanded to include users' intention to repurchase a smartphone to
build a model explaining the relations from the perceived value to retention through
satisfaction of technology innovation. These limitations should be our further works.
References
1. Kuo, Y.F., Wu, C.M., Deng, W.J.: The relationships among service quality, perceived val-
ue, customer satisfaction, and post-purchase intention in mobile value-added services.
Computers in Human Behavior 25(4), 887–896 (2009)
2. Kim, M.K., Park, M.C., Jeong, D.H.: The effects of customer satisfaction and switching
barrier on customer loyalty in Korean mobile telecommunication services. Telecommuni-
cations Policy 28(2), 145–159 (2004)
3. Han, S.H., Kim, K.J., Yun, M.H., Hong, S.W., Kim, J.: Identifying mobile phone design
features critical to user satisfaction. Human Factors and Ergonomics in Manufacturing &
Service Industries 14(1), 15–29 (2004)
4. Park, Y., Chen, J.V.: Acceptance and adoption of the innovative use of smartphone. Indus-
trial Management & Data Systems 107(9), 1349–1365 (2007)
5. Ling, C., Hwang, W., Salvendy, G.: Diversified users’ satisfaction with advanced mobile
phone features. Universal Access in the Information Society 5(2), 239–249 (2006)
6. Yuen, Y.Y., Yeow, P.H., Connolly, R.: Mobile phone satisfaction in Malaysia: a demo-
graphic analysis. International Journal of Mobile Communications 9(4), 341–358 (2011)
7. Ling, C., Hwang, W., Salvendy, G.: A survey of what customers want in a cell phone de-
sign. Behaviour & Information Technology 26(2), 149–163 (2007)
8. Dobrota, M., Milenković, I., Bulajić, M.: The Influence of Foreknowledge on Custom-
ers’Satisfaction With Mobile Operating Systems. Business Intelligence and Decision Mak-
ing in Management 251
9. Bayraktar, E., Tatoglu, E., Turkyilmaz, A., Delen, D., Zaim, S.: Measuring the efficiency
of customer satisfaction and loyalty for mobile phone brands with DEA. Expert Systems
with Applications 39(1), 99–106 (2012)
10. Bishop Jr., W.R.: Competitive intelligence. Progressive Grocer 63(3), 19–20 (1984)
11. Zeithaml, V.A.: Consumer perceptions of price, quality, and value: a means-end model and
synthesis of evidence. The Journal of Marketing, 2–22 (1988)
12. Sheth, J.N., Newman, B.I., Gross, B.L.: Consumption values and market choices: Theory
and applications, pp. 16–74. South-Western Pub. (1991)
13. Lai, T.L.: Service quality and perceived value’s impact on satisfaction, intention and usage
of short message service (SMS). Information Systems Frontiers 6(4), 353–368 (2004)
14. Lin, H.H., Wang, Y.S.: An examination of the determinants of customer loyalty in mobile
commerce contexts. Information & Management 43(3), 271–282 (2006)
15. Park, B.W., Lee, K.C.: A Pilot Study to Analyze the Effects of User Experience and De-
vice Characteristics on the Customer Satisfaction of Smartphone Users. Ubiquitous Com-
puting and Multimedia Applications, 421–427 (2011)
16. Park, S., Oh, D., Lee, B.G.: Analyzing user satisfaction factors for instant messenger-based
mobile SNS. In: Park, J.J., Yang, L.T., Lee, C. (eds.) FutureTech 2011, Part II. CCIS,
17. Lee, J.W.: Critical factors promoting customer loyalty to smartphone and mobile commu-
nications service providers (2011)
18. Bearden, W.O., Netemeyer, R.G., Haws, K.L.: Handbook of Marketing Scales, 3rd edn., p.
379. SAGE Publications (2011)
19. Tan, W.K., Yi-Der Yeh, S.J.C., Lin, Y.C., Kuo, C.Y.: How Consumers Assess Product’s
Features?: A Case Study of Product Features of Smartphone. In: Proceedings of the 6th In-
ternational Conference on Applied Mathematics, Simulation, Modelling, pp. 131–137.
World Scientific and Engineering Academy and Society, WSEAS (March 2012)
20. Power, J.D.: The Right Blend of Design and Technology is Critical to Creating an Excep-
tional User Experience with Smartphones and Traditional Mobile Devices. JD Power and
Associates (2011)
Semantically Integrated Business Applications
for Enterprise Resource Planning Systems
Patricia Kraft and Rainer Thome
Julius-Maximilian-University, Chair of Business Information Systems,

Wuerzburg, Germany
{pkraft,thome}@wiinf.uni-wuerzburg.de
Abstract. The development of enterprise resource planning (ERP) software or

enterprise information systems (EIS) has always been closely linked to the
discussion on integration. Over time, integration of software solutions has
improved considerably. This has been achieved through several changes made
in the past, but mostly it has become possible thanks to cutting-edge
technologies. A new approach is semantically integrated business applications
that enable the evolution of standard end-to-end software.
Keywords: Business applications, software development, enterprise resource

planning (ERP), integration, add-on, enterprise information systems.
1 Introduction
The development of solutions for supporting the processes and organization of

companies with enterprise resource planning (ERP) software has always been ac-
companied by a call for more integration. The ERP systems are the target of a great
many demands, all of which various advocates expect the systems to meet. Depend-
ing on the specific point-of-view, some demands seem more pressing than others.
These include calls for more flexibility, because in adapting to the real world, or-
ganizations are subject to constant change. But software should support change
rather than be so inflexible as to hinder it and reduce an organization’s competitive-
ness. Costs are also an issue. Enterprise software is not a primary driver of profit.
As a result, companies strive to keep the cost of their information system as low as
possible. A third point of conflict is functionality. Organizations’ user departments
raise this point because to them, the system is only beneficial if it can model all
relevant corporate processes. But additional functionality increases complexity,
compromising a system’s transparency and rendering it inflexible. What is more,
functionality is the main cost driver. Integration receives the weakest representation
in this discussion. It is often assigned the lowest priority because its advocates are
few and have little influence. In an ERP system, integration (of technical aspects
and of applications) is the key to the smooth operation and modeling of processes,
decreased maintenance costs and consistent data pools. Integration can also include
412 P. Kraft and R. Thome
the look and feel of the user interface, providing consistency for the user. But these
points are not concrete, cannot be implemented immediately, are difficult to quanti-
fy, are not a priority for management and are therefore at a disadvantage when
compared to the interests previously mentioned.
This paper will aim to answer the question: What changes will cloud-based solu-
tions bring to ERP software and how will these changes influence integration of soft-
ware development long-term? First, an overview of the research done on integration
of ERP solutions will be presented and the most recent developments in cloud-based
ERP systems described. Next, semantically integrated applications, an integrated
method for developing business application software, will be introduced.
2 State of the Art
FINK and HANSEN/NEUMANN define the term enterprise resource planning (ERP)
system as a multiple-component business application system for end-to-end informa-
tion processing. These systems are used to model and execute processes in all user
departments of an organization (e.g. Production, Procurement, Accounting, etc.). In
addition to other manufacturers, such as Oracle and Microsoft, SAP is the worldwide
market leader for business application software both for large companies and for
small and mid-size businesses [1][2].
Development in the area of integration of business application software up to the
present can be divided into phases or blocks along a timeline. HANSEN and
NEUMANN created a breakdown (Table 1) that divides development into five dis-
tinct phases. These five phases reveal development toward increasingly open plat-
forms, which can communicate with each other via the Internet and provide a network
that extends beyond an organization’s borders. In 2005, when this hypothesis was
posed, cloud solutions were not prevalent for ERP software. This thought has only
begun to push its way into the conversation since a range of offerings in this sector
has been launched onto the market and gained widespread popularity.
Software as a service can be described as a software delivery and business model
by which the software vendor enables customers use of the standard software applica-
tions via the Internet as a service. In contrast to traditional on-premises software ap-
plications, the customer does not acquire a software license, but subscribes to the
service. The software itself is hosted by the software vendor and does not require
implementation or deployment at the customer site. The specific pricing model for
SaaS is commonly based on a usage metric, such as per user per month.” [3]. Solu-
tions also exist as part of the platform as a service model. In this case, the object of
examination is not just software but a platform that unites multiple offerings and ven-
dors selling their cloud solutions and services. In other words, the platform is actually
the model of a partner ecosystem that delivers additional offerings for the solutions
available [4].
Semantically Integrated Business Applications 413
Table 1. Integration phases of enterprise software [2]
Phase 1 Phase 2 Phase 3 Phase 4 Phase 5

1960-1980 1980-1990 1990-2000 2000-2010 2010-2020
Isolated func- Application Internal busi- Cross- Extensive

Area
tions areas ness processes company corporate

processes networking
Customer- One manufac- One manufac- One manufac- Combination
specific pro- turer’s pro- turer’s pro- turer’s proof several
gramming, prietary appli- prietary appli- prietary appli- manufactur-
classic data cation soft- cation soft- cation soft- ers’ applica-
organization ware on a few ware on many ware on many tion software
IT technology
for proprie- proprietary proprietary proprietary or on open-

tary platform platforms platforms open plat- source plat-
forms forms
Terminals Terminals Client-server Client-server SOA system,
connected to connected to system system, Inter- web services
mainframe mainframe net-based
computers computers
Text-oriented Text-oriented Graphic UI Web-based UI Web-based
UI UI (browsers) UI (browsers)
Up to now, the prevailing systems on the ERP market have been on-premises solu-
tions. The customer installs, maintains and adds to these, if more functionality is de-
sired. These add-ons are normally intended for a single customer, since each customer
has his or her own requirements. Add-ons are either programmed by the customers
themselves, or customers hire a software manufacturer to develop a solution. This
gives rise to a number of problems: it compromises future ERP software releases, on
the one hand, and on the other, customer-specific programming of the solution causes
maintenance costs to skyrocket. All changes to the ERP system that go beyond those
made via customization settings (adjustment of parameters) hamper flexibility and
complicate maintainability. This knowledge leads to the demand for standard software
and the adjustment of organization and processes to the ERP solution within its scope
of performance. The continuation of this thought results in Continuous System Engi-
neering (CSE) by THOME, the ongoing development of application software in inte-
raction with the changes in corporate organization [5].
In the past decade, the issue of integration has been examined by a number of
authors but never defined completely and conclusively. Many specialists in the field
have attempted to quantify integration, and have illuminated it from various angles.
MERTENS, for example, makes a distinction between the range, direction, object and
scope of integration [6]. HAHN advocates a different approach. He distinguishes
among data, model, and process integration. These levels can be closely or loosely
linked to one another [7]. JANSEN and HUFGARD introduce the business aspect.
HUFGARD et al. maintain that the term business integration includes technical as-
pects of information processing, as well as organizational and marketing aspects [8].
JANSEN’s “six target levels of integration” comprise strategic, organizational, human
resources, cultural, operative and external levels of integration [9]. This diversity of
definitions makes clear the complexity of the term integration and its far-reaching
influence on an organization’s processes. In sum, one can say that for integrated busi-
ness processes, it is absolutely essential to be able to:
• access all relevant data

• save it without redundancy immediately after its creation,
• update software to state-of-the-art levels, and
• communicate digitally with other business partners across systems and company
borders.
The user may not be impacted by changes in the system or in the look and feel of the
user interface and yet, adjustments made necessary by organizational or business-
related changes must be integrated into the system quickly and seamlessly.
3 Cloud ERP Systems Today

When comparing simple, cloud-based solutions to their locally installed competitors,
one might wonder whether saving data and applications on external servers makes a
difference with respect to integration. A cursory glance can lead to suspicions that
cloud-based solutions are merely a marketing campaign and that the flashy new
façade of cloud products conceal only externally-hosted variations of existing solu-
tions. At first glance, the disadvantage of saving data in the cloud is that customers
lose data sovereignty over their most valuable corporate information. But this argu-
ment can be refuted by the simple reflection that most organizations that purchase
cloud applications are small and midsized businesses that do not have the necessary
resources to operate their own high-security data centers. In this case, the data is
probably safer in the hands of an external service provider than on the company’s
own premises. Another question has to do with workflow. When a web browser is
used, it might seem as though users are more remote than when they use an in-house
solution. But really, accessing an ERP system via the Internet is simpler and more
secure than accessing an internal solution that is purposely blocked off from the ex-
ternal world and can be accessed only remotely via secure VPN entry points, if at all.
The result is that, for example, sales reps can better access a cloud solution from a
customer’s premises than access a solution installed solely within their own company.
SaaS improves networking and availability of all corporate data.
Regardless of the availability of and security issues concerning these solutions, the
software structure holds the key to groundbreaking changes. Popular solutions on the
market, such as NetSuite and Salesforce.com all offer the benefits of a web applica-
tion, but behind their façades, they are nothing more than conventional software
packages, i.e. software as a service [10]. But SAP’s new Business ByDesign solution
has a completely different structure; it is a whole new type of application.
At the heart of SAP Business ByDesign is Business Configuration, a catalog for

setting all parameters of the ERP solution. It forms the basis for the application’s
configuration, the modeling of each customer’s requirements. Users answer simple
Yes or No questions to create, check and implement organization characteristics relat-
ing to system functionality. And then, depending on the answers to the questions, the
application areas are parameterized with the help of a fine-tuning tool. It recognizes
soft factors in the organization, migrates data and creates structures. This includes, for
example, defining various models for employee working hours and migrating custom-
er master data via an XML interface [11] [4].
Another unique component of ByDesign is the cloud-based development environ-
ment, Business ByDesign Studio. It can be used to develop add-ons that go beyond what
is offered in the standard solution, to set up a link to external software, and to customize
the standard software components. These two types of functionality differ very little
from capabilities available in on-premises solutions. Customer-specific programming
makes it possible to create each ERP solution in line with customer needs and to add
functions. However, making changes in ByDesign via public solution models (PSM) is
a completely different story. PSM contains a collection of data fields, methods and
structures that programmers (even external programmers) can access via the ByDesign
Studio development environment, in order to program additional functionality. PSM
content adapts to future releases, which ensures that a one-off solution will still work
after a release upgrade. Whenever a partner company develops an add-on for Business
ByDesign, it automatically becomes part of the overall solution. The new components
are sold in the SAP Store (a platform like an app store that offers add-on solutions
as applications). When customers purchase a solution, it appears in their business
configuration and they can make immediate use of it [4].
4 Infrastructure for Business Application Development

and Integration of Applications
The approach just described for developing add-ons for a cloud-based solution – fus-
ing add-ons with the host’s infrastructure and enabling them to become part of the
overall solution – offers a higher degree of integration from the perspective of all
stakeholders than the approach previously used to develop add-ons for on-premises
solutions. For the developer, using SAP software development kits (SDK) closely
links new functions to the overall solution. The development environment is based on
the PSM settings and contains specific patterns for creating the user interface that
correspond to the solution and its outward appearance. The user cannot tell which
screen is part of the standard solution and which has been added. The development
environment ensures that all requirements are met for an integrated, compatible solu-
tion that will also work with future release upgrades.
For the customer, this approach improves business flexibility; the system offers
multiple setting options, and add-ons from the SAP Store can easily be downloaded to
it. The vast number of customers it benefits also makes it less expensive than compa-
ny-specific developments for on-premises solutions. And, for the first time it has
made possible an objective comparison of alternative business applications on the

market and their respective providers. If a solution is not available, a customer can
ask an SAP partner company to create it. If a solution becomes obsolete because, for
example, an organization has changed its structure, the user can very simply deacti-
vate the solution. As a result, an organization is not weighed down with superfluous
coding from one release to the next.
The provider and platform hosts face higher infrastructure costs by providing the
solution, and lower maintenance costs with respect to new releases. “The SaaS vendor
provides multiple tenants, one tenant per customer, based on a shared technology
stack including hardware infrastructure, database and application” [3]. Integration of
infrastructure is seamless because the tenants share the same hardware. Because all
users are forced to upgrade to the new release during the two-to-three-month transi-
tion phase, a maximum of two releases are maintained – the current and the future
release. The entire infrastructure is state-of-the-art and no legacy systems need to be
maintained. All stakeholders are in the same boat. Web services support communica-
tion with external programs. They unite the communications structures and ensure
semantically-correct transmission of data. Communication between various different
business partners, all with separate ERP systems, is done this way, ensuring seamless
communication across company and system borders.
If one compares this to the degree to which ERP systems are integrated today,
marked discrepancies become apparent. The transition from the on-premises world to
that of cloud-based solutions with the option to add semantically integrated business
applications is comparable to the step from terminal computers to desktop computers.
A stand-alone, partially linked solution becomes a completely integrated standard
solution that grows in proportion to user demands.
In sum, semantically integrated business applications offer the following advantages:
• Development takes place within an SDK provided by the manufacturer,

• The user cannot distinguish between the standard solution and the add-on, with
respect to the look and feel of the user interface,
• The add-on is an integral part of the functionality within the basic ERP solution,
• The new function does not compromise future upgrades for the overall solution,
• The solution is activated and configured along with the default functions (e.g. via
Business Configuration),
• All customers who have the default system can use the solution,
• The solution is sold at a marketplace that is equally accessible to all customers.
Systems that meet all of the above-mentioned criteria are considered semantically
integrated business applications for cloud-based ERP solutions.
References
1. Fink, A., et al.: Grundlagen der Wirtschaftsinformatik, p. 207. Physica, Heidelberg (2005)
2. Hansen, H., Neumann, G.: Wirtschaftsinformatik 1, p. 529. Lucius & Lucius, Stuttgart
(2005)
3. Bandulet, F., et al.: Software-as-a-Service as Disruptive Innovation in the Enterprise Ap-

plication Market. In: Benlian, A., et al. (eds.) Software-as-a-Service. Anbieterstrategien,
Kundenbedürfnisse und Wertschöftungsstrukturen, pp. 16–17. Gabler, Wiesbaden (2010)
4. Schneider, T.: SAP Business ByDesign Studio. Application Development, pp. 24–28. Ga-
lileo Press, Boston (2012)
5. Thome, R.: Grundzüge der Wirtschaftsinformatik, pp. 149–159. Pearson, Munich (2006)
6. Mertens, P.: Integrierte Informationsverarbeitung 1 - Operative Systeme in der Industrie, p.
2. Gabler, Wiesbaden (2007)
7. Hahn, A.: Integrationsumgebung für verteilte objektorientierte Ingenieursysteme, pp. 7–23.
Universität Paderborn Heinz Nixdorf Inst., Paderborn (1998)
8. Hufgard, A., et al.: Business Integration mit SAP Lösungen, pp. 15–16. Springer, Berlin
(2005)
9. Jansen, S.: Mergers & Aquisitions: Unternehmensakquisitionen und -kooperationen. Eine
strategische, organisatorische und kapitalmarkttheoretische Einführung, p. 330. Gabler,
Wiesbaden (2008)
10. Salesforce: Salesforce.com., http://www.salesforce.com/company/
11. Hufgard, A., Krueger, S.: SAP Business ByDesign: Geschäftsprozesse, Technologie und
Implementierung anschaulich erklärt, pp. 154–177. SAP Press, Bonn (2012)
Towards a Next Generation Universally Accessible
‘Online Shopping-for-Apparel’ System
Kasper Kristensen, Nanna Borum, Line G. Christensen, Henrik W. Jepsen,

Jacob Lam, Anthony L. Brooks, and Eva Petersson Brooks
Aalborg University Esbjerg, Niels Bohrs vej 8, Esbjerg, Denmark

{tb,ep}@create.aau.dk
Abstract. This paper reports initial research findings from on-going longitudin-
al participatory design studies within a national (Danish) funded project to real-
ize a gesture-controlled ‘Online shopping-for-apparel’ system – A Virtual
Dressing Room (VDR). A product that reduces customer purchase returns, re-
portedly up to 40%, which is a huge burden to the clothing industries as shop-
ping percentile of sales online continues to increase, is targeted. Three studies
are reported where results cumulate to highlight the need for continued research
to realize a next-generation system to improve the user experience of online
shopping for apparel where conclusions point to the need for adaptive user in-
terface improvements. Unforeseen was that wheelchair-bound public especially
responded positively to the potentials for the concept due to their limited mobil-
ity in shopping and this accessibility aspect can be a significant future market.
Keywords: Online shopping system, purchase returns, e-shopping experiences.
1 Introduction
The United Kingdom Office of National Statistics reveals that the average weekly
value for Internet retail sales in June 2012 was estimated to be £493.3 million1. It is
predicted that by 2018, 35% of clothing sales will be via the Internet2. However, cur-
rent figures show that between 15-40% of apparel purchased online (e-shopping) is
returned because customers deem they do not fit3 or do not look right4. Trust policies
that safeguard customer purchases enabling such returns are influential and, in some
cases, these policies are reported as being abused whereby a purchase is received,
worn, and then returned with full credit or refund. In the clothing industry, such
high volume of returned apparel is economically disastrous and it is in respect of this
1
http://www.ons.gov.uk/ons/rel/rsi/retail-sales/june-2012/
stb-june-2012.html
2
Heikki Haldre, a founder of Fits.me, an online fitting room service that creates custom
virtual mannequins for people to dress with clothes before buying them.
3
http://online.wsj.com/article/
SB10001424052702304724404577293593210807790.html
4
http://www.imrg.org
Towards a Next Generation Universally Accessible ‘Online Shopping-for-Apparel’ System 419
problem that the project is directed. Thus, the main goal of this work is to realize an
online shopping-for-apparel system in the form of a Virtual Dressing Room (VDR)
where contemporary camera-based ICT is used to reduce consumer returns.
In order to develop a turnkey solution, a participatory design (PD) approach has
been applied in line with Brandt [1]. Fundamentally PD represents a set of theories,
practices and studies whereby user communities play a substantive role in activities
that can lead to the creation of software and hardware computer technologies and their
application in real-world contexts [2-4]. The benefits of adopting such an approach
for the design of the VDR system include a better understanding of the reasoning
behind online purchasing behavior. This is important particularly given that these
numbers are expected to rise even more in the future with predictions of online sales
in Western Europe to increase at a 10% compound annual growth rate over the next 5
years5. Findings from the PD field studies are periodically fed into the technical part-
ners process pipeline to support development decisions. This paper presents the result
from the two initial phases of the design process, namely the preparation and the in-
cubation phases (c.f. Seifert et al. [5]), which represent the initial inspirational and
definition phases where the problem was defined and potential solutions identified.
2 Related Work
In the clothing industry many fit technologies have been tried and flopped. For exam-
ple, retailers have created virtual mannequins for customers to dress and set up
full-body scans in stores. It was found that shoppers are reluctant to use systems that
require much effort and time unless a special purchase is the goal. It is acknowledged
that for women's clothing in particular, sizing is difficult for many complex issues
including the psychological. One brand's size 12 is another's 10 or 14, primarily be-
cause fashion labels shape their clothes so differently, using their own closely guarded
specifications to create patterns. Online retailers often ask shoppers to consult "fit
charts" or type in body measurements, which can be time-consuming. However, the
True Fit system reports increased sales and reduced returns up to 30% (premier de-
nims). Consumers create a profile of age, height, weight and body shape. Then the
customers select items that fit well from their own closets and identify the brands,
styles and sizes to True Fit. Another personalized online shopping approach by Stitch
Fix 6 involves female customers filling out a similar profile form whereafter online
photos of clothing and accessories are rated to give the company an idea of the client’s
taste. NoMoJeans 7 takes precise customer measurements with a 3D body scanner
which are kept on a data base for future purchases.
This paper reports initial trendspotting (field) and participatory design (PD) stu-
dies, conducted to unfold tendencies, key patterns and personas related to on- and
offline shopping. The trendspotting included structured interviews with different user
5
http://www.forrester.com
6
https://www.stitchfix.com
7
http://www.nomojeans.com
420 K. Kristensen et al.
segments representing consumers and retailers/industry. The consumer segment were

women 18-30 years of age (n=15). Furthermore, observations of consumer shopping
behavior was conducted in combinations with focus group interviews with females
18-35 years of age (n=7), expert interviews (sales staff) (n=3), and document analysis
(magazines, shopping forums, statistics, industry reports).
A field study was conducted at a public trade fair in Copenhagen, which included
video observations, questionnaires and unstructured interviews (n=11) where the free-
ly available LazyLazy8 system was used. Building on these studies a PD study was
undertaken at a shopping mall in Denmark’s second city Aarhus. This included video
observations, questionnaires and a focus group (n=31; females n=20). Following, a
low fidelity prototyping approach was used in order to get useful design feedback
from students attending Aalborg University Esbjerg, Denmark.
3 Retailer/Industry Trend Spotting/State-of-the-Art Study
Five selected companies were visited to determine state of the art in the industries
associated with apparel shopping in order to supplement the PD research. These were:
1. Inition - http://inition.co.uk - a preliminary prototype was presented based on a

single Kinect live mirroring the customer (similar to LazyLazy). The system was
demonstrated in combination with interviewing of the system creator and VP sales.
Clothes are selected from an image menu and a superimposed overlay positioned
onto the live video feed, thus, in this system there is no need to disrobe.
2. Shape Analysis - http://shapeanalysis.com - a proprietary multi-camera system (4)
marketed in partnership with [TC]2 in USA. The customer needs to disrobe for
precision scanning. Their latest systems offer a choice between a single Kinect
(2D) or four Kinect system (3D – 360 degrees). The software generates an avatar
(wireframe but can superimpose own face or selected clothing) giving precise mea-
surements. The system was demonstrated by creating the researcher’s body scan.
Furthermore, interviews with the creator and marketing person were conducted.
3. Bodymetrics - http://www.bodymetrics.com - a proprietary multi-camera (4) sys-
tem with installations in leading stores in London (Selfridges – where
tested/interviewed staff and clients) and USA (Bloomingdales). The latest system
uses single or multiple Kinect cameras for boutiques and home uses. The software
generates an avatar that mirrors customer rotation in real-time. The interviewer
viewed the system together with the CEO/creator’s representative and VP sales.
Furthermore, staff (2) and customers (2) were interviewed. Customers com-
plained of a need to initilly disrobe for scanning, then having to dress to get cloth-
ing from staff in the public area, and then to go to a fitting room to disrobe again to
test apparel feel and fit. This process is being changed using a female only area.
4. Holition - http://www.holition.com – as (1)
8
http://www.lazylazy.com
5. Cristina Holm, Personal Image Stylist - http://www.cristinaholm.com - a four-

camera scanner setup with software generating an avatar (static generated for pre-
cise measures – as 2). The [TC]29 NX16 scanner is located in a private studio
where made-to-measure exclusive clothing is designed and sold. [TC]2 company
launched a more affordable Kinect-based solution in February 201210. The system
was demonstrated with resultant scan obtained. The owner/stylist, was interviewed
alongside the system inventor. The 3D avatar can be rotated by a mouse and perso-
nalized with an image of the client’s face.
The protype by Holiton/Inition was without accurate measures and used flat screen
monitors in landscape mode. All of the companies are developing a single Kinect for
home use and multiple systems targeting boutiques, custom tailors/designers, etc.
3.1 Customer Trend Spotting Study
The customer trend spotting study resulted in a shopping behavior analysis and perso-
nas in terms of fictional characters. The shopping behavior analysis identified the
following main trends:
─ Fusion between online and offline shopping. The customers would like to see big
screens in the stores where it is possible to link to the online webshop.
─ Service level. Some of the customers considered online shopping convenient and
stressed that offline service should be improved to give value to offline shopping.
─ Convenience. The customers preferred online shopping due to that it is not bound
to time and space so they will, thereby, save time; they do not need any transporta-
tion; and they do not need to wait in any line.
─ Feel, touch and realism. The customers stressed that the feeling of the clothes, the
touch, and the realistic trying-on experience are desired features that enhance the
shopping.
─ Online tools enhancing offline shopping experiences. The customers suggested that
online tools, such as size and fit information, texture detail, matching of clothes to
the individual’s specific body type would enhance the offline shopping experience.
Furthermore, social media, such as FaceBook, where the customer can share pho-
tos of the potential purchase and receive immediate feedback, would also enhance
the offine shopping.
Furthermore, the customer trendspotting study showed that personal issues such as
sizes in not purely a technical question for virtual dressing room solutions, it consti-
tutes a highly emotional issue. Some of the customer suggested that this issue should
be handled with humor, for example the body could be replaced with sketchy models
or 3D avatars attached with humor-directed comments, e.g. “yes, you fit into this
model, but you cannot breath....”.
9
http://www.tc2.com
10
http://www.tc2.com/newsletter/2011/113011.html
Four personas, fictional characters, emerged from the customer trendspotting

study:
─ The practical non-shopper, represented by Anna, a teacher, 29 years of age. She
works roughly 37 hours per week. Buys clothes one (1) to two times per year,
mostly in stores. She always checks the quality in a store before ordering clothes
online. She buys for practical reasons, mainly out of necessity. She can go shop-
ping by herself but also with friends to have a good day out. I like to try on the
clothes and I like the social experience with my friends. I would definitely also buy
clothes online if I could afford it. Since I cannot afford it very often, I choose to
make special occasions out of the real shopping experience.
─ The lone shopper, represented by Jacob, university student, 23 years of age. Jacob
works as a weekend waiter and buys clothes in stores and online. He usually goes
for offers and mainly shops alone as he finds it a bit boring. I mostly need func-
tional clothing and in the end it is me who should feel comfortable wearing the
clothes. Others should not decide what clothes I should wear.
─ The social shopper, represented by Tina, senior college student, 20 years of age.
Tina works one or two days per month in a bookstore. She lives with her parents
and prefers to buy clothes in stores and often with her friends. She uses 1000 DKK
per month on clothes (on average). She looks for the brands and good offers when
shopping. Tina shops on average 12-20 times per year and often this is combined
with a day with her friends. It is just one of those things that is more fun to do to-
gether with friends combined with café-visits.
─ The spoiling-mom shopper, represented by Linda, physiotherapist running own
clinic, 44 years of age. Linda works roughly 37 hours per week having two child-
ren. She most often buys clothes for others than for herself. When she buys clothes
for herself, it is moslty for practical reasons. Linda enjoys buying clothes in offline
shops, but occasionally buys online. When buying for others, she is very aware of
fashion and trends. When I buy clothes for others, I do so because I see a style in
the stores which I think would fit the person, then the price doesn’t matter that
much.
3.2 Field Study

Shortly after the trendspotting study focusing on industry and costomers (i.e. sections
3.1 and 3.2) an initial field study with public interviewees was conducted at a major
Scandinavian Trade Fair event that attracted more than 7000 attendees11 to the Health
& Rehab Scandinavia at the Bella Center, Copenhagen 22-24 May 2012.
Questioning of attendee trends in shopping (online/offline) was conducted from a
large stand where videos of current state-of-the-art systems and simulations were used
to question public opinions of online shopping for apparel systems (figure 1). Wheel-
chair bound interviewees made clear their need for improved means to purchase
clothes without having to leave their homes. The perceived ease of use (PEOU)
and perceived usefulness (PU) toward enhancement and reduced effort of an online
11
http://www.liftup.dk/index.php?id=130,247,0,0,1,0
Towards a Next Generation Uniiversally Accessible ‘Online Shopping-for-Apparel’ System 423
apparel purchasing system were seen as key factors for wheelchair-bound consum mers.
However, it is clear that the UI needs to be adaptive to overcome challenges suchh as
illustrated in figures 2 and 3 that highlight interface/setup design considerations.
Fig. 1. Different statte-of-the-art virtual dressing room solutions questioned
Fig. 2. Current UI problems (left; height/handedness) (right; distance view/operation)
Figure 2 (left) illustratess a participant who was in a wheelchair, which placed hhim
lower in the picture than otther participants. As a result, he could not reach the buttton
at the top. He used solely hish right hand to make selections, which provided difficcul-
ties when trying to reach th he buttons on the left side of the screen. Younger childdren
had similar issues with reacching as they were also positioned lower in the picture ((due
to their height). The figure 2 right image illustrates a participant who was standing too
far away from the camera, which was necessary in an attempt to see her lower boody
and the selected to preview w trousers. As a result, she had trouble reaching the butttons
at the sides of the screen an nd the image was too small because of the distance. Figgure
3 (left) illustrates how the participant had difficulties interacting with the system be-
cause the buttons were actiivated by motion (of other people) in the background. T The
right image (figure 3) illusttrates persistent single-handed interaction and body twisst to
operate, even though particiipants complained it uncomfortable.
Fig. 3. Current UI probleems (left; background triggering) (right; single handedness)
3.3 Participatory Desig

gn Studies
Following on from the abo ove-mentioned investigations, two teams of university stu-
dents conducted participatoory design studies at a leading shopping mall in Aarhhus,
Denmark’s second city. An n online retail boutique system called LazyLazy was deem-
onstrated to offer research
h observations of an online apparel system use. The ddata
from this study led to UI investigation via focus groups of a created test imagee of
possible improvements to thhe commercial product’s user interface.
LazyLazy is an online sh hopping mall where customers can purchase via the Innter-
net (screen capture examplles figures 2 & 3). The observed demonstration set up for
the studies included a flatsscreen TV with a camera used by the software, and ttwo
cameras for ethnographic analysis, i.e. the prototype sessions, video observatioons,
Fig. 4. The general set

s up of the user study at Bruun’s Galleri, October 2012
Towards a Next Generation Uniiversally Accessible ‘Online Shopping-for-Apparel’ System 425
questionnaires, and unstructured interviews (figure 4). The results were used to genner-
ate an improved User Interrface (UI) being proposed (figure 5) that was tested by fo-
cus groups of independent university
u students.
The results indicated thaat some of the users asked why it was impossible to m move
the clothes left and right when
w they could move them vertically. They requested the
option to select other clothees directly from the system, without having to return to the
main page (which in the LazyLazy
L system requires going back to the computer and
using the mouse. The systtem should be able to remember the size of an adjussted
piece of clothing, as well ass recognize the size of the user in general. Furthermore, the
users would like to be ablee to view accessories (e.g. watches, hats, glasses) togetther
with the clothes. The UI sh hould allso offer the possibility to view the back of a piiece
of apparal (especially sincee some have patterns or pictures on the back). Some ussers
also asked for the possibilitty of using voice control to interact with the system.
Fig. 5. Existing interface (left)

( and proposed interface (right) as tested by focus groups
Figure 5 shows a propossed change to the LazyLazy UI with button relocation and
two steps combined into on ne. Three focus group interviews were conducted withh 13
participants, approximatelyy 30 minutes per group. Feedback was that the existting
interface was considered ass less overwhelming for new users; incrementaly takes the
users through step-by-step;; and is very guiding. The proposed interface was connsi-
dered as allowing modificcations of all aspects in one step; buttons were easierr to
reach; and looked nicer. Booth were judged to need improved and larger text. The ppre-
ferred design was 3/10 in faavor of the proposed change of UI design.
4 Conclusions
Online shopping means no ot having to deal with crowds or pushy/inattentive saales

personnel. It also saves money on travel expenses (car, petrol, parking…) and easee of
getting the goods through your
y door via direct postage/freight delivery that meanss no
‘mountain mule’ shopping (or public transport) with numerous bags encumbering the
person. Thus, potentials for an optimal system are huge, however, many constraaints
limit such optimization, mostly feel and fit aspects. The goal of this project is to ad-
vance the state of the art via an optimal dynamic system that would reduce purchased
apparel returns.
Design options in such work include for the user to control a matched avatar
through body gesture. However, this has not been implemented in the VDR project
where the focus is on a mirrored representation with a superimposed overlay using
camera recognition and mapping.
A challenge, known and strengthened according to the PD studies to date, is the feel
and fit aspects. In line with this are the technical challenges involved in cloth dynamics
(mirrored) simulation where the virtual matches the physical such that interactive rate
correspondences between the 2D and 3D views and the simulation addresses geometric
nonlinearity and frictional contact while remaining stable even under rapid user input.
The Sensitive Couture tool [6] claims to address such issues by combining techniques
including (i) fast prediction of 3D forms from cached shapes using sensitivity analysis
and generalized moving least squares, (ii) fast invisible re-meshing using positive mean
value coordinates to accommodate arbitrary revisions of the pattern boundary, and (iii)
stable and accurate cloth modeling using an isometric bending model, a modified St.
Venant- Kirchhoff membrane element, and progressive refinement.
PD outcomes also point to size of monitoring screen being problematic with a need
to balance distance to view full body comfortably and ease of interaction with UI
buttons. To help eliminate issues such as illustrated in figures 2 & 3, future research is
posited to explore using large screen portrait orientation (vs traditional landscape
orientation) monitors that enable real-time 3D auto stereoscopy12.
Wheelchair bound interviewees made clear their need for improved means to pur-
chase clothes without having to leave their homes. The perceived ease of use (PEOU)
and perceived usefulness (PU) toward enhancement and reduced effort of an accessi-
ble online apparel purchasing system were key factors when wheelchair-bound con-
sumers were questioned. This highlighted how a market opportunity is clearly evident
for people in wheelchairs to augment accessibility when purchasing for self or part-
ners, family, or friends.
Conflicts of Interest Statement. The authors have no conflict of interest to disclose.
Acknowledgements. The Virtual Dressing Room project is financed by the Danish

National Advanced Technology Foundation, Denmark.
The researchers want to acknowledge the host companies in UK (section 3.1); Lazy-
Lazy; and Bruuns Galleri, Aarhus.
References
1. Brandt, E.: Designing Exploratory Design Games. A Framework for Participation in Parti-
cipatory Design? In: Proceedings from Participatory Design Conference 2006, pp. 57–66.
ACM (2006)
12
http://www.tridelity.com
2. Buur, J., Bødker, S.: From Usability Lab to “Design Collaboratorium”: Reframing Usability
Practice. In: Proceedings of DISoo: Designing Interactive Systems: Processes, Practices,
Methods, & Techniques, pp. 297–307 (2000)
3. Buur, J., Jensen, V.M., Djajadiningrat, T.: Hands-only Scenarios and Video Action Walls;
Novel Methods for Tangible User Interaction Design. In: Proceedings of DIS 2004: Design-
ing Interactive Systems: Processes, Practices, Methods & Techniques, pp. 185–192 (2004)
4. Muller, J.M.: Participatory Design: The Third Space in HCI. In: Proceedings of the Partici-
patory Design Conference 2000, pp. 147–155 (2000)
5. Seifert, M.C., et al.: Demystification of Cognitive Insight: Opportunistic Assimilation and
the Prepared-Mind Persopective. In: Sternberg, R.J., Davidson, J.E. (eds.) The Nature of In-
sight. The MIT Press, Cambridge (1995)
6. Umetani, N., Kaufman, M.D., Igarashi, T., Grinspun, E.: Sensitive couture for interactive
garment modeling and editing. ACM Trans.Graph 30(4), Article 90 (2011)
Digital Menu Boards as Influencer for Healthy Eating
Anicia Peters1 and Brian Mennecke2
Iowa State University

1
1620 Howe Hall,
2
3313 Gerdin Building,
Ames IA 50011, USA
{anpeters,mennecke}@iastate.edu
Abstract. This paper reports on research designed to examine the role of

embedded videos in digital menu boards and it’s influence on consumer
decision-making. The study’s results showed that the embedded videos caught
the attention of decision makers and healthy food images have an effect on
choice; however, these variables do not interact. The results for the main effects
show that technology can aid decision-making during the information search
and purchase/choice stages.
Keywords: Digital signage, digital menu boards, decision-making, healthy

eating, information overload, video boards.
1 Introduction
A common challenge for advertisers is to capture the attention of consumers.

Nevertheless, this challenge is growing more daunting as consumers are presented
with an increasing number of advertisements from various sources and via numerous
media. For example, ads are presented to prospective customers on signs in public
spaces, on mobile phones, and on computers and a common result is that the intended
recipient of the message copes with this deluge by ignoring most of what they are
exposed to. This coping strategy is called “display blindness” [14] and an unfortunate
result is that the consumer may throw aware the good with the bad by ignoring
information that might be useful in helping him make better decisions [7].
This problem is most pronounced in public spaces and at retail venues where signs
are used both for advertisement and also for information dissemination. One solution
that has arisen in the last few decades is to use digital signage, which because of lower
costs and increasing capabilities, offers advertisers the ability to present information in
more dynamic and vivid ways with the goal of attracting the attention of consumers [3].
Digital signage used for advertising using consists of networked flat panel display
screens capable of displaying multimedia and high-speed Internet feeds [3].
For example, a number of fast-food retailers such as Wendy’s and McDonalds
have deployed dynamic digital menu boards [18] as is seen in figure 1 to attract
customer attention and promote their products and promotions. These displays often
prove to be quite effective. For example, Wendy’s reported an upsurge in sales due to
Digital Menu Boards as Influencer for Healthy Eating 429
the use of menu boards at the point of sale [18]. Of course, while fast-food venders
may have initially been motivated to use these technologies to increase sales or push
high-margin products, with the increasing interest in healthy eating and forthcoming
requirements for presenting nutritional information [15] it is likely that digital menus
will also provide potential benefits beyond increasing the bottom line. For example,
schools have begun to use digital signage in their lunchrooms and dining centers [8,
16] in an effort to promote healthy eating. Given this, we suggest that these
technologies could be useful in enticing consumers to make healthier eating choices
in point of sale settings. Yet, little has been done to examine whether and how
signage influences decision making at the point of sale; therefore, our focus is on
examining whether these menu boards can be used to influence healthy choices.
Fig. 1. A dynamic digital menu board at Wendy's
2 Research Question
Our research question focuses on whether the video in dynamic digital menu boards
can influence consumer decision in such a way that consumers will make healthier
food choices when confronted with video food ads of healthy food choices.
3 Literature Review
3.1 Information Overload

One of the characteristics of the modern world is the fact that technology enables
businesses, governments, and other people to “reach out and touch” us as we carry out
our daily activities. In some cases, we “subscribe” to services that “touch” us with
news feeds, promotions, or personal messaging, but as often as not we are also
presented with unsolicited messages that arrive via smartphones or through displays
we encounter in our environment. The cumulative effect of this information deluge is
that recipients of this messaging are often overloaded, which makes filtering through
the noise to hear the “signal” buried within the noise increasingly difficult and this
can also negatively affect decision making [1]. In other words, because consumer’s
have a limited working memory capacity and are often pressed for time, they are
likely to experience “information overload” [1, 13], which occurs “…when too much
430 A. Peters and B. Mennecke
information affects a person and the person is unable to recognize, understand or

handle this amount of information…” [7].
Consumers tend to develop coping strategies in reaction to excess stimuli. For
example, some of the coping strategies that are suggested in the literature are that
consumers process information less in-depth, they ignore less relevant information,
they delay or abandon reaching a decision, and they make less optimal decisions [7,
11]. Other studies have found that consumers often learn to ignore stimuli such as
signage and displays [14]. An important question is, what can break through to the
consumer? In the context of digital signage displays in public venues, one solution
that many advertisers are using is to imbed video, sound, movement, and other vivid
elements into displays with the goal of catching the involuntary selective attention of
consumers [10, 14]. Of course, an important empirical question yet to be addressed is
what factors in a display will be effective in overcoming display blindness and
garnering the attention of customers? Our research addresses this question in part by
examining one dimension of vividness, the movement of objects on the display.
3.2 Consumer Decision-Making
To examine the question of the role of vividness in overcoming display blindness, we

examine the role of movement in a context similar to the point of sale context of fast-
food restaurants. We first encountered digital signage used for purveying information
about food at such an establishment, and the venue is ideal for examining whether and
how the display characteristics influence attitudes and behaviors. Of course, the
decision about whether and what to eat is complex and will be influenced by more
than merely a sign in a restaurant. For example, consumers go through multiple stages
when deciding whether and what to purchase. One of the standard models of
consumer decision making is the Engel, Blackwell and Miniard (EBM) model, which
offers seven stages consisting of need recognition, search, pre-purchase alternative
evaluation, purchase, consumption, post consumption evaluation and divestment [6].
The EBM model distinguishes between individual differences, environmental
influences and psychological processes that influence consumer decision-making
through information processing, learning and attitude and behavior change.
While the EBM model is useful in a number of contexts, not all consumers go
through all the decision-making stages and decision rules. Specifically, stages are
used or skipped depending on the degree of complexity of the choice and the
consumer’s involvement in the process during each stage [6, 9, 19]. Additionally, a
number of other factors influence which stages are processed, how they are processed,
and how the processing influences outcomes. For example, problem size is important
and is determined based on the number of options to be evaluated within the available
time. Furthermore, time pressure causes consumers to adapt their decision strategies
according to the amount of time available for a decision [2]. Furthermore, information
format plays a significant role in decision making as shown by studies suggesting that
complex information structures on display causes increased cognitive load, thus
influencing the way that information is processed during each stage [1, 2].
The format of display features in particular can have an important influence on

how information is processed. Many menu boards are structured with items presented
in a list and therefore the consumer can view all of the information simultaneously,
which should have a positive impact on interpretation and decision-making [1].
Nevertheless, many menus are still cluttered with too much information, which
makes interpretation difficult [12]. One way that has been presumed to draw attention
to relevant content is to increase the vividness of contents on the display [10, 12].
Nevertheless, while movement, color, and similar vivid elements have been shown to
be influential in contexts such as web site ads, little has been done to identify how
these elements influence decision making in the context of menu boards.
3.3 Healthy Eating

Fast food restaurants have recently come under fire because of the association
between fast food consumption and obesity [4]. Nevertheless, many of the items sold
at fast food restaurants are quite healthy and if venders could encourage increased
consumption of these items not only would the health of their patrons improve, but so
too would the vender’s bottom line given that healthy items often have higher
margins.
Digital menu displays offer potential to encourage healthy eating. For example,
digital menus can be used to display nutritional information along with information
about the product. In fact, restaurants in the US are now required to display
nutritional information on their menus as required by the 2010 Health Care Reform
Act [15]. Nevertheless, displaying more information on an already overcrowded
menu board will undoubtedly increase the information load of consumers and might
result in exactly the opposite outcome of what is desired; that is, consumers might
become blind to this nutritional information. Understanding the role of information
content and presentation format will be important in helping venders present
nutritional content to consumers and potential in increasing healthy eating behaviors.
4 Framework and Research Model
We considered the basic five stages in consumer decision-making as presented by [5]

as: a) need recognition, b) search, c) alternatives evaluation, d) choice/purchase, and
e) outcomes. We also considered other factors such as the type of consumer,
situational factors, relevance of the ad, personal preferences, motivation, information
overload, and display layout conditions. We examined hypotheses associated only
with the information search stage, the alternatives evaluation stage, and the
purchase/choice stage.
To derive the hypotheses, it is useful to present them in the context of the five
basic consumer decision-making stages [5] as discussed above. We offer the research
model in Figure 2 to summarize our hypotheses as we discuss in detail below.
Fig. 2. Research model
Need Recognition Stage: We will not consider the need recognition stage in our
research because the scenario in which our subject is placed in our study is at the
Point-of-Purchase (POP) (i.e., the subject is presented with the video menu board
immediately before the decision is made). Future studies could examine the role that
post-consumption evaluation plays in the need recognition stage.
Search Stage: This stage is characterized by an external and internal search for
information. The internal search source is prior knowledge, whereas the external
search will be primarily limited to the video menu as this is typically installed at the
Point-of-Purchase (POP). The nature of the information search is influenced by goals,
time, financial resources, and sensory information. Information processing (i.e.
exposure, attention, comprehension, acceptance, and retention) is a very important
part of the search stage [6]. We expect that subjects will pay attention to the video ads
because it serves to declutter the displays and attract selective attention through its
vividness, colors and moving imagery. Because video ads are in line of sight at the
POP, consumers might be unable to screen out the video completely and by holding
the consumer’s attention, comprehension and acceptance will be greater. Memorable
video ads will also more likely be retained in memory and serve as a future internal
memory trigger when consumers recall information from previous exposure to the
ads. Our hypotheses for this stage are thus:
H1: A video in a dynamic digital menu board will draw more attention than an image
in a static menu board.
H1a: Moving imagery via video in a dynamic digital menu board will be more salient
than an image in a static menu board.
Evaluation of alternatives: According to previous studies [1, 2], variables such as

problem size, time pressure, attribute correlation, completeness of information,
information format, and comparable versus noncomparable choice all influence the
evaluation of alternatives. Because the video ad will be evaluated alongside static
text/graphics and internal memory, we propose the following hypotheses:
H2: The video in a dynamic digital menu board will lead to the evaluation of fewer
alternatives than an image in a static menu board.
H2a: The video in a dynamic digital menu board will reduce the consumer’s
perception of decision complexity more than an image in a static menu board.
H2b: The video in a dynamic digital menu board will lead to higher levels of
involvement than an image in a static menu board.
Choice/Purchase: A consumer may employ several decision-strategies or
combinations thereof depending on the situation or depending on the choice
complexity [1, 2]. Although choice complexity is influenced by a number of factors
such as the number of alternatives and attributes, the processing difficulty, the
uncertainty, the characteristics of the person, and the social context (e.g., family
decision-making versus individual choices) [1], the video ads will play a significant
role in influencing the final decision. Therefore, the following hypotheses are
proposed.
H3: The video in a dynamic digital menu board will become a reference heuristic for
decisions more than an image in a static menu board.
H3a: The availability bias of the video in a dynamic digital menu board will increase
the likelihood of the featured item being purchased more than an image in a static
menu board.
H3b: The persuasiveness of the video in a dynamic digital menu board will increase
the likelihood of the featured item being purchased more than an image in a static
menu board.
Outcomes/Post-purchase alternative evaluation: When a consumer evaluates the
product after consumption, this can influence future decision-making by associating
the video ad with a positive experience. The video will then, in turn, serve as positive
recommender for future purchase decisions. We did not test this stage.
5 Methods
We conducted a 2 x 2 factorial study with independent variables of health condition

(healthy, unhealthy) and visuals (video, static). We simulated a menu board consisting
of still images and video clips of two unhealthy items, two healthy items, and eight
additional food items displayed in plain text.
A total of 134 undergraduate students participated in the research, of which 40%

were male. The video clips consisted of moving images of selected food items (i.e.,
they were randomly shown either less healthful items such as chips and cookies or
more healthful items such as peanuts or raisins). The food items were selected and
classified by having a panel of subjects rate the items on appeal and perceived
healthfulness. Movement in the videos was created using the Ken Burns effect, which
involves panning across the object to show details and to create a sense of movement.
Subjects were given 20 seconds to view the menu choices (figure 3) with the video
clips and make a selection of a food product to eat. Following their selection, subjects
were asked to complete a survey.
Fig. 3. Menu board simulation during study
6 Results
ANOVAs were conducted on composite structures to test our hypotheses. Covariates

measuring habitual choices and time pressure were included in the analysis.
Information Search Stage (H1 and Ha): Main effects were found for both visuals
(p=.000) and healthy choices (p=.007), but no support for an interaction between
visual and healthy choices were found. The habitual choice covariate had an influence
on the effect size.
Hypothesis 1 (the video will catch the observer’s attention more than static images)
showed main effects for visuals (p=.008). Since the effect size for healthy condition
was low (R2 =.144), the power analysis (.491) showed the healthy condition would
have shown a main effect if there were more subjects. The habitual choice covariate
had an influence on the main effect. No support for interaction was found.
H1a (motion of video attracts attention more than static images) showed main
effects for visuals (p=.000) and healthy choices (p=.005). The habitual choice
covariate had an influence on the effect size. However, there was no support for the
interaction between visuals and healthy choices.
Alternatives Evaluation Stage (H2, 2a, 2b): For this stage, no main effects or an
interaction effect was encountered. The covariate habitual choice increased the effect
slightly to R2=.071. H2 was not supported. No main effects were determined for
either visuals (p=.341) or healthy choices (p=.418). No interaction effect was
observed. H2a (reduction of perception of decision complexity) showed no main

effects for either visuals (p=.581) or healthy choices (P=.940) and there was also no
interaction effect. H2b (high involvement level) showed a main effect for only visuals
(p=.007) and none for healthy choices (p=956). No interaction effects were
demonstrated.
Purchase/Choice Stage: This stage showed that consumer choice can be influenced
by video (main effect p=.020) and by healthy options (main effect p=.017).
Nevertheless, no interaction effects were found. None of the covariates had a
significant effect on the group hypothesis for the purchase stage. H3 (reference
heuristic) showed no main effects for visuals (p=.118) or healthy choice (p=.057). No
interaction effects were evident. H3a (availability bias) was not supported at all with
no main effects (visuals: p=.216 and healthy choice: p=.133) and no interaction
effects. Nevertheless, two covariates, “time was too short” and “habitual choice,”
influenced the effect size (R2=.173). A power analysis revealed a larger sample size
might increase the significance of the main effect healthy choice. H3b
(persuasiveness of vividness effect) showed main effects for visuals (p=.007) and
healthy choices (p=.014), but no interaction effects was demonstrated.
Full model ANOVA: The full model showed main effects for both visuals (p=.003)
and healthy choices (p=.014), but again no interaction effect was observed. Habitual
choices showed a significant influence on the effect size (R2=.167).
7 Discussion
For the overall model, our ANOVA results showed main effects for visuals (p=.003)
and healthy choices (p=.014) with a significant covariate for habitual choices, but
failed to show an interaction effect. The smaller samples size and the low effect sizes
resulted in no significant interaction effect; therefore, a larger sample size might show
results for the interaction.
Another explanation for the lack of interaction is that subjects did not consider the
health aspects of the snacks, as their choices did not correspond with their self-
reported health-consciousness. Nevertheless, their general perceptions of whether
they chose a healthy item agreed with their self-report health rating of the item.
Another factor influencing the results relates to the fact that subjects mostly made
habitual choices, i.e. they selected food items that they usually eat (64%) and usually
those that they usually like to eat (83%). As much as this indicates a familiarity bias
in the study, it also indicates a habitual decision-making style, which means that they
did not go through pre-defined steps in the decision process [9, 19]. By using this
decision-making style, subjects seemed to have used prior knowledge to guide their
decisions and blocked new information or unfamiliar products [9, 17]. This situation
was further fueled by the fact that subjects’ involvement levels were very low as they
participated in the study primarily because they would earn research credits. The
habitual decision-making style is also evident in the alternatives evaluation stage
hypotheses where no main effects were found for either visuals or healthy eating.
The good news is that the video and health aspects of the food had significant
effects during the information search stage and the purchase/choice stage, which
indicates that this technology can be used to influence decision-making.
8 Conclusion
Although we found no evidence for an interaction effect between video use in digital
menu boards and healthy eating, our results demonstrate main effects for video and
healthy eating choices. We thus suggest that the videos can be used to aid decision-
making and influence healthy food choices.
References
1. Bettman, J., Johnson, E., Payne, J.: Consumer decision making. In: Robertson, T.S.,
Kassarjian, H.H. (eds.) Handbook of Consumer Behavior, pp. 50–84. Prentice Hall Inc.,
Englewood Cliffs (1991)
2. Bettman, J., Luce, M., Payne, J.: Constructive consumer choice processes. Journal of
Consumer Research 25, 187–217 (1998)
3. Burke, R.: Behavioral effects of digital signage. Journal of Advertising Research 49(2),
180–186 (2009)
4. Drewnowski, A., Darmon, N.: The economics of obesity: dietary energy density and
energy cost. American Journal of Clinical Nutrition 82(1), 2655–2735 (2005)
5. Engel, J., Blackwell, R., Kollat, D.: Consumer Behavior, 3rd edn. Dryden, Hinsdale (1978)
6. Engel, J., Blackwell, R., Miniard, P.: Consumer Behavior. The Dryden Press, Harcourt
Brace College Publishers (1995)
7. Eppler, M., Mengis, J.: The concept of information overload: A review of literature from
Organization Science, Accounting, Marketing, MIS, and Related Disciplines. The
Information Society, An International Journal 20(5), 1–20 (2004)
8. Glasgow City Council serves up healthy eating message to city’s children using ONELAN
digital signage, http://www.mxdigitalsystems.com/documents/
Onelan_Glasgow_Healthy_Eat_case_study.pdf
9. Hoyer, W.: An examination of consumer decision making for a common repeat purchase
product. Journal of Consumer Research 11, 822–829 (1984)
10. Huang, E.M., Koster, A., Borchers, J.: Overcoming assumptions and uncovering practices:
When does the public really look at public displays? In: Indulska, J., Patterson, D.J.,
Rodden, T., Ott, M. (eds.) PERVASIVE 2008. LNCS, vol. 5013, pp. 228–243. Springer,
Heidelberg (2008)
11. Jacoby, J., Speller, D., Kohn, C.: brand choice behavior as a function of information load.
Journal of Marketing Research 11, 63–69 (1974)
12. Jansson, C., Marlow, N., Bristow, M.: The influence of color on visual search times in
cluttered environments. Journal of Marketing Communications 10, 183–193 (2004)
13. Malhotra, N.: Reflections on the information overload paradigm in consumer decision
making. Journal of Consumer Research 10, 436–440 (1984)
14. Müller, J., Wilmsmann, D., Exeler, J., Buzeck, M., Schmidt, A., Jay, T., Krüger, A.: Display
blindness: The effect of expectations on attention towards digital signage. In: Tokuda, H.,
Beigl, M., Friday, A., Brush, A.J.B., Tobe, Y. (eds.) PERVASIVE 2009. LNCS, vol. 5538,
15. Rosenbloom, S.: Calorie Data to be posted at most chains. The New York Times (March
23, 2010),
http://www.nytimes.com/2010/03/24/business/24menu.html
16. School District of Philadelphia Uses Digital Signage to Promote Get Healthy Philly,
http://www.digitalsignageconnection.com/3School-District-of-
Philadelphia-Uses-Digital-Signage-to-Promote-Get-Healthy-
Philly86
17. Stijn, M., Van Osselaer, J., Alba, J.: Consumer learning and brand equity. Journal of
Consumer Research 27, 1–16 (2000)
18. The Buzz: Installation spotlight: POP appeal,
http://svconline.com/digitalsignage/features/
digital_signage_wendys_0603/
19. Wright, P.: Consumer choice strategies: Simplifying vs Optimizing. Journal of Marketing
Research XII, 60–67 (1975)
Reverse Business Innovations – The Impact of ERP
Software Upgrades on Organizations
Stefanie Rauff1 and Andreas Hufgard2

1
University of Wuerzburg
[email protected]
2
University of applied sciences Aschaffenburg
[email protected]
Abstract. Vendors of enterprise software continue to evolve their products and

offer user organizations the latest functionality in new releases and upgrades.
But this means companies are faced with the challenge of selecting features that
add value, and trying to assess how implementation would impact them. Not
every new function impacts all employees in the same way. Depending on the
specific organizational structure, implementation will affect a varying number
of users. And depending on a company’s system landscape, implementation of
new functions might require a range of IT activities and affect many processes.
The authors have developed a concept that uses an organization’s system
usage data to deliver decision support. Various aspects of the upgrade can be
assessed with the aid of software tools (analyses and evaluations of new func-
tionality) to determine relevancy, benefits, profitability, strategic goals and im-
pact. In this article we will present a portion of this concept: A model that can
assess how new functionality would impact a business, if implemented.
Keywords: SAP ERP, upgrade, new release functions, decision support, usage
analysis.
1 Research Topic
An enterprise system (ES) is defined as a standardized software package whose aim is

to integrate the whole of an organization’s data [1]. ES development can be described
in two waves [2]: The first consists of procurement, configuration and implementa-
tion. The second, also commonly referred to as the post-implementation phase, focus-
es on achieving continuous improvement and driving value. Questions central to this
phase include how to gain greater benefit from the investment, how to manage and
enhance the ES and how the ES will impact the organization.
1.1 Current Research in the Post-implementation Phase

The first wave has been widely researched [3-8]. A number of surveys and case stu-
dies have been done on the implementation wave, with respect to strategic options,
Reverse Business Innovations – The Impact of ERP Software Upgrades on Organizations 439
failure avoidance, identification of alignment issues and improved implementation

methodologies [9].
Recently, research has moved beyond implementation to the second wave [2]. Re-
search topics include: the impact of ES implementation on financial performance [10]
and [11]; standardization and trust relations [12]; legitimization of certain practices in
the organization [13]; continuous changes in business processes [14]; potential impact
of enterprise system implementation and its effects on the organization [15]; user
acceptance and expectations [16]; the impact of ES implementation on job satisfaction
[17] and on the organization [18], and how employees’ social network structures af-
fect post-implementation success of enterprise systems [19]. In short, they deal main-
ly with various aspects of implementation and how these affect organizations.
This paper also concerns the second wave, but is not devoted to the impacts of
implementation. Its focus is on the variety of new functions regularly provided by
enterprise software vendors in upgrades.
In this paper we will examine specific types of innovations that are characterized
as follows:
• They are functional innovations

• and are included in an upgrade for enterprise software.
To set this specific type of innovation apart, we invented the term “upgrade
innovation.”
In this paper, the research question is: How can we help organizations evaluate the
impact of upgrade innovations?
1.2 Research Methodology
The model was developed in the following steps:
1. The current situation and problem were analyzed (section 1)

2. A prototype was created (sections 2 and 3)
3. The model was validated (section 4)
An analysis of relevant literature revealed that next to no theory exists on upgrades.

The subject of untapped potential in business upgrades has been neglected entirely.
We created a prototype and identified assessment criteria (relevancy, cost-benefit
analysis, impact and strategic goals) according to which upgrade innovations were to
be assessed. This paper will present the issue of impact assessment.
2 Description of Method
We identified a lack of research on evaluating the relevancy of new functions offered

in upgrades, and on system-supported impact assessments prior to an upgrade.
A vast number of organizations that use enterprise systems are confronted with the
same problem. This is why it makes sense to build an analysis tool for this purpose
440 S. Rauff and A. Hufgard
rather than let each company review all innovations by itself. The authors of this pa-
per are currently collaborating with SAP to create the technology and content neces-
sary for this tool.
2.1 Reverse Business Innovations
Our goal was to develop a model that would help organizations assess the impact of
upgrade innovations relevant for them. To this end, we looked for criteria that would
enable objective inference of an upgrade innovation’s impact. In other words, we posed
the question: Under which conditions would this functionality have an effect on the
company? We developed an approach we dubbed “Reverse Business Innovations.”
Figure 1 shows how the system-based usage analysis and the rule-based impact
assessment are interrelated.
Fig. 1. RBI model
In preparation for the analysis, we examined and classified the upgrade innovations
to gain insight into the new functionality offered. For each innovation, we identified
criteria that enabled us to infer the impact of the innovation, i.e. what technical re-
quirements an organization would have to fulfill to enable us to conclude how the
innovation would impact it. We then filed away this information in the “innovation
catalog,” a reference for determining, for example, which indicators need to be mined
from an organization’s system usage data.
When the analysis is conducted for an organization, the usage indicators are
extracted from the company’s live ERP system. They help determine impact
based on the innovation catalog. The results enable us to assess the impact an upgrade
innovation has on an organization and so, immediately identify quick wins.
2.2 Impact Assessment

In their maintenance model, NG et al. indicate that a gap analysis must be conducted
between the new upgrade and the existing software version in order to identify the
impact on an organization, and finally, to create a budget, a schedule and a plan for
employees [6].
Organizations are often short of resources [20], so an impact assessment can help
them identify upgrade innovations that require only very few resources.
We defined three distinct aspects of the term “impact”:
• Effect on users: The number of employees who will use the new functionality,
• Effect on IT: This can include anything from a simple change request to new
hardware or software,
• Effect on processes: This considers the range of processes for the organization
Employees
An upgrade always entails change, so it is essential to incorporate routine change
management activities, such as risk management assessments, training, and know-
ledge transfer [21] and [3]. In a qualitative study, Zhao found a correlation between
engaging employees in the upgrade and offering them training, and an upgrade’s suc-
cess [22]. So we can conclude that it is essential to consider the impact on the em-
ployees of an organization when contemplating an upgrade.
This means it is important to conduct an assessment of the number of people im-
pacted by the upgrade innovation. We defined three categories for impact on users:
• Few users affected
• (Individual) user communities affected
• Many users affected
IT
According to Brose, the extent of the risk posed by an innovation can be determined
by the following two factors [23]:
1. The degree of technical difficulty, because it is a measure of the ability to solve a
problem that occurs, especially during the required development time.
2. The technical complexity, because it is a measure of the diversity and number of
individual activities that must be coordinated.
Since upgrade innovations are not products or services the company itself has devel-
oped, but software tested and supplied by a vendor of a business-related information
system, we can disregard the first factor when assessing upgrade innovations. But the
second factor involves the number of technical measures that must be implemented
and how complex these are, and so, must be included in the impact assessment.
We defined three categories for technical measures:
1. A single change request
2. The IT resources required (developers, project managers) for a change project
3. New hardware or software required
Normally this value is set by the manufacturer of enterprise software, because it is not
dependent on the specifics of any one organization.
Processes
Companies are predominantly organized by processes. And these might range across
several user departments or even across companies. Organization by processes also
enables problems to be solved across functions [24].
Therefore, when examining impact, it is important to look at the effects of upgrade
innovations on the specific process in question. We defined three categories for im-
pact on processes:
1. Affecting isolated processes
2. Affecting several processes
3. Affecting a large number of processes
Normally this value is set by the manufacturer of enterprise software, because it is not
dependent on the specifics of any one organization.
3 Validation
We will conclude our paper with a validation of the model based on a concrete up-
grade innovation from SAP CRM.
We created a prototype in a rule-based expert system specifically for the new up-
grade functionality in SAP CRM added between Release 5.0 and 7.0 EHP 2. We then
validated it on the system of a manufacturing company that operates worldwide. The
CRM system examined is strategically crucial for this user organization. Approx-
imately 4,000 users from Europe, Asia and North America actively work within it and
it links three ERP systems.
The company currently uses Release SAP CRM 7.0 EHP 1. They last upgraded be-
cause of system consolidation with a subsidiary whose CRM system used EHP 1.
The pilot organization agreed to the analysis because decision-makers thought
some of the necessary SAP CRM areas may not have been activated. In addition, they
were interested in obtaining new functionality available in Release 7.0 EHP 2 without
much added expense.
Usage analysis results showed that during the analysis period, 232 campaigns were
used by 23 employees. Leads, trade promotions and marketing plans were not used at
all. Trade promotions are used to depict marketing activities in collaboration with
partners. Marketing plans track the activities in Marketing at a higher strategic level
and for a longer period than trade promotions and campaigns.
The marketing calendar is an upgrade innovation that offers users a graphic over-
view of all marketing projects according to their duration. Projects can also be created
directly on the graphic.
Under impact it shows that for IT, the marketing calendar has very little impact
(see Figure 2), because only one function (i.e. the business function) has to be
activated. The users affected are the employees that create the marketing projects
(campaigns, trade promotions and marketing plans). They total 23. The marketing
calendar’s impact on processes has been rated medium, because processes concerning
trade promotions and marketing plans might also be affected.
Fig. 2. Impact assessment for the marketing calendar
Overall, impact assessment reveals that several impact areas are affected: 23 users
and a medium impact on processes.
4 Discussion of Results
The newly developed model delivers organization-wide decision support that deter-
mines the impact of an upgrade innovation – automatically and according to rules.
In sum, the new model offers a number of benefits to user organizations, and also
to manufacturers of enterprise software.
User organizations benefit most by the use of automated assessment when compar-
ing their as-is situation to their (potential) situation following an upgrade:
• They save time otherwise spent gathering and assessing information for an up-
grade.
• They require less external consulting and so, reduce costs.
• Consultants’ advice/recommendations are replaced by computer-based activities
and so, reducing the number of errors made.
• Recommendations are easier to understand because detailed data and explanations
are included in the results.
• The computer-based analysis is more thorough than an analysis conducted by hu-
man consultants. It also recommends upgrade innovations from earlier releases that
may have become relevant under current company conditions.
• The tool can be used on local systems and so, is always available. It can also be
employed in several user organizations simultaneously.
• The expert system delivers much faster results than human experts can.
• The tool contributes to long-term transparency, because it is very simple to docu-
ment the results in detail.
The consequence of all above-mentioned points is that user organizations will begin
upgrading sooner to higher releases and the upgrade innovations will be put to use
and their potential tapped sooner. In some cases, the upgrade innovation can replace a
customer-specific add-on with a standard function. All of these consequences would
result in far lower costs for maintaining the enterprise software.
What is more, vendors of enterprise software would also benefit from the availabil-
ity of this kind of tool:
• User organizations would no longer rely as much on the expensive maintenance of

old releases if they had an incentive to upgrade.
• User organizations would require less consulting for the same or better service, and
so, improve their cost structure (copying software costs nothing).
• Software vendors would enjoy improved customer loyalty because organizations
would be able to benefit more from their enterprise software (and changing soft-
ware would be more expensive and time-consuming).
In light of the prevalence of enterprise software, the wealth of practical applications

for the model presented in this paper is clear. This tool could greatly reduce costs for
both user organizations and manufacturers, and could improve the integration of up-
grade innovations in organizations.
This same methodology can also be applied toward new products not included in
an upgrade, but that require additional licensing costs.
In conclusion, it is essential to address the issue of how the analysis results can ac-
celerate and prepare user organizations for the impending implementation and appli-
cation of any given upgrade innovation.
References
1. Davenport, T.H.: Putting the enterprise into the enterprise system. Harvard Business Re-
view, 121–131 (July/August 1998)
2. Kræmmergaard, P., Koch, C.: Managing ERP after going-live. In: Proceedings of the 9th
International Conference European Operations Management Association, Copenhagen
(2002)
3. Esteves, J., Pastor, J.: Towards the unification of critical success factors for ERP imple-
mentations. In: Proceedings of the 10th Annual BIT Conference (2000)
4. Dong, L., Neufeld, D., Higgins, C.: The Iceberg on the Sea: What do you see. In: Proceed-
ings of the 8th Americas Conference on Information Systems, Paper 124, pp. 857–864
(2002)
5. Jacobs, F., Bendoly, E.: Enterprise Resource Planning: Development and Directions for
Operations Management Research. European Journal of Operational Research 2(146),
233–240 (2003)
6. Ng, C., Gable, G., Chan, T.: An ERP Maintenance Model. In: Proceedings of the 36th
Hawaii International Conference on System Sciences (2003)
7. Moon, Y.: Enterprise Resource Planning (ERP): a review of literature. International Jour-
nal of Management and Enterprise Development 4(3) (2007)
8. Maedche, A., Mueller, B.: Enterprise Systems - A Research Agenda (2012),

http://ssrn.com/abstract=2033145,
http://dx.doi.org/10.2139/ssrn.2033145
9. Al-Mashari, M.: Enterprise resource planning (ERP) systems: a research agenda. Industrial
Management & Data Systems 1(103), 22–27 (2003)
10. Poston, R., Grabski, S.: Financial impacts of enterprise resource planning implementa-
tions. International Journal of Accounting Information Systems 2, 271–294 (2001)
11. Hunton, J., Lippincott, B., Reck, J.: Enterprise resource planning systems: comparing firm
performance of adopters and nonadopters. International Journal of Accounting Information
Systems 4, 165–184 (2003)
12. Locke, J., Lowe, A.: Standardisation and trust relations in ERP implementation: reflections
from a case investigation. In: 28th Annual Congress of the European Accounting Associa-
tion, Goeteborg, Sweden, May 18-20 (2005)
13. Dillard, J., Ruchala, L., Yuthas, K.: Enterprise resource planning systems: a physical ma-
nifestation of administrative evil. International Journal of Accounting Information
Systems 6, 107–127 (2005)
14. Davenport, T.H., Brooks, J.D.: Enterprise systems and the supply chain. Journal of Enter-
prise Information Management, 17(1), 8–19 (2004)
15. Rikhardsson, P.: Kræmmergaard: Identifying the impacts of enterprise system implementa-
tion and use. International Journal of Accounting Information Systems 7(1), 36–49 (2006)
16. Liang, H., Saraf, N., Hu, Q., Yajiong, X.: Assimilation of Enterprise Systems: The Effect
of institutional pressures and the mediating role of top management. MIS Quarterly 31(1),
59–87 (2007)
17. Morris, M.G., Venkatesh, V.: Job Characteristics and Job Satisfaction: Understanding the
role of ERP Implementation. MIS Quarterly 34(1), 143–161 (2010)
‐
18. Volkoff, O., Strong, D.M., Elmes, M.B.: Understanding enterprise systems enabled inte-
gration. European Journal of Information Systems 14(2), 110–120 (2005)
19. Sasidharan, S., Santhanam, R., Brass, D.J.: The Effects of Social Network Structure on
Enterprise Systems Success: A Longitudinal Multilevel Analysis. Information Systems
Research, 1–21 (2011)
20. Cooper, R.: Top oder Flop in der Produktentwicklung. 2. Aufl. Wiley-VCH Verlag, Wein-
heim (2010)
21. Hernández, J., et al.: The SAP R/3 handbook. 3. Aufl. McGraw-Hill, New York (2006)
22. Zhao, F.: ERP Upgrade Report (2007)
23. Brose, P.: Planung, Bewertung und Kontrolle technologischer Innovationen. Erich Schmidt
Verlag, Berlin (1982)
24. Krüger, W.: Grundfragen der Bildung und Integration organisatorischer Subsysteme. In:
Franz, X., et al. (eds.) Allgemeine Betriebswirtschaftslehre Bd.2: Führung. 8. neubearbei-
tete und erweiterte Aufl., Lucius & Lucius, Stuttgart, pp. 127–157 (2001)
Repository-Based Implementation of Information
Pyramid: A Study Based on an ERP Case Study
Hans-Jürgen Scheruhn1, Daniel Ackermann1, Roman Braun1, and Ulrich Förster2

1
Business Informatics, Media Informatics, Hochschule Harz, Friedrichstr. 57-59,
38855 Wernigerode, Germany
{dackermann,hscheruhn,u27077}@hs-harz.de
2
Atos Werner-von-Siemens-Platz 1, 30880 Laatzen, Deutschland
[email protected]
Abstract. ERP case studies have an important impact on the transfer of know-
ledge between software vendor, lecturer and user. This paper describes to which
extent ERP case studies can be used to demonstrate a repository based inte-
grated approach for modeling and implementing the entire information pyramid
in the context of the Business Process Management life cycle. The study is
based on the Global Bike Inc. enterprise model repository created with ARIS
Business Designer for SAP by Software AG. The information models have
been exemplarily synchronized with SAP Solution Manager repository and im-
plemented with SAP ERP respectively with SAP Netweaver BI.
Keywords: ERP case study, ARIS Business Designer, Business Process Man-
agement life cycle, SAP Solution Manager, SAP Netweaver BI, Sustainabilty
Balanced Scorecard.
1 Problem Identification
ERP case studies are used for the transfer of knowledge in both the academic and the
non-academic sector [17]. The authors state a lack of integration of strategic, tactical
and operational business information models (entire Information Pyramid) as well as
a lack of integration of its implementation with corresponding IT-Systems (entire
business process life cycle) based on ERP case studies [21].
Scheruhn et al. [24] compared three different ERP case studies on the as-is status
with each other regarding information model structures, information model types and
information object types of the corresponding enterprise models [5; 30]. The compari-
son based on the GBI enterprise information model repository resulted in a common
ERP case study framework which was worked out structurally in an evaluation matrix
on the operational level.
A study about the possible enterprise information model structure of an integrated
model based implementation of ERP case studies (to-be status / extended by future
business objectives of tactical and strategic level) with the intention to create a common
framework as repository for structured transfer of knowledge and for support of model
based implementation in both the academic and the non-academic sector is yet missing.
Repository-Based Implementation of Information Pyramid 447
2 Objective of a Solution
ERP case studies support students learning the handling of different functional areas
in a practical way [29]. Several ERP case studies exist, for example Global Bike Inc.
[32], SSB Inc. [8], SAP TERP10 exercises at IDES AG [29], SAP Business ByDe-
sign´s ALMIKA Holding [14; 15] and OPM´s IDES AG [23] which provide an intro-
duction to different ERP systems of SAP.
This paper aims to investigate the information model based teaching of implement-
ing ERP case studies (to-be status) in approach to further benefits. Consecutively
existing ERP case studies (as-is status) are to be analyzed by students of business
administration and/or business informatics as a basis for possible improvements (to-
be status) and its implementation in the context of the business process life cycle [23].
The paper identifies all necessary information models based on an extended evalua-
tion matrix created by the authors which can be used as a framework for further case
study enterprise model or company enterprise model comparisons (e.g. ALMIKA Hold-
ing or VW Financial Service [31]). The integration of aspects of business and informa-
tion technologies, of strategic, tactical and operational points of views as well as the
integration of “actors”, “business processes, business services and information flow” up
to its implementation with ERP and BI systems in the context of the information pyra-
mid in one single enterprise model is to be demonstrated [25].
3 Design and Development
3.1 Constructs
Software Reference models like the SAP Solution Manager Business Blueprint Repo-
sitory [13] and/or Industry Reference models like Handels-H [2] provide support for
construction processes of an individual enterprise model [5]. The enterprise models
developed and evaluated in this paper are considered as integrated software-industry
reference models which – in this case - depict typical company processes of the engi-
neering industry based on software, e.g. SAP ERP. The IPR (Indus-
try.PerformanceREADY) model is a good example [27] for such an integrated
reference model approach which deals very successful in the non-academic sector.
Compared to the GBI (to-be) enterprise model it covers most of the information mod-
els and objects of the evaluation matrix (see 5.) and additionally is connected with
best practice KPI provided by APQC[1].
According to Scheruhn (Fig.1), the Business Process Management (BPM) life
cycle encompasses five phases [23]. The main focus in this paper will be the integra-
tion of phases 2 and 3 whereas the previous paper [24] investigated phases 2 and 4.
Phases 2-5 are supported (but not fully integrated) by the SAP Solution Manager and
SAP ERP. SAP BI covers all phases with data seen as important constructs of
processes.
448 H.-J. Scheruhn et al.
Management
Portal
1.Strategic
planning
Enterprise
Controller 5. Process- 2. Process-
Architect
Portal controlling Design
Portal
4. Process- 3.Implemen-
execution tation
Customer / Integration
Employee Portal Portal
Fig. 1. Process life cycle with five phases by Scheruhn and Information Pyramid
3.2 Information Pyramid

The information pyramid is divided into three levels (Fig.1). The tactical level con-
nects the strategic and the operational level [9]. The considered ERP case studies
mainly cover the operational level in the context of the horizontal and vertical integra-
tion. In this study the tactical and strategic level on top also need to be considered.
For implementing the corresponding tactical and strategic information models the
authors additionally evaluated SAP Netweaver BI.
3.3 Four IT Integration Layers

In this research business processes are regarded as sub models of a model system,
which includes: Subject-oriented modeling [26], Process-oriented modeling, Service-
oriented modeling [3], Enterprise data modeling [6].
This leads to four horizontal IT integration layers (1.Presentation, 2.Processes,
3.Functions and 4.Data) of the enterprise model (see Fig. 2) describing an IT system
like SAP ERP or SAP Netweaver BI in its interaction with a specific enterprise. A
vertical decomposition of each of the four IT integration layers leads to a model
hierarchy of four additional levels shown in figure 2. These hierarchical levels can
be mapped with the four process levels of the SCOR reference model [28] as well
as partly with the three levels of the SAP Solution Manager Business Blueprint
Architecture [13].
3.4 Methods
First a hypothesis is formulated which has to be reviewed in the course of this investi-
gation [12] based on the implementation of the entire information pyramid exempli-
fied with SAP ERP and Netweaver BI regarding the corresponding GBI enterprise
model.
Hypothesis: The available GBI ARIS to-be enterprise model can be fully integrated
into SAP ERP and SAP Netweaver BI including all necessary information models to
form a common information model framework repository for structured transfer of
knowledge and for support of model based implementation in both the academic and
the non-academic sector.
Fig. 2. GBI enterprise model for implementing SAP NW BI and for implementing SAP ERP [25]
Thus, it has to be proven to which extent SAP ERP and SAP Netweaver BI are
able to interchange information model structures, information model and object types
with ARIS Business Designer (BD) respectively SAP Solution Manager. These re-
quirements will be worked out and compared in an evaluation matrix based on the
following assumptions:
• SAP ERP and SAP NW BI applied together cover the entire Information pyramid
and the entire Process life cycle
• SAP NW BPM is not considered because of focus on Core application without
automization of Business Processes spanning different applications
• Single information models (phase two of Process life cycle) of the GBI enterprise
model repository are provided completely but separately by SAP ERP and SAP
NW BI
• Single information models of the GBI enterprise model repository based on ARIS
Business Designer (BD) are entirely integrated in one enterprise model
• Information models of GBI based on ARIS BD and based on SAP ERP respective-
ly SAP NW BI are equal but not directly integrated
• SAP Solution Manager serves as an interface between the information models of
ARIS BD and SAP ERP respectively SAP NW BI
Area MM MM MM PP SD HCM
Purchase Order Incoming Invoice Goods Received Production Order Customer Order Business Event
Business Object BUS2012 BUS2081 MKPF BUS2005 BUS2032 PDOTYPE_E
Success Objectives Reduce Purchase Reduce Costs for Decrease Costs for Decrease Production Increase Turnover Reduce Costs fo
Perspective Order Costs Payment Target Raw Material Costs
Overdraft
KPI Purchase Oder Costs Overdraft Costs as-is/ Costs of Raw Material Production Costs as-is / Turnover Growth as-is / Training Costs as
as-is/to-be to-be as-is/to-be to-be to-be be
Target Ratio 5 % less costs (than 5 % less costs (than 5 % less costs (than 5 % less costs (than 10 % more turnover 5 % less costs (t
Best Practice) Best Practice) Best Practice) Best Practice) (than previous year) Practice)
Customer Objectives Increase Internal Increase Internal Increase Internal Increase Internal Increase Customer Increase Internal
Perspective Delivery Reliability Delivery Reliability Delivery Reliability Delivery Reliability (PP) Satisfaction Reliability (HCM)
(MM) (MM) (MM / Warehouse)
KPI Delivery Reliability as- Delivery Reliability as- Delivery Reliability Delivery Reliability as- Customer Satisfaction Delivery Reliabilit
is/to-be is/to-be (Warehouse) as-is/to-be is/to-be as-is / to-be as-is/to-be
Process Objectives Improve Purchase Improve Incoming Improve Goods Receip t Improve Production Improve Customer Optimize Busines
Perspective Order Process Invoice Process Process Order Process Order Process Process
KPI Purchase Order Incoming Invoice Goods Received Production Order Customer Order Event set-up cyc
Process cycle time Process cycle time Process cycle time Process cycle time Process cycle time
KPI QM Code Purchase QM Code Incoming QM Code Goods QM Code Production QM Code Customer QM Code Busine
Order Invoice Receipt Order Order
Potential Objectives Expand MM Workflow Expand MM Workflow Expand MM Workflow Expand PP Workflow Expand SD Workflow Expand HCM WF
Perspective
Objectives Increase MM Skills Increase MM Skills Increase MM Skills Increase PP Skills Increase SD Skills Increase HCM Sk
KPI Number of Workflows Number of Workflows Number of Workflows Number of Workflows Number of Workflows Number of Workf
as-is/to-be as-is/to-be as-is/to-be as-is / to-be as-is / to-be is / to-be
Employee Skill Sets as- Employee Skill Sets as- Employee Skill Sets as- Employee Skill Sets as- Employee Skill Sets as- Employee Skill S
KPI is / to-be is / to-be is / to-be is / to-be is / to-be is / to-be
Fig. 3. Area Department Balanced Scorecards and corresponding SAP Business Objects
3.5 Instantiation
The GBI (Global Bike Inc) model company is situated in the manufacturing industry. It
is a multinational enterprise producing and distributing bicycles. GBI is an international
group with subsidiaries in Dallas, Miami and San Diego in the United States of America
and Heidelberg and Hamburg in Germany. The GBI enterprise model and its implemen-
tation have been taught at the Hochschule Harz in a degree course of Business Informat-
ics in 2012/2013 and its results have been documented by the students in a weblog [11],
which contents can be found in Chapter 4. Relevant SAP ERP models used in this en-
terprise model are Materials Management (MM), Production Planning (PP), Sales and
Distribution (SD), Human Capital Management (HCM) and Financial Accounting (FI).
GBI 2.1 case study is hosted by SAP University Alliance [18]. SAP Solution Manager
7.1 as well as SAP ERP ECC 6.0 and SAP NW BI 7.0 (7.3 in progress) are hosted by
SAP UCC Magdeburg [19]. As one of the most recognized performance measurement
frameworks [20], the balanced scorecard is implemented on the strategic level based on
the establishment of cause and effect chains between strategic objectives respectively
key performance indicators. To improve the enterprise performance in all dimensions of
sustainability (economic, ecologic and social) the pillars of sustainability are integrated
in the successful realization of the enterprise strategy [4]. All department Balanced Sco-
recards of the GBI repository are connected via internal customer relationships [11] and
organized as cost centers (except sales) pursuing department IT-strategies (expand im-
plementation of workflows). To achieve a successful realization of the department ba-
lanced scorecards (economic pillar), business rules are implemented in terms of Quality
Management codes [10]. The Compliance of these business rules and a constant
comparison with best practices can be used as a key performance indicator in internal
processes. This is done by a 100% integration of the strategic (top of fig. 2) with the
operational models (bottom of fig. 2).
4 Demonstration
This section presents results of the first demonstration of above-mentioned ideas to
implement an enterprise model repository based on the GBI model company. The
corresponding blog “Online Process Management at GBI” has been posted in the SAP
Community Network and currently consists of 4 (5 and 6 in progress) parts, which
address following aspects of the enterprise model:
Online Process Management at GBI GBI Business Process Management Life Cycle, GBI Enterprise Model, GBI Sustainability
/ Part 1 [11] Balanced Scorecard
Online Process Management at GBI Harz Roadmap at SAP Solution Manager, GBI Business Blueprint goes ARIS
/ Part 2
Online Process Management at GBI Business Workflow and NW BPM as a major step towards process automation at GBI,
/ Part 3 How the Workflow works, Workflow Builder challenge
Online Process Management at GBI WebDynpro goes iPhone, Responsive Frameworks at GBI
/ Part 4
Online Process Management at GBI Providing and consuming WebServices at GBI
/ Part 5 (in progress)
Online Process Management at GBI BI at GBI
/ Part 6 (in progress)
Beginning with the WebDypro Context Diagram for future business the presenta-
tion design is capable of reacting to trends like mobile and ubiquitous computing.
For WebDynpro there are possibilities to design themes with stand-alone software,
for instance with an Eclipse-Plugin (ThemeEditor32). On small high resolution dis-
plays information presented with "Portal Theme Editor" themes can result in a poor
readability because of pixelation and small font sizes. It should be possible to imple-
ment a basic design reacting to different browser sizes.
This could be reached by taking advantage of technologies like responsive frame-
works (e.g."Twitter Bootstrap" [33]), which is based on CSS3, Html5 and Javascript.
At the present those frameworks are capable of "reacting" to user devices from smart-
phones up to large desktop by delivering the correct font size and layout. This respon-
siveness is an approach of web design providing an optimal viewing experience.
Like in Wroblewski [34] and Marcotte [35] described, the responsive design me-
thod scales the html content for a wide range of displays in a very ergonomic way.
The integration of basic responsive technologies in WebDynpro could lead to even
more user acceptance in terms of readability and overview. As a general example
Figure 4 shows the same content on 3 different devices.
Fig. 4. Resp
ponsive layout presented on 3 different displays
5 Evaluation Matrrix
An examination of the Enteerprise Meta-Model (EMM) by Gudas [5] shows 8 differrent

core objects (Actor, Processs, Function, Information, Business Rule, Event, Objecttive
and Material Flow), of wh hich SAP Solution Manager is able to implement the ffirst
four, one per IT integration
n layer(fig.2). Thus, the Solution Manager offers a permmea-
bility of 50 % of object typ
pes considering the Enterprise Meta-Model by Gudas. T The
following Evaluation Matriix extends the EMM with reference to information moodel
and information object typees and its instances found in SAP ERP (compare bottom m of
Fig. 2) and SAP BI (compare top of Fig. 2) that are necessary for implementing the
to-be status of GBI.
A comparison of the inteegrated GBI ARIS repository in its to-be status with its ddis-
tributed information modell repository based on SAP ERP and BI shows 80/300 %
o data layer). If combined, they are able to display all
distribution (overlapping of
aspects of the informatio on pyramid (SAP ERP on the operational level, S SAP
Netweaver BI on the tacticcal and strategic level) and all (ERP four, BI five) of the
process life cycles by Scheeruhn. All the four IT-integration layers as well as the ffour
levels of hierarchy can be integrated into one ARIS BD repository. In summary it can
be said:
- 8 (EMM) + 3 (GUI, Money,

M Location) object types found in SAP ERP (80%) and
SAP BI (30%) with 10 % overlapping of data layer = 11 object types
- All 11 object types inteegrated (related to each other) in 14 model types in one
single GBI enterprise model
m in ARIS
- Only 50 % of object tyypes (in both EMM and EMM+) and 30% of model typees
(EMM+) of ARIS integ grated with SAP ERP / BI because of SAP Solution Mann-
ager restrictions:
• Business Blueprint not connected with SAP BI: (no Objectives, KPI)
• Business Blueprint coverrs only 3 of 4 hierarchy levels: (no Events, Business Ruules,
Money flow/material floow of ERP)
• Business Blueprint coverrs no object instances (e.g. no locations of enterprise)
Table 1. Evaluation Matrix GBI (to-be)

Enterprise Models GBI to-be GBI SAP ERP SAP NW
[ARIS BD ] [SAP Solu- BI
tion Manager]
Model structure Level of information pyramid 1-3 3 3 1,2
Process life cycle phases by Scheruhn 2,4 2-5 2-5 1-5
4 IT integration layers ● ● ● ●
Levels of hierarchy 4 3 4 4
Model types instances Shown in Figure 2 ● (1,0) ◔ (0,3) ◕ (0,8) ◔ (0,3)
Object types instances Ref. to Process Governance Matrix by Sche- ● (1,0) ◑ (0,5) ◕ (0,8) ◑ (0,5)
ruhn [22]
Model type instances/ Organization Chart ● ◑ ● ○
1.Presentation layer
Function Role Allocation Diagram ● ● ● ○
2.Process layer E-Business Scenario Diagram ● ◑ ● ○
Company and Department Balanced Score- ● ○ ○ ●
cards
Business Workflow of Business Rules eEPCs / ● ◑ ● ○
BPMN Process Diagrams
Workflow Container Flow ● ○ ● ○
3.Function layer Business Services Tree ● ● ● ○
Objective Diagram ● ○ ● ○
UML Class Diagram ● ○ ● ○
4.Data layer Info Cube, Dimension and Facts Data as Data ● ○ ○ ●
Warehouse Structure Diagrams
Data Source Model ● ○ ● ●
KPI Allocation Diagram ● ○ ○ ●
System Organization Customizing Diagram ● ◑ ● ○
Mask Diagram and WebDynpro Context ● ◑ ● ○
Object type instances/ Roles, Positions, Jobs ● ● ● ○
1.Presentation layer
2.Process layer Strategy, Perspective, Strategic Objectives , ● ○ ○ ●
KPI
Scenarios, Business Processes ● ● ● ○
3.Function layer Process Steps as Methods of SAP Business ● ◔ ● ○
Objects and External WebService
4.Data layer Master, Transaction and Customizing Data ● ● ● ●
Masks (GUI) ● ● ● ○
System Organizational Units (Locations) ● ◑ ● ◑
Keys, Foreign Keys, Describing Attributes ● ○ ● ●
Key Figures, Characteristics ● ○ ○ ●
Money ● ◔ ● ●
Material ● ◕ ● ●
Degree of Integration Integration Model / Object Types seen from 100% 30 / 50% 0% 0%
ARIS Business Designer
Integration seen from SAP Solution Manager - 100% 100% 0%
Legend : ◔ : available only via navigation into SAP ERP

◑ : no interconnection of objects (org structure, Flow of money/material, GUI I/O)
◕ : incomplete (no persons, no instantiation)
6 Conclusion and Outlook

The evaluation matrix identifies all necessary information models on the strategic,
tactical and operational level to create a common framework as a repository for struc-
tured transfer of knowledge and for support of model based implementation in both
the academic and the non-academic sector. First, all of the necessary model and object
type instances found in SAP ERP and SAP Netweaver BI can be (partly manually)
represented and integrated in one single enterprise model repository based on ARIS
BD. Second, as they have to be imported from the ARIS BD through SAP Solution
Manager, which serves as a highly integrated and powerful interface, only a fraction
of the models and objects can be implemented. Seen from ARIS BD only about 30 %
of the considered model type instances and only about 50 % of the object type in-
stances can be synchronized with SAP Solution Manager, with no possibility to
connect strategic or tactical model and object types. This fact is showing a need for
improvement of integration of enterprise model repositories. The matrix therefore can
be used for possible future improvements to show necessary steps in the creation of
repository frameworks.
All information available of the GBI case study has been considered in the enterprise
model. Concerning the non-academic sector it has to be noticed that the TOGAF [30]
enterprise architecture reference model possibly covers more objects of a real enterprise
architecture. Actually, the GBI enterprise model covers some aspects of Architecture
Principles, 100% of Business Architecture, about 70 % of Information System Architec-
ture, no Technology Architecture and about 50 % of Architecture Realization.
“Considering the implementation of SAP projects at external customers of ATOS,
it can be said, that the ERP system with its processes and master data (level 2 and 4)
plays the dominant role. A mapping in ARIS, besides the positive effect of visualiza-
tion may also represent a chance to already show possible future extensions of the
Solution Manager.” [36] The authors strongly believe that the enterprise information
model based on ARIS BD also can help to indicate SAP HANA database relevant
areas in SAP ERP and SAP BI since most foreign keys columns are integrated.
References
1. American Productivity & Quality Center, Best Practice Knowledge Base (2013),
http://www.apqc.org/APQC-knowledge-base
2. Becker, J., Schütte, R.: Handelsinformationssysteme, Redline, Frankfurt (2004)
3. Becker, A.: Nutzenpotentiale und Herausforderungen Serviceorientierter Architekturen.
Gabler Verlag, Wiesbaden (2011)
4. Hahn, T., Wagner, M.: Sustainability Balanced Scorecard. von der Theorie zur Umsetzung,
Universität Lüneburg (2001)
5. Gudas, S., Lopata, A.: Approach to Enterprise Modelling for Information Systems Engi-
neering (2005)
6. Houy, C., Fettke, P., Loos, P., van der Aalst, W., Krogstie, J.: Geschäftsprozessmanage-
ment im Großen in Wirtschaftsiformatik 06/11. Gabler Verlag (2011)
7. Hufgard, A., Krüger, S.: SAP Business by Design: Geschäftsprozesse, Technologie und
Implementierung anschaulich erklärt: Galileo Press, Bonn (2012b)
8. Magal, S., Word, J.: Essentials of Business Processes and Information Systems. Wiley &
Sons (2009)
9. Mertens, P., Bodendorf, F., König, W., Picot, A., Schumann, M.: Grundzüge der Wirt-
schaftsinformatik. Springer, Berlin (2005)
10. Nüttgens, M. et al: Governance and Sustainability in Information Systems. Managing the
Transfer and Diffusion of IT. In: IFIP WG 8.6 International Working Conference, Ham-
burg, (2011)
11. Online Process Management at GBI / part 1 to part 4
http://scn.sap.com/community/uac/blog/2012/10/29/
online-process-management-at-gbi-part-1
12. Peffers, K.: A Design Science Research Methodology for Information Systems Research.
Journal of Management Information Systems 24(3), 45–77 (2007)
13. SAP Bibliothek: Business Blueprint
(2012a), http://help.sap.com/saphelp_sm32/helpdata/de/2a/
62c33af63ae93ae10000000a11402f/content.htm
14. SAP: Business ByDesign – Referenzsysteme - Szenario-Überblick Strategische Bezugs-
quellenfindung, Rel. 3.5.ss (2012a)
15. SAP: Business ByDesign: Unternehmensprofil für ALMIKA Vers.3.5. (2012b)
16. SAP: Strategische Bezugsquellenfindung - Referenzsystem für SAP Business ByDesign,
Rel. 3.5. (2012c)
17. SAP:TERP10 (2012d), http://www.sap.com/germany/services/
education/globaltabbedcourse.epx?context=[[|terp10|||||]]|
18. SAP UCC Magdeburg (2013), http://www.sap-ucc.com/
19. SAP University Alliance: GBI 2.1 ARIS Repository (to-be status) (2013), http://sap-
schulungen.hs-harz.de/GBI_21_Sollzustand_en/index.htm
20. Schaltegger, S., Bennett, M., Burritt, R.: Sustainability Accounting and Reporting. Sprin-
ger, Dordrecht (2006)
21. Scheer, A.-W., Habermann, F.: Making ERP a Success. Communications of the
ACM 43(4), 57–61 (2000)
22. Scheruhn, H.-J.: Model Governance Matrix: ALMIKA (2012c), http://hscheruhn.hs-
harz.de/lehre/alle/Modell-Matrix_horizontale_Navigation.pdf
23. Scheruhn, H.-J.: Online-Prozessmanagement mit SAP ERP. In: Scheruhn, H.-J., Papenfuß,
D., Funk, B., Niemeyer, P. (eds.) Modellierung und Implementierung von Geschäftspro-
zessen in verteilten Systemen, Dr. Kovac, pp. 141–177 (2010)
24. Scheruhn, H.-J., Sicorello, S., Weidner, S.: Repository-based ERP case studies: A study
about chances and benefits of agile case study development. In: Abramowicz, W., Domin-
gue, J., Węcel, K. (eds.) BIS 2012 Workshops. LNBIP, vol. 127, pp. 186–197. Springer,
Heidelberg (2012)
25. Scheruhn, H.-J., Thielert, B., Weidner, S.: GBI-Unternehmensmodell/ Business Process
Analysis/Soll-Zustand (2012b), http://hscheruhn.hs-
harz.de/lehre/alle/Intro_ERP_Using_GBI_Slides_BPA_de_v2.1_
Sollzustand_ARIS_final.pdf
26. Schmidt, W., Fleischmann, A., Gilbert, O.: Subjektorientierter Ansatz in Prozessmanage-
ment: Praxis der Wirtschaftsinformatik. Heft 266 4/09 (2009)
27. Software AG: Industry.PerformanceREADY (2013), http://www.softwareag.com/
de/service/sap_consulting/ipr/default.asp
28. Supply Chain Council: SCOR-Referenzmodell (2012),
http://supply-chain.org/scor
29. TERP10: SAP ERP - Integration of Business Processes (2012),
http://www.terp10.de/pages/deutsch/kurs.php
30. TOGAF: Module 7 TOGAF Content Metamodel (2013), http://www.togaf.org/
togafSlides91/TOGAF-V91-M7-Metamodel.pdf
31. Volkswagen Financial Services AG (2013), http://www.vwfsag.de/content/
sites/vwcorporate/vwfsag_de/de/home.html
32. Weidner, S., Magal, S., Word, J.: Global Bike Inc. Version 2.1. (2011)
33. Twitter Bootstrap, http://twitter.github.com/bootstrap/
34. Wroblewski, L.: Mobile First. A Book Apart, New York, New York, USA (2011)
35. Marcotte, E.: Responsive Web Design. A Book Apart, New York, New York, USA (2011)
36. Förster, U: Interview at ATOS Hannover (January 28, 2013)
Agility Based on Stakeholder Interaction – Blending
Organizational Learning with Interactive BPM
Christian Stary1, Werner Schmidt2, and Albert Fleischmann3

1
University of Linz, Freistädterstraße 315, A-4040 Linz, Austria
[email protected]
2
University of Applied Sciences Ingolstadt, Esplanade 10, D-85049 Ingolstadt, Germany
[email protected]
3
Metasonic AG, Münchner Str. 29, D-85276 Pfaffenhofen, Germany
[email protected]
Abstract. Highly dynamic adaptation of interactive work procedures not only

requires structuring mechanisms, but also engaging stakeholders. As it touches
business operation, not only user interface designs are challenged, but rather
underlying business logic and data management issues. An inclusive perspective
and interactive development support can be provided by Subject-oriented
Βusiness Process Management (S-BPM). It enables stakeholder involvement
throughout an entire engineering cycle based on interaction specifications of all
involved stakeholders. In this paper we propose to use S-BPM representations
for both, capturing the business processes that are modified (i.e. content), and the
process of modifying them along Organizational Learning (OL) cycles (i.e.
adaptation). Since validated S-BPM models can be executed automatically from
each stakeholder perspective, such an integrated BPM-OL approach allows
structuring and implementing agility in operation under direct control of
stakeholders.
Keywords: Agility, Organizational Learning Framework, Organizational

Memory, Subject-oriented Business Process Management.
1 Introduction
Being forced to structural flexibility (agility) either by product or service

management, customer or partner demands, organizations enter the stage of learning,
both on the level of individuals and beyond [cf. 3]. Stakeholders play an important
role for triggering and performing learning processes [cf. 7]. Ideally, they could drive
change processes on both, the level of individual and organizational learning [cf. 2].
Subject-oriented Business Process Management (S-BPM) [cf. 4] relies on
stakeholders and their capabilities constructing behavior specifications. Empirical
results indicate positive effects on organizational velocity [cf. 5].
Various stakeholders can profit from a seamless support of modeling and execution
[cf. 1], concerning user-interface design, business-logic, and business objects:
employees assigned to functional roles in business processes, organization developers
Agility Based on Stakeholder Interaction – Blending Organizational Learning 457
while negotiating and mediating, and IT specialists when working with stakeholders.
Either starting with process analysis and proceeding with modeling, validation, and
execution, S-BPM models reflecting organizational change need to be guided by a
reference model on a higher level to ensure valid semantic representations und mutual
understanding. However, expressing modeling-relevant information in terms of
intuitive representation is crucial for effective BPM, otherwise modeling problems
well known from other BPM developments, such as BPMN [cf. 6], might occur.
In section 2 a reference model for effective change management stemming from
OL is reviewed. In addition, S-BPM with respect to modeling activities is recaptured,
revealing basic requirements for understanding the use of the S-BPM modeling
approach. Both inputs are required for informed model construction and respective
learning designs on lower and higher level OL. As S-BPM allows for seamless round
trip engineering due to its aligned representational and execution capabilities, the
stakeholder-driven Organizational Learning models can form an effective frame of
reference for agile BPM. In section 3 an S-BPM implementation of the OL approach
is exemplified, providing respective S-BPM representations. Further studies to
blending OL with BPM are sketched in the conclusion of the paper.
2 Foundations
In the following we describe how stakeholders can trigger learning processes to

reconfigure business operation. We also review Subject-Oriented Business Process
Management with respect to modeling and execution as a baseline.
2.1 Organizational Learning Framework

For intertwining individual knowledge creation and collective learning processes in
[9] an experiential learning cycle has been detailed (see figure 1) where individual
knowledge creation serves as input to organizational learning processes. Change on
the collective layer is triggered through designing processes, and followed by access,
experience, and assessment on the individual level, before being shared in work
communities:
• Design: Stakeholders express their specific view onto information structures or

business processes, according to their individual experiences.
• Implement: The resulting information or models can be embedded into actual task
situations. Interactive artefacts could enable hands-on-experience for task
accomplishment (archetyping).
• Observe: Stakeholders observe, in particular when interactive artefacts are used,
possible effects the executed tasks have on the situation and the organization.
• Assess: If the results fit individual expectations or meet individual demands for
change, the concerned items serve as input for the learning process on the
collective level. If further process refinements or modifications are required the
cycle starts again.
458 C. Stary, W. Schmidt , and A. Fleischmann
In order to transfer the individually elicited knowledge to the organizational level

several activities facilitate interaction, collective exploration, and reflection:
• Annotating information and process models. Basically, all created evidence for
change needs to be documented by its proposer(s). It can then become available in
some kind of Organizational Memory (OM) accessible for all stakeholders. The
OM is the core of a support system, as it is used to store not only originally
proposed structures, but also updates and versions. Each stakeholder can then
express individual concerns and formulate individual inputs in a context-sensitive
way, namely through annotating information. Annotations comprise comments,
associations, and supplements to the created change request.
• Deploying and sharing representations (models). In order to share created
evidence with other stakeholders, all generated information needs to be kept.
Stakeholders should be able to share their findings with others. In this way
individual perspectives on a process proposal can be taken by others.
Organizational Learning can be considered as a swinging pendulum between

individuals and affected stakeholders, providing inputs for change and reflecting
created information (structures) collectively before putting it to operation. First,
individuals take the role of content providers, before stepping into the role of
receptors or respondents, reflecting content and inputs on the individual or collective
layer, respectively.
A step or cycle is complete, once a modified information model, content or process
descriptions have been negotiated as a commonly agreed basis for organizational
development, and can become effective on the collective level. Stakeholders then take
the role of receptors embodying information.
Figure 1 shows the fundamental structure, activities, knowledge management
system (KMS) components, and relationships. Individual stakeholders provide inputs
or start discussions, leading to information they consider being of relevance to be
documented. Once created, content can be refined by other stakeholders, modified or
supplemented through sharing different inputs until proceeding on the collective level.
Information and process models are initially generated on an individual level. At
some point, either immediately after creating or re-modeling, or after archetyping the
input provider makes his/her model public. Then, others could step in and provide
annotations to reflect the model collectively. Once a version is recognized by the
involved stakeholders to become operational (i.e. put to production), the build time
has been completed, and a novel way of running the business at hand is ready to be
deployed. Build time has come to an end, run time can start. This step is indicated by
‘transfer’ transition to the Organizational Memory (OM). At that time additional
implementation steps, such as additional validation with respect to stakeholder
involvement, might be performed. After completing detailed implementation, the
business operation actually changes, however still in an unembodied way, as new
work practices need to be experienced in actual work settings realtime to become
fully embodied. Hence, the roles stakeholders take in that context are still
experimenter and explorer. The roles indicate the resonance space created now for all
involved stakeholders to embody novel work processes or information structures that
are stored in the Organizational Memory. Once they have embodied novel work
practices (indicated in the figure by the transition ‘Embodiment’) they enter the
individual ‘Creating and Reflecting’ cycle on the individual level. Even in case
stakeholder experience difficulties in following novel work practices they should be
able to act in a constructive way in the next learning step. In this way, the framework
in figure 1 addresses explicitly the interplay between individual and collective
learning. Both aspects are considered essential for interactive organizational change.
The model enables considering each of them from dedicated perspectives, and their
interfaces along the Organizational Learning life cycle.
Creating & Reflecting
observe
Embodiment
implement assess
design
Annotating
& Sharing
Information and Information
Archetyping Process Model Build Time
Transfer
Run Time
Organizational Memory
Fig. 1. The operational frame of reference and resulting KMS components for change support
as given in [9]
2.2 Subject-Oriented Business Process Modeling and Execution

Subject-oriented business process models aim to bridge the gap between describing,
documenting, and processing collaborative work processes. When expressing their
understanding of work behavior stakeholder only have to understand interaction as the
process of exchanging messages between actors and systems - which they are capable
once they collaborate and use mail, in particular e-mail. Models play a crucial role in
S-BPM. They are generated or processed, respectively, along a chain of various
activities:
• eliciting and representing knowledge about work and its organization

• analyzing process descriptions with respect to specifity, accuracy, completeness
• validating models whether they could be executed in a stringent way

• executing models allowing hands-on experience of specifications
• embodying processes in organizational and technological settings of organizations
• monitoring with respect to expected achievements
• simulating in order to explore alternative or novel ways of task accomplishment
These bundles of activities represent fundamental phases of business process

development. Once in each phase of development all stakeholders can participate due
to their modeling capabilities, the transparency and traceability of organizational
development processes can be ensured or even increased. A key enabler to this
respect is the capability of support tools to execute validated process models (cf.
www.metasonic.de).
We now briefly review the creation of subject-oriented representations. It starts
with the identification of process-specific roles involved in the process, the subjects,
and the messages exchanged between them. When sending messages, the required
data is transmitted from the sender to the receiver. Thus, with a message indicating
the intention to go on a business trip, e.g., ‘request’, sent by an employee to the
supervisor, among other things the start and end date are transmitted. The behavior of
communication partners, such as the supervising manager is complementary. For
instance, messages sent by the employee are received by the manager, and vice versa.
The manager therefore waits first in a receiving state for a business trip request from
the employee. For each work procedure or business case, the
1. subjects involved in the process,

2. interactions taking place between them
3. messages they send or receive during each interaction, and
4. behavior of the individual subjects
are described as they represent the essential elements of a subject-oriented model. The
description of a subject determines the order in which it sends and receives messages,
and performs internal functions. Its behavior thus defines the order in which the
subject processes which activities: sending or receiving, or services that are defined
on the corresponding objects. Services are used to assign a specific meaning to the
individual steps captured by a subject behavior model. They are triggered
synchronously, i.e., a subject does not enter the corresponding next state, unless the
used service has been also completely processed.
Figure 2 exemplifies a model on the left side. The screen on the right side shows a
snapshot of the screen displayed when the model on the left side is executed as a
workflow using the Metasonic Suite (www.metasonic.de). Once all ingredients have
been added, the model can be validated and executed without further transformations.
In the figure the current state is the receiving state as also highlighted in the model –
an employee receives an answer to his/her request. The corresponding inbox can also
be visually displayed in terms of user interface widgets as known from model-based
development [cf. 8]. It requires the assignment of user-interface elements in the
course of instantiating subject specifications.
Fig. 2. Model Representation and Execution
Once those subject models have been completed that are involved in a work
process, an organization-wide representation has been created. Its execution allows
from each subject (i.e. stakeholder) perspective to experience the entire process
interactively.
3 (Re-)Structuring Operations ‘On-the-Fly’
We demonstrate how the framework given in section 2.1 can be implemented in a

seamless BPM environment. Figure 3 gives the principal interaction structure of the
OL process with input provider, stakeholder and organizational memory (OM) as
Fig. 3. Subjects involved in Organizational Learning

subjects and several types of messages they exchange. As subjects represent roles, at
run time one person can act in different roles, such as an input provider and
stakeholder commenting proposals for change provided by others.
For technical systems, such as the organizational memory, it is also decided at the
time of implementation, which technology is going to be used at run time. In S-BPM
the organizational implementation is distinguished from the technical, in order to
capture both implementation aspects [4].
According to the OL life cycle (Figure 1) initially a stakeholder familiar with a
work procedure (in the role of input provider) triggers change (Figure 4). He prepares
some content to be considered as a new work pattern, for example a process model,
such as the behavior of an employee applying for a business trip, or some other input
triggering change. The content is sent to the OM as a request for change and then, the
input provider waits for feedback, which may subsequently lead to changes until
content has been approved by the other stakeholders. Finally the input provider
embodies the new work procedure in his daily operation and acts according to it until
a new learning cycle is initiated by him or other stakeholders and leads to another
change of work practice.
Processing travel expense reimbursements for sales persons can serve as an
example. The responsible stakeholder, e.g., an accountant, receives electronic forms
containing data like name and organizational unit of the sales person, travel details,
money spent etc. In order to book the reimbursement the accountant needs to retrieve
the relevant cost center id of the organizational unit from a list. As this takes him 2
minutes for every single instance his idea is to let the sales people fill in the right id
into the form. This would only take them some seconds because they usually know
their cost center id. So the accountant provides a new version of the form including a
field for the cost center id.
Fig. 4. Behavior of the subject ‘input provider’

Fig. 5. Behavior of the subject ‘stakeholder’ responding to provided inputs
Fig. 6. Behavior of the subject Organizational Memory
Other stakeholders may, beside themselves suggesting changes to the OM, share
the provided content, and make annotations (Figure 5). These annotations refer to the
originally provided content, are sent to the Organizational Memory and through it
become available to the input provider who may modify the change request.
The Organizational Memory receives all inputs and informs other stakeholders
about the input (Figure 6). It also stores incoming annotations and provides versioning
for further iterations. In our example stakeholders like sales people and the owner of
the process express their acceptance of the modified form which then can become
effective in operation. In another case some of them could suggest further
improvement by extending the electronic form by services to automatically retrieve
personal and cost center data from databases once the short employee id is entered.
This could lead to an iteration, e.g., collecting the stakeholder’s opinions on whether
to start with the additional field or go straight for the more sophisticated change with
the IT services.
The behavior diagrams have been simplified to demonstrate the interaction flow
being the distinct feature of the approach. After validation it finally enables the
automated execution of the behavior diagrams.
Hence, the resulting learning and business operation occurs in a choreographic
way. This is of dual importance: On one hand, actors and systems may interact in
parallel, however, timely synchronized. On the other hand, organizational learning is
an intervention on demand, rather than prescribed, which results in non-intrusive
change processes with respect to the technically informed business.
4 Conclusion
Meeting the requirement of many organizations to reconfigure their business

processes dynamically, we followed a blended BPM-OL approach. A procedural
framework for individual and collective changes and the operational representation
for executing business processes have been intertwined, utilizing the capabilities of
Subject-oriented Βusiness Process Management (S-BPM). It provides a stakeholder
perspective for modeling processes and organizational development. Re-configuring
running business processes is enabled in seamless round trips.
Field studies still need to be set up to validate the results empirically in the context
of BPM projects, as the current tool support is undergoing major improvements (see
www.i2pm.net – Open S-BPM). Of particular importance is investigating existing
role concepts, as known from BPM, e.g., process owner, and their impact on the roles
identified for organizational learning, such as input provider. It is likely that mutual
mappings need to be defined to perform change management on-the-fly effectively.
References
1. Avital, M., Te’eni, D.: From Generative Fit to Generative Capacity: Exploring an Emerging
Dimension of Information Systems Design and Task Performance. Information Systems
Journal 19(4), 345–367 (2009)
2. Böhmann, T., Burr, W., Herrmann, T., Krcmar, H. (eds.): Implementing International
Services: A Tailorable Method for Market Assessment, Modularization, and Process
Transfer. Gabler, Wiesbaden (2011)
3. Chen, M.T.: The Impact of Organizational Agile, Organizational Learning and Innovation
on Value Creation, Ph.D., Department of Business Administration, Shih Hsin University,
Taipei (2012)
4. Fleischmann, A., Schmidt, W., Stary, C., Obermeier, B.E.: Subject-oriented Business
Process Management. Springer, Heidelberg (2012) http://link.springer.com/
book/10.1007/978-3-642-32392-8/page/1
5. Nakamura, S., Tan, T., Hirayama, T., Kawai, H., Komiyama, S., Hosaka, S., Nakamura, M.,
Yuki, K.: CGAA/EES at NEC Corporation, Powered by S-BPM: The Subject-Oriented
BPM Development Technique Using Top-Down Approach. In: Schmidt, W. (ed.) S-BPM
ONE 2011. CCIS, vol. 213, pp. 215–231. Springer, Heidelberg (2011), doi:10.1007/978-3-
642-23471-2_15
6. Recker, J.: Opportunities and Constraints: The Current Struggle with BPMN. Business
Process Management Journal 16(1), 181–201 (2010)
7. Rouse, W.B. (ed.): Enterprise Transformation: Understanding and Enabling Fundamental
Change. Wiley, Hoboken (2006)
8. Stary, C.: TADEUS: Seamless Development of Task-based and User-oriented Interfaces.
IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans 30(5),
509–525 (2000)
9. Stary, C., Fleischmann, A.: Evidence-based Interactive Management of Change. Knowledge
Management & E-Learning: An International Journal (KM&EL) 3(2), 170–200 (2011)
http://www.kmel-journal.org/ojs/index.php/
online-publication/article/view/31/83
Analysis of Customer Preference through Unforced
Natural Passive Observation
Terumasa Tajima1, Yusuke Iida1, and Toshikazu Kato2

1
Graduate School of Science and Engineering, Chuo University, 1-13-27 Kasuga,
Bunkyo-ku, Tokyo, 112-8551 Japan
2
Faculty of Science and Engineering, Chuo University, 1-13-27 Kasuga,
Bunkyo-ku, Tokyo, 112-8551 Japan
[email protected]
Abstract. In our former research, customer’s preference has been estimated by

passive observation of shopping behavior, e.g. customer’s “look” and “touch”.
It takes much time to understand their preferences form the log. We need quick-
ly to build up the preference model to perform suitable recommendation for a
new customer. For this reason, we will propose an active observation mechan-
ism that detects customer’s unforced natural behavior to information through
ambient devices such as speakers and electric displays. This mechanism also
analyzes customer’s preference on features and their values of commodities,
which enables the system to estimate the rate of preference to an unknown
product. We have experimented on ten university students. We had them eva-
luate the thirty-six Shirts. We used these evaluations for precision evaluations
in naive Bayes classifier. We used the leave-one-out cross-validation. As the re-
sult, we have achieved the average precision in the estimating preferences by
naive Bayes classifier is 71%.
Keywords: Passive and active observation, Recommendation system, Machine

classifier, Digital signage, Decision making support
1 Introduction
Customer’s preference has been analyzed on shopping logs in a real shop as well as
page view logs in web shopping sites [1][2][3][4][5]. These approaches require mass
data for each individual to build personal preference model, otherwise the analysis is
just on the mass market as an average preference of customers. We have developing a
smart shop system that analyzes each customer’s preference to commodities observ-
ing his unforced natural behavior in the shop [6]. We have adopted passive observa-
tion of the customer’s behavior, such as “look”, “touch”, and “take” a commodity in a
shelf, which requires some period to save log. We will also adopt active observation
of the customer’s response to the digital signage, such as gaze and ignore, to capture
log more quickly without forcing him to answer the messages.
Analysis of Customer Preference through Unforced Natural Passive Observation 467
2 Discussion on Passive Observation
In our former research, customer’s preference has been estimated by passive observa-
tion of shopping behavior, e.g. customers “look” or “touch”. As this observation can
only be captured their behavior without forcing replies, it takes much time to under-
stand their preferences form the log. Quickly we have to build up the mechanism of
preference model to perform preferable recommendation for a new customer. For this
reason, we will propose on active observation mechanism that detects customer’s
unforced natural behavior to information through ambient devices such as speakers
and electric displays. This mechanism also analyzes customer’s preference for fea-
tures and their values of commodities, which enables the system to estimate the rate
of preference to an unknown product.
3 Active Observation
The aim of our research is to propose active observation system that can estimate
customers’ preferences from small load. This section explains three things. First is a
way of observing behaviors that enables us to estimate customer’s preferences.
Second is a method to estimate customer’s preferences from behaviors. Third is a
method to understand customers’ preference from small load. We described the over-
view process of active observation in Fig. 1.
3.1 A Way of Observing Behaviors
As devices of the active observation system, we adopted ambient display as digital

signage devices with a function to detect a human face and its direction.
3.2 Estimating a Preference from Behaviors
Our former research found correlation between the gaze time and the rate of prefe-
rence. It considers the subjectivity of gaze behavior by customer analysis [7]. The
gaze time is the total duration of watching information on the ambient device. This
observation system captures gaze time of each customer. Thereby, we estimate the
customer’s preference, “like” or “dislike”, to be compare the mean and the variance of
gaze time. For example, Fig 1 (a) (a’) shows that the system estimates red polo shirt
“like” and blue turtleneck shirt “dislike”.
3.3 Analyzing Customer’s Preference
We enable the system to estimate the rate of preference to unknown products. For that
analyzing and learning algorithm, we have embedded content based filtering to
468 T. Tajima, Y. Iida, and T. Kato
system. We have used the naive Bayes classifier. Thereby, we can get an evaluation
for the product which a customer hasn’t evaluated. For example, Fig 1 (b) (b’) shows
that the system computed a posterior probability with product property of polo shirt
and turtleneck shirt. Thereby, it estimated rates of preference the similar products in
the database. On the other hands, content based filtering is recommending only simi-
lar products which customer evaluates as high. However, the customer’s over all pre-
ference cannot have been analyzing.
We explain model making and a classification by naive Bayes. The way of calcu-
lating naive Bayes probabilistic model wants you to refer to documents [8]. The prob-
ability model for a classifier is a conditional model. Conditional model is dependent
class variable, Class, and feature variables, sample, by eq. (1).
| (1)
Furthermore, using Bayes' theorem, we obtain the following expression (2).
P (Class ) P ( sample | Class )

P ( Class | sample ) = ∝ P (Class ) P ( sample | Class ) (2)
P ( sample )
The feature variable, sample, is a set of the several feature variables

through . Each feature is conditionally independent of every other fea-
ture for j i given the class variable. We obtain the following expression
(3).
P(sample| Class) = P(word1 ∧ ....word i | Class) = ∏ P(word i | Class) (3)

i
Thus, we had classified unknown samples by eq. (4).
Classmap = arg max P(Class)∏ P(word i | Class) (4)

Class
i
We have embedded a metric based on entropy to estimate rates of recommendations.

For example, Fig 1 (c) (c’) shows that the system computes posterior probability of
property on product (P (red|like), P (red|dislike) ). Entropy, H, are calculated from
posterior probability, P, by eq. (5).
1 1 (5)
Therefore, we calculated the entropy of each product and decided a next candidate for
recommendation from entropy. We combine entropy with naive Bayes classifier.
Thereby, we can analyze the customer’s preference with short duration and a small
load.
䛱
䛱䜑
Fig. 1. Image of the flow to estimated customer’s preference
(a)(a’) … We estimated the customer’s preference, “like” or “dislike”, to compare the mean
and the variance of gaze time.
(b)(b’) … We computed a posterior probability with product property and estimated similar
product in the database.
(c)(c’) … We calculated the entropy of each product and decided a next candidate for recom-
mend from entropy
4 Experiment
We had been experimenting on the effectiveness of the two processes. First is effec-
tiveness of the estimated precision of the customer’s preference by naive Bayes clas-
sifier. Second is effectiveness of the customer’s preference estimate number of times
by the entropy.
4.1 Estimating a Preference by Naive Bayes Classifier

We have experimented on ten university students (21- 24 years-old, in Tokyo). We
had them look and touch thirty-six Shirts. Also, we had them evaluate the Shirts. We
used these evaluations for precision evaluations in naive Bayes classifier. We used the
leave-one-out cross-validation.
4.2 Reduction of the Estimate Duration by Entropy

We have experimented on three university students (21- 24 years-old, in Tokyo).
We had them look one hundred and four Shirts. Furthermore, we had subjects eva-
luate one hundred and four Shirts. We used these evaluations in analysis and precision
evaluations. We compared the number of times to preference estimate when using the
entropy with does not use it.
5 Result
5.1 Estimating a Preference by Naive Bayes Classifier
The result is as follows (Table1).
Table 1. Precision by naive Bayes classifier
Subject No Precision(％)
1 61.1
2 75.0
3 69.4
4 72.2
5 61.1
6 66.7
7 80.7
8 80.6
9 80.6
10 72.2
average 71.1
As the result, the average precision in the estimating preferences by naive Bayes
classifier is 71%. Therefore, we think that the estimating preferences by naive Bayes
classifier are effective. Hence, we think that we enable the system to estimate the rate
of preference to the unknown products.
5.2 Reduction of the Estimate Duration by Entropy

First of all, we explain the figure of this chapter in Fig. 2.
Fig. 2. Example of the figure
(x) … The upper bound that we can preference estimate in naive Bayes
(a) … Relationship between the number of times to product recommendation and preference
estimate rate when using the entropy.
(b) … Relationship between the number of times to product recommendation and preference
estimate rate when to not use the entropy.
The result is as follows (Fig. 3, Fig.4 and Fig. 5).

As analysis, we analyzed the change of the preference estimate rate when a product
recommended by evaluation of the subject. We applied the analysis three times and
averaged the results every subject.
Fig. 3. Analysis result of subject A
Fig. 4. Analysis result of subject B
Fig. 5. Analysis result of subject C

Furthermore, we gathered up all the results of the subject in Fig. 6.
Fig. 6. Preference estimate when using the entropy with does not use it (united in subjects)
As the result, the learning using the entropy may achieve a high preference esti-
mate rate in a short duration. As much analysis results, the learning to use the entropy
is shorter than learning not to use. Hence, we think that we can analyze the customer’s
preference with short duration and a small load.
6 Conclusion
Our purpose has developing a smart shop system that analyzes each customer's prefe-
rence to commodities observing his unforced natural behavior in the shop. In our
former research, customer's preference has been estimated by passive observation of
shopping behavior, e.g. customer’s “look” and “touch”. It takes much time to under-
stand their preferences form the log. We need quickly to build up the preference
model to perform suitable recommendation for a new customer. For this reason, we
proposed an active observation mechanism that detects customer’s unforced natural
behavior to information through ambient devices such as speakers and electric dis-
plays. This mechanism also analyzes customer's preference on features and their val-
ues of commodities, which enables the system to estimate the rate of preference to an
unknown product. We had been experimenting on the effectiveness of the two
processes. First is effectiveness of the estimated precision of the customer's prefe-

rence by naive Bayes classifier. Second is effectiveness of the customer's preference
estimate number of times by the entropy. As the result, we have achieved the average
precision in the estimating preferences by naive Bayes classifier is 71%. We can ex-
pect to analyze that we can analyze the customer's preference with short duration and
a small load by using the entropy.
Acknowledgement. This work was partially supported by JSPS KAKENHI grants,

"Effective Modeling of Multimodal KANSEI Perception Processes and its Applica-
tion to Environment Management" (No. 24650110), "Robotics modeling of diver-
sity of multiple KANSEI and situation understanding in real space" (No. 19100004)
and TISE Research Grant of Chuo University, "KANSEI Robotics Environment".
References
1. Fayyad, U.M., Wierse, A., Grinstein, G.G.: Information Visualization in Data Mining and
Knowledge Discovery. Morgan Kaufmann (2001)
2. Eirinaki, M., Vazirgiannis, M.: Web mining for web personalization. ACM Transactions on
Internet Technology (TOIT) 3(1), 1–27 (2003)
3. Yukihiro, A.: In-Store Buying Behavior and Retail Management(1). Journal of Business
Administration of Kwansei Gakuin University 32(4), 117–146 (1985)
4. Yukihiro, A.: In-Store Buying Behavior and Retail Management(2). Journal of Business
Administration of Kwansei Gakuin University 33(1), 163–179 (1985)
5. Nakayama, A., Tsurumi, H.: Analysis of Customer’s Multiple Purchase Behaviors and In-
fluences of Store Arrangement in Department Store by INDSCAL. Jpn.J. of Behaviorme-
trics 32(2), 201–208 (2005)
6. Imamura, N., Ogino, A., Kato, T.: PatternAnalysis of Customer Behavior Using Ubiquitous
Interface. In: Proceedings of the Annual Conference of JSKE, vol. 8(3), pp. 741–747 (2009)
7. Iida, Y., Kato, T.: Presumption of interest by active sensing using measurement of natural
reaction to recommendation. In: Proceedings of the Annual Conference of JSKE, vol. 13, p.
26 (2011)
8. Manning, C.D., Raghavan, P., Schütze, H.: Introduction to Information Retrieval. Cam-
bridge University Press (2004)
The Development of an Innovative Design Process
for Eco-efficient Green Products
Shiaw-Tsyr Uang1 and Cheng-Li Liu2
1
Department of Industrial Engineering and Management,
Minghsin University of Science & Technology, Hsinchu, Taiwan
2
Department of Industrial Management, Vanung University, Taoyuan, Taiwan
[email protected]
Abstract. Along with the technological advances and the rapid flow of
information, people continue to create more innovative technology products in
order to make life more convenience and comfortable than the past. However,
huge industrial processes bring significant changes in the environment, the
earth's resources are getting to dry up, the natural environment and ozonosphere
have been destroyed, and leading serious threats to the survival of the Earth's
species. Therefore, in recent years, environmental awareness began to be
respected, and people explore green issues, hoping to make the earth resources
and living things can be survival sustainable. This research attempts to integrate
39 engineering parameters of TRIZ with 7 eco-efficient elements to construct a
green product design matrix, and use the patents of LED lights to verify the
validity of this innovative green product design process. The green product
design process of this research may allow designers to generate green products
by improving eco-efficient elements as well as resolving conflict design issues,
and assist organizations to increase competitiveness and to achieve sustainable
development goals.
Keywords: Theory of Inventive Problem Solving (TRIZ), Eco-efficient, Green

Design, Product Innovation, Light Emitting Diode (LED).
1 Introduction
Along with the technological advances and the rapid flow of information, people
continue to create more innovative technology products in order to make life more
convenience and comfortable than the past. Product designers usually concentrate on
design issues such as product efficacy, cost, appearance, usability, and so on.
However, huge industrial processes bring significant changes in the environment, the
earth's resources are getting to dry up, the natural environment and ozonosphere have
been destroyed, and leading serious threats to the survival of the Earth's species
[7][20]. Product designers are now facing brand-new challenges than before. In recent
years, environmental awareness began to be respected, and people explore green
issues, hoping to make the earth resources and living things can be survival
476 S.-T. Uang and C.-L. Liu
sustainable. Under these growing environmental pollutions, many countries have

implemented environmental protection laws to reduce the environment impact of
industry. For instance, WEEE (Waste Electrical and Electronic Equipment Directive),
RoHS (Restriction of Hazardous Substances Directive) and EuP (Energy Using
Product) organizations under the auspices of the European Union limits the
environmental pollution caused by products [2][19]. Product designers should thus
provide eco-friendly solutions to product design problem through innovative methods
and tools to help implement eco-friendly designs [1][14].
As defined by the World Business Council for Sustainable Development
(WBCSD) [20], “eco-efficiency is achieved by the delivery of competitively priced
goods and services that satisfy human needs and bring quality of life, while
progressively reducing ecological impacts and resource intensity throughout the life-
cycle to a level at least in line with the Earth’s estimated carrying capacity.” In short,
eco-efficiency is concerned with creating more product values with less negative
environmental impacts.
In addition, eco-efficiency can be implemented along the entire value chain of a
product or service [4][9][17]. For some companies, the main harmful effects on the
environment with which they are associated actually occur outside their own
organizations – either upstream in the raw material generation and supplier processing
phases, or downstream in the product use or disposal phases. In light of this, WBCSD
pointed out eco-efficiency can be achieved through seven key eco-efficient elements
as below (REDUCES, in short) [20]. It will produce high eco-efficient products or
services while improving these seven eco-efficient elements.
1. Reduce material intensity

2. Energy intensity minimized
3. Dispersion of toxic substances is reduced
4. Undertake recycling
5. Capitalize on use of renewable resources
6. Extend product durability
7. Service intensity is increased.
Theory of Inventive Problem Solving was developed by Genrich Altshuller and his
colleagues, and is now being developed and practiced as an innovative tool
throughout the world. "TRIZ" is the acronym for this theory in Russian [3][5][6].
TRIZ is a dialectic way of thinking in finding a suitable solution while facing a design
predicament [10] [18].
The contradiction matrix of TRIZ provides designers which of the 40 inventive
principles have been used most frequently to solve a problem that involves a
particular contradiction [11]. For example, if you need a static object to be longer
without becoming heavier. The improving feature of this contradiction is “04.Length
of non-moving object” and the worsening feature is “02.Weight of non-moving
object”. As indicated in Table 1, we may use inventive principles “35.parameter
changes”, “28.mechanics substitution”, “40.composite materials”, and “29.pneumatics
and hydraulics” to solve this design problem.
The Development of an Innovative Design Process for Eco-efficient Green Products 477
Table 1. Part of the TRIZ contradiction matrix
The purpose of this research is to integrate engineering parameters of TRIZ with

eco-efficient elements to construct a green product design process, and use the patents
of LED lights to verify the validity of this green product design process.
2 Method
Research procedures are introduced in the following sections.

First, in order to transfer 39 engineering parameters of TRIZ into 7 eco-efficient
elements, literature reviews [1][8][12][13][18][21][22][23] are conducted to find out
the relationship between them. For example, reducing material intensity of a product
may be achieved by changing product’s features, such as weight, size, shape, and so
on. Table 2 demonstrates the relationship between 39 engineering parameters and 7
eco-efficient elements based on literature reviews.
Next, Table 2 is used to transform/reduce the original TRIZ 39×39 contradiction
matrix into 7×39 green design matrix. Since the matrix is downsized in this study, the
number of suggested inventive principles increases within each cell. Hence, inventive
principles are sorted descending by their frequencies within each cell of the green
design matrix. And, inventive principles with frequencies less than two are eliminated
from the table. Table 3 is part of the green design matrix constructed by this study.
Most frequently used inventive principles are shown in the cells under various
combinations of design contractions.
Table 2. Part of the relation table between engineering parameters and eco-efficient elements
Table 3. Part of the proposed green design matrix

The innovative design process of this research is depicted in Figure 1. Product

designers may consider and select eco-efficient elements according to their priorities,
and then analyze the worsening feature under this condition. After that, the improving
eco-efficient element and worsening feature can be used to locate inventive principles
in green design matrix. Finally, designers may adopt suggested inventive principles to
design/improve products.
Fig. 1. Innovative design process for eco-efficient product
The validity of the above innovative green product design process is verified and
case-studied by existent patents of LED lights. Detailed description of two patents is
illustrated in the Results section.
The proposed green design matrix is examined by existent patents. Our cases used to
evaluate the validity of the proposed green design model are selected from the Taiwan
Patent Database.
First one (Patent No M348049) is an indicator light on the bike (Figure 2) [15].
One piezoelectric module is installed under the bike saddle. While riding a bike, the
induced vibration drives the piezoelectric module to generate lighting which can be
used to alert other road users.
Fig. 2. Diagram of Taiwan Patent No M348049 [15]
This patent wants to minimize energy intensity; however, any imbedded devices
may cause waste of substance. Hence, as shown in Table 3, green design matrix
suggests inventive principles 35, 10, 18, 27, 28, 31 can be applied to solve this design
problem. This patent obviously adopts inventive principle “18.mechanical vibration”
as suggested.
Table 4. Part of the green design matrix used to illustrate Patent No M348049
Second one (Patent No I327632) is a multi-functional LED camping light (Figure

3) [16]. Fixtures can be easily setting to change light direction according to users’
preferences.
Fig. 3. Diagram of Taiwan Patent No I327632 [16]
This patent wants to reduce material intensity while camping; however, it may
cause inconvenience of use. Hence, as shown in Table 4, green design matrix suggests
inventive principles 02, 13, 15, 32, 16, 25, 35 can be applied to solve this design
problem. This patent obviously adopts inventive principle “15.dynamicity” by
adjusting setting as suggested.
Table 5. Part of the green design matrix used to illustrate Patent No I327632
Case studies of the existence patents demonstrate the validity of this green product
design matrix. The proposed green design matrix may simplify the process of
selecting the most appropriate inventive principle to solve a specific contradiction
between eco-efficient element and engineering feature.
4 Conclusion
With raising environmental protection awareness, product designers are now facing
brand-new challenges than before. Product designers should thus provide eco-friendly
solutions to product design problem through innovative methods and tools to help
implement eco-friendly designs. Present research attempts to integrate 39 engineering
parameters of TRIZ with 7 eco-efficient elements to construct a green product design
matrix, serving as an innovative green design tool. Case studies of two LED patents
verify the validity of the proposed innovative green product design process. The
proposed green design matrix may simplify the process of selecting the most
appropriate inventive principle, and then to accelerate design problem solving. To
sum up, the green product design process of this research may allow designers to
generate green products by improving eco-efficient elements as well as resolving
conflict design issues, and assist organizations to increase competitiveness and to
achieve sustainable development goals.
Acknowledgement. The authors would like to thank the National Science Council of
the Republic of China for financially supporting this work under Contract No.
NSC101-2221-E-159-002.
References
1. Chang, H.-T., Chen, J.L.: The conflict-problem-solving CAD software integrating TRIZ
into eco-innovation. Advances in Engineering Software 35, 553–566 (2004)
2. Chu, C.-H., Luh, Y.-P., Li, T.-C., Chen, H.: Economical green product design based on
simplified computer-aided product structure variation. Computers in Industry 60, 485–500
(2009)
3. D’Anna, W., Cascini, G.: Supporting sustainable innovation through TRIZ system
thinking. Procedia Engineering 9, 145–156 (2011)
4. Dangelico, R.M., Pontrandolfo, P.: From green product definitions and classifications to
the Green Option Matrix. Journal of Cleaner Production 18, 1608–1628 (2010)
5. Fresner, J., Jantschgi, J., Birkel, S., Barnthaler, J., Krenn, C.: The theory of inventive
problem solving (TRIZ) as option generation tool within cleaner production projects.
Journal of Cleaner Production 18, 128–136 (2010)
6. Hou, J., Su, D.: A customer-manufacturer-competitor orientation model for product life
cycle analysis based on QFD, AHP/ANP and TRIZ. International Journal of Design
Engineering 1(1), 104–124 (2007)
7. Intergovernmental Panel on Climate Change (IPCC), http://www.ipcc.ch/
8. Jones, E., Harrison, D.: Investigating the use of TRIZ in eco-innovation. The TRIZ Journal
(2000), http://www.triz-journal.com
9. Kobayashi, H.: A systematic approach to eco-innovative product design based on life cycle
planning. Advanced Engineering Informatics 20, 113–125 (2006)
10. Li, S.-L., Huang, H.-H.: Applying TRIZ and fuzzy AHP to develop innovative design for
automated manufacturing systems. Expert Systems with Applications 36, 8302–8312
(2009)
11. Lim, I., Sheng, S., Kok-Soo, T.: Eco-efficient product design using theory of inventive
problem solving (TRIZ) principles. American Journal of Applied Sciences 7(6), 852–858
(2010)
12. Liu, C.-C., Chen, J.-L.: Development of product green innovation design method. In:
Proceedings of Eco Design, pp. 168–173 (2001)
13. Liu, C.-C., Chen, J.-L.: Inventive principles and evaluation method for eco-innovative
design of products. Environmentally Conscious Design and Inverse Manufacturing,
369–376 (2003)
14. Pujari, D., Wright, G., Peattie, K.: Green and competitive influences on environmental new
product development performance. Journal of Business Research 56, 657–671 (2003)
15. Taiwan Patent No M348049,
http://twpat2.tipo.gov.tw/tipotwoc/tipotwkm?@@1083144416
16. Taiwan Patent No I327632,
http://twpat2.tipo.gov.tw/tipotwoc/tipotwkm?@@1083144416
17. Tseng, H.-E., Chang, C.-C., Li, J.-D.: Modular design to support green life-cycle
engineering. Expert Systems with Applications 34, 2524–2537 (2008)
18. Uang, S.-T., Liu, C.-L., Chang, M.: A Product Design Approach by Integrating Axiomatic
Design and TRIZ. In: Marcus, A. (ed.) Design, User Experience, and Usability, HCII
2011, Part II. LNCS, vol. 6770, pp. 225–233. Springer, Heidelberg (2011)
19. Vezzoli, C., Sciama, D.: Life cycle design: from general methods to product type specific
guidelines and checklists: a method adopted to develop a set of guidelines/checklist
handbook for the eco-efficient design of NECTA vending machines. Journal of Cleaner
Production 14, 1319–1325 (2006)
20. World Business Council for Sustainable Development (WBCSD),
http://www.wbcsd.org/home.aspx
21. Yang, C.J., Chen, J.L.: Accelerating preliminary eco-innovation design for products that
integrates case-based reasoning and TRIZ method. Journal of Cleaner Production 19,
998–1006 (2011)
22. Yang, C.J., Chen, J.L.: Forecasting the design of eco-products by integrating TRIZ
evolution patterns with CBR and simple LCA methods. Expert Systems with
Applications 39, 2884–2892 (2012)
23. Zhang, J., Shang, J.: Research on developing environmental protection industry based on
TRIZ theory. Procedia Environmental Sciences 2, 1326–1334 (2010)
Principled Ways of Finding, Analysing and Planning
for Communicative Overhead in Interaction Technology
for Fashion Industry
Jason Yang, Sean Rintel, and Stephen Viller
School of ITEE, The University of Queensland, QLD 4072 Australia

{jyang,viller}@itee.uq.edu.au, [email protected]
Abstract. In this paper, we identify and analyse the problems associated with
communicative overheads of a fashion design and manufacturing workflow. We
conduct a multi-stage qualitative study to investigate where the rich multiple
channels of communication afford and constrain the workflow during remote
collaboration. From this study, we define what communication channels we
have in our system. We then use the video data that we had collected through
user testing, together with the feedback from the video-assisted stimulated re-
call interviews, to identify three kinds of communicative overheads in our sys-
tem: novelty of technology, ongoing constraints and operational problems. We
analyse each communicative overhead with examples from our video data, and
conclude that there are various overheads, some that participants may overcome
through learning and familiarity with the system, some that may not be easily
overcome, and some that may not even be unique to a remote collaborative en-
vironment, as they also occur in a face-to-face collaborative environment. We
believe that offering richer communication channels does not necessarily equate
to greater efficiency in the collaborative process, and that designers and devel-
opers of collaborative systems need to investigate the effect of communicative
overheads before introducing new channels to their system, as these overheads
may decrease the efficiency of any collaborative experiences.
Keywords: Coordination, workflow, communicative overhead, channel, remote

collaboration, mash-up technologies, qualitative study.
1 Introduction
As communication technology advances, new communication channels emerge.

These channels can assist coordinating activities, joint problem solving and sharing
artefacts of work, but they also raise the potential for new communicative overheads.
Communicative overhead is anything that participants could account for as extra
communicative work, such as taking more time explain something, increased inci-
dence of repair, or increased levels of meta-communication. In this paper we explore
three sources of communicative overheads: novelty of the technology, ongoing
constraints, and operational problems.
Principled Ways of Finding, Analysing and Planning for Communicative Overhead 485
2 Related Work
The choice of communication channel can be determined by both social presence

theory and media richness theory. Short, Williams and Christie first used the term
‘social presence’ [1] to describe the absence of nonverbal communication which
reduced social presence in mediated communication. As computer-mediated commu-
nication (CMC) has evolved, the social presence is said to become a way for the indi-
viduals to represent themselves and exchange messages, and how those messages are
interpreted by others in an online environment [2]. Media richness theory suggests
that the effectiveness of the communication is dependent on the selection of an appro-
priate medium [3], based on the degree of ambiguity users are willing to accept in any
given communication situation [4]. It is a channel-based approach to the selection of
an appropriate technology for a given situation.
Burgoon proposed the principle of interactivity framework that offers understand-
ing of the interrelationships of verbal and nonverbal cues within CMC [5]. The
principle of interactivity assumes that the degree of interdependent, contingent, parti-
cipative, and synchronous interaction provided by a communication interface or
experienced by interlocutors, or both, will affect the social judgments and task
performance. [5] [6].
According to the 3C collaboration model (communication, coordination and coop-
eration) [7], individuals need to communicate and negotiate with others in order to
make decisions and exchange information. They also need to minimise conflict to
prevent loss of communication while they are coordinating and organising themselves
to work together. “The need for renegotiating and for making decisions about unex-
pected situations that appear during cooperation may demand a new round of com-
munication, which will require coordination to reorganize the tasks to be executed
during cooperation.” [8]. Neale and Carroll interpreted that coordination can some-
times be viewed as overhead or operating cost while completing interactive group
activities [3].
The challenge of supporting effective collaborative group work in a distributed
context depends on the level of communicative overheads that are embedded within
each communication channel. Communicative overheads have an impact upon inte-
ractivity, especially when they are seen as getting in the way of the task. That being
said, it is overly simplistic to equate communicative overheads with a negatively
valanced sense of trouble or that they are necessarily show-stoppers to adoption [9].
In this paper, we are focused on finding, analysing and planning for communica-
tive overheads within a Textual Visual Tangible Multi-touch (TVTM) system [10] in
a remote collaboration setting. The TVTM system consists of two networked multi-
touch tables that present users with a shared common workspace. Each table offers an
array of simple design editing tools, a basic pen tool for writing and annotating, and a
live webcam for demonstrating physical artefacts.
486 J. Yang, S. Rintel, and S. Viller
3 Previous Study
We previously identified different modes of interaction, such as the manipulation of

physical artefacts, as opposed to textual, visual or tangible encoded virtual representa-
tion of these artefacts. We also identified collaboration issues in the textile industry
through interviews and observational studies, in terms of classification of context in a
fashion design environment [11].
Based on our pilot studies of collaboration, and the use of currently available tech-
nologies such as natural user interface (NUI) technology [12] and tangible user inter-
face (TUI) technology [13] [14] in a fashion design environment, we have developed
the TVTM system; a ‘mash-up’ system from a selection of currently available tech-
nologies that will allow us to utilise the multi-touch and tangible aspects of the tech-
nology, which is more effective than a single standalone technology such as video
conferencing [15].
We have also conducted an experiment that required the solving of a design prob-
lem using a traditional fax machine versus our TVTM system between two remote
locations. The results showed our TVTM system was significantly more efficient, and
minimised communicative overheads.
4 Qualitative Study
We have conducted a multi-stage qualitative study to investigate and evaluate the use
of the TVTM system and hardware for supporting remote interaction and collabora-
tion in the fashion industry, and also to find and evaluate communicative overheads
within communication channels. This study also explored the phenomenon of the
gesture-based interaction around the multi-touch gesture technology. For our first
stage, we were evaluating the TVTM software with pairs of design students undertak-
ing collaborative tasks. Stage 2 of the study used the video recording from stage 1 to
explore and review the interactions in more detail in terms of communicative over-
heads. Stage 3 of the study involved a game playing process to explore common
interaction themes in the video data.
4.1 Methods and Settings
We conducted our user testing with a group of female students in their second and
final years of a fashion design course at a local TAFE collage. There were eight par-
ticipants in total and each user testing session was conducted in pairs, giving four
groups in total.
The first stage of the study was to evaluate the participants undertaking a design
critique collaborative task (which is a part of their course assessment). Each
participant was assigned a TVTM system to complete the task. The participants were
shown a short video clip detailing the functionality of the interface, and operating
instructions prior to each user testing session.
The participants were encouraged to “think out loud” during the entire session, and
the interaction between the participants was studied and captured by recording video
during each testing session. At the conclusion of the first stage of the study, we re-
viewed our observation notes together with the recorded footage to create targeted
sets of questions for the video-assisted stimulated recall interviews. This style of in-
terview technique uses video footage to assist participants to recall their experience of
user testing during post-event interviews.
We began conducting the second stage one week after we had completed our stage
1 study. We requested our participants come back for the interviews in the same pairs
as in stage 1 of the study. The entire interview procedure was described to the partici-
pants, followed by their viewing a video clip of their specific user testing session.
While the participants were watching video footage of their participation in the
first stage, we began a series of questions directly related to some of the interactions
that were observed during their user testing session. The majority of these questions
were related to why there were certain interactions with the system or with their part-
ner during the user testing.
We asked questions related to the participants’ thought processes at a specific mo-
ment during the user testing, to obtain a clear understanding of their design (thought)
process while they were critiquing each other’s work. We also asked the participants
to describe their experience during the entire collaborative process, and obtained
feedback regarding the usability of the TVTM system.
4.2 Defining Communication Channels
Our TVTM system offers a mash-up of multiple communication channels. These

channels include an audio channel, a shared screen visual channel, a live webcam
video feed channel, a multi-touch & gesture enabled visual interface channel, a virtual
keyboard input channel and a hand writing/drawing input channel.
4.3 Results – Discovering Overheads

Through observation and analysis of the video data that was obtained, it was deter-
mined that there were three different kinds of overheads. They are the novelty of the
technology, ongoing constraints, and operational problems. We will discuss each kind
of overhead, and relate it to some specific events that were observed in the video data
and the participants’ responses from the video assisted stimulated recall interviews.
Novelty of the Technology

Example 1: Participant A was trying to convey to participant B a fabric image she
wanted for her alternative swimwear design. Participant A was using her finger to tap
on the image thinking it would bring up the image that she selected (Fig. 1), partici-
pant A said “I don’t think it is working”. In the meantime, participant B did not know
what was happening, so participant B asked participant A “what is going on? what
are you trying to do?”.
It appeared as though Participant A was trying to select the image using the tradi-
tional ‘point and click’ method with which she was familiar. During this period, Par-
ticipant B was unaware of the problem that Participant A was experiencing. The in-
terviewer then encouraged Participant A to think out loud so that Participant B could
be aware of the situation.
Participant A: ”I realised that I needed to drag the images out of the image selec-
tion panel in order to bring up the fabric image. I then explained my thought process
to participant B while dragging the image out to the work area.”
Fig. 1. Participants tapped on the screen first, but nothing happened, a few seconds later she
realised she needed to drag the image out
Example 2: Participant C appeared to be exploring the functionality of the TVTM

system as she was repeatedly tapping and browsing on the screen, working through
the available tools to select the most appropriate.
The interviewer asked Participant C to confirm what she was trying to do at that
time.
Participant C: “I did not know which tools I needed so I was trying to find another
way to show participant D my alternative design so she can understand it”
In this particular example, participant C was unable to recall what design tools the
TVTM system offered. Participant C had to click on each design tool in turn, to re-
member the function of each tool, so as to identify the best tool to illustrate the alter-
native design to participant D. While participants are exploring and trying to familia-
rise and learn the system, it causes communication delay/overhead.
Ongoing Constraints
Participant E was browsing through some images, she repeatedly browsing up and
down using her finger on the screen.
Participant E: “Can you see what I am doing?”

Participant F: “Yes I can but I have no idea what you are doing though.”
Participant E continued to browse the images.
Participant F: “What are you doing (Participant E’s name!)”
In this example, participant F was watching participant E’s interaction with the
TVTM system through the shared screen ability of the system. Participant F did
not know what participant E was trying to do, since participant E did not initially
verbalise her thought process. There are many examples in our video data where simi-
lar situations occurred. This is due to the fact that if one of the participants worked
‘individually’ and ‘silently’ (for example searching an image) and did not “think out
loud”, then the other participant had no clue about what is happening. This particular
type of overhead is not unique to this environment using the TVTM system however,
it could also occur with participants in a face-to-face collaborative environment.
Operational Problems (Hardware & software)

Example: Participant I was showing an actual fabric sample over the webcam to par-
ticipant J. The live webcam video feeds were transmitted over to the system. Partici-
pant J was having trouble seeing the live video feed (Fig 2) so she said, “could you
lift it up a bit?”
Fig. 2. Participant I (left) raised the fabric samples up towards the webcam
The interviewer then asked Participant I to clarify the nature of the problem she
was experiencing.
Participant J: “the image on the screen is too small, so I asked participant I to lift it
up so the camera can zoom in a bit”
In this example, there was a hardware and software limitation, as the resolution of
the webcam did not provide sufficient quality of live video feed. Participant J re-
quested participant I to lift the fabric closer to the webcam, so the image would ap-
pear larger, with more fine detail visible. This lost time represents an overhead during
the collaboration.
Novelty can potentially be overcome through training, experience and time. Ongo-
ing constraints and operational problems, which can be either hardware or software
related, can also introduce additional overheads.
5 Discussion
From our previous observational study on a clothing factory, we noticed clothing

designers were still using both telephone and facsimile to communicate with their
sub-contracted manufacturers in China.
Communication by telephone can be classified as ‘mediated interaction’ [5]. It is

leaner than face-to-face communication as it removes all the visual information while
maintaining verbal and audio cues to communication.
Facsimile can be classified as a written channel; for example, clothing designers
were able to send out a design alteration specification quickly over to the overseas
manufacturer, especially when there is a time difference between the two remote loca-
tions. However, the facsimile communication channel produces overheads. As we
showed in our previous study, it was not efficient in a situation like collaborating on a
problem-solving task in two remote locations.
The communicative overhead inherent with a facsimile communication channel is
primarily due to asynchronous communication. Facsimile also lacks important proper-
ties, such as the ability to manipulate text and graphic images [16]. From our inter-
view with the clothing designers, they usually received a facsimile response from the
remote manufacturer within normal business hours. Sometimes they sent and received
facsimiles that contained the same graphic images that had previously been facsi-
miled, along with additional hand written notes, or annotations around the images on
the original facsimile.
In this study, we have attempted to identify the communicative overheads asso-
ciated with the deployment of a mash-up of rich multiple channels within our TVTM
system. This mash-up of multiple channels provides both affordance and constraint to
the workflow. In this context, the richer communication channels afford greater effi-
ciency and interactions for remote collaboration, whereas the communicative over-
head is the constraint.
In real world settings, people who engage in collaboration apply various channels
of interactions to cope with these complexities of workflow; in most cases, overhead
or operating cost is involved [1, 3]. If the group collaboration suffers from a high
level of overheads, it is possible that the time lost due to these overheads may not be
regained by the use of the richer communication channels, leading to a potentially less
efficient collaborative experience, and possible abandonment of the joint activities.
Designers and developers need to consider that the notion of overheads is related to
the notion of cognitive overload, which can be identified by tracking participants’
thought processes while they are completing a task.
New channels may not necessarily create overheads or eliminate existing over-
heads. From the results of our study, we identified an overhead not directly attributa-
ble to the use of any specific channel, which may equally occur during face-to-face or
remote collaboration.
Designers and developers of CMC systems can bring new channels into use for
particular tasks, however overheads have an influence on how effective new channels
are for users working remotely. Therefore they should not necessarily expect an im-
mediate improvement to workflow.
We conclude that our study has shown that offering richer communication chan-
nels does not necessarily equate to greater efficiency in the collaborative process.
References
1. Short, J., Williams, E., Christie, B.: The social psychology of telecommunications, 195p.
Wiley, London (1976)
2. Lowenthal, P.R.: The evolution and influence of social presence theory on online learning.
Online education and adult learning: New frontiers for teaching practices. IGI Global, Her-
shey (2009)
3. Neale, D.C., Carroll, J.M., Rosson, M.B.: Evaluating computer-supported cooperative
work: models and frameworks. In: Proceedings of the 2004 ACM Conference on Comput-
er Supported Cooperative Work. ACM (2004)
4. Daft, R.L., Lengel, R.H.: Organizational information requirements, media richness and
structural design. Management Science 32(5), 554–571 (1986)
5. Burgoon, J.K., Bonito, J.A., Ramirez, A., Dunbar, N.E., Kam, K., Fischer, J.: Testing the
interactivity principle: Effects of mediation, propinquity, and verbal and nonverbal modali-
ties in interpersonal interaction. Journal of Communication 52(3), 657–677 (2006)
6. Walther, J.B., Gay, G., Hancock, J.T.: How do communication and technology researchers
study the internet? Journal of Communication 55(3), 632–657 (2006)
7. Ellis, C.A., Gibbs, S.J., Rein, G.: Groupware: some issues and experiences. Communica-
tions of the ACM 34(1), 39–58 (1991)
8. Fuks, H., Alberto, B.R., Marco, A.G., Lucena, C.J.: Applying the 3C model to groupware
development. International Journal of Cooperative Information Systems 14(02n03),
299–328 (2005)
9. Rintel, E.S.: Conversational management of network trouble perturbations in personal vi-
deoconferencing. In: Proceedings of the 22nd Conference of the Computer-Human Interac-
tion Special Interest Group of Australia on Computer-Human Interaction. ACM (2010)
10. Yang, J., Muhlberger, R., Viller, S.: TVTM: A case study and analysis of 3 virtual
representations to support remote collaboration within the fashion industry. In: 2010,
School of Information Technology and Electrical Engineering (ITEE). The University of
Queensland,
11. Yang, J., Dekker, A., Muhlberger, R., Viller, S.: Exploring virtual representations of phys-
ical artefacts in a multi-touch clothing design collaboration system. In: Proceedings of the
21st Annual Conference of the Australian Computer-Human Interaction Special Interest
Group: Design: Open 24/7. ACM (2009)
12. Jacob, R.J., Girouard, A., Hirshfield, L.M., Horn, M.S., Shaer, O., Solovey, E.T.,
Zigelbaum, J.: Reality-based interaction: a framework for post-WIMP interfaces. In:
CHI-Conference 2008. Association for Computing Machinery Inc. (2008)
13. Hinckley, K., Pausch, R., Proffitt, D., Kassell, N.F.: Two-handed virtual manipulation.
ACM Transactions on Computer-Human Interaction (TOCHI) 5(3), 260–302 (1998)
14. Patten, J., Ishii, H., Hines, J., Pangaro, G.: Sensetable: a wireless object tracking platform
for tangible user interfaces. In: Proceedings of the SIGCHI Conference on Human Factors
in Computing Systems. ACM (2001)
15. Yang, J., Viller, S., Rintel, S.: Outsourcing: Mashing up design methods and technologies
in the fashion industry. In: Proceedings: PIN-C 2012 (2012)
16. Sproull, L., Kiesler, S.: Connections: New ways of working in the networked organization.
MIT press (1992)
Continuing On-Premise or Adopt On-Demand?
An Empirical Study of ERP Adoption in SMEs
Fan Zhao and Elias Kirche
Florida Gulf Coast University, FL, USA

{fzhao,ekirche}@fgcu.edu
Abstract. Vendors are now expanding into the small to mid-size organizational
markets with simplified, less-risk, less-reward systems However, SMEs, who
adopted ERP on-premise solutions before, are now facing a dilemma: continuing
with ERP on-premise upgrades or switch to on-demand solutions. This study sur-
veyed Chief Information Officers of SMEs with respect to indicators of ERP
adoptions. Cost, reduced demand for own IT resources, outage/accessibility and
performance were found to be the most critical and important factors to assess
ERP adoptions for SMEs.
Keywords: ERP, Cloud Computing, Adoption, Upgrade, on-demand.
1 Introduction
ERP solutions have evolved from reorder point and materials management focused
systems (ROP and MRP) based on mainframes to responsive and integrated resource
planning systems based on client-server architectures and web platforms. Through
data standardization and process integration, ERP systems have the potential to facili-
tate communications and co-ordination, enable the centralization of administrative
activities, reduce IS maintenance costs and increase the ability to deploy new IS func-
tionality [11]. ERP systems are adopted for strategic, technical, and operational rea-
sons including: providing an integrated enterprise-wide application with real-time
data access available across the entire organization, simplifying and standardizing
systems and business processes, and replacing legacy systems. When they are well
implemented, ERP systems are able to bring operational, managerial, strategic, infor-
mation technology (IT) infrastructure and operational benefits to their customers [25].
The motivations which have become more dominant include pressure to keep up with
competitors on a global scale, mergers and acquisitions and the need for restructuring,
and combining and integrating business processes. ERP systems have spread rapidly
among organizations. According to Gartner [10], worldwide spending on enterprise
application software will total $120.4 billion in 2012, with ERP the largest application
software market expected to reach $24.9 billion.
The fast changes in technologies during the last decade dramatically altered the na-
ture of the demand requiring more reliable services, rapid customization and on time
delivery of goods and services, forcing many small and medium-sized enterprises
(SMEs) to adjust their business strategies and to adopt a better technology to increase
Continuing On-Premise or Adopt On-Demand? An Empirical Study of ERP Adoption 493
the productivity of business processes. For SMEs, the advantages gained by adopting
ERP systems are obvious. Gattiker and Goodhue [11] describe the ERP benefits in
following aspects: (1) improving the integration of information flow between sub-
units; (2) centralizing the administrative activities, such as accounts payable and pay-
roll; (3) reducing costs of system maintenance; (4) increasing the ability to deploy
new IS functionality; (5) enabling transformation from inefficient business processes
to accepted best practices. With these insights, an ERP system seems to be one logical
alternative to be pursued by SMEs [21]. And with the majority of large companies
having already implemented ERP systems, today ERP vendors, such as SAP and
Oracle, are shifting their focus towards SMEs with scaled-down (and less expensive)
versions of ERP solutions. Other important reasons have encouraged the interest of
ERP vendors towards SMEs. These include the need for SMEs to be integrated into
the supply chain at risk of becoming less competitive, the large market represented by
the number of SMEs (90 percent of all businesses in the USA are SMEs, Sledgia-
nowski et al. [26]) compared to the number of large enterprises, and the advancement
in technologies with corresponding decrease in hardware costs. Another factor in
support of ERP systems is that often information visibility is an important aspect in
supply chain performance and the lack of an adequate technology will debilitate the
SME’s ability to adapt to changes imposed by customers or suppliers with unfavora-
ble impact in its competitiveness. However, the use of commodity software, such as
ERP systems, may force a more rigid structure on a SME and thus weaken its flexibil-
ity and agility to adapt as a competitive advantage. Therefore, it is critical to SMEs
to investigate their performance after the implementation of ERP systems.
In recent years, after many SMEs adopted various ERP packages from different
ERP vendors, a new technology model, called Cloud Computing, became possible for
sourcing ERP systems with the potential to have a significant impact for SMEs’ com-
petitiveness. Cloud computing is an emerging business and technology concept to
support an on-demand delivery of computing, storage and applications over the Inter-
net. A recent IDC report shows global revenue in cloud software market reaching
$22.9 billion and it will grow to $67.3 billion in 2016 [18]. This projection includes
revenue generated by the shift from on-premise to on-demand providers as well as by
the planning and architecture behind the shift.
Cloud computing is a model for enabling convenient, on-demand network access to
a shared pool of applications and resources and may be offered in one of the follow-
ing service models [4, 16, 17, 28, & 29]:
• Software as a service (SaaS): providing software subscription services

• Storage as a service: providing remote storage resource services
• Database as a service: providing remotely hosted database services
• Information as a service: providing remotely hosted information services
• Process as a service: providing business processes based on remote resources
• Testing as a service: providing testing services for local or remote systems
• Platform as a service (PaaS): providing a complete platform to support application
development, interface development, database development, storage, information
and testing
• Infrastructure as a service (IaaS): providing a service to access computing re-
sources remotely
494 F. Zhao and E. Kirche
• Security as a service: providing core security services remotely over the Internet
• Integration as a service: providing a complete integration stack service
Specifically for ERP systems, Beaubouef [2] indicated that are three cloud models for
adoption:
• Software as a service (SaaS): a subscription model for small customers who share
hardware.
• Hosted ERP: a typical solution for large customers who have separate hardware
and instances.
• Hybrid ERP: a combination solution that maintains on-premise software as well as
integrated a degree of on-demand services.
As high-speed internet has become more accessible to businesses, cloud models have
helped businesses with limited IT resources take advantage of technology to improve
business processes. Typically, a firm would rent licenses or access the software from
an application service provider (ASP) that actually runs on servers or devices owned
and maintained by the ASP. These large datacenter facilities with redundant layers
of power and data security are often too expensive for SMEs to have under the on-
premise model. Also, the ASP is generally responsible for maintaining and updating
the software, and often includes some level of support for users in the monthly fee.
This model has certain benefits for SMEs that can attenuate some of the problems
associated with maintaining purchased software in-house. For example, SaaS allows
ASP to maintain their offerings consistently by automating testing, monitoring, main-
tenance and upgrades without sending out constant updates that need to be applied by
end users. Also, SaaS allows smaller firms with limited (or no) IT staff to benefit
from the economies of scale and efficiencies implemented by the ASP. Equally im-
portant, SaaS allows companies to pay to use the software they need, without making
a huge investment in IT infrastructure for servers, software, etc., by "renting" access
to what they need and paying monthly, quarterly or annually.
Additionally, according to Gartner’s report [1], more and more companies are
considering on-demand services in different applications (Figure 1).
Fig. 1. How SaaS Is Developed by Application

Consequently, SMEs with existing ERP on-premise packages will be facing a new
challenge when a new version of the on-premise ERP is released or when they require
a substantial IT infrastructure investment. The question for management is whether
it will decide in favor of the ERP upgrade or switch to an on-demand ERP. To date,
there is no empirical research that has been done to evaluate and compare the two
options for SMEs. Therefore, the purpose of this study is to investigate the on-premise
and on-demand options for SMEs sourcing ERP solutions and summarize the key
factors influencing decision making of the ERP future solutions among CIOs and IT
managers.
2 Theoretical Background
The primary purpose of ERP upgrades is to take advantages of new technologies and
business strategies to ensure that the organization keeps up with the latest business
development trends. Therefore, the decision to upgrade ERP is usually not driven by
code deterioration or anticipated reduction in maintenance costs alone. According to
an AMR research [27], 55% of upgrades were voluntary business improvements trig-
gered by the need for new functionality, expansion or consolidation of systems; 24%
of upgrades were triggered by technology stack changes; 15% of upgrades were
forced by de-support of the running version of software to avoid vendor support ter-
mination [7]; and 6% of upgrades were triggered by bug fixes or statutory changes.
The cost of ERP upgrades is high [20]. Swanton [27] mentioned that the cost of each
upgrade includes: 50% of the original software license fee and 20% of the original
implementation cost per user, which means over $6 million dollars for a 5,000-user
system. Typically, each ERP upgrade requires eight to nine months of effort with a
team the equivalent of one full-time employee per 35 business users. The ERP-
adopting organization does not have to develop and re-write the ERP system itself but
rather it replaces (or upgrades) the old version with a readily available new version
from the ERP vendor. However, a lack of experience may cause the costs and length
of the upgrade project to approach or even exceed those of the original ERP imple-
mentation effort. Every three years, a major ERP upgrade and several small upgrades
are typically needed to keep the system running smoothly. Organizations will spend a
significant amount of money on each ERP upgrade project, which may place the sys-
tem out of reach (or postponed) for many SMEs with limited financial resources and
shortage of skilled personnel.
However, cloud computing supports an on-demand delivery of ERP over the Inter-
net with competitive implications for SMEs. Typically, ERP functions will be served
out of the cloud, and the license fees and/or maintenance fees will be charged based
on the use of the software, such as number of users and ERP functions/modules.
Companies will no long need to purchase the software and to keep IT staff, which will
make the initial investigation of EPR less expensive. This means SMEs now can
access the same technology until recently only available to larger firms, and provide
the same or superior level of services than larger competitors with the on-premise
ERP classic model. According to Gheorghe and Lupasc [12], cloud ERP solution
will save about 30% project time after analyzing the critical path of both classic and
cloud ERP solutions. Table 1 summarizes the benefits derived from cloud computing
ERP systems.
Table 1. Benefits of On-demand ERP solutions
Benefits of on-demand ERP Reference

Short implementation cycles [12]; [13];
Low entry costs [19]; [13]; [3]
Reduced demand for own IT resources [13]; [3]
Elasticity/Flexibility of services [19]; [13]; [23]
Scalability [19]; [10]
Focus on core business [13]
However, there are also several challenging issues for the adopters of on-demand
ERP, and cost is always an important one. Although companies can adopt on-
demand ERP solutions at a much lower initial cost and lower total cost of ownership
comparing with the traditional ERP solutions, the on-going monthly or annual pay-
ment will result in a higher budget cost because of the subscription payment method
[22]. Therefore, the subscription method is of critical importance for this technology
model to be sucessfuly adopted and widely accepted by SMEs. Table 2 summarizes
the challenging issues derived from cloud computing ERP systems. As more com-
panies who adopted on-premise ERP solutions are considering on-demand solutions, a
research study is needed to give guidelines to SMEs when they are faced with this
dilemma.
Table 2. Challenging Issues of On-demand ERP solutions
Challenging Issues of on-demand Reference

ERP
Outage/Accessibility [15]; [22]
Security [15]
Performance [15]
Integration [15]; [22]
Cost [15]; [22]
Control of the system [22]
Flexibility of customization [22]
This study seeks to provide a comprehensive understanding of two ERP solutions:

on-premise and on-demand, collecting CIOs’ perspectives in different SMEs accord-
ing to their experiences from their organizations’ adoptions of ERP solutions.
3 Research Methodology
The survey questionnaire was sent to CIO/IT managers in 272 SMEs, who are cur-
rently using on-premise ERP solutions. Company size is measured by the number of
employees [6]. The objective is to conduct an analysis to confirm the underlying
structure of the factors involved in the on-demand and on-premise ERP adoption de-
cisions. In this study, an SME is one with fewer than 500 employees. A total 63
survey responses were obtained after removing 7 incomplete questionnaires. The
distribution of the respondents was as follows: gender (male: 76%, female: 24%),
current EPR systems in use come from many vendors, such as SAP, Microsoft, Sy-
spro, NetSuite, etc. The number of employees range from 36 to 483 (64% small com-
panies and 36% mid-size companies). A wide variety of industries were represented
in the responses (shown in Table 3).
Table 3. Industry Summary
Industry Sector Number of companies

Manufacturing 18
Information technology 19
Retail 5
Health Services 16
Insurance 5
Total 63
To substantiate our measurement model for the factors identified from the litera-
ture (Tables 1 and 2), content validity was evaluated by a group of experts. Six ex-
perts from two universities and four organizations were asked to read the survey and
decide whether in their opinion it measured what its name suggested. The survey
questionnaires were improved according to comments from the experts. The final
survey was returned to the experts and approved by all. The coefficient alpha [8] to
measure the internal consistency reliability of the survey instrument ranged from 0.73
to 0.85, indicating good reliability according to the acceptable level (0.7) suggested
by Bryman [5]. An item-to-total correlation was conducted to evaluate the convergent
validity. The literature suggests that a measurement variable may be considered for
removal if the item-to-total correlation is less than 0.4. One question was removed
because the item-to-total correlation was lower than 0.4. Construct validity is one of
the most significant scientific advances of modern measurement theory and practice
[14]. It measures how well the instrument measures the concept of interest. Confir-
matory factor analysis (CFA) was conducted to examine construct validity. All but
one construct that emerged from the factor analysis showed high construct validity.
The overall results of the analysis confirmed the factors in the ERP adoption solutions
presented in the literature. Additionally, the survey asked the respondents to rank
the factors in order of importance. The results are explained in the next section.
Survey results focused on factors involved at ERP adoptions between on-premise and
on-demand solutions. For each factor, a brief description and a 5-level rating scale
ranging from extremely critical and important in ERP adoption evaluation to neither
critical nor important in ERP adoption selections was provided. The numerical anc-
hors for the 5-level rating scale are indicated in Table 4.
Table 4. Survey Averages by Factor
FACTOR AVERAGE (1 – 5 scale)

Cost 4.93
Reduced demand for own IT resources 4.51
Outage/Accessibility 4.46
Performance 4.07
Control of the system 3.85
Elasticity/Flexibility of services 3.78
Short implementation cycles 3.49
Security 3.24
Scalability 3.03
Integration 2.70
Flexibility of customization 2.46
Focus on core business 2.23
According to the average rating in Table 4, the results indicate that there are top
four factors, which is extremely critical and important in ERP adoptions for SMEs,
identified by CIOs facing to the options of on-premise or on-demand solutions: cost,
reduced demand for own IT resources, outage/accessibility and performance. The
second group, which is critical and important, includes five factors: control of the
system, elasticity/flexibility of services, short implementation cycles, security and
scalability. The rest of the factors, integration, flexibility of customization, and focus
on core business, receive low rating from CIOs.
There are several important results summarized from data. First, besides the three
factors (cost, reduced demand for own IT resources, and performance) that are
commonly regarded as critical success factors in ERP selection, outage/accessibility
is rated as one of the highest factors. It is obvious that system outage is closely
correlated with on-demand adoptions and may represent a high concern for SMEs
with intentions of adopting this ERP solution. In fact, after the outage issue of
Microsoft Azure, CIOs realized the importance of system accessiblity for on-demand
solutions. Second, security was emphasized by several researchers in their studies of
cloud computing [15, 24]. However, according to our data, CIOs only rated security
in the second importance group. Third, most of the small companies prefer adopting
on-demand solutions while mid-size companies are more hesitante to switch from
current system to on-demand ERP.
The results of this study provides SMEs with valuable knowledge that might
prompt them to make significant changes in their current ERP-related processes and
the acquisition of enterprise packaged software, which could result in substantial ben-
efits. The findings have important theoretical and practical implications. Our under-
standing of the ERP related issues in SMEs helps both researchers and practitioners
understand how ERP system adoption and implementation impact the competitiveness
of SMEs. To researchers, this study shows the relationship between on-premise and
on-demand ERP adoptions and give the possibility in future research to expand the
model. To practitioners, this study shows some critical issues of ERP adoption for
SMEs. It allows CIOs and other IT managers to think thoroughly before they make
decisions in ERP adoptions. Additionally, guidelines are provided to IT managers
regarding the ERP adoption issue between on-premise and on-demand models. Fur-
thermore, the ERP vendors will benefit from this study in developing better ERP solu-
tions tailored to SMEs specific requirements, and improve the cost structure of their
cloud models for higher adoption rates among SMEs.
References
1. Anderson, E., Britz, B., Bell, W.: User Survey Analysis: Using Cloud Services for Mis-
sion-Critical Applications (September 28, 2012),
http://www.gartner.com/id=2182216 (retrieved February 3, 2013)
2. Beaubouef, B.: Cloud ERP – New Dog, Same Fleas (May 22, 2011),
http://gbeaubouef.wordpress.com/2011/05/22/cloud-erp/ (retrieved
February 5, 2013)
3. Bhardwaj, S., Jain, L., Jain, S.: Cloud Computing: A Study of Infrastructure as a Service
(IaaS). International Journal of Engineering and Information Technology 2(1), 60–63
(2010)
4. Branzei, O., Vertinsky, I.: Strategic pathways to product innovation capabilities in SMEs.
Journal of Business Venturing 21, 75–105 (2006)
5. Bryman, A.: Social Research Methods. Oxford University Press, New York (2004)
6. Chow, S., Holden, R.: Toward an understanding of loyalty: The moderating role of trust.
Journal of Managerial Issues 9(3), 275–298 (1997)
7. Craig, R.: Laurier enterprise system upgrade. In: Proceedings of International Conference
of Information Systems, Charlotte, pp. 654–662 (1999)
8. Cronbach, L.: Coefficient alpha and the internal structure of tests. Psychometrika 16,
297–334 (1951)
9. Furht, B., Escalante, A.: Handbook of Cloud Computing. Springer Science + Business
Media, New York (2010)
10. Gartner, Inc. Stanford, Connecticut, USA. Press release: Global IT Spending Forecast
(June 20, 2012), http://www.gartner.com/newsroom/id/2055215 (retrieved)
11. Gattiker, T., Goodhue, D.: Understanding the plant level costs and benefits of ERP: will
the ugly duckling always turn into a swan. In: Proceedings of the 33rd Hawaii Internation-
al Conference on System Science, Hawaii (2000)
12. Gheorghe, G., Lupasc, I.: Crises Solutions for SMEs Case Study: ERP Cloud Vs. Classic
Solution. In: International Conference “Risk in Contemporary Economy” Galati, Romania,
SIIIth edn., pp. 149–156 (2012)
13. Karabek, M., Kleinert, J., Pohl, A.: Cloud Services for SMEs - Evolution or Revolution?
Business + Innovation, 26–33 (2011)
14. Kerlinger, F.: Foundations of Behavioral Research. Rinehart and Winson, Fort Worth
(1986)
15. Kim, W.: Cloud Computing: Today and Tomorrow. Journal of Object Technology (Janu-
ary/February 2009)
16. Kim, W., Kim, S., Lee, E., Lee, S.: Adoption Issues for Cloud Computing. In: Proceedings
of the 7th International Conference on Advances in Mobile Computing and Multimedia,
pp. 2–5. ACM, New York (2009)
17. Linthicum, D.: Cloud Computing and SOA Convergence in Your Enterprise. Addison
Wesley, Boston (2010)
18. Mahowald, R., Sullivan, C.: Worldwide Saas and Cloud Software 2012-2016 Forecast and
2011 Vendor Shares. International Data Corporation (IDC) (2012),
http://www.idc.com/getdoc.jsp?containerId=
236184#.USTXoh3kseB (retrieved)
19. Marston, S., Li, Z., Bandyopadhyay, S., Zhang, J., Ghalsasi, A.: Cloud computing - The
business perspective. Decision Support Systems 51(1), 176–189 (2011)
20. Montgomery, N.: Build your business case for upgrades by adding functionality. Computer
Weekly, 16 (2004)
21. Olson, K., Satre, P.: ERP for SMEs – is proprietary software an alternative? Business
Process Management 13(3), 379–389 (2007)
22. Panorama Consulting Group.: SaaS ERP vs. on-premise ERP software: Six key differentia-
tors (March 12, 2009),
http://searchcloudapplications.techtarget.com/news/1350684/
SaaS-ERP-vs-on-premise-ERP-software-Six-key-differentiators
(retrieved February 20, 2013)
23. Rittinghouse, J., Ransome, J.: Cloud Computing: Implementation, Management, and Secu-
rity. CRC Press (2009)
24. Saeed, I., Juell-Skielse, G., Uppstrom, E.: Cloud Enterprise Resource Planning Adotions:
Motives & Barriers. In: International Conference on Research and Practical Issues of En-
terprise Information Systems (2011)
25. Shang, S., Seddon, P.: A comprehensive framework for classifying the benefits of ERP
systems. In: Proceedings of the Americas Conference on Information Systems, Long
Beach, CA (2000)
26. Sledgianowski, D., Tafti, M., Kierstead, J.: SME ERP system sourcing strategies: a case
study. Industrial Management and Data Systems 108(4), 421–436 (2008)
27. Swanton, B.: Build ERP upgrade costs into the business change program – not the IT
budget. Computer Weekly, 28–28 (September 21, 2004 )
28. Velte, A., Velte, T., Elsenpeter, R.: Cloud Computing, A Practical Approach. McGraw-
Hill (2010)
29. Wang, L., von Laszewski, G., Younge, A., He, X., Kunze, M., Tao, J., Fu, C.: Cloud
Computing: A Perspective Study. New Gneration Computing 28, 137–146 (2010)
Author Index
Abdelnour-Nocera, José III-146 Barroso, João V-39

Abed, Mourad V-141 Beldad, Ardion III-371
Abou Khaled, Omar IV-157, IV-167, Bell, Carrie III-285
V-19 Bellini, Pierfrancesco III-259
Acharya, Subrata II-26, III-3 Belluati, Maurizio III-277
Ackermann, Daniel III-446 Benetti, Angelo II-428
Ahn, Jong-gil II-249, III-87 Benford, Steve II-376
Akihiro, Ogino V-372 Bengler, Klaus II-578, II-596
Akiyoshi, Masanori I-159, II-444 Bergmann, Francine B. III-117
Alexandris, Christina IV-3, IV-13 Bergmans, Anne II-541
Alfredson, Jens I-221 Bernhaupt, Regina I-521
Almutairi, Badr IV-23 Bevan, Nigel I-281
Altaboli, Ahamed I-549 Beyerer, Jürgen IV-252
Alves, Aline Silva III-324 Bilgen, Semih I-310
Alves, Paulo Henrique Cardoso II-3 Billman, Dorrit I-193
Amandi, Analı́a A. III-107 Bim, Sı́lvia Amélia I-3
Amer, Hossam III-352 Blezinger, David V-10
Amin, Rahul III-97 Böck, Ronald V-301, V-381
An, Kwang-Ok V-3 Borsci, Simone I-166, I-203
Anastasiou, Dimitra IV-32 Borum, Nanna III-418
Andujar, Marvin II-335 Boscarioli, Clodis I-3
Angelini, Leonardo II-531, V-19 Bourimi, Mohamed III-39
Angulo, Julio III-10 Boy, Guy Andre IV-629
Ann, Sangmin V-441 Brat, Guillaume I-290
Arantes, Ana Cláudia Costa II-370
Braun, Andreas IV-147
Ariya, Kanako I-275
Braun, Roman III-446
Armentano, Marcelo G. III-107
Brejcha, Jan I-13
Arning, Katrin III-49
Bröhl, Christina III-127
Arnold, Todd I-593
Brooks, Anthony L. III-418
Arya, Ali IV-215
Brouwers, Aurélie III-136
Asaba, Nobutake I-565
Brunk, Sören V-46
Atia, Ayman II-561
Brunner, Simon IV-388
Attard, Judie V-122
Bruno, Ivan III-259
Burghardt, Manuel I-176
Bach, Cédric I-521
Burzacca, Paolo I-241
Badioze Zaman, Halimah IV-523
Bützler, Jennifer III-127
Baguma, Rehema III-249
Bahn, Sangwoo IV-561, IV-594, IV-610,
IV-618 Cagiltay, Nergiz E. I-310
Baranak, Andrew III-285 Câmara, Marcela IV-439
Barbé, Jérôme IV-429 Campos, José Creissac I-421
Barbosa, Simone Diniz Junqueira I-3, Canedo, Arquimedes I-350
I-460, IV-439 Cantú, Andrea II-316
Barnett, Julie I-166 Caon, Maurizio II-531, IV-167, V-19
502 Author Index
Cardozo, Aline Teodosio dos Santos Crowe, John II-36, II-189

II-370 Cruciol, Leonardo L.B.V. IV-648
Carrino, Francesco II-531, IV-157 Cusseau, Guillaume II-46
Carrino, Stefano II-531, IV-167
Caruso, Mario IV-637 Dahlbäck, Nils II-586
Catarci, Tiziana IV-637 Das, Amitava V-310
Cederholm, Henrik V-403 Davaasuren, Enkhbat IV-196
Cerrato-Pargman, Teresa II-464 David, Bertrand V-29
Chakraborty, Joyram II-13 de Abreu Braz, Priscilla Fonseca IV-439
Chalon, René V-29 de Araújo, Fábio Rodrigo Lopes II-3
Chamberlain, Paul I-22 Debattista, Jeremy V-122
Chan, Susy III-183 de Carvalho, Alfredo Veiga II-3
Chang, Hsien-Tsung III-359, IV-177 Dehlinger, Josh II-66
Chang, Hsin-Chang II-606, II-654 Delomier, Florent V-29
Chang, Tsen-Yao III-237 de Lucena, Carlos Alberto Pereira
Charoenpit, Saromporn II-343 II-127, II-144, II-370
Chen, C.L. Philip V-236 de Lucena, Carlos José Pereira
Chen, Huan-Ting IV-177 II-3, II-370
Chen, Po Chun V-132 De Marsico, Maria II-351
Chen, Wei-Ju II-20 Dempster, Peter III-269
Chen, Weiqin IV-186 Deng, Xiaoming V-227
Chen, Xiuli I-193 de Oliveira, Helvecio Siqueira II-428
Cheng, Cheng IV-243 de Oliveira, João Batista S. III-117
Cheng, Mingkai IV-243 de Oliveira, Káthia Marçal I-211
Cheng, Yun-Maw II-20 DePalo, Philip II-56
Cheung, Yiu-ming V-246 de Souza, Clarisse Sieckenius I-320
Chien, Tsung-Tien IV-177 Dey, Anind K. V-92
Ching, Joanna Sin Sie II-181 Dharma, Anak Agung Gede IV-568,
Chizari, Mahmoud V-262 IV-578
Cho, HyunKyoung I-32 Di Bitonto, Pierpaolo II-484
Cho, Yongjoo V-339 Dicke, Christina II-551
Choh, Ikuro IV-348 Di Giovanni, Pasquale III-342
Choi, Heesun IV-618 Dixon, Jeremy II-66
Christensen, Line G. III-418 Dixon, Shannan DeLany II-66
Chu, Chi Nung II-386 Dörner, Ralf I-381
Chu, Kimberly IV-750 Drury, Jill L. IV-658
Chuang, Ming-Chuen I-540 Duan, Fuqing V-206, V-216, V-280
Chung, Byung Do III-402 Duong, Anh-Duc IV-513
Chung, Ching-Wen III-199
Chung, Donghun II-258 Ebert, Achim I-371
Cincotti, Febo IV-637 Ebisawa, Yoshinobu IV-205
Cirilo, Elder José Reioli II-3 Edirisinghe, Chamari IV-601
Clamann, Michael IV-551 Ehrhart, Stefan I-371
C. Mac, Khoi-Nguyen IV-67 Eibl, Maximilian IV-706, V-196
Coats, Brian II-26 Ekandem, Josh I. II-335
Coleti, Thiago Adriano I-184 Engel, Jürgen I-300, I-401
Corbett, Brendan IV-561 Eriksson, Jeanette V-403
Craig, Claire I-22 Erturan, Yusuf Nasuh I-310
Craven, Michael P. II-36, II-189 Esat, Ibrahim II-76, V-262
Crebolder, Jacquelyn I-109 Eshraghi, Mona II-76
Author Index 503
Eshraghi, Saba II-76, V-262 Göbel, Matthias I-61

Eskildsen, Søren II-361 Gomez, Randy IV-408
Espinoza, Matı́as II-299 Gonçalves, Gil IV-678
Etemad, S. Ali IV-215 Gong, Yang II-94
Euman, Rob III-146 Gopalakrishnakone, Ponnampalam
IV-601
Fain, Brad III-285 Goto, Yukio IV-722
Fallman, Daniel I-128 Gotoda, Naka II-84
Fang, Xiaowen II-150, III-183 Gotoh, Takuto II-216
Farhadi-Niaki, Farzin IV-215 Gotoh, Tomomi IV-235
Faria, Jose V-39 Greene, Kristen K. IV-449
Feary, Michael I-193 Grinsell, Jon II-46
Federci, Stefano I-203 Gris, Iván IV-97
Fekry, Mohamed II-561 Grønli, Tor-Morten V-56
Feng, Shih-Yi IV-177 Gruss, Sascha V-474
Fernandes, Hugo V-39 Grüter, Barbara II-323
Ferreira, Juliana Jansen I-320 Gu, Zhenyu I-555
Ferreira, Sérgio IV-678 Guercio, Elena III-277
Fischer, Arthur IV-147 Guo, Ping V-236
Fleischmann, Albert I-330, III-456 Guo, Weimin IV-243
Forbrig, Peter I-51, I-300, I-340
Förster, Ulrich III-446 Hachimura, Kozaburo I-611
Frajhof, Leonardo II-127, II-370 Hamdy, Aya II-561
Franke, Thomas II-612 Hammer, Jan Hendrik IV-252
Fritsch, Lothar III-68 Han, Chia Yung IV-291
Froehlich, Elaine IV-668 Handschuh, Siegfried V-122
Fuchs, Christian IV-678 Hansen, Jarle V-56
Fuglerud, Kristin Skeide I-41 Harata, Miho V-319
Fukumoto, Kiyotaka IV-205 Harley, Linda III-285
Funke, Alexandra V-46 Harris, Helen II-571
Furuhashi, Takeshi IV-740 Harun, Afdallyna II-376
Furukawa, Hiroshi III-156 Hasegawa, Koyo II-647
Hashiyama, Tomonori I-480
Gabbard, Joseph L. III-217 Hashizume, Ayako I-68
Gabillon, Yoann I-211 Hattori, Masatsugu IV-462
Gadahad, Pallavi Rao II-181 Haubner, Nadia I-381
Gailliot, Ava IV-135 Hayakawa, Seiji I-77, I-119, I-137, I-500,
Galatas, Georgios IV-43 I-531
Galindo, Michel I-51 Hayami, Takehito IV-312, IV-320,
Gambäck, Björn V-310 IV-330
Gao, Dekun IV-225 Hayashi, Kazuya IV-320
Gao, Yue V-153 Hayashi, Yugo III-20
Garcia, Franco Eusébio II-229 He, Ning V-153, V-216
Germanakos, Panagiotis II-418 Heikkinen, Kari II-239, V-431
Ghedira, Khaled V-141 Heinroth, Tobias III-59
Ghinea, Gheorghita V-56 Helmbrecht, Magnus II-578
Gilbert, Juan E. II-335, III-97 Henschen, Lawrence J. II-107, IV-688
Giménez, Rafael I-256 Herberhold, Carlotta III-381
Go, Kentaro I-77, I-119, I-137, I-500, Herdin, Christian I-300, I-401
I-531, IV-456 Hermann, Fabian I-256, III-29, V-122
504 Author Index
Hermann, Sven I-350 Iwata, Mitsuru I-480

Hernández Rubio, Erika IV-301 Iwata, Naoko I-146
Heupel, Marcel III-39 Izumi, Tomoko III-295, V-329
Hijikata, Yoshinori III-314
Hilborn, Olle V-403 Jackson, France III-97
Himmel, Simon III-49 Jakus, Grega II-551
Hiramatsu, Yuko II-398 Jander, Hans I-221
Hirata, Ichiro I-361 Jeng, Tay Sheng V-132
Hishina, Masateru I-565 Jeon, Myounghoon IV-49
Ho, Yi-Lun III-359 Jeon, Wooram IV-594
Hölscher, Christoph V-10 Jepsen, Henrik W. III-418
Hori, Maiya V-391 Jerg-Bretzke, Lucia V-474
Horne, Sara II-76 Ji, Yong Gu I-103
Hörold, Stephan I-85, I-391 Jiang, Yonghan IV-243
Howes, Andrew I-193 Jin, Lianwen V-254
Hrabal, David V-474 Jochems, Nicole III-127
Hsieh, Min-Chih III-166 Johnson, Chris II-20
Hua, Lei II-94 Johnson, Michael D. II-408
Huang, Chi-Yu III-199 Jokela, Timo II-101
Huang, Lixiao IV-561 Jokinen, Kristiina IV-32, IV-262
Huang, Xinpeng I-480 Jonsson, Ing-Marie II-586
Huang, Yi-Pai IV-506 Joo, Jaekoo V-62
Huang, Yu Ting II-386 José, Rui I-421
Hufgard, Andreas III-438 Ju, Hehua V-171
Hui, Mingqi V-498 Jung, Tzyy-Ping V-448
Humayoun, Shah Rukh I-371
Hung, Jian-Yung II-606, II-654 Kaber, David B. IV-551
Hung, Matthew III-285 Kaji, Ken V-372
Huseyinov, Ilham N. II-391 Kakusho, Koh IV-340
Hwang, Sheue-Ling II-606, II-654, Kalwar, Santosh Kumar II-239, V-431
III-166, III-193 Kambayashi, Yasushi IV-235
Hwang, Wonil IV-587 Kamei, Katsuyuki IV-722
Hwangbo, Hwan I-103 Kampf, Constance III-388
Hyden, Douglas IV-135 Kanayama, Naofumi I-480
Kanenishi, Kazuhide II-503
Ichino, Junko I-480 Karunanayaka, Kasun IV-601
Ifukube, Tohru II-398 Kashima, Tomoko II-444
Iida, Yusuke III-466 Katagami, Daisuke IV-340
Ingold, Rolf IV-157, IV-167 Kato, Toshikazu III-466, V-411
Inoue, Satoshi II-647 Kawabe, Akhiro III-295
Ioannou, Nikolaos IV-358 Kawahara, Tatsuya IV-408
Ishihara, Makio V-163 Kawakami, Hiroshi III-301
Ishihara, Manabu II-474 Kawakatsu, Hikaru IV-491
Ishihara, Yukio V-163 Kellen, Vince III-183
Itakura, Naoaki IV-225 Kesdoğan, Doğan III-39
Iteya, Satoru V-411 Khaled, Omar Abou II-531, V-186
Ito, Atsushi II-398 Khong, Chee Weng IV-750
Ito, Kyoko V-421 Kida, Takahiro IV-235
Itou, Junko III-174 Kim, Da-Hey V-3
Iwai, Yoshio V-270, V-391 Kim, Dongkeun V-339, V-441
Author Index 505
Kim, Dongsoo IV-587 Le, Hoang-An IV-67

Kim, Eun Yi V-362 Leal, Jeremy I-109
Kim, Gerard Jounghyun II-249, III-87 Leal Ferreira, Simone Bacellar III-324
Kim, Hyo Chang I-103 Lee, Eui Chul V-346
Kim, Jongbae V-3 Lee, Hyunji II-258
Kim, Jonghwa V-441 Lee, Jaehong IV-368
Kim, Minyoung V-339 Lee, Ja Young IV-610
Kim, Youngwon II-249 Lee, Ju-Hwan IV-49
Kinoshita, Yuichiro IV-456 Lee, Julia C. IV-688
Kirakowski, Jurek IV-87 Lee, Unseok IV-281
Kirche, Elias III-492 Lehmann, Simon I-381
Klein, Andreas III-396 Lekkas, Zacharias II-418
Klein, Gary L. IV-658 Lemme, Diana IV-698
Kleindienst, Jan IV-59 Leotta, Francesco IV-637
Kline, Keith III-285 Lepreux, Sophie I-211
Knecht, Christian III-29 Lewis, Richard I-193
Ko, Deajune V-346 Li, Haidong V-82
Kobayashi, Noriyuki III-307 Li, Yan IV-568
Kobayashi, Tomoyuki IV-378 Li, Yueqing IV-594
Koh, Yoon Jeon III-402 Li, Yu-Wen III-359, IV-177
Kohls, Niko II-163 Liang, Chao V-171, V-206
Kohno, Izumi I-146 Liang, Xueyi V-82
Kojima, Kazuaki IV-398 Lietz, Holger V-196
Kojima, Takatsugu IV-340 Lietzenmayer, Ryan IV-135
Komatsu, Tsuyoshi I-585 Lim, Brian Y. V-92
Kosnik, David E. II-107 Lim, Ji Hyoun IV-368
Kouroupetroglou, Georgios IV-358 Lima, Gabriel Vial Correa II-3
Kraft, Patricia III-411 Lima, Tania II-428
Krause, Michael II-596 Limbrecht-Ecklundt, Kerstin V-301
Krems, Josef F. II-612 Limongelli, Carla II-434
Kristensen, Kasper III-418 Lin, Ming-Hui IV-506, V-483
Kroes, Lies II-117 Lin, Qing-Wen III-193
Krömker, Heidi I-85, I-391 Lin, Ray F. III-199
Kryssanov, Victor III-20 Lin, Yuan-Pin V-448
Ku, Pei-Ying II-606 Lin, Yu-Ting V-483
Kühn, Romina IV-698, V-46 Lindley, Craig V-403
Kümmerling, Moritz I-411 Liu, Cheng-Li III-475
Kunc, Ladislav IV-59 Liu, Hai-Lin V-82
Kuno, Yuki V-72 Liu, Rong V-206
Kuo, Chih-Chung II-606, II-654 Liu, Shijing IV-561, IV-618
Kuramochi, Toshiya III-314 Liu, Yijun V-498
Kuribara, Takuro IV-416 Liu, Yikun IV-658
Kurosawa, Toshifumi IV-272, IV-469 Løvborg Jensen, Kasper III-332
Kurosu, Masaaki I-68, I-95 Lombardi, Matteo II-434
Long, Zhiying V-498
Labský, Martin IV-59 Longo, Lucia III-277
Laine, Juha II-101 Lorite, Salvador I-231
Lalanne, Denis II-531, IV-388, V-186 Lou, Jian I-555
Lam, Jacob III-418 Lu, Ke V-153, V-216
Laterza, Maria II-484 Lucid, Brian IV-668
506 Author Index
Luderschmidt, Johannes I-381 Mitsuishi, Takashi I-565

Luo, Bin V-227 Miwa, Shotaro V-178
Miyaji, Chikara II-84
Ma, Tao IV-291 Miyaji, Yutaka I-620
Ma, Wenqi IV-551 Mizuno, Tota IV-225
Macek, Tomáš IV-59 Modesto, Débora Maurmo III-324
Maekawa, Yasuko II-134 Mohagheghi, Amir II-76
Majewski, Martin IV-147 Mollard, Régis IV-429
Majima, Yukie II-134 Mondin, Fabio Luciano III-277
Makedon, Fillia IV-43 Mont’Alvão, Claudia Renata II-127,
Maki, Atsushi V-411 II-144, II-370
Malagardi, Ioanna IV-13 Monteiro, Ingrid IV-439
Manssour, Isabel H. III-117 Morandini, Marcelo I-184, I-249
Manthey, Robert V-196 Moreau, Guillaume IV-530
Marani, Alessandro II-434 Moreira, Samuel I-421
Marcus, Aaron I-13 Moriwaka, Makoto IV-320
Marinc, Alexander IV-147 Morreale, Patricia II-335
Märtin, Christian I-300, I-340, I-401 Morrissey, Kellie IV-87
Martin, Jennifer L. I-166 Motta, Gustavo H.M.B. II-197
Martin, Jim III-97 Mouchere, Harold IV-77
Martinie, Célia I-51, I-290 Mourlas, Constantinos II-418
Martins, Paulo V-39 Mugellini, Elena II-531, IV-157, IV-167,
Massey, Adam II-36, II-189 V-19, V-186
Matsui, Tatsunori IV-398 Munemori, Jun III-174
Matsumoto, Shimpei II-444 Muñoz, Adolfo I-231
Matsuura, Kenji II-84 Murakami, Satoru I-565
Matsuyama, Takashi IV-408 Murata, Atsuo IV-312, IV-320, IV-330
Mattar, Nikita V-102 Murata, Kazuyoshi IV-462, IV-491
Mayas, Cindy I-85, I-391
Mbatha, Blessing II-454 Nagano, Hiromi V-319
Mbatha, Mbali II-454 Nakagawa, Koji II-84
McDougall, Siné I-575 Nakamura, Tetsuaki IV-126
Mecella, Massimo IV-637 Nakamura, Yutaka III-156
Medjkoune, Sofiane IV-77 Nakatani, Yoshio III-295, V-329
Meixner, Gerrit I-411 Nakatsu, Ryohei IV-601
Mele, Maria Laura I-203 Nam, Chang S. IV-561, IV-594, IV-610,
Mello, Erick II-197 IV-618
Mendori, Takahiko II-521 Neris, Vânia Paula de Almeida II-229,
Meneses Viveros, Amilcar IV-301 III-227
Mennecke, Brian III-428 Nesi, Paolo III-259
Mhd Pauzi, Muhammad Asyraf IV-750 Nguyen, Vinh-Tiep IV-67
Micheals, Ross J. IV-449 Nguyen-Huynh, Duy-Hung IV-513
Miike, Katsuaki I-565 Nicholas, Michael II-640
Miles, Robert II-36 Nieminen, Marko II-101
Min, Wensheng IV-540 Nishida, Shogo III-314, IV-378, V-421
Minker, Wolfgang III-59 Nishino, Kazunori II-511
Miranda, Pedro Augusto da Silva e Souza Nishiuchi, Yusuke II-521
II-3 Noguchi, Anna IV-469
Mita, Yuusaku IV-416 Noor, Nor Laila Md. II-376
Mito, Kazuyuki IV-225 Norcio, Anthony F. II-173
Author Index 507
Nothdurft, Florian III-59 Pengnate, Supavich I-593

Nouri, Elnaz II-266 Perez Pelaez, Mariano IV-348
Nouri, Jalal II-464 Peters, Anicia III-428
Novick, David IV-97 Petersson Brooks, Eva III-418
Novotny, Philipp V-454 Petitrenaud, Simon IV-77
Nunes, Fátima de Lourdes dos Santos Pfaff, Mark S. IV-658
I-184 Pham, Truong-An IV-67
Philippow, Ilka I-510
Oami, Takuma IV-568 Pichelmann, Stefan II-612
Obata, Akihiko III-307 Pierantoni, Felipe II-127, II-144
Obata, Yuhki IV-568 Pino, Alexandros IV-358
Oblaender, Vera IV-706 Platt, Donald IV-629
Ochi, Keita IV-330 Plischke, Herbert II-163, V-454
Oe, Tatsuhito IV-469, IV-712 Ponsa, Pere I-231
Ogawa, Hitoshi III-20 Porras, Jari II-239, V-431
Oh, Hyung Jun I-103 Potamianos, Gerasimos IV-43
Ohiro, Tomoya V-329 Prates, Raquel Oliveira I-3
Ohkawa, Yuichi I-565 Price, Chandler III-285
Ohkura, Michiko I-585, II-343, II-647
Ohmori, Kosuke V-421 Rabe, Felix IV-117
Ohori, Kotaro III-307 Rabie, Osama II-173
Okada, Akira V-353 Rahmanivahid, Pooyan II-76, V-262
Okada, Naoki III-314 Raine-Fenning, Nicholas II-36, II-189
Okawa, Jun V-372 Rampoldi-Hnilo, Lynn II-309
Oki, Maho III-209 Ramsbrock, Jens II-621
Okuuchi, Keita IV-340 Randall, Tania I-109
O’Malley, Claire II-376 Raposo, Alberto IV-439
Ono, Yuichi II-474 Rauch, Marta II-276
Orfgen, Marius I-411 Rauff, Stefanie III-438
Otmane, Samir IV-530 Ray, Jerry III-285
Owen, Charles B. V-464 Rebenitsch, Lisa V-464
Özcebe, Esra I-310 Rehm, Matthias I-431, II-361, III-332
Ozturk, Elif II-408 Rekimoto, Jun V-112
Reppa, Irene I-575
Palanque, Philippe I-51, I-290 Riccio, Angela IV-637
Pang, Natalie II-181 Ridi, Antonio IV-157
Paolucci, Michela III-259 Rigas, Dimitrios IV-23
Pappu, Aasish IV-107 Rintel, Sean III-484
Paredes, Hugo V-39 Ritter, Marc V-196
Park, Jae Heon III-402 Rivera, Ismael V-122
Park, Jongsoon III-217 Robichez de Carvalho, Gustavo II-3
Park, Kyoung Shin V-339 Rocha, Mario Sandro II-428
Park, Kyung Eun II-56 Rodeiro-Iglesias, Javier I-441, I-490
Park, Sangin V-346, V-441 Rodil, Kasper III-332
Parthasarathy, Rangarajan II-150 Rodrigues, Kamila III-227
Pastor, Enric I-231 Røssvoll, Till Halbach III-68
Paternò, Fabio I-241 Romano, Marco III-342
Paulick, Peyton II-163, V-454 Roselli, Teresa II-484
Peeters, Jeroen I-128 Rossano, Veronica II-484
Peng, Xiaobo II-408 Rudnicky, Alex IV-107
508 Author Index
Ruffieux, Simon V-186 Shahid, Suleman II-117, II-541

Rukavina, Stefanie V-474 Shaw, Heather IV-668
Ryu, Taebeum IV-368 Shaw, Terry III-97
Shen, Yang Ting V-132
Saito, Yukou II-647 Sherry, Lance I-193
Sakairi, Takeo IV-722 Shiba, Haruya II-521
Sakamoto, Kiyomi V-353 Shibuya, Yu IV-462, IV-491
Sakamoto, Maki IV-126 Shiizuka, Hisao I-601
Sakamoto, Takehiko IV-730 Shikanai, Nao I-611
Sakamoto, Yuichiro IV-469 Shimizu, Yuichiro IV-126
Sakashita, Seiji V-353 Shin, Yunhee V-362
Sakata, Nobuchika IV-378 Shizuki, Buntarou IV-272, IV-416,
Sakoda, Masayuki II-134 IV-469, IV-712, V-72
Sakurai, Yoshihisa II-84 Shoji, Hiroko V-372
Samaras, George II-418 Siegert, Ingo V-301, V-381
Sampanes, Anthony Chad II-284 Siew, Siang-Ting I-470
Sampanthar, Kes II-292 Siio, Itiro III-209
Sánchez, Jaime II-299 Silveira, Milene S. I-3, III-117
Sandnes, Frode Eika II-20 Simione, Luca IV-637
Santos, Oscar Francisco dos I-249 Simões, Fabiana I-460
Santos, Thebano Almeida II-428 Singh, Satinder I-193
Sarathy, Rathindra I-593 Siriwardana, Sanath IV-601
Sasaki, Shiori II-493 Sloan, David I-41
Sato, Ayaka V-112 Snyder, Michele II-309
Sato, Kosuke V-270 Soares, Evandro II-428
Satoh, Hironobu II-521 Sodnik, Jaka II-551
Scerri, Simon III-29, V-122 Soga, Masato II-134
Scheruhn, Hans-Jürgen III-446 Son, Jae Pyo II-631
Schettini, Francesca IV-637 Song, Bifeng IV-540
Schieck, Ava Fatah gen. V-10 Song, Sutao V-491
Schierz, Christoph II-163 Song, Yeong-Tae II-56
Schlegel, Thomas IV-698, V-46 Sotero, Gabriel I-460
Schlick, Christopher M. III-127 Soui, Makram V-141
Schmidt, Michael IV-479 Sousa, João IV-678
Schmidt, Werner I-330, III-456 Souza Filho, Guido II-197
Schnädelbach, Holger II-36 Sowmya, Arcot II-631
Schneidermeier, Tim I-176 Spiegel, Götz III-396
Scholz, Sebastian C. I-451 Sridharan, Seshadri IV-107
Schuller, Andreas I-256, III-29, V-122 Srivastava, Sumit IV-497
Schwaller, Matthias IV-388 Stary, Christian I-330, III-456
Schwarz, Markus J. V-454 Steinhoff, Camie I-265
Sciarrone, Filippo II-434 Sterbini, Andrea II-351
Sebillo, Monica III-342 Still, Jeremiah D. I-265
Segers, Mariël III-371 Stillwater, Tai II-640
Segura, Vinı́cius I-460 Stocklöw, Carsten IV-147
Seki, Hironori II-647 Storz, Michael V-196
Seki, Makito V-178 Sturm, Christian III-352
Selvarajah, Kirusnapillai II-36, II-189 Sugano, Shohei I-620
Seo, Ki-Young V-339 Sugiyama, Masayuki IV-456
Servières, Myriam IV-530 Sun, Ming IV-107
Author Index 509
Sung, Tae-Eung V-362 Tsai, Shih-Lung III-78

Suzuki, Ayaka IV-469 Tsai, Teng-Yao IV-506
Suzuki, Ryo IV-348 Tsai, Tsai-Hsuan III-359
Tsai, Yi-Chien III-199
Tabak, Feride S. II-391 Tseng, Steven I-540
Tajima, Terumasa III-466 Tsianos, Nikos II-418
Takahashi, Katsumi I-77, I-119, I-137, Tsukada, Koji III-209
I-500, I-531 Tsumori, Shin’ichi II-511
Takahashi, Nobumichi II-647 Tsunoda, Koichi II-398
Takahashi, Shin IV-730 Tsuruda, Yu V-391
Takano, Kosuke II-493 Tung, Fang-Wu III-237
Takao, Shizuka III-174 Tung, Tony IV-408
Takemura, Noriko V-270 Tzemis, Evangelos IV-358
Takeno, Hidetoshi I-159
Takimoto, Munehiro IV-235 Uang, Shiaw-Tsyr III-475
Tamada, Takashi IV-722 Ueda, Kazutaka II-398
Tamborello, Franklin P. IV-449 Ueda, Yoshihiro I-77, I-119, I-137,
Tamura, Kazuki IV-740 I-500, I-531
Tan, Jun-Wen V-474 Uetsugi, Raku IV-312
Tanaka, Jiro IV-196, IV-272, IV-281, Ushio, Shuhei II-647
IV-416, IV-469, IV-712, IV-730, V-72
Tang, Honglin IV-243 Valla, Massimo V-122
Tanikawa, Kyohei III-314 Valverde, Lauralee II-408
Tano, Shun’ichi I-480 Vedhara, Kavita II-36, II-189
Tauchi, Hikaru II-398 Viard-Gaudin, Christian IV-77
Tavares, Tatiana A. II-197 Vilimek, Roman II-578, II-621
Tawatsuji, Yoshimasa IV-398 Viller, Stephen III-484
Teixeira, Cesar Augusto Camillo III-227 Vitiello, Giuliana III-342
Teixeira-Faria, Pedro M. I-441, I-490
Temperini, Marco II-351 Wachsmuth, Ipke IV-117, V-102
Theng, Yin-Leng II-181 Wahab, Norshahriah IV-523
Thiel, Simon III-29 Wallach, Dieter I-451
Thome, Rainer III-411 Walter, Steffen V-301, V-474
Tian, Yun V-280 Wang, Hongmei IV-135
Ting, Chih-Hung IV-506 Wang, Jan-Li III-193
Togawa, Satoshi II-503 Wang, Liang V-206
Tokdemir, Gul I-310 Wang, Lingyun (Max) I-350
Tokumaru, Masataka V-319 Wang, Pengwei V-254
Tomimatsu, Kiyoshi IV-568, IV-578 Wang, Qian V-216
Tomiyama, Ken I-620 Wang, Weizhong V-280
Tong, Xin II-207 Wang, Wenzhong V-227
Torenvliet, Gerard I-109 Wang, Yao-lien V-483
Tornero, Josep I-231 Wang, Yijun V-448
Tortora, Genoveffa III-342 Wang, Yixue V-153
Tran, Minh-Triet IV-67, IV-513 Wang, Yuping V-246
Traue, Harald C. V-474 Wang, Zhaoqi V-227
Traum, David II-266 Wästlund, Erik III-10
Tripathi, Ramesh Chandra IV-497 Watanabe, Keita V-112
True, Nicholas I-128 Watanabe, Shintaro V-178
Truong, Chau Thai IV-513 Webb, Erika Noll II-316
510 Author Index
Weber, Gerhard IV-479 Yan, Fu V-289

Weber, Julian II-621 Yanagida, Koji I-77, I-119, I-137, I-500,
Wee, William IV-291 I-531
Wei, Chun-Shu V-448 Yang, Chao-Yang II-20
Wei, Ming-Hsuan II-654 Yang, Jason III-484
Weigang, Li IV-648 Yang, Lei V-216
Wen, Chao-Hua III-78 Yang, Weixin V-254
Wendemuth, Andreas V-301, V-381 Yang, Yifang V-246
Wendler, Stefan I-510 Yang-Mao, Shys-Fan V-483
Wenzel, Christopher II-46 Yao, Li V-491, V-498
Whang, Mincheol V-346, V-441 Yasu, Hiroko I-146
Wilcock, Graham IV-32 Yasumura, Michiaki V-112
Winckler, Marco I-51, I-521 Yazdifar, Mahshid II-76, V-262
Winschiers-Theophilus, Heike III-332 Yazdifar, Mohammadreza V-262
Wolff, Christian I-176 Yeo, Alvin W. I-470
Wolff, Marion IV-429 Yildiz, Alparslan V-270
Wong, Chui Yin IV-750 Yildiz, Ekrem I-310
Wong, Man Leong IV-750 Yin, Yabo V-280
Woodjack, Justin II-640 Yoon, Joonsung I-32
Worpenberg, Annika II-323 Yoshida, Masanobu II-521
Wu, Guanshang II-207 Yoshikawa, Takuto IV-416, IV-469
Wu, Qiong II-207 Yoshikawa, Tomohiro IV-740
Wu, Siju IV-530 Yoshimura, Hiroki V-391
Wu, Tong-Ying II-20 Yu, Der-Jang I-540
Wu, Zhanwei I-555 Yu, Jiamin I-555
Wu, Zhongke V-280 Yue, Yong V-19
Yun, Myung Hwan IV-368
Xiaohui, Tan V-289
Xiong, Chenlin V-254 Zaki, Michael I-340
Xu, Bingxin V-236 Zaman, Tariq I-470
Xu, Lele V-498 Zeng, Wen-Jun IV-506, V-483
Xuesong, Wang V-289 Zetali, Karwan II-464
Zhan, Yu V-491
Yabe, Takao II-398 Zhang, Hang V-498
Yajima, Hiroshi II-216 Zhang, Jiacai V-491
Yalvac, Bugrahan II-408 Zhang, Rushao V-498
Yamaguchi, Takehiko IV-561 Zhang, Xin V-254
Yamaguchi, Takumi II-521 Zhang, Yugang IV-540
Yamane, Shohei III-307 Zhao, Fan II-46, III-492
Yamaoka, Toshiki I-361 Zhaolao, Lu V-289
Yamashiro, Mitsuo II-474 Zhong, Zhengyang V-254
Yamashita, Kuniko V-353 Zhou, Mingquan V-280, V-289
Yamazaki, Kazuhiko I-77, I-119, I-137, Zhou, Xuefu IV-291
I-500, I-531 Ziefle, Martina III-49

Kurosu (2013) HCI User and Contexts of Use

Uploaded by

Copyright:

Available Formats

Kurosu (2013) HCI User and Contexts of Use

Uploaded by

Document Information

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

Kurosu (2013) HCI User and Contexts of Use

Uploaded by

Copyright:

Available Formats

Masaaki Kurosu (Ed.

15th International Conference, HCI International 2013

ISSN 0302-9743 e-ISSN 1611-3349

Library of Congress Control Number: 2013941394

CR Subject Classification (1998): H.5, H.3, H.4, K.3, K.6

LNCS Sublibrary: SL 3 – Information Systems and Application, incl. Internet/Web

The 15th International Conference on Human–Computer Interaction, HCI In-

• 10th International Conference on Engineering Psychology and Cognitive

• Identity, Privacy and Trust

The remaining volumes of the HCI International 2013 proceedings are:

May 2013 Constantine Stephanidis

Program Chair: Masaaki Kurosu, Japan

Human Interface and the Management of Information

Program Chair: Sakae Yamamoto, Japan

Hans-Jorg Bullinger, Germany Mark Lehto, USA

Michale Smith, USA Kim-Phuong Vu, USA

Engineering Psychology and Cognitive Ergonomics

Program Chair: Don Harris, UK

Universal Access in Human–Computer Interaction

Program Chairs: Constantine Stephanidis, Greece,

Julio Abascal, Spain Georgios Kouroupetroglou, Greece

Virtual, Augmented and Mixed Reality

Program Chair: Randall Shumaker, USA

Waymon Armstrong, USA Mark Livingston, USA

Program Chair: P.L. Patrick Rau, P.R. China

Online Communities and Social Computing

Program Chairs: A. Ant Ozok, USA,

Program Chairs: Dylan D. Schmorrow, USA,

Digital Human Modeling and Applications in Health,

Program Chair: Vincent G. Duﬀy, USA and Russia

Design, User Experience, and Usability

Program Chair: Aaron Marcus, USA

Anthony Faiola, USA Jennifer McGinn, USA

Distributed, Ambient and Pervasive Interactions

Program Chairs: Norbert Streitz, Germany,

Human Aspects of Information Security, Privacy and Trust

Program Chairs: Louis Marinos, ENISA EU,

Claudio Agostino Ardagna, Italy Ronald Leenes, The Netherlands

Julien Touzeau, France Claire Vishik, UK

Maysoon Abulkhair, Saudi Arabia Vassilis Kouroumalis, Greece

The 16th International Conference on Human–Computer Interaction, HCI

Identity, Privacy and Trust

Identity Management through “Proﬁles”: Prototyping an Online

An Empirical Investigation of Similarity-Driven Trust Dynamics in a

The di.me User Interface: Concepts for Sharing Personal Information

Trust and Privacy in the di.me Userware . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

From Living Space to Urban Quarter: Acceptance of ICT Monitoring

The Impact of Explanation Dialogues on Human-Computer Trust . . . . . . 59

Trustworthy and Inclusive Identity Management for Applications in

Towards a Design Guideline of Visual Cryptography on Stereoscopic

Assessing the Impact of Latency and Jitter on the Perceived Quality of

The Eﬀects of Negative Interaction Feedback in a Web Navigation

Automatic Layout Generation for Digital Photo Albums:

Icon Design for Older Users of Project Management Software . . . . . . . . . . 127

Search Strategies in Hypermedia Navigation and Spatial Abilities:

Data Visualisation, User Experience and Context: A Case Study from

A Pedestrian Navigation Method for User’s Safe and Easy

The Eﬀect of Information Quantity on Cbp Interface in the Advanced

Development of Communication Support System at Mealtimes Using

Improving User Performance in Conditional Probability Problems with