0% found this document useful (0 votes)

115 views319 pages

Systems Thinking in Practice - Applications of The Event Analysis of Systemic Teamwork Method (PDFDrive)

Uploaded by

dyanaguirre

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

115 views319 pages

Systems Thinking in Practice - Applications of The Event Analysis of Systemic Teamwork Method (PDFDrive)

Uploaded by

dyanaguirre

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 319

Systems Thinking

in Practice
Applications of the Event Analysis
of Systemic Teamwork Method
Transportation Human Factors: Aerospace,
Aviation, Maritime, Rail, and Road Series
Series Editor
Professor Neville A Stanton
University of Southampton, UK

Automobile Automation
Distributed Cognition on the Road
Victoria A. Banks, Neville A. Stanton

Eco-Driving
From Strategies to Interfaces
Rich C. McIlroy, Neville A. Stanton

Systems Thinking in Practice

Applications of the Event Analysis of Systemic Teamwork Method
Neville A. Stanton, Paul M. Salmon, Guy H. Walker

For more information about this series, please visit: https://www.crcpress.com/

Transportation-Human-Factors/book-series/CRCTRNHUMFACAER
Systems Thinking
in Practice
Applications of the Event Analysis
of Systemic Teamwork Method

Neville A. Stanton
Paul M. Salmon
Guy H. Walker
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2019 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-138-09787-2 (Hardback)

International Standard Book Number-13: 978-1-315-10468-3 (eBook)

This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have
been made to publish reliable data and information, but the author and publisher cannot assume responsibility
for the validity of all materials or the consequences of their use. The authors and publishers have attempted to
trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if
permission to publish in this form has not been obtained. If any copyright material has not been acknowledged
please write and let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmit-
ted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented,
including photocopying, microfilming, and recording, or in any information storage or retrieval system, with-
out written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration
for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate
system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used
only for identification and explanation without intent to infringe.

Library of Congress Cataloging‑ in‑ Publication Data

Names: Stanton, Neville A. (Neville Anthony), 1960- author. | Salmon, Paul

M., author. | Walker, Guy, author.
Title: Systems thinking in practice : applications of the Event Analysis of
Systemic Teamwork method / Neville A. Stanton, Paul Salmon, and Guy Walker.
Description: First edition. | Boca Raton, FL : CRC Press/Taylor & Francis
Group, 2018. | Series: Transportation human factors: Aerospace, aviation,
maritime, rail, and road series | Includes bibliographical references and
index.
Identifiers: LCCN 2018026593| ISBN 9781138097872 (hardback : alk. paper) |
ISBN 9781315104683 (ebook)
Subjects: LCSH: Systems engineering. | Teams in the workplace.
Classification: LCC TA168 .S68 2018 | DDC 658.4/022011--dc23
LC record available at https://lccn.loc.gov/2018026593

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com
Contents
Preface.................................................................................................................... xiii
Authors....................................................................................................................xvii
Board Members and Affiliations.............................................................................xix

Section I Overview of EAST

Chapter 1 The Event Analysis of Systemic Teamwork (EAST) Method............... 3

With Nicholas J. Stevens
Background and Applications...............................................................4
Domain of Application..........................................................................5
Application in Land Use Planning and Urban Design (LUP & UD)...... 5
Procedure and Advice...........................................................................5
Step 1: Define Analysis Aims........................................................... 5
Step 2: Define the Task/System Under Analysis.............................. 6
Step 3: Data Collection..................................................................... 6
Step 4: Transcribe Data....................................................................6
Step 5: Construct Task Network.......................................................6
Step 6: Conduct Social Network Analysis........................................8
Step 7: Construct Information Networks.......................................... 8
Step 8: Construct Composite Networks............................................ 8
Step 9: Analyse Networks................................................................8
Advantages............................................................................................9
Disadvantages...................................................................................... 10
Related Methods.................................................................................. 10
Approximate Training and Application Times................................... 10
Reliability and Validity....................................................................... 10
Flowchart............................................................................................. 10
Tools Needed....................................................................................... 11
Example............................................................................................... 11
References........................................................................................... 15

Section II Applications of EAST

Chapter 2 EAST in Air Traffic Control............................................................... 19

With Chris Baber, Linda Wells, Huw Gibson and Daniel P. Jenkins
Introduction......................................................................................... 19
Command and Control................................................................... 19
v
vi Contents

Distributed Cognition..................................................................... 19
Distributed Situation Awareness....................................................20
Beyond Ethnography...................................................................... 21
The Air Traffic Control Work Setting................................................. 23
Charts and Standard Routes........................................................... 23
Flight Data Strip............................................................................. 23
Flight Data Strip Bay...................................................................... 23
Radar Display................................................................................. 23
Aircraft Call Signs.........................................................................24
Communications.............................................................................24
Distributed Cognition Methodology: The Importance of Methods........ 24
Descriptive vs. Formative Methods................................................24
Method Integration.........................................................................25
Air Traffic Control Scenarios..............................................................25
Applying the Method......................................................................26
Representing Distributed Cognition...............................................26
Task Networks................................................................................26
Social Networks.............................................................................. 27
Propositional Networks..................................................................28
Application to Air Traffic Control......................................................28
Analysis of Agents in the Distributed Cognition System...............28
Facilitating Technology.................................................................. 31
Control Architecture....................................................................... 31
Systemic Situational Awareness..................................................... 33
Temporal Aspects of Command and Control in Air Traffic
Control............................................................................................ 35
Conclusions......................................................................................... 35
Acknowledgements............................................................................. 36
References........................................................................................... 36

Chapter 3 EAST in Military Command and Control.......................................... 39

With Rebecca Stewart, Daniel P. Jenkins, Linda Wells and
Chris Baber
Introduction......................................................................................... 39
Description of Command and Control Scenarios............................... 39
Army Land Warfare and the Combat Estimate.............................. 39
Data Collection...............................................................................40
Description of the East Method.......................................................... 41
The Importance of Methods........................................................... 41
Descriptive vs. Formative Methods................................................ 41
Method Integration......................................................................... 42
Situational Awareness.................................................................... 42
Theoretical Basis............................................................................44
Findings...............................................................................................44
Contents vii

Coordination Demand Analysis (CDA)..........................................44

Communications Usage Diagram (CUD).......................................46
Advantages and Disadvantages of Existing
Communications Media............................................................. 48
Social Network Analysis (SNA)..................................................... 48
Activity Stereotypes.................................................................. 52
Facilitation of Network Links.................................................... 52
Calculation of Social Network Metrics..................................... 53
Operation Sequence Diagrams (OSD)............................................ 54
Propositional Networks (PN)......................................................... 56
Conclusions......................................................................................... 58
Acknowledgements............................................................................. 61
References........................................................................................... 61

Chapter 4 EAST in Energy Distribution Operations........................................... 65

With Daniel P. Jenkins, Chris Baber and Richard McMaster
Introduction......................................................................................... 65
Example Case Study: A Human Factors Analysis of Civilian
Command and Control................................................................... 67
Scenario 1: Switching Operations Scenario (Barking).............. 68
Scenario 2: Maintenance Scenario (Tottenham)....................... 68
Methodology....................................................................................... 69
Design............................................................................................. 69
Participants..................................................................................... 69
Materials......................................................................................... 70
Procedure........................................................................................ 71
Results................................................................................................. 72
Task Networks................................................................................ 72
Social Networks.............................................................................. 72
Information Networks.................................................................... 73
Additional EAST Analyses............................................................ 74
Summary............................................................................................. 77
References........................................................................................... 81

Chapter 5 EAST in a Submarine Control Room.................................................. 83

Analysing Distributed Cognition........................................................ 83
System Properties of EAST................................................................. 85
Data Collection and Analysis.............................................................. 86
Task Network Analysis........................................................................ 89
Social Network Analysis (SNA)..........................................................92
Information Network Analysis............................................................ 93
Combining Network Models............................................................... 95
Task and Social Networks..............................................................97
viii Contents

Information and Social Networks..................................................99

Information and Task Networks................................................... 100
Combined Task, Social and Information Networks..................... 100
Summary and Conclusions................................................................ 100
Acknowledgements........................................................................... 106
References......................................................................................... 106

Chapter 6 EAST in Railway Maintenance......................................................... 109

With Huw Gibson and Chris Baber
Introduction....................................................................................... 109
Development of the East Methodology............................................. 110
Layer 1 – Data Collection Methods.............................................. 111
Layer 2 – Analysis Methods......................................................... 112
Hierarchical Task Analysis (HTA).......................................... 112
Coordination Demand Analysis (CDA)................................... 114
Communications Usage Diagram (CUD)................................ 115
Social Network Analysis (SNA).............................................. 115
Layer 3 – Representational Methods............................................ 115
Operation Sequence Diagram (OSD)...................................... 116
Propositional Network (PN).................................................... 116
Summary of Methods................................................................... 117
Structure of EAST Methodology................................................. 118
Procedure.......................................................................................... 119
Application of the Method to Live Data........................................... 119
Background................................................................................... 119
Scenario 1 – Planned Maintenance Activities......................... 121
Scenario 2 – Emergency Engineering Work........................... 122
Scenario 3 – Ending a Track Possession................................. 122
Results and Discussion...................................................................... 122
Analysis Methods......................................................................... 122
Task Networks......................................................................... 122
CDA......................................................................................... 124
CUD......................................................................................... 125
SNA......................................................................................... 126
Representational Methods............................................................ 127
Scenario Process Charts (OSD)............................................... 127
Propositional Networks (PN)................................................... 127
Summary...................................................................................... 129
Conclusions....................................................................................... 129
Further Insights............................................................................ 129
Wider Context............................................................................... 131
Acknowledgements........................................................................... 131
References......................................................................................... 131
Contents ix

Chapter 7 EAST at Road Intersections.............................................................. 135

With Michael G. Lenné and J. Ashleigh Filtness
Statement of Relevance..................................................................... 135
Introduction....................................................................................... 135
EAST Intersection Case Study..................................................... 136
Methodology..................................................................................... 138
Design........................................................................................... 138
Participants................................................................................... 138
Materials....................................................................................... 138
Procedure...................................................................................... 139
Results............................................................................................... 141
Task Networks.............................................................................. 141
Social Networks............................................................................ 144
Situation Awareness Networks..................................................... 147
Discussion......................................................................................... 151
Reducing Complexity................................................................... 151
Performance Limitations.............................................................. 152
System Redesign........................................................................... 153
Conclusion......................................................................................... 154
References......................................................................................... 154

Chapter 8 EAST in Elite Women’s Cycling Teams........................................... 157

With Clare Dallat and Amanda Clacy
Introduction....................................................................................... 157
Elite Women’s Cycling................................................................. 158
Event Analysis of Systemic Teamwork........................................ 159
EAST Analysis of Elite Women’s Cycling................................... 159
Methods............................................................................................. 159
Participants................................................................................... 159
Materials....................................................................................... 160
Procedure...................................................................................... 160
Results............................................................................................... 161
Task Network................................................................................ 161
Social Network............................................................................. 162
Situation Awareness Network....................................................... 162
Conclusion......................................................................................... 163
What Does the Analysis Tell Us About DSA and Teamwork
in Elite Women’s Cycling Teams?................................................ 163
What Are the Implications for Optimising Performance in
Elite Women’s Cycling Teams?.................................................... 165
What Are the Implications for Future EAST Applications in
Elite Sport?................................................................................... 166
References......................................................................................... 166
x Contents

Chapter 9 EAST in Automated Driving Systems.............................................. 169

With Victoria A. Banks
Introduction....................................................................................... 169
Levels of Automation and the Role of the Driver......................... 169
Method.............................................................................................. 171
Results............................................................................................... 173
Task Networks.............................................................................. 173
Social Networks............................................................................ 175
Information Networks.................................................................. 179
Discussion......................................................................................... 181
Conclusions....................................................................................... 182
Acknowledgements........................................................................... 183
References......................................................................................... 183

Chapter 10 EAST in Future Road Transportation Systems................................. 187

With Victoria A. Banks, Gary Burnett and Setia Hermawati
Introduction....................................................................................... 187
Method.............................................................................................. 188
Results............................................................................................... 190
Identification of System Agents.................................................... 190
Task Networks.............................................................................. 190
Social Networks............................................................................ 193
Information Networks.................................................................. 196
Discussion......................................................................................... 199
Conclusions....................................................................................... 201
References......................................................................................... 201

Section III Future Developments in EAST

Chapter 11 STAMPING on EAST: Adding a Control Network to EAST

to Examine the Safety Controls in the Railway Level Crossing
System Lifecycle ..............................................................................207
With Gemma J. M. Read, Natassia Goode, Eryn Grant, Clare
Dallat, Tony Carden and Anjum Naweed
Introduction.......................................................................................207
Research Context: Railway Level Crossings.....................................208
Part 1: Integrating EAST and STAMP.............................................209
STAMP and Control Theory........................................................209
Integrating EAST and STAMP: A Network of Controls............. 212
Part 2: Applying the Integrated Method to RLX Safety
Management...................................................................................... 213
Contents xi

Methodology................................................................................. 213
Data Inputs............................................................................... 213
EAST Network Development.................................................. 213
Network Analysis.................................................................... 214
Results............................................................................................... 216
Task Network................................................................................ 216
Social Network............................................................................. 216
Information Network.................................................................... 216
Control Network........................................................................... 219
Discussion..................................................................................... 219
Implications for Railway Level Crossing Safety
Management................................................................................. 223
Conclusion......................................................................................... 223
References.........................................................................................224

Chapter 12 The EAST ‘ Broken-Links’ Approach: Assessing Risk in

Sociotechnical Systems..................................................................... 227
With Catherine Harvey
Introduction....................................................................................... 227
Case Study of Hawk Missile Simulation Training............................ 229
Analysis of Networks........................................................................ 232
Results............................................................................................... 233
Social Network............................................................................. 233
Task Network................................................................................ 234
Information Network.................................................................... 235
Broken-Links Analysis...................................................................... 237
Discussion......................................................................................... 241
Conclusions.......................................................................................246
Acknowledgements...........................................................................246
References .........................................................................................246

Chapter 13 From CWA to SNA: Modelling Future Flight Decks....................... 251

With Don Harris and Alison Starr
Introduction....................................................................................... 251
Design Approach............................................................................... 254
Aircraft Component..................................................................... 254
Ground-Based Component........................................................... 255
System ‘Mirror’............................................................................ 256
Modelling and Analysis of System Configurations........................... 256
Cognitive Work Analysis (CWA)................................................. 258
Work Domain Analysis (WDA)................................................... 258
Functional Purposes..................................................................... 258
Values and Priority Measures....................................................... 259
xii Contents

Physical Objects........................................................................... 259

Object-Related Processes.............................................................260
Purpose-Related Functions...........................................................260
Control Task Analysis (ConTA)................................................... 261
Social Organisational Cooperation Analysis – Contextual
Activity Template (SOCA-CAT).................................................. 262
Social Network Analysis.............................................................. 263
Conclusions....................................................................................... 270
References......................................................................................... 271

Chapter 14 Future Directions for the Event Analysis of Systemic Teamwork........275

Introduction....................................................................................... 275
Normal Performance as a Cause of Accidents.................................. 275
Accident Prediction........................................................................... 278
A Systems Approach to Prediction............................................... 279
East and Accident Prediction....................................................... 279
Migration towards Safety Boundaries...............................................280
Mapping Migration....................................................................... 281
Systems Concepts.............................................................................. 281
Human Factors and Ergonomics Methods in Design........................ 282
Conclusions....................................................................................... 285
References......................................................................................... 287
Index....................................................................................................................... 291
Preface
This book has arisen from the desire to share our insights into the practical applica-
tion of the Event Analysis of Systemic Teamwork (EAST) framework. Over the past
decade, we have applied EAST in many domains and have been impressed by the
insights that the products of the analysis afford. EAST offers a systemic and system-
atic approach to the analysis, design and evaluation of sociotechnical systems. As a
formative method, we have used it to design new concepts of operations, new teams
and new ways of working. As a summative method, we have used it to gain new
insights into existing systems and current ways of working.
The EAST method was initially developed for a programme of research into
command and control funded by the UK Ministry of Defence called the Human
Factors Integration Defence Technology Centre. This research began in 2003 and
ended in 2012.
The total funding was for £30 million over 10 years, and one of our projects was
to provide Human Factors advice for new networked architectures for command and
control. The idea was that potentially all military systems across all of the forces (from
the infantry soldier to the moving platforms [on land, in the air and at sea] to the joint
operations headquarters) could be connected like the World Wide Web is for civilian
activities.
This kind of connectivity meant we had to look at command and control in a
completely new way. Rather than develop completely new methods, we developed
EAST by integrating existing methods, as is explained in this book. Prof Chris Baber
and I decided to use network methods to investigate these command and control
networks. This began with investigations into both civilian and military examples
of command and control, some of which are presented in this book. The approach
has developed into the EAST method as we know it today. Over the years, we have
developed and refined the approach. What we particularly like about the network
methods is that they are both scalable and systemic in nature. We can analyse the
networks both qualitatively and quantitatively (using network statistics, as demon-
strated in this book).
We imagine people will approach this book with different purposes. For those
new to the EAST approach, we advise reading Chapter 1 and then finding a chap-
ter with your domain of interest. For those familiar with EAST, go straight to
the chapters that are of interest. For researchers and those keen on extending
the EAST approach, the final section of chapters on developments and future
directions will be of most interest. Each of the chapters is intended to be read
as a stand-alone article, so there is some inevitable repetition on the overview
of EAST (although different emphases and approaches to the method are taken).
Experts in EAST can skip over those sections. For those who use EAST and find it
useful in their work, we say, welcome to the tribe of happy EASTers. That EASTer
tribe already comprises the authors in this book, and we are aware of other groups
using EAST around the world. All of the chapters were led by one or more of the
lead authors of the book with the exception of Chapters 9 and 10, which were both

xiii
xiv Preface

led by Victoria Banks. We are grateful for the contributions and insights from our
co-authors and the progress that has been made with the method since its original
conception.
The EAST development journey is not at the end yet. There have been recent
developments that have extending the use of EAST. EAST, as a systems method, has
fared well, as systems approaches are very much in favour in the Ergonomics and
Human Factors world. We have had some successes linking EAST to other systems
methods, such as Systems Theoretic Accident Model and Process |(STAMP) and
Cognitive Work Analysis (CWA). We have also used EAST in formative ways to pre-
dict system network resilience. The journey does not end here, however, and we are
sure that we and others will continue to develop and extend the approach and apply
it to even more domains of application.
Neville A. Stanton
Professor of Human Factors Engineering
University of Southampton

THERE ARE ALSO THESE BOOKS:

Stanton, N. A., Salmon, P. M., Walker, G. H., Baber, C. and Jenkins, D. (2005). Human
Factors Methods: A Practical Guide for Engineering and Design (first edition).
Ashgate: Aldershot.
xvi Preface

Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork. Ashgate: Aldershot.
Stanton, N. A., Salmon, P. M., Rafferty, L. A., Walker, G. H., Baber, C. and Jenkins, D.
(2013). Human Factors Methods: A Practical Guide for Engineering and Design (sec-
ond edition). Ashgate: Aldershot.
Authors
Professor Neville Stanton, PhD, DSc, is a chartered psychologist, chartered ergon-
omist and chartered engineer. He holds the Chair in Human Factors Engineering in
the Faculty of Engineering and the Environment at the University of Southampton
in the United Kingdom. He earned degrees in Psychology, Applied Psychology and
Human Factors and has worked at the Universities of Aston, Brunel, Cornell and
MIT. His research interests include modelling, predicting, analysing and evaluating
human performance in systems as well as designing the interfaces and interaction
between humans and technology. Professor Stanton has worked on the design of
automobiles, aircraft, ships and control rooms over the past 30 years on a variety
of automation projects. He has published 40 books and over 300 journal papers on
Ergonomics and Human Factors. In 1998, he was awarded the Institution of Electrical
Engineers Divisional Premium Award for research into System Safety. The Institute
of Ergonomics and Human Factors awarded him the Otto Edholm Medal in 2001, the
President’s Medal in 2008 and the Sir Frederic Bartlett Medal in 2012 for his contri-
butions to basic and applied ergonomics research. The Royal Aeronautical Society
awarded him and his colleagues the Hodgson Prize and Bronze Medal in 2006 for
research on design-induced, flight-deck error published in The Aeronautical Journal.
The University of Southampton awarded him a Doctor of Science in 2014 for his sus-
tained contribution to the development and validation of Human Factors methods.

Professor Paul Salmon holds a Chair in Human Factors and is creator and director
of the Centre for Human Factors and Sociotechnical Systems at the University of
the Sunshine Coast. He currently holds a prestigious Australian Research Council
Future Fellowship and has almost 15 years’ experience in applied Human Factors
research in a number of areas, including defence, transportation safety, sports and
outdoor recreation and disaster management. Professor Salmon currently leads
major research programmes in the areas of road and rail safety, identity theft and
cybersecurity and led outdoor recreation accidents. He has co-authored 14 books,
over 180 peer-reviewed journal articles and numerous conference articles and book
chapters. He has received various accolades for his contributions to research and
practice, including the Australian Human Factors and Ergonomics Societies 2016
Cumming Memorial medal, the UK Ergonomics Society’s Presidents Medal, the
Royal Aeronautical Society’s Hodgson Prize for best research and paper and the
University of the Sunshine Coast’s Vice Chancellor and President’s Medal for
Research Excellence. Professor Salmon’s current research interests relate to extend-
ing Human Factors and Sociotechnical Systems theory and methods to support the
optimisation of systems in many areas. Specific areas of focus include accident pre-
diction and analysis, systems thinking in transportation safety, the development of
systemic accident countermeasures, human factors in elite sports and cybersecurity.

Professor Guy Walker works within the Institute for Infrastructure and Environment
at Heriot-Watt University in Edinburgh. He lectures on Human Factors and is the

xvii
xviii Authors

author/co-author of over 100 peer-reviewed journal articles and 13 books. He and

his co-authors have been awarded the Institute for Ergonomics and Human Factors
President’s Medal for the practical application of Ergonomics theory and the Peter
Vulcan Prize for best research paper by the 2013 Australasian Road Safety Research
Conference. In 2011, he also won Heriot-Watt University’s Graduate’s Prize for inspi-
rational teaching. Prof Walker earned a BSc Honours degree in Psychology from the
University of Southampton and a PhD in Human Factors from Brunel University. His
research interests are wide ranging, spanning driver behaviour and the role of feed-
back in vehicles, using Human Factors methods to analyse black-box data recordings
and the application of sociotechnical systems theory to the design and evaluation
of civil engineering systems through to safety, risk and reliability. His research has
featured in the popular media, from national newspapers, TV and radio through to
an appearance on the Discovery Channel.
Board Members and Affiliations

Chris Baber Huw Gibson

School of Electronic, Electrical Human Factors Specialist
& Computing Engineering Rail Safety and Standards Board
University of Birmingham London, UK
Birmingham, UK
Natassia Goode
Victoria A. Banks Centre for Human Factors and
Human Factors Engineering Sociotechnical Systems
Transportation Research Group University of the Sunshine Coast
University of Southampton Queensland, Australia
Southampton, UK
Eryn Grant
Gary Burnett Centre for Human Factors and
Human Factors Research Group Sociotechnical Systems
University of Nottingham University of the Sunshine Coast
Nottingham, UK Queensland, Australia

Tony Carden Don Harris

Centre for Human Factors and Human Systems Integration Group
Sociotechnical Systems Coventry University
University of the Sunshine Coast Coventry, UK
Queensland, Australia
Catherine Harvey
Human Factors Research Group
Amanda Clacy
University of Nottingham
Centre for Human Factors and
Nottingham, UK
Sociotechnical Systems
University of the Sunshine Coast Setia Hermawati
Queensland, Australia Human Factors Research Group
University of Nottingham
Clare Dallat Nottingham, UK
Centre for Human Factors and
Sociotechnical Systems Daniel P. Jenkins
University of the Sunshine Coast DCA Design International Ltd
Queensland, Australia Warwick, UK

Ashleigh J. Filtness Michael G. Lenné

Design School Accident Research Centre
Loughborough University Monash University, Clayton Campus
Loughborough, UK Victoria, Australia

xix
xx Board Members and Affiliations

Richard McMaster Alison Starr

Senior Human Factors Consultant The National Composites Centre
Abbott Risk Consulting (ARC) Ltd Bristol and Bath Science Park
London, UK Bristol, UK

Anjum Naweed Nicholas J. Stevens

Appleton Institute for Behavioural Science Centre for Human Factors and
Central Queensland University Sociotechnical Systems
Queensland, Australia University of the Sunshine Coast
Queensland, Australia
Gemma J. M. Read
Centre for Human Factors and Rebecca Stewart
Sociotechnical Systems Lockheed Martin
University of the Sunshine Coast Langstone Technology Park
Queensland, Australia Havant, UK

Paul M. Salmon Guy H. Walker

Centre for Human Factors and School of the Built Environment
Sociotechnical Systems Heriot-Watt University
University of the Sunshine Coast Edinburgh, UK
Queensland, Australia
Linda Wells
Neville A. Stanton BAE Systems
Human Factors Engineering Somerset, UK
Transportation Research Group
University of Southampton
Southampton, UK
Section I
Overview of EAST
1 The Event Analysis of
Systemic Teamwork
(EAST) Method
With Nicholas J. Stevens

In this book, we describe a series of studies that apply the systems thinking approach
using the Event Analysis of Systemic Teamwork (EAST) method. Systems thinking
is a contemporary approach that has currency within the discipline of human fac-
tors. It aims to understand and improve safety and performance in complex socio-
technical systems. Human factors issues are increasingly being examined through
the systems thinking lens (Karsh et al. 2014; Salmon et al. 2017; Walker et al.
2017). In line with this, since the turn of the century, a range of Human Factors
methods have either been developed or have experienced a resurgence in popular-
ity. These include systems analysis frameworks, such as Cognitive Work Analysis
(CWA) (Vicente 1999) and EAST (Stanton et al. 2008); accident analysis methods,
such as AcciMap (Svendung and Rasmussen 2002), the Systems Theoretic Accident
Model and Processes (STAMP) (Leveson 2004), and the Functional Resonance
Analysis Method (FRAM) (Hollnagel 2012); and systems design methods, such as
the MacroErgonomic Analysis and Design method (MEAD) (Kleiner 2006) and the
Cognitive Work Analysis Design Toolkit (Read et al. 2016).
The aim of this book is to demonstrate how one of these methods, EAST, can be
used to provide in-depth analyses of performance and safety in complex sociotechni-
cal systems (STS). The systems thinking approach involves taking the overall system
as the unit of analysis, looking beyond individuals and considering the interactions
between humans and between humans and artefacts within a system. This view also
encompasses factors within the broader organisational, social or political system in
which behaviour takes place. Taking this perspective, behaviours emerge not from
the decisions or actions of individuals but from interactions between humans and
artefacts across the wider system. At the most basic level when examining STS, the
descriptive constructs of interest can be distilled down to simply

• Why: the goals of the system, sub-system[s] and actor[s]

• Who: the actors performing the activity, including humans and technologies
• When: when activities take place and which actors are associated with them
• Where: where activities and actors are physically located
• How: (how activities are performed and how actors communicate and col-
laborate to achieve goals).

3
4 Systems Thinking in Practice

To assist researchers and practitioners explore these constructs, EAST (Stanton

et al. 2013) offers a comprehensive framework for the design, evaluation and analysis
of complex sociotechnical systems. As well as offering a description of the activ-
ity performed within a particular system, the approach provides methods that can
be used to develop an in-depth analysis of the constraints that shape agent activ-
ity within the system. Sociotechnical systems scenarios are often so complex and
multi-faceted, and analysis requirements so diverse, that various methods need to
be applied as one method in isolation cannot cater for the scenario and analysis
requirements. Building on a long history and tradition of methods integration in
human factors research and practice (Stanton et al. 2005), EAST (Stanton et al. 2008,
2013) provides an integrated suite of methods for analysing the performance of com-
plex sociotechnical systems. The framework supports this by providing methods to
describe, analyse and integrate three network-based representations of activity: task,
social and information networks. An overview of the EAST method is provided in
the remainder of this introductory chapter.

BACKGROUND AND APPLICATIONS

EAST (Stanton et al. 2008) provides a framework of methods that allows system
performance to be comprehensively described and evaluated. Since its conception,
the framework has been applied in many domains, including land and naval warfare
(Stanton et al. 2006; Stanton 2014), aviation (Stewart et al. 2008), air traffic control
(Walker et al. 2010), road transport (Salmon et al. 2014a) the emergency services
(Houghton et al. 2008) and elite cycling (Salmon et al. 2017). Within this book, the
application areas covered include

• aviation (Chapters 2, 12 and 13)

• command and control (Chapters 3 and 5)
• energy distribution (Chapter 4)
• rail transportation (Chapters 6 and 11)
• road transportation (Chapters 7, 9 and 10) and
• sport (Chapter 8)

Underpinning the approach is the notion that distributed teamwork can be

meaningfully described via a ‘network of networks’ approach, shown in Figure
1.1. Specifically, three networks are considered: task, social and information net-
works. Task networks describe the goals and subsequent tasks being performed
within the system. Social networks analyse the organisation of the system (i.e.
communications structure) and the communications taking place between the
actors working in the team. Finally, information networks describe the informa-
tion and knowledge (situation awareness) that the different actors use and share
during task performance.
Recent applications of the framework have also adopted a composite network
analysis approach whereby the three networks are integrated to show the relation-
ships between tasks, social interactions and information (Stanton 2014).
The Event Analysis of Systemic Teamwork (EAST) Method 5

FIGURE 1.1 Network of networks approach.

DOMAIN OF APPLICATION
EAST is a generic approach that was developed originally for the analysis of
teamwork in sociotechnical systems, but it has since been used to provide analy-
ses at the micro (Salmon et al. 2014a), meso (Stanton 2014) and macro (Stanton
and Harvey 2017) levels of sociotechnical systems. As such, it can be used in any
domain in which social and technical elements are working together in pursuit
of a common goal. The case study presented in this chapter, used to demonstrate
EAST, is based on an application within the area of land use planning and in
urban design (Stevens et al., 2018).

APPLICATION IN LAND USE PLANNING

AND URBAN DESIGN (LUP & UD)
LUP & UD is most often a product of multi-disciplinary approaches to complex
environments and sociotechnical systems (Stevens et al. 2018). Our projects and pro-
cesses require different resources from a variety of participants over a range of time
frames. Whilst we may not always define these approaches as teamwork, the EAST
approach has the potential to offer critical insights into more effective and efficient
cooperative processes and project performance.

PROCEDURE AND ADVICE

Step 1: Define Analysis Aims
First, the aims of the analysis should be clearly defined so that appropriate scenarios
are used and relevant data are collected. In addition, not all components of the EAST
framework may be required, so it is important to clearly define the aims at this point
to ensure that the appropriate EAST methods are applied.
6 Systems Thinking in Practice

Step 2: Define the Task/System Under Analysis

Next, the task (or tasks) or scenario (or scenarios) under analysis should be clearly
defined. This is dependent upon the aims of the analysis and may include a range
of tasks or one task in particular. It is normally standard practice to develop a
Hierarchical Task Analysis (HTA) (see Stanton 2006) for the task under analysis
if sufficient data and subject matter expert (SME) access are available. This is use-
ful later on in the analysis and is also enlightening, allowing the analyst to gain an
understanding of the task before the observation and analysis begins.

Step 3: Data Collection

Once the aims of the analysis are clearly defined, the next step involves collect-
ing targeted data about the system and its behaviour. The specific data collected are
dependent on the analysis aims and the resources available; however, data collection
for EAST typically involves observations, concurrent verbal protocols, structured or
semi-structured interviews (e.g. the Critical Decision Method [CDM]; Klein et al.,
1989), walkthrough analysis and documentation review (e.g. incident reports, stan-
dard operating procedures).
The observation step is often the most important part of the EAST procedure.
Typically, a number of analysts are used to observe the system or scenario under
analysis. All activities involved in the scenario under analysis should be recorded
along an incident timeline, including a description of the activity undertaken, the
agents involved, any communications made between agents and the technology
involved. Additional notes should be made where required, including the purpose of
the activity observed; any tools, documents or instructions used to support activity;
the outcomes of activities; any errors made; and also any information that the agent
involved feels is relevant. In addition, it is useful to video record the task and record
verbal transcripts of all communications, if possible.
Once the task under analysis is complete, each ‘key’ agent (e.g. scenario com-
mander, agents performing critical tasks) involved should be subjected to a CDM
interview. This involves dividing the scenario into key incident phases and then
interviewing the actor involved in each phase using a set of pre-defined CDM probes
(e.g. O’Hare et al. 2000; see also Chapter 4 for more information on the CDM).

Step 4: Transcribe Data

Once all of the data are collected, it should be transcribed in order to make it compat-
ible with the EAST analysis phase. An event transcript should then be constructed. This
should describe the scenario over a timeline, including descriptions of activity, the actors
involved, any communications made and the technology used. In order to ensure the valid-
ity of the data, the scenario transcript should be reviewed by one of the SMEs involved.

Step 5: Construct Task Network

The first analysis step involves constructing a task network. Prior to this, the ini-
tial HTA should be reviewed and refined based on the data collected during step 3.
The Event Analysis of Systemic Teamwork (EAST) Method 7

The data transcription process allows the analyst to gain a deeper and more
accurate understanding of the scenario under investigation. It also allows any dis-
crepancies between the initial HTA scenario description and the actual activity
observed to be resolved. Typically, activities in complex sociotechnical systems
do not run entirely according to protocol, and certain tasks may have been per-
formed during the scenario that were not described in the initial HTA description.
The analyst should compare the scenario transcript to the initial HTA and add any
changes as required.
Constructing the task network involves identifying high-level tasks and the rela-
tionships between them and creating a network to represent this. Some general rules
around the construction of EAST networks are presented in Table 1.1.

TABLE 1.1
Analysis Rules Regarding the Relationships Between Nodes Within EAST
Networks

Network Nodes Relationships Examples

Task network Represent high-level Represent instances The nodes ‘Identify legal
tasks that are required where the conduct of constraints’ and ‘Identify
during the scenario one high-level grouping site and zoning’ are
under analysis. of tasks (i.e. task linked because the
High-level tasks are network node) zoning cannot be
typically extracted from influences, is established until the site
the sub-ordinate goals undertaken in has been legally
level of the HTA combination with or is identified
dependent on another
group of tasks
Social network Represent human, Represent instances The nodes ‘Urban planner’
technological, or where agents within the and ‘community’ are
organisational agents social network interact linked as the planner
who undertake one or with one another during needs to communicate
more of the tasks the scenario under with and understand the
involved in the scenario analysis local community if an
under analysis (as informed analysis of the
identified in the HTA site is to be established
and task network)
Information Represent grouped Represent instances The nodes ‘views’ and
network categories of where information ‘topography’ are linked
information that is influences other as the establishment of
required by agents when information or is used views requires
undertaking scenario in combination with appropriate topography
under analysis (as other information in the
identified in the task and network during the
social network) scenario under analysis
8 Systems Thinking in Practice

Step 6: Conduct Social Network Analysis

A Social Network Analysis (SNA) (Driskell and Mullen 2004) is used to analyse the
relationships (e.g. communications, transactions) between the agents involved in the
scenario under analysis. This involves first creating a social network matrix show-
ing the relationships between agents followed by a social network diagram which
provides a visual representation of the social network. Typically, the direction (i.e.
from actor A to actor B) frequency, type and content of associations are recorded. It is
normally useful to conduct a series of SNAs representing different phases of the task
under analysis (using the task phases defined during the CDM part of the analysis).

Step 7: Construct Information Networks

The final step of the EAST analysis involves constructing information networks
(see Chapter 7 for a full description) for each scenario phase identified during the
CDM interviews. Following construction, information usage should be defined for
each actor involved via shading of the information elements within the propositional
networks.

Step 8: Construct Composite Networks

Composite networks are used to explore the relationships between tasks, agents
and information (Stanton 2014). As such, composite networks are constructed
by combining the different networks. For example, a task by agents network can
be constructed by combining the task and social network to show which tasks
are undertaken by which agents. This involves assigning a colour to the different
agents within the social network and shading each node within the task network
to show which agent performs that particular task. Useful composite networks to
construct include

• Task by agents network (combined task and social network);

• Information by agents network (combined information and social network);
• Task and associated information network (combined task and information
network);
• Information by agents and tasks network (combined task, social and infor-
mation network).

Once the EAST networks are complete, it is pertinent to validate the outputs using
appropriate SMEs and recordings of the scenario under analysis. Any problems iden-
tified should be corrected at this point.

Step 9: Analyse Networks

An important component of EAST analyses involved using network metrics to anal-
yse the task, social and information networks. This enables analysis of the struc-
ture of the networks and identification of key nodes (e.g. tasks, agents, information)
The Event Analysis of Systemic Teamwork (EAST) Method 9

within the networks. Three popular network analysis metrics have previously been
used to interrogate EAST networks:

1. Network Density (overall network): Network density represents the level

of interconnectivity of the network in terms of relations between nodes.
Density is expressed as a value between 0 and 1, with 0 representing a
network with no connections between nodes and 1 representing a network
in which every node is connected to every other node (Walker et al. 2011).
Higher density values are indicative of a well-connected network in which
tasks, agents, information and controls are tightly coupled.
2. Sociometric Status (individual nodes): Sociometric status provides a mea-
sure of how ‘busy’ a node is relative to the total number of nodes within
the network under analysis (Houghton et al. 2006). In the present analy-
sis, nodes with sociometric status values greater than the mean socio-
metric status value plus one standard deviation are taken to be the ‘key’
(i.e. most connected) nodes within each network. These nodes represent
either key tasks, agents, pieces of information or controls. For example,
in the case of the social network, the node with the highest sociometric
status is the agent that is the most interrelated with other agents based on
communication.
3. Centrality (individual nodes): Centrality is used to examine the standing
of a node within a network based on its geodesic distance from all other
nodes in the network (Houghton et al. 2006). Central nodes represent
those that are closer to the other nodes in the network as, for example,
information passed from one to another node in the network would travel
through less nodes. Houghton et al. (2006) point out that well-connected
nodes can still achieve low centrality values as they may be on the
periphery of the network. For example, in the case of the social network,
nodes with higher centrality status values are those that are closest to all
other agents in the network as they have direct rather than indirect links
with them.

ADVANTAGES
• The analysis produced is extremely comprehensive and activities are anal-
ysed from various perspectives.
• The analysis is both qualitative (networks) and quantitative in nature (net-
work analysis metrics).
• Composite networks enable analysts to explore the relationships between
tasks, agents and information.
• The use of network analysis metrics enables analysts to identify key tasks,
agents and information.
• The framework can be used both retrospectively and predictively to fore-
cast system behaviour (e.g. Stanton and Harvey 2017).
• The framework approach allows methods to be chosen based on analysis
requirements.
10 Systems Thinking in Practice

• EAST has been applied in a wide range of different domains for various
purposes. The approach is generic and can be used to evaluate activities in
any domain.
• Various Human Factor (HF) concepts can be examined, including distrib-
uted situation awareness, distributed cognition, decision making, teamwork
and communications.
• It uses structured and valid HF methods and has a sound theoretical
underpinning.

DISADVANTAGES
• When undertaken in full, the EAST framework is a very time-consuming
approach.
• The use of various methods ensures that the framework incurs a high train-
ing time.
• In order to conduct an EAST analysis properly, a high level of access to the
domain, task and SMEs is required.
• Some parts of the analysis can become overly time-consuming and labori-
ous to complete.
• Some of the outputs can be large, unwieldy and difficult to present in
reports, papers and presentations.
• Reliability and validity have not yet been formally tested.

RELATED METHODS
EAST uses HTA, social network analysis and information networks. Various
methods can be used to collect the data required to construct task, social and
information networks, including concurrent verbal protocol analysis, the CDM
(Klein et al. 1986), observation, documentation review and content and thematic
analysis.

APPROXIMATE TRAINING AND APPLICATION TIMES

Due to the number of different methods involved, the training time associated with
the EAST framework is high. Similarly, application time is typically high, although
this is dependent upon the task under analysis and the scope of the analysis.

RELIABILITY AND VALIDITY

Due the number of different methods involved, the reliability and validity of the
EAST methods are difficulty to assess. Indeed, they have not yet been formally tested.

FLOWCHART
A flowchart showing the main phases of EAST is shown below, separated into the
three phases of data collection, data analysis and representation methods (Figure 1.2).
The Event Analysis of Systemic Teamwork (EAST) Method 11

FIGURE 1.2 Flowchart showing phases and associated methods for EAST.

TOOLS NEEDED
Normally, video and audio recording devices are used to record the activities under
analysis. A drawing software package such as Microsoft Visio is also typically used
to reproduce the networks. The HTA tool or other task analysis tools can be used to
support development of the task network. The AGNA social network analysis soft-
ware tool is typically used to quantitatively analyse the networks, and the Leximaner
thematic analysis tool can be used to construct information networks directly from
verbal or communications transcripts (e.g. Salmon et al. 2014b).

EXAMPLE
In the following example, EAST was used to examine a generic land use and urban
planning site analysis process (Stevens et al., 2018). Task, social and information
networks were constructed to describe the key tasks, agents and information used
during site analysis. Initially, a task network was constructed based on a HTA of a
generic site analysis process.
As shown in Figure 1.3, 11 key interrelated tasks were identified. The task
network is a dense one with many interdependencies between tasks, suggesting
that the tasks required are tightly coupled. In particular, the tasks of analysing
the neighbourhood and determining circulation patterns are the most connected
within the task network, suggesting that they are central to the overall site analy-
sis process.
A social network diagram was constructed based on identifying which agents
are required to communicate with each other during the site analysis process (see
12 Systems Thinking in Practice

FIGURE 1.3 Task network of the generic site analysis.

Figure 1.4). The social network demonstrates that there are 19 distinct agents involved
in the site analysis process. This reliance on multiple actors to assemble the required
information for the site analyses is central the work within LUP & UD disciplines.
In contrast to the well-connected task network, however, there are few connections
between the agents within the social network, suggesting that the network of agents
involved in site analysis are loosely coupled.
The social network suggests that, while many actors are necessary for the assem-
bly of the required information, they are most often working independently, while
a central organisational agent coordinates their responses. This central agent is the
urban planner and designer, with the social network diagram revealing that the most
connected agents within the site analysis process are urban planners and designers,
with connections to all other agents involved in the process. Indeed, urban planners
and designers are the only agent in the social network to have connections with more
than four other agents. This indicates that urban planners and designers are the key
agents within the site analysis process, with the local council being the next most
connected. The structure of the social network diagram suggests that there may some
simplistic interventions that could improve the site analysis process. Logical inter-
ventions would be to attempt to increase the connectivity of the network through
incorporating a requirement for further communication between agents and to reduce
the load placed on urban planners and designers.
The Event Analysis of Systemic Teamwork (EAST) Method 13

FIGURE 1.4 Social network of the agents in the site analysis process.

An information network showing the information required during the site analy-
sis process was constructed based on the HTA (see Figure 1.5). According to the
information network, multiple sources of information are required to complete the
site analysis process, ranging from information on locations, topography, views,
drainage, climate and utilities to traffic, urban form, sensory elements and commer-
cial, retail, residential and community functions.
Examining the connectedness of different nodes within the network suggests
that there are various critical pieces of information required, including locations,
commercial functions, neighbourhood context, road hierarchy, residential functions,
retail functions, pathways and travel times.
To demonstrate the composite network function of EAST, the task and social net-
works were combined to show which agents are involved in the 11 key site analysis
tasks. The task by agent network is presented in Figure 1.6.
14 Systems Thinking in Practice

FIGURE 1.5 Information network for the site analysis task.

The composite task and social network shows the allocation of tasks across dif-
ferent agents. Again, the critical role of the urban planners and designers is high-
lighted, as they are the only agent that is constant across all 11 site analysis tasks.
In addition, the network demonstrates the key requirement for multi-stakeholder
input across the tasks. For example, tasks such as ‘Identify human and cultural ele-
ments’ and ‘Identify natural physical features’ have eight and seven agents involved,
respectively. When this is considered with the social network, which shows that
there are few agents communicating with one another through the site analysis pro-
cess, it suggests that the process could be made more efficient through the addition
of mechanisms designed to enhance communications and interactions between all
of the stakeholders involved. The remainder of this book will show applications of
EAST to a variety of domains as well as some developments and extensions to the
approach.
The Event Analysis of Systemic Teamwork (EAST) Method 15

FIGURE 1.6 Combined task, social and information network showing site analysis tasks by
agent.

REFERENCES
Driskell, J. E. and Mullen, B. (2004). Social network analysis. In: N. A. Stanton, A. Hedge E.
Salas, H. Hendrick and K. Brookhuis (Eds.), (2005). Handbook of Human Factors and
Ergonomics Methods (pp. 58.1–58.6). Taylor & Francis: London.
Hollnagel, E. (2012). FRAM: The Functional Resonance Analysis Method: Modelling
Complex Socio-Technical Systems. Ashgate: Aldershot.
Houghton, R. J., Baber, C., Cowton, M., Walker, G. H. and Stanton, N. A. (2008). WESTT
(Workload, Error, Situational Awareness, Time and Teamwork): An analytical pro-
totyping system for command and control. Cognition, Technology and Work, 10 (3),
199–207.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P., Stewart, R. and Walker,
G. (2006). Command and control in emergency services operations: A social network
analysis. Ergonomics, 49 (12–13), 1204–1225.
Karsh, B. T., Waterson, P. and Holden, R. J. (2014). Crossing levels in systems ergonomics:
A framework to support ‘mesoergonomic’ inquiry. Applied Ergonomics, 45 (1), 45–54.
Klein, G. A., Calderwood, R., and MacGregor, D. (1989). Critical decision method for elicit-
ing knowledge. IEEE Transactions on Systems, Man, and Cybernetics, 19 (3), 462–472.
Kleiner, B.M. (2006). Macroergonomics: Analysis and design of work systems. Applied
Ergonomics, 37 (1), 81–89.
Leveson, N. (2004). A new accident model for engineering safer systems. Safety Science, 42
(4), 237–270.
Read, G. J. M., Salmon, P. M. and Lenné, M. G. (2016). When paradigms collide at the road–
rail interface: Evaluation of a sociotechnical systems theory design toolkit for cognitive
work analysis. Ergonomics, 59 (9), 1135–1157.
Salmon, P. M., Lenné, M. G., Walker, G. H., Stanton, N. A. and Filtness, A. (2014a). Using the
Event Analysis of Systemic Teamwork (EAST) to explore conflicts between different
road user groups when making right hand turns at urban intersections. Ergonomics, 57
(11), 1628–1642.
16 Systems Thinking in Practice

Salmon, P. M., Lenné, M. G., Walker, G. H., Stanton, N. A. and Filtness, A. (2014b). Exploring
schema-driven differences in situation awareness across road users: An on-road study
of driver, cyclist and motorcyclist situation awareness. Ergonomics, 57 (2), 191–209.
Salmon, P. M., Walker, G. H., Read, G. J., Goode, N. and Stanton, N. A. (2017). Fitting meth-
ods to paradigms: Are ergonomics methods fit for systems thinking? Ergonomics, 60
(2) 194–205.
Stanton, N. A. (2006). Hierarchical task analysis: Developments, applications and extensions.
Applied Ergonomics, 37, 55–79.
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57 (3), 403–418.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork. Ashgate: Aldershot.
Stanton, N. A. and Harvey, C. (2017). Beyond human error taxonomies in assessment of risk
in sociotechnical systems: A new paradigm with the EAST ‘broken-links’ approach.
Ergonomics, 60 (2), 221–233.
Stanton, N. A., Salmon, P. M., Rafferty, L. A., Walker, G. H., Baber, C. and Jenkins, D.
(2013). Human Factors Methods: A Practical Guide for Engineering and Design (sec-
ond edition). Ashgate: Aldershot.
Stanton, N. A., Salmon, P. M., Walker, G. H., Baber, C. and Jenkins, D. (2005). Human
Factors Methods: A Practical Guide for Engineering and Design (first edition).
Ashgate: Aldershot.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon, P.,
et al. (2006). Distributed situation awareness in dynamic systems: Theoretical devel-
opment and application of an ergonomics methodology. Ergonomics, 49 (12–13),
1288–1311.
Stevens, N., Salmon P. M., Walker, G. H. and Stanton N. A. (2018). Human Factors in Land
Use Planning and Urban Design: Methods, Practical Guidance, and Applications.
CRC Press: Boca Raton, FL.
Stewart, R., Stanton, N. A., Harris, D., Baber, C., Salmon, P., Mock, M., Tatlock, K., Wells,
L. and Kay, A. (2008). Distributed situation awareness in an airborne warning and
control system: Application of novel ergonomics methodology. Cognition, Technology
and Work, 10 (3), 221–229.
Svendung, I. and Rasmussen, J. (2002). Graphic representation of accident scenarios: Mapping
system structure and the causation of accidents. Safety Science, 40 (5), 397–417.
Vicente, K. (1999). Cognitive Work Analysis: Towards Safe, Productive and Healthy Work.
Lawrence Erlbaum Associates: Mahwah, NJ.
Walker, G. H., Salmon, P. M., Bedinger, M. and Stanton, N. A. (2017). Quantum ergonomics:
Shifting the paradigm of the systems agenda. Ergonomics, 60 (2), 157–166.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. M. and Jenkins, D. P.
(2010). From ethnography to the EAST method: A tractable approach for representing
distributed cognition in air traffic control. Ergonomics, 53 (2), 184–197.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2011). Cognitive compatibility of motorcy-
clists and car drivers. Accident Analysis & Prevention, 43 (3), 878–888.
Section II
Applications of EAST
2 EAST in Air Traffic
Control
With Chris Baber, Linda Wells,
Huw Gibson and Daniel P. Jenkins

INTRODUCTION
Command and Control
In command and control scenarios, there is a common goal (comprising interact-
ing sub-goals), there are multiple individuals who need to communicate and coor-
dinate with each other in order to attain these goals and, increasingly, there are
ever more complex ways of facilitating this process with technology. Command
and control, at the most generic level, can be viewed as a form of management
infrastructure for planning and organisation (Harris and White 1987). It involves
the exercise of authority and direction by properly designated individual(s) over
assigned resources, as well as planning, directing, coordinating and controlling how
those resources are deployed (Builder et al. 1999). ‘Command’ can be viewed as
the definition of overall system objectives or goals, whereas ‘control’ is the manage-
ment of process and activities that lead to the achievement of these objectives (or
compensate for changes in the environment that hinder their achievement). Many
contemporary sociotechnical systems involve authority, planning, directing and
coordinating and can be considered as forms of command and control. Air Traffic
Control (ATC) is one example.

Distributed Cognition
From a distributed cognition perspective (e.g. Hutchins 1995a), the task of ATC can
be viewed as a form of ‘“computation” to maintain separation between aircraft in a
region of airspace’ (Fields et al. 1998, p. 86). It is further argued that the computa-
tions do not reside solely in the heads of individual controllers; instead, they are dis-
tributed across the entire ATC system, comprising numerous controllers, teams and
technical artefacts. The essence of distributed cognition is on ‘how [these computa-
tions] transcend the boundaries of the individual actor’ (Rogers 1997, p. 1; Hutchins
1995a; Hollan et al. 2000).
The language of representational states is used to describe the visible and external
manifestations of various ‘environmental contributions’ to the total system (Rogers
and Ellis 1994; Fields et al. 1998). Representational states subsume the full range of

19
20 Systems Thinking in Practice

observable interactions between people and artefacts, as well as the resulting states
(and state changes) that arise from the various ‘computations to maintain separation
between aircraft’. For example, an observable interaction might be a controller issu-
ing an instruction to an aircraft. The resulting state might be a corresponding change
in the aircraft’s representation on the radar display.
In command and control situations, these computations and representational
states interact. A change in representational state leads to further computations, fur-
ther representational states and further computations. But whilst these simple low-
level mechanisms can be multiplied in simple ways to form the total ATC system,
the high-level function, the system’s aggregate behaviour, can be highly complex
and adaptive (Chalmers 1990). Phenomena ‘wherein complex, interesting high-
level function is produced as a result of combining simple low-level mechanisms in
simple ways’ are referred to as ‘emergence’ (Chalmers 1990, p. 2). A key emergent
property of ATC is the so-called ‘picture’ or, in ergonomics parlance, situational
awareness (SA).

Distributed Situation Awareness

The ability to sense changes in representational states, understand them and then
perform some kind of computation based on them not only describes the essence
of distributed cognition but also that of SA. At an individual level, SA is about
the psychological processes, and information in working memory required, for
developing ‘the picture’ (e.g. Endsley 1995; Bell and Lyon 2000). The controller’s
picture is not the same as the radar display; it does not arise solely from any one
physical or human component but from the interaction of many such components.
It arises both from the parts of the ATC task that have been overtly designed (i.e.
radar displays and other prescribed forms of communication) and those parts of the
task that have not been overtly designed (i.e. the clicking of new flight data strips,
which informs controllers that a new aircraft is about to enter their sector). Because
the picture, or SA, arises from these myriad individual components, yet its totality
cannot be predicted solely from any one of them individually, it can be referred to
as emergent.
Individualistic approaches to SA dominate ergonomics, but whilst they may be
appropriate for tasks that are performed by individuals in isolation, few complex
tasks are performed entirely independently of others (Perry 2003). In systems terms,
SA is what helps entire sociotechnical systems such as ATC to be orientated towards
and ‘tightly coupled to the dynamics of the environment’ (Moray 2004, p. 4). A dis-
tributed cognition perspective applied to command and control scenarios requires a
shift from traditional notions of SA that focus on the individual (e.g. Endsley 1995)
to those that focus on the system (e.g. Sandom 2001; Gorman et al. 2006; Salmon
et al. 2008). At face value, the response to this might be the concept of team SA (e.g.
Salas et al. 1995; Perla et al. 2000) but even here there are problems. ‘The degree to
which every team member possesses the SA required for his or her responsibilities’
(Endsley 1995, p. 39) runs into difficulties when confronted with the twin concepts
of ‘overlapping SA’ (i.e. portions of SA that are identically shared between people,
normally represented as a Venn diagram) and ‘compatible SA’ (i.e. that which is
EAST in Air Traffic Control 21

‘not’ overlapping between two or more people, but which fits together like a jigsaw;
Salmon et al. 2008). The fundamental problem stems from a tacit assumption that
the ‘situation’ can be defined as a single, objective, external reality and that the goal
of the people operating within the situation is to respond to all features appropriately.
This is problematic on three counts:

1. There are many aspects of command and control scenarios that require the
individual to make judgements and interpretations (so the assumption of the
‘objective reality’ of a situation is not always valid).
2. There are multiple sub-goals and multiple views of the situation (so the idea
of a single reality is not valid either).
3. Diﬀerent agents within the system use diﬀerent representational states to
inform and support their work, so the notion that there can be a single view
of the situation (as opposed to several interlocking views) is also not easily
supported.

One way to resolve the mismatch between mainstream thinking in SA and distrib-
uted cognition of the sort encountered in complex systems, such as ATC, is to con-
sider one of the relatively invariant properties of it: information. According to Bell
and Lyon (2000): ‘all aspects of momentary SA are eventually reducible to some
form of [. . .] information in working memory’ (p. 42). Information, in the SA sense,
refers to what, in distributed cognition language, are called representational states.
The question to ask is whether ‘working memory’ is the only place where such
states can be represented. Distributed cognition would suggest not. It suggests that
non-human artefacts can create, manage and share such states (to some extent at
least), meaning that the technical aspects of a sociotechnical system will be con-
tributing to the exchange of representational states too. The totality of this will be a
form of systems level awareness that is not traceable to any one individual and is not
consistent with a distributed cognition view of the world, nor resides exclusively in
the minds of humans. Thus, not only is the individual-level ATC ‘picture’ emergent,
so too is the systems level ‘picture’.
Another consideration is that representational states can be promulgated around
the system with very little in the way of overt communication. One of the great
strengths of expert operators is their ability to chunk information, to abstract and
pattern match, to develop a high level of awareness from relatively little information
in the world (e.g. Chase and Simon 1973). The update of a representational state for
one agent might lead to partial updating of that used by another. For example, as an
aeroplane moves across a sector, its route is plotted on the ATC displays and its posi-
tion updated dynamically; if there is no definable risk, then the updating happens
automatically and without the need for intervention.

Beyond Ethnography
Whilst there is much to say in favour of blending some of the ideas connected
with SA and distributed cognition, there are several problematic issues when it
comes to actually applying either approach. It should be clear by now that purely
22 Systems Thinking in Practice

individualistic approaches to analysis may not capture all the required information
about complex command and control scenarios. It is also the case that the tradi-
tional experimental approach may lack predictive efficiency in the face of significant
numbers of emergent behaviours. Distributed cognition, focusing as it does ‘on the
material and social means of the construction of action and meaning’ (Hollan et al.
2000, p. 178), employs ethnography to understand how information is used to sup-
port decision making, how it is represented and how it is manifest in the physical
and social world (e.g. Rogers and Ellis 1994; Hutchins and Klausen 1996; Hutchins
and Holder 2000; Hutchins and Holder 2001). Ethnography is a form of naturalistic,
qualitative description based on observation. True ethnography places a requirement
to live as a member of the ‘tribe’ for an extended period (possibly weeks or months)
and to undertake the tasks and rituals of the tribe. The understanding gained from
this participatory observation normally forms the basis of a report written from a
first-person perspective as a participant observer. In the field of distributed cogni-
tion, ethnographic methods are deployed. The problems with ethnography in terms
of applying it in practice are as follows:

1. The outputs of ethnographic approaches (as they are used currently)

remain couched at a qualitative, often highly discursive level of analysis
(e.g. Hutchins 1995b; Hutchins and Klausen 1996; Hutchins and Palen
1997). Such outputs may not be easily reconciled with a predominantly
engineering, or at least non-social science, audience (McMaster and Baber
2005).
2. Depending on the level of analysis, having the necessary requirement for an
observer to be ‘imbued’ in the culture of the scenario can, in some cases, be
incompatible with objectivity and validity in measurement (e.g. Hutchins
1995b).
3. Ethnographic analyses are not always easily amenable to generalisation,
concerned as they are with specific and localised scenarios and phenom-
ena. It is certainly difficult to imagine some analyses being amenable to
repetition.

These issues limit the practical value of distributed cognition in command and
control settings and serve as a barrier to employing what is potentially a useful
and enlightened approach to the analysis of complex systems. A response to this
problem, however, appears to lie in the fundamentals of the distributed cognition
approach itself. By characterising complex systems using the language and meta-
phors of cognitive science, many of the phenomena of interest, such as communica-
tions, shared awareness and other determinates of decision superiority, are rendered
physically and observably manifest. That is to say, they no longer reside just in the
heads of individual actors; they become manifest in the way that information is repre-
sented, modified, communicated and shared. The purpose of this chapter is to deploy
the Event Analysis for Systemic Teamwork (EAST) method as a way of capturing
these phenomena, to show that there are ways to leverage the favourable theoretical
perspective of distributed cognition that are also appropriate and relevant to systems
designers.
EAST in Air Traffic Control 23

THE AIR TRAFFIC CONTROL WORK SETTING

A case study from the ATC work domain is used to illustrate the complementary
approach to ‘doing’ distributed cognition. The primary strategic objective of ATC is
safety. This is achieved by providing instructions on height, speed and route to aircraft
pilots so that individual aircraft maintain legally mandated physical separation criteria
(typically 3 or 5 miles horizontally and 1000 feet vertically; Civil Aviation Authority
2004). A secondary tactical objective is to optimise the routing of aircraft so that they
take oﬀ and land at prescribed times and follow the most expeditious routes to desti-
nations. These strategic goals translate into the following operational level activities:

• Keeping in radio and/or radar contact with aircraft.

• Instructing aircraft in relation to speed, altitude and direction.
• Providing information to aircraft about weather conditions.
• Ensuring that minimum distances are maintained between aircraft.
• Handling unexpected events, emergencies and unscheduled traffic.

Underlying these activities are several key non-human artefacts. Like their human
counterparts, they contain, represent and modify information and are part of a sys-
tem that a distributed cognition perspective is able to model. Fields et al. (1998) pro-
vide a cogent summary of the key informational artefacts contained within the ATC
work domain. These are as listed below.

Charts and Standard Routes

These can be seen as a prescription of how certain manoeuvres should be made.
The information is shared between controllers and pilots as well as being stored on
aircraft flight management systems (Fields et al. 1998).

Flight Data Strip

This can be seen as a representation of the projected state of an aircraft. It takes the
form of a paper strip that contains information on aircraft height, speed, heading and
call sign. This information is modified by the controller by hand when instructions
are issued and confirmation of receipt is received (Fields et al. 1998).

Flight Data Strip Bay

Whereas the individual flight data strip is a projected state of an individual aircraft,
the totality of data strips that refer to a region of airspace provides a ‘schematic
model’ of air traffic progress (Fields et al. 1998, p. 87).

Radar Display
‘The radar screen provides controllers with a snapshot of the current horizontal [. . .]
locations of aircraft’ annotated with individual aircraft call signs and height (Fields
et al. 1998, p. 88).
24 Systems Thinking in Practice

Aircraft Call Signs

These are the unique alpha-numeric codes given to individual aircraft. Aircraft call
signs serve ‘as an important (indeed the only) means of coordinating the information
represented in [. . .] various media’, such as the flight strip and radar display (Fields
et al. 1998, p. 88).

Communications
These represent, in a practical sense, the ‘mediation of control’. Fields et al. (1998)
argue that communications can also be seen as a ‘system of representations’ (p. 4).
Of critical importance is that communications ‘are also situated in the network of
artefacts and information, and are [only] made comprehensible by reference to a
larger context of shared representations’ (p. 88).

DISTRIBUTED COGNITION METHODOLOGY:

THE IMPORTANCE OF METHODS
Explicit methods lie at the heart of ergonomics as a discipline, enabling the practi-
tioner to vary their approach between scientist (i.e. testing and developing theories
of human performance using rigorous data collection and analysis techniques) and
practitioner (evaluating the eﬀects of change, developing best practice and, funda-
mentally, addressing real-world problems). Ergonomics methods are useful in the
scientist–practitioner model because of the structure and potential for repeatability
that they oﬀer over and above ethnography alone.

Descriptive vs. Formative Methods

This study uses the EAST method in an attempt to reconcile distributed cognition
with the methodological traditions of ergonomics. EAST is based on the integration
of seven individual methods, which in turn is a reflection of the multi-faceted nature
of the command and control. In other words, no one method can adequately describe
all of the degrees of freedom inherent in such a complex sociotechnical system. That
said, no such claim is made for the EAST method, but it can be argued that at least
some of the major human dimensions of the problem space can be explored by taking
a multi-method approach.
EAST is a descriptive method. It does not specify a formal architecture and what
‘should’ happen. Even though it uses normative methods such as task analysis and
process modelling, these are populated with data on what is actually observed. Neither
does EAST focus on constraints, boundaries and a problem space defined forma-
tively by the scope of what ‘could happen’, such as cognitive work analysis (CWA).
EAST focuses on what ‘did’ happen. CWA admits the possibility of non-linear and
emergent behaviour, whereas EAST is designed to identify specific instances of it.
Despite their descriptive vs. formative diﬀerences, EAST and CWA are both rep-
resentative of a shift in methodological thinking. They share two key aspects: both
acknowledge that complex sociotechnical systems require more than one approach
EAST in Air Traffic Control 25

(EAST comprises seven individual methodologies, CWA comprises five ‘phases’)

and both acknowledge that these perspectives are as interlinked as the complex
sociotechnical phenomenon under analysis (i.e. they are both systemic in nature).
This is a core principle of sociotechnical design (Clegg 2000).

Method Integration
The following formal methodologies combine to form EAST: hierarchical task
analysis (HTA) (Annett); coordination demand analysis; communications usage
diagram; social network analysis (SNA) (Driskell and Mullen 2005); propositional
networks; and an enhanced form of operation sequence diagram. A multiple method
approach has a number of compelling advantages. Not only does the integration
of existing methods bring reassurance in terms of a validation history, but it also
enables the same data to be analysed from multiple perspectives. Also, with over
200 existing methodologies to choose from (Stanton et al. 2005) there seemed little
pragmatic need to develop yet more. Of course, multiple interconnected methods
require greater eﬀort to analyse, but a companion to EAST, called workload, error,
situation awareness, tasks and time, is designed to help. This is a software tool that
greatly streamlines and simplifies the application of the method and it was used in
the current analysis (Houghton et al. 2006).
EAST is structured as follows. The HTA provides input into the analysis of
teamworking (CDA), communications usage (CUD) and the linkage (via communica-
tions) between agents (SNA). Data for the HTA are gathered from live observation
of the scenario. The output of all these methods (HTA, CDA, CUD and SNA) is
given a summary visual form by using an enhanced Operation Sequence Diagram
(OSD). Interview data, in the form of the critical decision method (CDM) (Klein and
Armstrong 2005), are used to create a network of linked ‘information objects’ or
representational states. This representation is called a propositional network and is
rather similar in concept to semantic networks. It is important to note at this point that
the purpose of this chapter is not to introduce the methodological intricacies of the
EAST method (the reader is referred to Chapter 1 for further detail) but to show the
eﬀect of method integration in terms of enabling a distributed cognition perspective
upon ATC.

AIR TRAFFIC CONTROL SCENARIOS

Data were collected from the ATC work domain between 28 and 30 June 2004 at a
major UK terminal and area control centre. Separate analyses were performed on
four discrete scenarios that were observed to take place repeatedly. By dividing up the
controller’s task in this way, it was possible to capture something of the dynamics of
the system. In practice, multiple aircraft are presenting themselves to the controller(s),
prompting them to engage interchangeably in one of the following four behaviours:

• Scenario 1: To bring aircraft inbound from a major air route into a hold-
ing stack and then pass them onto an aerodrome controller (the holding
scenario)
26 Systems Thinking in Practice

• Scenario 2: To deal with aircraft that have left the holding stack and are
en route to airfield(s) but have yet to enter the final approach phase (the
approach scenario)
• Scenario 3: To deal with over-flying aircraft in such a way as to avoid
conflict with the holding stack and other en-route aircraft (the over-flight
scenario)
• Scenario 4: To deal with departing aircraft in such a way as to avoid conflict
with the holding stack (the departure scenario)

Applying the Method

The first stage of the EAST method involves a two-step process of observation and
interview (Annett 2005). This is similar in some respects to the ethnographically
based approaches currently used in distributed cognition research, but here there is
not a particular requirement for the observer to be directly ‘imbued’ in the scenario.
The focus instead is on unobtrusive observation. The bulk of the analysis derives from
live audio feeds, which detail who is communicating to whom and about what. These
data are supplemented by the observers’ notes and by in-depth technical critique and
insight provided by subject matter experts. Whilst observational techniques provide
information on the observable artefacts of interaction, they produce limited data on the
representational states internal to individuals. It is to the CDM (Klein and Armstrong
2005) that relevant insights can be provided. The CDM is a semi-structured interview
technique that uses cognitive probes in order to elicit information on expert decision
making. This was administered to each participant in relation to each scenario.

Representing Distributed Cognition

The data collection methods provide information on the activities performed within
ATC, how those activities are facilitated and how non-human artefacts participate
in the joint cognitive system by containing, representing and transforming infor-
mation. Using more conventional ethnographic techniques, these facets of a sce-
nario would be described, exhaustively, in a first-person written account. While this
can capture the nuances of the situation as it was observed and is often illustrated
with diagrams and photographs of the work setting, it is not easy to summarise the
account in a manner that brings out common features. In this chapter, the EAST
method is used to present multiple views of the situation in such a way as to comple-
ment an ethnographic account, if such an analysis were to be performed (which, in
a lot of ergonomics studies, would be unlikely). The EAST method maps the tasks
being performed (in a task network), the communications between agents (in a social
network) and the representational states being sensed (in a propositional network).
Figure 2.1 illustrates the relationships between these three views.

Task Networks
HTA is a means of describing a system in terms of a structured hierarchy of goals
and sub-goals with feedback loops (Annett 2005). Its appropriateness in this instance
EAST in Air Traffic Control 27

FIGURE 2.1 The three network-based approaches provide a detailed characterisation of a

complex sociotechnical system.

can be put down to two key points. First, it is inherently flexible, and the approach
can be used to describe any system. Second, it can be used for many ends, from per-
son specification and training requirements to error prediction, team performance
assessment and system design. Key to its present application is its ability to model
the temporal relations between tasks and the external conditions upon which task
activity is cued. The task network, specifically, is a way of representing the detailed
task analysis in terms of the interaction of higher-level goals. This technique allows
complex task analyses to be easily compared according to diﬀerences in overall task
structure and type. The task analysis/network forms the foundation for developing
insightful social and propositional networks.

Social Networks
SNA is particularly compatible with the distributed cognition perspective because
it ‘focuses on the relationships among actors embedded in their social context’
(Driskell and Mullen 2005, p. 58). Furthermore, it can represent the technologi-
cal mediation of communication and networks in which some of the nodes are
non-human. A social network is a way to represent individuals or teams who are
linked by communications to each other and to subject that network to mathematical
analysis using Graph Theory (Driskell and Mullen 2005). Two mathematical indices
are used in the present analysis, namely centrality (i.e. a numeric ranking allowing
key agents in the network to be identified) and density (the interconnectivity of the
network as a whole). The metrics reveal certain important characteristics of the net-
works to be revealed, in turn allowing comparisons between them.
28 Systems Thinking in Practice

Propositional Networks
Propositional networks offer a novel and effective means of visualising represen-
tational states as held/experienced by the individuals at work within the system.
The data used to construct them are based on the outputs of the CDM, in which
a content analysis of the interview transcripts permits representational states to
be extracted and causal links between them defined. Representational states take
the form of specific knowledge objects within these networks, which, in turn, are
analogous to propositions, that is, entity or phenomena about which an individual
requires information in order to act effectively. The propositional network offers
four perspectives.

1. They do not diﬀerentiate between diﬀerent types of representational state

(e.g. information related to states, people or ideas); therefore, from a design
perspective, they do not constrain assessments to consideration of existing
configurations of people and states but rather to the required representa-
tional states associated with a scenario (Stanton et al. 2005; Walker et al.
2006).
2. The network shows the totality of information used in the scenario (within
the constraints of the data collection techniques used) regardless of whether
agents in the scenario are human or technical.
3. Shared SA can be accessed from the CDM, in which multiple agents can be
attributed to common knowledge objects/states within the network.
4. The dynamic aspects of SA can also be captured by animating the propo-
sitional network. This is achieved by highlighting active and non-active
knowledge objects/states occurring in diﬀerent task scenarios.

These individual network-based outputs not only meet the need to go beyond
reductionism (and focus on interconnections as well as parts) but they can also be
linked to provide several diﬀerent perspectives on the scenario.

APPLICATION TO AIR TRAFFIC CONTROL

Analysis of Agents in the Distributed Cognition System
The social network representation provides a particularly powerful example of dis-
tributed cognition, in particular, the idea of a joint cognitive system of collaborating
human and non-human agents. The SNA of all the actors participating in the ATC
task shows that there are 21 agents in the network, 14 human and 7 non-human
(shown as shaded in Figure 2.2), joined by 22 communication links. Note that this
network has been simplified somewhat. The number of aircraft that an individual
controller would be expected to handle could be significantly larger than the four
shown for illustration in Figure 2.2.
The metrics ‘status’ and ‘centrality’ are used to identify key agents in the sce-
nario, and these are shown in Table 2.1.
Between them, they indicate the degree of connectedness an individual agent
has and, consequently, the amount of influence that agent has on the performance
EAST in Air Traffic Control 29

FIGURE 2.2 Social network diagram illustrating agents in the air traffic control (ATC)
work domain linked by communication (emphasis denotes non-human agent).

of the network as a whole. Key agents can be defined as having high status and high
connectedness. The analysis indicates that the controller (as one might expect) has
high status and centrality in the network, but the metric also reveals that the most
significant agent is in fact non-human; the ATC technical system (which is the tech-
nical infrastructure and assemblage of ground aids, communications systems, radar
displays, etc.). A great deal of information is received, stored, displayed or otherwise
transformed by this (non-human) agent. This is a similar finding to that of Houghton
et al.’s (2006) analysis of police control rooms (in which the central logging com-
puter was also the most central agent).
In addition to the distribution between collaborating human and non-human
agents, cognition is also distributed in a geographical sense, between, for example,
aircraft in flight within the sector, controller(s) based at the control centre and con-
trollers based at airfield(s). Table 2.2 presents the number of times communication
links cross these geographical boundaries.
Geographical dispersion is further illustrated in geographical regions. The type of
communications media is physically constrained in cases where links between nodes
30 Systems Thinking in Practice

TABLE 2.1
Results of Social Network Centrality for
Individual Agents (in Descending Order)

Agent Centrality
ATC System 16.8
Controller 14.0
Heathrow Director 12.3
Desk Coordinator 12.3
Welin Controller 11.6
Adjacent Controller 11.3
Flight Strip Computer (Heathrow) 10.5
Senior Watch Assistant 10.2
FMS 1 10.0
FMS 2 10.0
FMS 3 10.0
FMS 4 10.0
Flight Strip Computer (West Drayton) 10.0
Group Supervisor 10.0
Deputy Watch Manager 10.0
Watch Manager 10.0
Flight Strip Assistant 9.5
Pilot 2 9.3
Pilot 3 9.3
Pilot 4 9.3
Pilot 1 9.1

cross from one shaded region to the other. For example, verbal communications are not
physically possible between controllers on the ground and pilots in flight without some
sort of facilitation by technical means (such as radio). There remain a host of other local
conditions that may also require some form of technological mediation. For example,
individual controllers have to remain at their workstations during their shift in order to
communicate with controllers who are not immediately adjacent, the telephone is used
and/or the agent acting in the role of desk coordinator may have to facilitate.

TABLE 2.2
Number of Communications Links That Exist
between Geographically Disperse Locations

From/To Number of Links

Aircraft to Airfield 3
Aircraft to Traffic Control Centre (TCC) 6
TCC to Airfield 3
EAST in Air Traffic Control 31

Facilitating Technology
Figure 2.4 shows how ATC operations are facilitated by seven types of communica-
tion media.
Face-to-face communication comprises verbal communications. Face-to-face
communication also contains a visual component. Controllers will point and demon-
strate visually to aid in understanding instructions. Telephone and radio technology
facilitates voice communication. Radio also facilitates the dissemination of data,
as do data network facilities. Written communication is dominated by the flight
data-strip aspect of the task, although machine interfaces (e.g. the radar display)
also contain written information as well as visual representations. Table 2.3 provides
a summary by crossing communications modality with communications media to
provide a technology/modality matrix.
ATC is a highly evolved and proceduralised work domain, and the prominent role
of implicit communications is also noted. The CDM interview gives access to some of
these unobservable artefacts of interaction, one of which is a system of passing aircraft
between sectors that does not rely on explicit verbal exchanges. The system works by
putting aircraft into a particular position in the new sector at a pre-agreed height, speed
and heading. As soon as the controller sees an aircraft in this position, he/she knows it
is being passed onto them, and they can take control of it. A wider awareness is gained
via open channel radio communications and instructions to aircraft overheard from
other controllers. Several contextual factors, such as the high tempo of operations and
the proceduralised nature of the task, enable this level of shared understanding and
this form of communication to take place. From a distributed cognition perspective, it
can be seen how a comparatively lengthy verbal exchange is often replaced with a far
simpler visio-spatial task (facilitated by external technological artefacts such as the
radar display and flight data strips; Hutchins and Klausen 1996). These are captured
in the social networks. In other words, the presence of a link does not necessarily have
to connote just an overt form of communication but also implicit ones.

Control Architecture
Although Figures 2.2 through 2.4 are visually complex, it is possible to discern cer-
tain features of the social network. It can be noted that there is a diverging hierarchy

TABLE 2.3
Communications Modality/Technology Matrix

Technology
Modality Phone Radio Network Strips MMI In-Person
Verbal
Written
Data
Visual/Other
32 Systems Thinking in Practice

FIGURE 2.3 Geographical dispersion of human and non-human agents in the air traffic
control (ATC) work domain.

from the Watch Manager level downwards. This hierarchy splits into the Group
Supervisor level (and downwards towards individual controllers) and the Senior
Watch Assistant level (and downwards towards Flight Strip Assistants). According
to Dekker (2002), this so-called ‘split architecture’ ‘provides some of the benefits
of centralised planning with tactical adjustments to new information from subordi-
nate units’ (p. 5). At this level of ‘air traffic management’, the inherent complexity
appears to justify a degree of hierarchical sub-division so that aspects of complexity
can be spread across agents. The possible trade-oﬀ in situations where the state of
the world is changing rapidly (a fast tempo), and where decisions need to be enacted
quickly, is that any delay in the dissemination of information through a hierarchy
becomes critical (Dekker 2002).
At the tactical or ‘air traffic control’ level, in cases where one agent requires some
form of assistance, or to avoid the degradation of aircraft separation, the controller
can interact sideways through the structure to an adjacent controller so that issues can
be resolved and task load shared or re-distributed quickly. This pattern of communi-
cations is known as peer to peer or a negotiation architecture. In these instances, each
agent is more or less independently responsible for a defined area (Dekker 2002) and
is able to react promptly to rapid changes in the environment.
EAST in Air Traffic Control 33

FIGURE 2.4 Social network overlain with the media that facilitates communication.

A further characteristic of the scenario (from a structural perspective) is the

amount of information sharing. This relates back to the ATC system and its dis-
plays (which subsumes the entire radar, aircraft identification, warning, communi-
cation and ground systems in general) being the highest-ranking agent in terms of
‘centrality’. Overlain across these two interleaved command architectures (split and
negotiated) is the provision of high-quality information to facilitate decision making
within them. This overall configuration, known as a negotiation architecture with
information sharing, is an emerging paradigm in several alternate domains, where
it is referred to as network centric. The benefits of this architecture are agility, fast
response and the ability to quickly organise and re-organise. In the theoretical work
by Dekker (2002), the negotiation architecture with information sharing proved to be
the most eﬀective in high-tempo tasks.

Systemic Situational Awareness

Systemic SA is modelled using propositional networks (Figure 2.5). From these net-
works, it is possible to divine certain structural characteristics of the knowledge base
that underpins eﬀective SA for the total ATC task.
Central nodes in the network are identified using a centrality metric based simply
on divining nodes with five or greater links. Table 2.4 presents an analysis of core
knowledge objects for each individual scenario based on this criterion. By repre-
senting and simplifying SA in this manner, the table shows that depending on the
34 Systems Thinking in Practice

FIGURE 2.5 Knowledge network representing the entire knowledge base for the air traffic
control (ATC) work domain. Knowledge that is active during task enactment represents situ-
ational awareness at the systems level.

TABLE 2.4
Summary of Key Knowledge Objects (KO) that are Active
During Different Phases of Air Traffic Control Process

Knowledge Object Departure Over-Flight Holding Approach

Pressures
Runway
Radar
Stack
Aircraft
Flight Level
Strip
Position
Read-back
Instruction
Acknowledge
Call Sign
Flight Plan
Plan
Options
Sector

Note: The table demonstrates changes in systems level state of situational awareness.
EAST in Air Traffic Control 35

task (which occurs at diﬀerent points in time), not only does the type of information
change, but also, by implication, so does the structure of it. This information can be
used to explore and understand the information needs of controllers and, moreover,
how one might go about supporting them.

Temporal Aspects of Command and Control in Air Traffic Control

The controller will typically switch backwards and forwards among the four scenar-
ios detailed above. As they do so, not only do the type and structure of information
change (as shown in Table 2.4) but so also do the characteristics of the task and social
networks. To visualise the temporal aspects of the interrelations between the net-
works requires an alternate form of representation. Some form of animation seems a
likely candidate for bringing this facet to life, but this and further refinements to the
method, though eminently feasible, are within the purview of future work.

CONCLUSIONS
The purpose of this chapter has been to show how the EAST method puts ergonomic
analyses in touch with the distributed cognition perspective, rendering the output
much more tractable than comparable ethnographic techniques. Although this chap-
ter is necessarily couched at a summary level of analysis, the key characteristics
of the ATC work domain are identified in Table 2.5. Within the table is a checklist

TABLE 2.5
Key Characteristics of the Air Traffic Control Work Domain and the
Network Approach in which Detailed Systems Level Insights Reside

Specified in Detail by:

Task Social Knowledge
Key Characteristic Network Network Network
The coordination of individuals and teams to achieve a
common goal (comprising separate though interacting
sub-goals, knowledge and situational awareness [SA])
The technologically mediated geographical dispersion of
‘agents’ (from controllers located in area control centres, or
actual airfields, to specific aircraft in controlled airspace)
A substantial technology infrastructure (providing high-quality,
accurate, real-time information via seven different
communications modalities), which supports the dispersion and
representation of knowledge and a systems level state of SA
A high degree of information sharing (where most agents
have access to parts of the technology infrastructure)
A highly evolved and proceduralised mode of operation
(involving implicit communication between and among
separate actors in the scenario)
36 Systems Thinking in Practice

used to sign-post where the system designer will find detailed insights into the type,
nature and structure of these distributed cognition artefacts.
This chapter has focused on a systems level description of the ATC scenario and
serves to illustrate that this description can be achieved with the EAST method in live
settings. The strength of this descriptive level of analysis can be summarised as follows:

• The methods avoid bias by focusing on objective and manifest phenomena.

• The methods are applicable to any domain, and the results gained are com-
parable across and within domains.
• The results are graphical and easily interpreted, yet amenable to further
summarisation using tables and numerical indices.
• The summary level is underpinned by considerable detail that can be
explored further in the context of system design.
• It is consistent with existing narrative approaches to distributed cognition
analysis.

As well as providing a descriptive level of analysis, the real potential of the

method lies in its ability to offer predictive insights, that is, to ‘model’ complex
sociotechnical systems. In theory, it should be possible to subject the networks to
known changes (in task, social or information structure) and to derive outputs as to
the effect of these under various performance contexts. For example, an alteration
in the command structure may influence the type of communications available to an
agent, therefore affecting the type of information that is able to be communicated to
other agents, leading to a deleterious affect on SA (or indeed vice versa). The model-
ling aspect of this work is at a nascent stage but shows promise for further develop-
ment. What can be communicated about the current application of the method to
live data is that the highly relevant theoretical perspective of distributed cognition is
within the reach of systems designers and ergonomists.

REFERENCES
Annett, J. (2005). Hierarchical task analysis. In: N. A. Stanton, A. Hedge, K. Brookhuis, E.
Salas and H. Hendrick (Eds.), Handbook of Human Factors and Ergonomics Methods
(pp. 33.1–33.7). CRC: London.
Bell, H. H. and Lyon, D. R. (2000). Using observer ratings to assess situation awareness. In:
M. R. Endsley (Ed.), Situation Awareness Analysis and Measurement (pp. 129–146).
Lawrence Erlbaum Associates: Mahwah, NJ.
Builder, C. H., Bankes, S. C. and Nordin, R. (1999). Command Concepts: A Theory Derived
from the Practice of Command and Control. Rand: Santa Monica, CA.
Chalmers, D. J. (1990). Thoughts on emergence. Available at: http://consc.net/notes/emer-
gence.html (Accessed April 2005).
EAST in Air Traffic Control 37

Chase, W. G. and Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 55–81.

Civil Aviation Authority (2004). CAP 413: Radiotelephony Manual (edition 15). Safety
Regulation Group, Civil Aviation Authority: Gatwick, UK.
Clegg, C. W. (2000). Sociotechnical principles for system design. Applied Ergonomics, 31,
463–477.
Dekker, A. H. (2002). C4ISR Architectures, Social Network Analysis and the FINC
Methodology: An Experiment in Military Organisational Structure. DSTO Electronics
and Surveillance Research Laboratory: Edinburgh, Australia.
Driskell, J. E. and Mullen, B. (2005). Social network analysis. In: N. A. Stanton, A. Hedge,
K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors and
Ergonomics Methods (pp. 58.1–58.6). CRC: London, UK.
Endsley, M. R. (1995). Toward a theory of situation awareness in dynamic systems. Human
Factors, 37 (1), 32–64.
Fields, R., Wright, P., Marti, P. and Palmonari, M. (1998). Air traffic control as a distrib-
uted cognitive system: A case study of external representations. In: T. R. G. Green,
L. Bannon, C. P. Warren and J. Buckley (Eds.), ECCE9: Proceedings of the Ninth
European Conference on Cognitive Ergonomics (pp. 85–90). EACE: Limerick,
Republic of Ireland.
Gorman, J. C., Cooke, N. J. and Winner, J. L. (2006). Measuring team situation aware-
ness in decentralized command and control environments. Ergonomics, 49 (12–13),
1312–1325.
Harris, C. J. and White, I. (1987). Advances in Command, Control & Communication
Systems. Peregrinus: London, UK.
Hollan, J., Hutchins, E. and Kirsh, D. (2000). Distributed cognition: Toward a new foundation
for human–computer interaction research. ACM Transactions on Computer Human
Interaction, 7, 174–196.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49 (12–13), 1204–1225.
Hutchins, E. (1995a). How a cockpit remembers its speeds. Cognitive Science, 19,
265–288.
Hutchins, E. (1995b). Cognition in the Wild. MIT Press: Cambridge, MA.
Hutchins, E. and Holder, B. (2000). Conceptual models for understanding an encounter with a
mountain wave. Proceedings of HCI-Aero 2000. Toulouse, France, September.
Hutchins, E. and Holder, B. (2001). What pilots learn about autoflight while flying on the line.
11th International Symposium on Aviation Psychology, 6 March.
Hutchins, E. and Klausen, T. (1996). Distributed cognition in an airline cockpit. In: Y.
Engestrom and D. Middleton (Eds), Cognition and Communication at Work (pp. 15–
34). Cambridge University Press: Cambridge, UK.
Hutchins, E. L. and Palen, L. (1997). Constructing meaning from space, gesture, and speech.
In: L. B. Resneck, R. Saljo, C. Pontecorvo and B. Burge (Eds.), Discourse, Tools and
Reasoning: Essays in Situated Cognition (pp. 23–40). Springer-Verlag: Heidelberg,
Germany.
Klein, G. and Armstrong, A. A. (2005). Critical decision method. In: N. A. Stanton, A.
Hedge, K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors
and Ergonomics Methods (pp. 35.1–35.8). CRC: London, UK.
McMaster, R. and Baber, C. (2005). Integrating human factors into systems engineering
through a distributed cognition notation. The IEE and MoD HFI Symposium on People
and Systems (pp. 77–83). IEE: London, UK.
Moray, N. (2004). Où sont les neiges d’antan? In: D. A. Vincenzi, M. Mouloua and P. A.
Hancock (Eds.), Human Performance, Situation Awareness and Automation: Current
Research and Trends (pp. 1–31). Lawrence Erlbaum Associates: Mahwah, NJ.
38 Systems Thinking in Practice

Perla, P. P., Markowitz, M., Nofi, A., Weuve, C., Loughran, J. and Stahl, M. (2000). Gaming
and shared situation awareness. Available at: http://www.thoughtlink.com/publication-
sGamingSSA00Abstract.htm/[Accessed April 2005].
Perry, M. (2003). Distributed cognition. In: J. M. Carroll (Ed.), HCI Models, Theories and
Frameworks (pp. 93–224). Morgan-Kaufmann: San Francisco, CA.
Rogers, Y. (1997). A Brief Introduction to Distributed Cognition. Interact Lab, University of
Sussex: Brighton, UK.
Rogers, Y. and Ellis, J. (1994). Distributed cognition: An alternative framework for analys-
ing and explaining collaborative working. Journal of Information Technology, 9 (2),
119–128.
Salas, E., Prince, C. and Baker, D. P. (1995). Situation awareness in team performance –
implications for measurement and training. Human Factors, 37, 123–136.
Salmon, P. M., Stanton, N. A., Walker, G. H. and Jenkins, D. P. et al. (2008). What really is
going on? Review of situation awareness models for individuals and teams. Theoretical
Issues in Ergonomics Science, 9 (4), 297–323.
Sandom, C. (2001). Situation awareness. In: J. Noyes and M. Bransby (Eds.), People in
Control: Human Factors in Control Room Design (pp. 51–68). Institution of Electrical
Engineers: London, UK.
Stanton, N. A., Salmon, P., Walker, G. H., Baber, C. and Jenkins, D. (2005). Human Factors
Methods: A Practical Guide for Engineering and Design. Ashgate: Aldershot.
Walker, G. H., Stanton, N. A., Gibson, H., Baber, C., Young, M. and Green, D. (2006).
Analysing the role of communications technology in C4i scenarios: A distributed
cognition approach. Journal of Intelligent Systems Special Issue: Information and
Communication Technology, 15 (1–4), 299–328.
3 EAST in Military
Command and Control
With Rebecca Stewart, Daniel P. Jenkins,
Linda Wells and Chris Baber

INTRODUCTION
Event Analysis for Systemic Teamwork (EAST) is a macroergonomic method for
extracting large-scale systems level data on the emergent properties of command
and control scenarios (e.g. Kleiner 2006). The method is based on a number of
observations: firstly, ‘effective team performance is not an automatic occurrence’
(Salas et al. 1995, p. 55); secondly, increased complexity in military operations gives
them the appearance of ‘different components and layers of subsystems with mul-
tiple non-linear interconnections that are difficult to recognise, manage and predict’
(Marashi and Davis 2005; Johnson 2005, p. 1); and thirdly, the interaction of compo-
nents and subsystems (teamworking + complexity) creates non-linear emergent prop-
erties at the level of the entire system. In other words, sociotechnical systems like
this can be more (or indeed, much less) than the sum of their ‘socio’ and ‘technical’
parts. The challenge is to find ways to exploit complexity and non-linearity in order
to ‘obtain a disproportionate leverage from a given action’ (Smith 2006, p. 40). Thus,
focusing on the interrelations between command and control’s component parts is
perhaps as important as the parts in isolation. So, by shifting the unit of analysis
from ‘technical’ to ‘human’ and shifting it again from ‘individuals’ to that of the
‘system’, and by deploying network-based methodologies as a form of non-linear
modelling, the data that EAST provides ultimately speak to this goal.

DESCRIPTION OF COMMAND AND CONTROL SCENARIOS

Army Land Warfare and the Combat Estimate
The focus of this chapter is on the application of the EAST method to military
command and control and the specific case of army land warfare. Land warfare
(and other services) rely on a highly evolved planning heuristic called the Combat
Estimate (or ‘the seven questions’), which forms the topic of current EAST analysis.
The Combat Estimate describes the process by which plans are made, expected out-
comes are defined and actions then have to be taken.
In broad terms, Questions 1 and 2 are concerned with the development of situ-
ational awareness concerning the spatial configuration of the battlespace and of mis-
sion objectives. The specific activities undertaken in Question 1 include Battlefield
39
40 Systems Thinking in Practice

Area Evaluation (BAE), which deals with the potential effects of the physical envi-
ronment on military operations; and Threat Evaluation, which involves assessing
the enemy’s capabilities and tactics. Question 2 is concerned with Mission Analysis
and the scrutiny of orders that have been received. Questions 4 to 7, in equally
broad terms, can be subsumed under the heading ‘Course of Action Development’.
Figure 3.1 shows how the seven phases of the Combat Estimate relate to each other
functionally and temporally. The diagram is a ‘task network’ based on the high-level
goals of a comprehensive Hierarchical Task Analysis (HTA) of the scenario. The
links between goals are specified by the HTA’s top level ‘Goal 0’.

Data Collection
Data for the EAST analysis were gathered by live observation of Command and Staff
Training (CAST) exercises at the British Army’s Land Warfare Centre in Warminster.

FIGURE 3.1 Task network for the observed military command and control scenario.
EAST in Military Command and Control 41

The exercises took place in a battlegroup command post set up on-site as it would be
deployed in the field. The command post was set to work within a scenario which, in
broad terms, required effects to be delivered to a large-scale enemy force passing in
one direction through the battlegroup’s area of operations. Enemy and friendly units
were simulated by a team of remotely located operators who provided a form of aug-
mented reality, supplying friendly radio traffic (simulating units that would ordinar-
ily be located in the field) and updating digital data pertaining to enemy movements,
reactions and counter-actions. A team of analysts and subject matter experts moni-
tored and transcribed video and audio feeds from the battlegroup headquarters. Key
personnel active in the scenario were further interviewed at key points in the scenario
using the Critical Decision Method (Klein and Armstrong 2005).

DESCRIPTION OF THE EAST METHOD

The Importance of Methods
The importance of ergonomics methods cannot be overstated (e.g. Stanton et al.
2005a, 2005b; Wilson and Corlett 1995). Explicit methods lie at the heart of ergo-
nomics as a discipline, enabling the practitioner to vary their approach between those
of scientist (i.e. testing and developing theories of human performance using rigor-
ous data collection and analysis techniques) and practitioner (evaluating the effects
of change, developing best practice and, fundamentally, addressing real-world prob-
lems). Ergonomics methods are useful in the scientist–practitioner model because of
the structure and potential for repeatability that they offer over and above informal
methods.

Descriptive vs. Formative Methods

The EAST method is based on the integration of seven individual Ergonomics meth-
ods. What appears to be a sizeable modelling endeavour is a reflection of the
multi-faceted nature of the command and control problem to which it is designed to
apply. No one method can adequately describe all of the degrees of freedom inher-
ent in such a complex sociotechnical system. No such claim is made for the EAST
method; however, it can be argued that at least some of the major human dimensions
of the problem space can be explored by taking a systems perspective on the diagno-
sis and description of ‘what is going on’ when command and control organisations
are set to work.
EAST is a descriptive method. It does not specify a formal architecture and what
‘should’ happen. Even though it uses normative methods like task analysis and pro-
cess modelling, they are populated based on what is actually observed. Neither does
EAST focus on constraints, boundaries and a problem space defined formatively by
what ‘could happen’ like Cognitive Work Analysis (CWA) (Vicente 1999): EAST
focuses on what ‘did’ happen. CWA admits the possibility of non-linear and emer-
gent behaviour: EAST is designed to identify specific instances of it.
Despite their descriptive versus formative differences, EAST and CWA are both
representative of a shift in methodological thinking. They share two key aspects: both
42 Systems Thinking in Practice

acknowledge that complex sociotechnical systems require more than one approach
(EAST is comprised of seven individual methodologies, CWA is comprised of five
‘phases’), and both acknowledge that these perspectives are as interlinked as the
complex sociotechnical phenomenon under analysis (i.e. they are both systemic in
nature). This is a core principle of sociotechnical design (Clegg 2000).

Method Integration
The following formal methodologies combine to form EAST: Hierarchical Task
Analysis (HTA) (Annett 2005), Coordination Demand Analysis (CDA) Burke 2005),
Communications Usage Diagram (CUD) Watts and Monk 2000), Social Network
Analysis (SNA) Driskall and Mullen 2005), Propositional Networks (PN) (e.g. Ogden
1987) and an enhanced form of Operation Sequence Diagram (OSD) (Kirwan and
Ainsworth 1992). A multiple method approach has a number of compelling advan-
tages. Not only does the integration of existing methods bring reassurance in terms
of a validation history but it also enables the same data to be analysed from multiple
perspectives. With over 200 existing methodologies to choose from (Stanton et al.
2005a) there seemed little pragmatic need to develop yet more, hence the approach
adopted, which was to integrate existing methods. The trade off, of course, is time.
Multiple interconnected methods require greater effort to analyse, but in some sense,
this is an artefact of the problem domain being analysed. Good news in this respect
comes in the form of a companion to EAST called WESTT. This is a software tool
that greatly streamlines and simplifies the application of the method and was used in
the current analysis (Houghton et al. 2007).
The HTA provides input into the analysis of teamworking (CDA), communica-
tions usage (CUD) and the linkage (via communications) between agents (SNA).
Data for the HTA are gathered from live observation and activity sampling as
mentioned above. The output of all these methods (HTA, CDA, CUD and SNA) is
given a summary visual form by using an enhanced Operation Sequence Diagram
(OSD). Interview data, in the form of the Critical Decision Method (CDM: Klein
and Armstrong 2005), are used to create the final part of EAST, a network of linked
‘information objects’ (or Propositional Networks; PN). This is a systemic, network-
based approach to the concept of Situation Awareness (SA). The debates surrounding
the concept of SA require more in-depth discussion in relation to EAST.

Situational Awareness
One of the key emergent properties from command and control scenarios, and one
of the major determinants of decision superiority, is the concept of situation aware-
ness (SA) Endsley 1995; Stanton et al. 2009; Salmon et al. 2008a). At an individual
level, SA is about simply ‘knowing what is going on’ (Endsley 1995). At a systems
level, SA enables decisions to be made in real time, and for sociotechnical systems
like military command planning to be orientated towards and ‘tightly coupled to
the dynamics of the environment’ (Moray 2004, p. 4). A distributed cognition per-
spective applied to command and control scenarios requires a shift from traditional
notions of SA that focus on the individual (e.g. Smith and Hancock 1995; Adams
EAST in Military Command and Control 43

et al. 1995; Bedny and Meister 1999; Endsley 1995). Whilst these approaches may
be appropriate for tasks that are performed by individuals in isolation, few complex
command and control tasks are performed entirely independently of others. The idea
of co-dependence in SA, of course, finds expression in several approaches to ‘team
SA’ (e.g. Perla et al. 2000; Salas et al. 1995). Broadly speaking, these approaches to
SA tacitly assume that the ‘situation’ can be defined as a single, objective, external
reality, and that the goal of the people operating within the situation is to respond to
all features appropriately. These approaches are problematic on three counts: first,
there are many aspects of command and control scenarios that require the individual
to make judgements and interpretations (so the assumption of the ‘objective reality’
of a situation is not always valid); second, there are multiple sub-goals and, there-
fore, multiple views of the situation (so the idea of a single reality is also not always
valid either); and third, as mentioned above, different agents within the system use
different pieces of information to inform and support their work, so the notion that
there can be a single view of the situation (as opposed to several overlapping views)
is not easily supported. Distributed cognition provides a way of coping with these
conceptual issues and of providing a systems level view of SA (Stanton et al. 2009).
This view rests on three key factors.

1. Firstly, a relatively invariant theoretical property of SA concepts is ‘informa-

tion’, in so far as ‘all aspects of momentary SA are eventually reducible to some
form of … information in working memory’ (Bell and Lyon 2000, p. 42). The
question is whether ‘working memory’ is an individual phenomenon (as implied
by Bell and Lyon 2000) or whether it is a ‘system-level’ representational state
(as implied by the notions of Distributed Cognition); for the purposes of the
current EAST analysis, the latter, more contentious, view is taken.
2. Secondly, the ‘information’ that underlies distributed SA is itself distrib-
uted across the entire system, including non-human artefacts that can cre-
ate, manage and share representational states. This means that the system
will be managing the exchange of aspects of representational states through
the passage of information between agents.
3. Third, there is often implicit transaction of information rather than a con-
scious hand-over or exchange. Thus, the update of a representational state
for one agent might lead to partial updating of that used by another, and this
might not be the result of communication that is managed by human agents
(or might not require detailed information processing by any of the agents).
For example, as an enemy force element moves across a sector, its route is
plotted on the various map displays and its position updated dynamically;
if there is little immediate risk, then the updating happens without a cor-
responding need for intervention or overt ‘awareness’.

The concept of SA, therefore, looks rather different when considered from a
‘distributed’ as opposed to an individual perspective. It is still about the dynamic
orientation of a system to its operational context, but the units of analysis centre
around information as opposed to individuals (and the psychological processes by
which the state of SA is achieved). Information is held, exchanged, represented and
44 Systems Thinking in Practice

transformed by human and non-human agents, its propagation is supported by what-

ever communication infrastructure is currently in place, managed by the underlying
control structure, and evaluated by the underlying command activity.
Information objects, defined as entities or phenomena about which an ‘agent’ in
the system requires information in order to act effectively, are extracted from the
CDM interview transcripts using content analysis. Causal links between objects are
established in order to create the propositional network. The totality of information
residing at the systems level, when modelled as an interconnected web of informa-
tion, can be viewed as ‘systemic SA’. Different parts of the system use/share different
items of information; different items of information, when active, relate to yet more
information that is connected to it; and the whole configuration of usage and sharing
in turn changes dynamically in response to the context.

Theoretical Basis
The combination of observation and interview, analysis and representational meth-
ods forms EAST. A more detailed description of how the specific method outputs are
derived is provided in the next section (as well as in Walker et al. 2006a). A summary
of EAST, and how its component methods relate to each other functionally, is shown
below in Figure 3.2.
EAST is a human-centred approach but one whose component methods are all
anchored to a common systems perspective. This is manifest in a number of under-
pinning ideas, principal among which is distributed cognition. Under this per-
spective, the ‘computations’ that comprise military command and control are not
the exclusive province of individuals (Rogers 1997; Hollan et al. 2000; Hutchins
1995); instead, they are distributed across the entire command and control system,
comprised of numerous individuals, teams and technical artefacts. The essence
of distributed cognition is on ‘how [these computations] transcend the boundaries
of the individual actor’, and to this end, the common language of representational
states is used. Representational states are visible and external manifestations of
various ‘environmental contributions’ to the total system (Fields et al. 1998; Rogers
and Ellis 1994). They subsume the full range of observable interactions between
people and artefacts, as well as the resulting states (and state changes) that arise.
In essence, the focus of the EAST method is on describing artefacts that relate to

1. the changes that are made to these representational states

2. their influence and promulgation around a distributed network of human
and non-human actors
3. how that influence, in turn, generates new changes to representational states.

FINDINGS
Coordination Demand Analysis (CDA)
According to Salas, Bowers and Cannon-Bowers (1995), ‘the military is growing
increasingly dependent on the ability of individuals to coalesce quickly into effective
EAST in Military Command and Control 45

FIGURE 3.2 Structure of the EAST method.

teams’ (p. 55). Therefore, it might be assumed that command and control scenarios
will be dominated by teamwork activities, but this supposition can be checked in
more detail. The Coordination Demand Analysis (CDA) procedure allows for the
identification of teamwork skills needed for smooth coordination among team mem-
bers. Individual tasks from the HTA were categorised into task or teamwork tracks
(e.g. Salas et al. 1995; Morgan et al. 1986). The teamwork tasks were then scored
against the CDA taxonomy of communication, situational awareness, decision mak-
ing, mission analysis, leadership, adaptability and assertiveness (Burke 2005). Each
CDA taxonomy item was scored from 1 to 3, where 1 is low coordination and 3 is
high coordination. From these individual scores, a ‘total coordination’ figure can be
derived, which is based on the mean of the component scores. In this scenario, the
scoring was derived from a focus group comprised of the analyst team and subject
matter experts. Each item was briefly discussed and its final score based on consen-
sus among the group. Overall, the mean total coordination score for the military
scenario is 1.2 (out of a maximum score of three). This score is broadly comparable
to civilian command and control domains like air traffic control (e.g. Walker et al.
2006a), railway signalling and safety operations (e.g. Walker et al. 2006b) and bulk
energy distribution (e.g. Salmon et al. 2008b). The mean coordination score was also
46 Systems Thinking in Practice

calculated for the seven main stages of the HTA (these represent the phases in the
military scenario), and the results are shown in Table 3.1.
The supposition that command and control activities have a prominent teamworking
component is well justified, as would be expected, yet different phases of the planning
process require different types and amounts of teamwork; the CDA provides a window
onto this. The scores for the individual coordination dimensions vary across the full
range of permissible values. It can be noted that communications and SA score consis-
tently highly, whereas decision making scores relatively low. This and other facets of
the teamworking profile could arise because the decision-making components of the
scenario are constrained by hierarchical patterns of interaction, pre-specified forms of
interaction and/or unitary decision rights (NATO 2006). This pattern differs from the
civilian examples mentioned above where the EAST method has been applied. Here,
the decision-making and planning phases tend to occur concurrently and continuously
(as opposed to a relatively discrete stage), and decision making scores more highly.

Communications Usage Diagram (CUD)

Current military command and control activities do not necessarily have a particu-
larly complex communications infrastructure in place compared to the various forms
of ‘Network Enabled’ versions of command and control seen in other domains.
Network Enabled Capability (NEC) is an emerging paradigm whereby a perva-
sive information infrastructure is created, which, combined with greater autonomy
and a focus on teams and task outcomes, offers considerable potential to impact
all the human and organisational parameters currently under analysis. For example,
experience in the energy distribution domain reveals a recent development that sees
‘peripatetic working’ (focusing on autonomous and mobile teams) and the extensive
use of tablet PCs and a dedicated information infrastructure (Salmon et al. 2008b).
Indeed, EAST is motivated to a great extent by the growing demands of this emerg-
ing paradigm. The value in assessing an ostensibly non-NEC military scenario like
the current one is to establish a form of baseline.
The advent of NEC technology like advanced communications and data net-
works has an immediate impact on communications; therefore, its proximal effects
are likely to be seen first in this particular sub-method, the Communications Usage
Diagram (CUD). In the current scenario, the military planning process involves ver-
bal dialogues between planning personnel in close proximity (during meetings) and
announcements in which the command staff will shout (e.g. timescales for upcoming
meetings or deadlines). Radio communications are also an integral part of mediating
voice communications and take the form of verbal dialogues using a standardised
radio telephony method. Communications within the command centre are, therefore,
conducted principally by voice. Having said that, the final group of communications
media are the operational materials produced out of the planning process, along with
the visual aids such as whiteboards and clear overlays that are used during it. All of
these communication methods can be analysed using the CUD.
The CUD contains a description of the activity conducted at each geographical
location, the communication between actors involved, the technology used for the
communications, the advantages and disadvantages associated with that technology
TABLE 3.1
CDA Analysis Results According to Task Phase
Display & Manage Combat Estimate Translate Products of Q1-7 Conduct Execute
Category Prepare Plan Information (Make Plan) into Operational Graphics Wargame Plan
Mean Communications 2.5 2 2.2 2 3
EAST in Military Command and Control

Mean SA 2 2 2.2 2 2
Mean DM 1.5 1.8 1 2
Mean MA 1 2 2.1 1 1
Mean Leadership 1.25 1.3 1.6 2 3
Mean Adaptability 2 2 3 2
Mean Assertiveness 1.8 1 2
Total Coordination 1.7 1.8 2.0 0 1.7 2.1
47
48 Systems Thinking in Practice

medium and a recommended technology, if there is one. The CUD method, there-
fore, is a structured way to represent communications within a scenario based on
task flow (and thus the HTA). For the purposes of this chapter, the method output has
been synthesised into a list of advantages and disadvantages (Table 3.2) followed by
a critique of communications usage.

Advantages and Disadvantages of Existing Communications Media

Critique of Communications Usage
The survey of advantages and disadvantages forms part of a structured means to
describe and critique existing communications technology. It should be added that
the critique is not indicative of any actual or proposed recommendation; rather, it is
a consideration of possible alternatives and issues based on the data collected.

Verbal Communications
A compelling advantage of verbal communications is the level of immediacy and
redundancy it provides. Carvalho (2006) states that, ‘operators use verbal exchanges
to produce continuous, redundant and recursive interactions to successfully construct
and maintain individual and mutual awareness’ (p. 51). There remains, however, the
possibility of psychological issues such as ‘group-think’ and bias in using this form
of communication in this context, although there are techniques available to help
overcome this (Janis 1982a, 1982b). It can be noted that verbal communications also
dominate in civilian command and control scenarios.

Network Enabled Solutions

Gaining situational awareness of the battlespace, a key aspect of the Combat
Estimate, requires a number of transformations to be undertaken in order to repre-
sent and to understand the state of the world. Voice, written and 2D imagery in the
planning process all need to be visualise, and then acted upon in 4D space (3D plus
time). Cognitive effort is required, therefore, to achieve adequate levels of SA from
which to develop and resource courses of action. Thus, there is a lot of ‘informa-
tion in the head’, which can be advantageous for SA. NEC approaches that embody
positional data, 3D representations of the battlespace and live updating of it directly
from the field embody considerably more ‘information in the world’. The task of
air traffic control, for example, would be virtually impossible to conduct at its cur-
rent tempo without the sort of visualisation and live updating provided by the radar
display, although even here, the representation is two dimensional. An issue that
emerged from the CUD and CDA methods was that the combat estimate ‘process’
was often just as important as the ‘outcome’. For example, a considerable degree of
teamworking was devoted to developing SA. In addition, scrutinising the supporting
communications methods using the CUD reveals advantages as well as prima-facia
disadvantages of what could be considered simplistic planning apparatus.

Social Network Analysis (SNA)

Social Network Analysis (SNA) is a means to present and describe the underlying
network structure of individuals or teams who are linked through communications
TABLE 3.2
Advantages and Disadvantages of Existing Communications Media
Media Advantages Disadvantages
In-Person Voice Physical verification that correct individual (and planning Possible detrimental role of social status/military rank
role) is being referred to Possible contextually related distractions (e.g. noise and general confusion in
Favourable role of non-verbal communications in aiding the command centre)
shared understanding Possible ambiguity in physically pointing out and referring to shared resources
Possible favourable role of military rank in face-to-face (e.g. plans and whiteboards)
communications Relatively static descriptions of a highly dynamic and spatially dispersed
Sharing of explanatory resources such as operational scenario
graphics/whiteboards etc.
Radio Communications Sound stable Intelligibility can be an issue with distortion/artefacts in radio comms.
Possible for communications to be recorded for post-hoc Language/accent ambiguities
EAST in Military Command and Control

analysis and training Unscheduled/ad-hoc presentation of comms.

Hands free Any informality/abbreviation in comms. relies on assumption of shared
Time saving with common abbreviations and meaning
nomenclature Relatively slow communications compared to other comms solutions
Enhanced intelligibility with common abbreviations and Translation from verbal domain to visio-spatial domain required (and vice
nomenclature versa)
Read-back provides validation of shared understanding Open channel radio comms. could permit simultaneous comms. on same
Open channel radio comms. aids shared SA among other frequency causing masking
units and members of planning staff Read-back can be out of synchronisation with current activities if sender/
Possibly favourable dilution of group-think/military rank recipient is slow to respond
artefacts on communication and comprehension errors Unfavourable dilution of favourable aspects of military rank
(Continued )
49
50

TABLE 3.2 (CONTINUED)

Advantages and Disadvantages of Existing Communications Media
Media Advantages Disadvantages
Operational Graphics Paper-based materials can be substantially degraded Static representations of typically dynamic scenarios
and Planning Aids without information loss Training load relatively high in the use of methods to overcome the
Relatively easy to derive with little extra training required disadvantages of representing dynamic 3D phenomena as 2D paper-based
in having to use a pen and paper representations
Complexity and dynamism of scenario can cause administrative bottlenecks.
Legibility of graphics and handwriting
Whiteboards and overlays potentially cumbersome to handle
Document tracking and administration potentially difficult
Systems Thinking in Practice
EAST in Military Command and Control 51

(Driskell and Mullen 2005). Social networks focus ‘on the relationships among
actors embedded in their social context’ (Driskell and Mullen 2005, p. 58). NEC
concepts embody a form of information commodification; thus, social networks can
be used to represent the technological mediation of communication and networks
where some of the nodes are non-human. The resulting network can then be subject
to mathematical analysis using Graph Theory in order to simplify it and express
its key features in a standardised form using a numeric index (Driskell and Mullen
2005). Two mathematical indices are used, namely centrality (i.e. a numeric ranking
allowing key agents in the network to be identified) and density (the interconnectiv-
ity of the network as a whole). Both of these metrics can be understood in relation
to other contextual factors to enable judgements to be made about what aspects of
the network configuration constrain or enhance performance. The metrics, being
emergent properties of the networks as well as a means to simplify them, permit easy
comparison between alternate domains.
The social network for military command and control defines seven key actors,
some of which have been grouped into sub-systems for simplicity (as illustrated
in Figure 3.3). The actors or nodes include the Higher Command Formation, the
Commander and Chief of Staff (COS) at the command headquarters, the ‘Principal’
Planning Staff (which subsumes individual roles such as G2 Intelligence and
EngGeo), other command staff (responsible for more general tasks and information

FIGURE 3.3 Social network for military command and control scenario (icons represent
the communications media that facilitates the linkage).
52 Systems Thinking in Practice

management), personnel in the field and the collection of graphics and planning
aids derived from the Combat Estimate process (an informational node that is
non-human).

Activity Stereotypes
The social network is dynamic and adaptive, with different nodes and links becom-
ing active under different activity stereotypes (revealed by detailed analysis of the
HTA). The activity stereotypes are as follows:

• Briefing or providing direction: the Commander is directing communica-

tions and information outwards to subordinate staff in a prescribed and
tightly coupled manner (particularly Questions 1 and 3 of the Combat
Estimate planning technique).
• Reviewing: the planning staff communicate in a more collaborative, peer-
to-peer manner, with mutual exchange of information and ad-hoc usage
of planning materials and outputs (in particular Questions 2 and 5 of the
Combat Estimate).
• Semi-autonomous working: members of the headquarters are working indi-
vidually on assigned tasks and become relatively loosely coupled in terms
of communication. The communication channels remain open but are used
in an ad-hoc, un-prescribed manner (this occurs at various points in all
phases of the Combat Estimate and the scenario more generally).

The temporal and task-based activation of agents and communications, in which

they assume different stereotypical configurations, is illustrated in Figure 3.4.

Facilitation of Network Links

Figure 3.3 also illustrates the communications media that facilitate the links between
nodes in the network. These are formally defined by the CUD method above. The
results are also summarised in Table 3.3 as a communications/modality/technology
matrix. Shading indicates where a specific communications technology is crossed
with a specific modality. The matrix appears to be relatively simple in the observed

FIGURE 3.4 Social networks illustrating the characteristics of a number of network arche-
types detected in the analysis (to be read in conjunction with Figure 3.3).
EAST in Military Command and Control 53

TABLE 3.3
Technology/Facilitation/Modality Matrix
Technology/Facilitation
Modality Radio Planning Aids In-Person Voice
Verbal
Visual
Written

Note: Shading Represents a Match Between Communica-

tions Technology and Communications Modality.

military scenarios. A heavy reliance on verbal information, which occurs in person

and via radio, was noted. There is also heavy reliance on visual communication, as
embodied by the various planning aids. In-person communications, it was noted,
also incorporate visually mediated communication as staff point and gesture at maps
and planning aids. Although simplistic compared to other civilian examples of NEC
(like air traffic control) the underlying social and communications system is undeni-
ably robust. Similar robustness might be regarded as a redundant and, indeed, an
inefficient feature in civilian contexts. Clearly, there are opportunities, for example,
to more rapidly acquire the state of SA through novel technology that does not neces-
sarily rely on verbal communications and manual updating of maps. Such a system
is realised in the civilian examples of air traffic control (Walker et al. 2010) and
energy distribution (Salmon et al. 2008b), in which the resulting social networks
show, in comparison, a much denser interconnection between actors using a more
diverse array of technology to facilitate the kind of redundant, recursive interactions
that are required in high-tempo tasks. Table 3.3 is based on Figure 3.3 (which in turn
is based on the CUD method above) and shows the communication modality that is
afforded by the various items of communications technology in the scenario. It is
interesting to note that the dominant communications modality is well supported;
it is, moreover, interesting to note that future digitisation of military command and
control is likely to expand the visual and written modalities somewhat more than it
will the verbal modality.

Calculation of Social Network Metrics

The ‘most central agents’ are revealed, by network mathematics, to be the principal
planning staff, followed by the commander and chief of staff (Table 3.4). In NEC
scenarios, it might be anticipated that the spread of centrality scores will be less
pronounced as a result of more devolved decision rights and peer-to-peer interac-
tion, and this is certainly evident in civilian examples. The network density figure
of 0.31 is suggestive of a moderate level of connectivity within the network and is
again comparable with civilian examples. The point here is that the total number of
available communications links is more or less the same, but they are configured
54 Systems Thinking in Practice

TABLE 3.4
Network Metrics Illustrating Centrality (Key
Agents in the Scenario) and Density (Network
Connectivity) for the Social Network as a Whole
Agent Agent Centrality Network Density
Higher Formation 0.89 0.31
Commander 1.11
COS 1.11
Other Command Staff 0.67
Principal Staff 1.33
Field Personnel 0.22
Planning Materials 0.67

differently in NEC paradigms. These links define the structure of the network and
also its function.
Table 3.5 presents the results of this analysis, showing how the properties of
the network change to reflect the stereotypical ways in which it is configured. The
change in network density for each activity stereotype is also indicative of a high
degree of reconfigurability. This appears to be a relatively unique feature of mili-
tary command and control. In civilian examples, the network density figures do not
change as dramatically as the task progresses through its distinct phases. The reason
is due to the complexity inherent in the operational context. Both civilian examples
operate within a relatively stable and placid environment in contrast to the turbulent
and complex military one.

Operation Sequence Diagrams (OSD)

The Operation Sequence Diagram (OSD) is the main descriptive summary repre-
sentation within the EAST method. It is an activity-based representation showing
who is performing what, when. Using colour coding, symbology and annotation,
the OSD can represent most of the critical features of the preceding methods on
one common representation. There is insufficient space to present the full OSD
analysis of the current scenario, but an opportunity remains to present a small
sub-set of the charts in order to reflect on the more distinctive overall features of
the analysis so far.
The sample OSD in Figure 3.5 illustrates many of the facets dealt with in ear-
lier methods. Teamworking and coordination is reflected not only by the explicit
colour coding of operations (darker shading denotes higher levels of total coordina-
tion) but also in the varying patterns of connectivity between and among operations
(represented by the symbols). The social network stereotypes are also represented
by the configurations of links and operations. For example, in the earlier phases of
the Combat Estimate (more Briefing activities), there is a focus on the hierarchical/
TABLE 3.5
Network Metrics Illustrating Centrality (Key Agents in the Scenario) and Density (Network Connectivity) for the Activity
Stereotypes of Briefing, Reviewing and Semi-Autonomous Working
Centrality Density
Agent Briefing Reviewing Semi-Autonomous Briefing Reviewing Semi-Autonomous
Higher Formation 0.03 0.20 0.13
EAST in Military Command and Control

Orders
Commander 0.33 0.67 0.33
COS 0.11 0.67 0.33
Other Command Staff 0.11 0.33 0.33
Principal Staff 0.11 0.67 0.33
Field Personnel
Planning Materials 0.33 0.67
55
56 Systems Thinking in Practice

FIGURE 3.5 Sections through complete OSD chart for military scenario showing (left to
right) semi-autonomous working, briefing and reviewing.

vertical flow of information, with the Commander producing it and other agents
receiving it. In the second phase (more Reviewing activities), the operations (and
agents) become closely interconnected, with decision-making components dominat-
ing (shown by the diamond symbol[s]). In the third and final phase (more Semi-
Autonomous working), the pattern of operations once again assumes a vertical/
hierarchical disposition. At the highest level, then, the command and control process
seems to assume a pattern of information retrieval, closely knit decision-making pro-
cesses and then dispersion and action.

Propositional Networks (PN)

From the CDM interview, it is possible to construct Propositional Networks (PN)
(an example of which is shown in Figure 3.6) to show the information that is related
to the scenario. The propositional network consists of a set of nodes that represent
sources of information, agents and objects that are linked through specific causal
paths (for example, the object [situation] ‘has’ the property of [updates] associated
with it, etc.). As mentioned earlier, these objects are extracted from the CDM inter-
view transcripts using content analysis. The deeper, more fundamental concept that
this method refers to is SA, an important concept in decision making, agility and
tempo. The advantage of the propositional network approach is that it represents
a way of modelling the information that comprises the state of SA, from an indi-
vidual as well as a systems perspective. In addition, because it is network based, it
meshes with the social and task networks that form the basis for the rest of the EAST
EAST in Military Command and Control 57

FIGURE 3.6 Illustration of propositional network representing a systems level view of

information but also temporally activated information for a particular task phase (shaded
objects). The overall network is presented for illustration; the ‘zoomed’ section shows some
of the detail.

method. From the PN, it is possible to identify the structure and temporal nature of
distributed SA (explained in full in Stanton et al. 2009) and the information under-
pinning decision making.
The summary table (Table 3.6) uses simple Graph Theory metrics (as used in
the social networks above) to summarise the visually complex network(s) into a
more tractable form. Based on an analysis of centrality, so-called ‘core informa-
tion objects’ can be defined for each phase in the scenario (a CDM interview was
carried out in relation to each phase, as was a separate PN). The table crosses each
phase of the Combat Estimate planning technique with the list of core information
objects. Shading denotes specifically what information objects are active in what
phase. These core information objects also feed back up to the CUD method ear-
lier. Their prescription enables an analysis of what information objects are shared
between what agents, and, therefore, require some form of communications technol-
ogy to mediate the sharing. In the CUD method, the appropriateness of this match
forms one aspect of the basis by which communications technology is, and can be,
critiqued (see Walker et al. 2006a for an in-depth treatment of how this specific
approach can be realised).
Eighteen key information objects can be identified. As the Combat Estimate plan-
ning process progresses through its distinct phases, it can clearly be seen that the
activation of these key objects changes. This is further indicative of changes in the
type and structure of the propositional network and in the type of situational aware-
ness possessed by individuals and the system as whole. This information can be
58 Systems Thinking in Practice

TABLE 3.6
Summary of Key Information Objects Active within Each Scenario

Key Information Objects Q1 Q2 & 3 Q4 Q5 Q6 Implementing Plan

Sub-units
Plan
Friendly Forces
Orders
Situation
Intention
Capability
Phases of Battle
Weapons
Enemy
Intelligence
Effects
Courses of Action
Uncertainty
Terrain
Position
Tactics
History
Total Information Objects 9 12 6 8 13 10

Note: Shading indicates what information is specifically active during what stage of the com-
bat estimate.

used to assess the extent to which the system is orientated towards the dynamics of
its situation as well as to identify information requirements for different tasks and
scenarios.

CONCLUSIONS
The principles of Human Factors Integration (HFI) encompass ‘a balanced develop-
ment of both the technical and human aspects of equipment provision. It provides a
process that ensures the application of scientific knowledge about human character-
istics through the specification, design and evaluation of systems’ (MoD 2000, p. 6).
The EAST method is couched firmly within this context. It is, at bottom, a way of
capturing and describing the human view of complex sociotechnical systems – not
what should or might happen when they are set to work but what actually does hap-
pen. Being based on the integration of seven existing ergonomics methodologies, the
outputs are structured, systematic and standardised. The validation history that is
EAST in Military Command and Control 59

associated with each of them brings the promise of repeatability and compatibility
across situations and even domains.
The aim of the present analysis has been to demonstrate the applicability of the
method within military contexts, which represent a particularly challenging and
complex command and control environment. The analysis has been couched at a
fairly coarse-grained level of analysis, anchored as it has been to the discrete phases
of the Combat Estimate planning process. As much (or as little) resolution can be
extracted depending on the questions being asked and analysis effort that actors are
willing to expend. For the purposes of presenting EAST’s human view of military
command and control, and how those insights map across to other domains, the pres-
ent level of analysis seems appropriate.
What does this application of the EAST method tell us? It tells us about the struc-
tural and temporal nature of the goals and tasks that have to be performed and how
they are interlinked and interdependent. The structure of the task network (like any
system) defines its function. In the case of the military, it is noted not just that there
is a defined ‘planning phase’ but also that it is relatively sequential and recursive
in nature. In the civilian domains that have been observed previously, planning, at
least at strategic and operational levels, tends to occur concurrently with the task –
‘on the fly’, so to speak. That said, in civilian domains, the sociotechnical system’s
combined experience (of which it is an evolutionary expression) and its generally
placid environment enables rule-based optimum means to ends to be learnt and rapid
decision making to take place.
It can be noted that the function of the social networks also changes; they are
reconfigurable (and readily reconfigured) to suit the task phase, often in ways that
are not anticipated beforehand. This is, therefore, an emergent property of the inter-
action between the task and social networks. The links in the social network are
informational, and the content of this information is represented by the PN, a rep-
resentation that captures systems level information and, we argue, SA. SA is once
again an emergent property of both the task and social networks. In particular, the
way in which the social network is configured dictates the informational constraints
placed on it by the facilitating communications technology. Overall, it is clear that
cognition is distributed between socio and technical agents in the scenario. This,
combined with the fact that the socio elements (the people) will adapt the technical
elements to suit their needs and preferences, contributes to a degree of self-organisa-
tion and emergence that is apparent through analysis with EAST.
EAST’s three main network representations, illustrated in Figure 3.7, are an
analytical response to complexity and emergence but from a uniquely human per-
spective. A key point is that EAST acknowledges the sociotechnical principle of
equifinality: that the same people and the same technology may reach the same
goal by entirely different means and from entirely different initial conditions. This
is not a conceptual inconvenience but actually one of the very promises of NEC.
The inviolable fact that humans will adapt to the techno-organisational properties
of a given system is what permits desirable emergent properties to ‘emerge’, like
systems level ‘shared awareness’, tempo, agility and self-synchronisation. EAST
seems to offer a way to capture and describe this in a way that deterministic (i.e.
linear) methods may not.
60 Systems Thinking in Practice

FIGURE 3.7 Summary of EAST’s three network-based perspectives and the insights into
emergent properties that they provide.

Driven off the higher-level network representations are a number of more explicit
methods that relate more directly to requirements capture. In the CUD, for example,
task and social networks combine to inform a structured assessment of communica-
tions technology. This analysis is further informed by the notions of distributed SA
that are embodied in EAST. In other words, it becomes possible to assess what pur-
pose a communication exchange is designed to serve, what information it is designed
to convey and, moreover, what would be the best means of facilitating that exchange.
In addition to this, the CDA method highlights certain important aspects of team-
working at different phases of the planning process. This sort of information is useful
in the design of digital command planning systems. For example, there would be lit-
tle point in providing ‘decision support’ for a planning phase that is actually revealed
as requiring more ‘mission analysis’. The benefits of the interlinked EAST method
are that changes in any of the three underlying network representations (task, social
and propositional networks) are reflected in each other and in the supporting methods
like CUD and CDA. The disadvantage of the EAST method, at least at face value, is
the amount of time it takes to perform such an analysis. The reader is referred to a
companion software tool called WESTT that greatly simplifies and streamlines the
application of EAST, automating many functions, notably the production of the vari-
ous task, social and knowledge networks (Houghton et al. 2007).
In conclusion, EAST is offered as a means to describe the dynamic, emergent
behaviour of complex sociotechnical systems from a human perspective. It appears
consistent with wider trends in Ergonomics towards method integration and systems
views, and it is eminently compatible with similar approaches like CWA. The aim of
the present high-level analysis has been to illuminate some of these wider issues and
EAST in Military Command and Control 61

the applicability of EAST to military contexts. More targeted forms of application

will enable EAST to provide more targeted and direct answers to design and pro-
curement questions in an age when the critical networks in NEC are not necessarily
technological but human.

ACKNOWLEDGEMENTS
This work from the Human Factors Integration Defence Technology Centre (HFI
DTC) was part-funded by the Human Sciences Domain of the UK Ministry of
Defence Scientific Research Programme.
The assistance of LtCol Allan Ellis is gratefully acknowledged as is the assis-
tance of the staff at the Land Warfare Centre and CAST in Warminster, in particular
Majors Forster and Sharkey.
Information on the Combat Estimate draws heavily on material used during the
experimental team’s observation of the Combat Estimate training noted above. In
particular, two presentations, the first entitled ‘Command and Staff Trainer (South) 1
WFR MiniCAST 2/3 Aug 05’ and the second ‘Wargaming: Mastering Your Enemy’
(based on the Army Field Manual Vol I (Combined Arms Operations) Part 2 (July
98) and 3(UK) Div Wargaming Aide Memoire), are used with permission.

REFERENCES
Adams, M. J., Tenney, Y. J. and Pew, R. W. (1995). Situation awareness and the cognitive
management of complex systems. Human Factors, 37, 85–104.
Annett, J. (2005). Hierarchical task analysis. In: N. A. Stanton, A. Hedge, K. Brookhuis, E.
Salas and H. Hendrick (Eds.), Handbook of Human Factors and Ergonomics Methods
(pp. 33.1–33.7). CRC: London, UK.
Bar Yam, Y. (1997). Dynamics of Complex Systems. Perseus: Jackson, TN.
Bedny, G. and Meister, D. (1999). Theory of activity and situation awareness. International
Journal of Cognitive Ergonomics, 3 (1), 63–72.
Bell, H. H. and Lyon, D. R. (2000). Using observer ratings to assess situation awareness.
In: M. R. Endsley (Ed.), Situation Awareness Analysis and Measurement 129–146).
Lawrence Erlbaum Associates: Mahwah, NJ.
Builder, C. H., Bankes, S. C. and Nordin, R. (1999). Command Concepts: A Theory Derived
from the Practice of Command and Control. Rand: Santa Monica, CA.
Burke, S. C. (2005). Team task analysis. In: N. A. Stanton A. Hedge, K. Brookhuis, E. Salas
and H. Hendrick (Eds.), Handbook of Human Factors and Ergonomics Methods (pp.
56.1–56.8). CRC: London, UK.
Carvalho, P. V. R. d. (2006). Ergonomic field studies in a nuclear power plant control room.
Progress in Nuclear Energy, 48, 51–69.
Clegg, C. W. (2000). Sociotechnical principles for system design. Applied Ergonomics, 31,
463–477.
DoD (1999). Realizing the Potential of C4I: Fundamental Challenges. National Academy
Press: Washington, DC.
Driskell, J. E. and Mullen, B. (2005). Social network analysis. In: N. A. Stanton, A. Hedge,
K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors and
Ergonomics Methods (pp. 58.1–58.6). CRC: London, UK.
Endsley, M. R. (1995). Toward a theory of situation awareness in dynamic systems. Human
Factors, 37 (1), 32–64.
62 Systems Thinking in Practice

Fields, R., Wright, P., Marti, P. and Palmonari, M. (1998). Air traffic control as a distrib-
uted cognitive system: A case study of external representations. In: T. R. G. Green,
L. Bannon, C. P. Warren and J. Buckley (Eds.), ECCE9: Proceedings of the Ninth
European Conference on Cognitive Ergonomics (pp. 85–90). EACE: Limerick, Ireland.
Harris, C. J. and White, I. (1987). Advances in Command, Control & Communication
Systems. Peregrinus: London, UK.
Hollan, J., Hutchins, E. and Kirsh, D. (2000). Distributed cognition: Toward a new foundation
for human–computer interaction research. ACM Transactions on Human Computer
Interaction, 7, 174–196.
Hollnagel, E. (1993). Human Reliability Analysis: Context and Control. Academic Press:
London, UK.
Houghton, R. J., Baber, C., Cowton, M., Walker, G. H and Stanton, N. (2007). WESTT
(Workload, error, situational awareness, time and teamwork): An analytical prototyp-
ing system for command and control. Cognition Technology and Work, 10 (3), 199–207.
Hutchins, E. (1995). Cognition in the Wild. MIT Press: Cambridge, MA.
Janis, I. L. (1982a). Groupthink. Houghton Mifflin: Boston, MA.
Janis, I. L. (1982b). Counteracting the adverse effects of concurrence-seeking in policy plan-
ning groups: Theory and research perspectives. In: H. Brandstatter, J. H. Davis and G.
Stocker-Kreichgauer (Eds.), Group Decision Making. Academic Press: London, UK.
Jones, C. (1993). Toward a Science of Command, Control, and Communications. American
Institute of Aeronautics and Astronautics: Washington, D.C.
Johnson, C. W. (2005). What Are Emergent Properties and How Do They Affect the
Engineering of Complex Systems? University of Glasgow, Department of Computing
Science: Glasgow, UK.
Kirwan, B. and Ainsworth, L. K. (1992). A Guide to Task Analysis. Taylor and Francis:
London, UK.
Klein, G. and Armstrong, A. A. (2005). Critical decision method. In: N. A. Stanton, A.
Hedge, K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors
and Ergonomics Methods (pp. 35.1–35.8). CRC: London, UK.
Kleiner, B. M. (2006). Macroergonomics: Analysis and design of work systems. Applied
Ergonomics, 37, 81–89.
Marashi, E. and Davis, J. P. (2005). A systems approach to resolving complex issues in a
design process. Workshop on Complexity in Design and Engineering, 10–12 March,
Glasgow, Scotland, pp. 140–149.
MoD (2000). Human Factors Integration: An Introductory Guide. HMSO: London, UK.
MoD (2005a). Network Enabled Capability (JSP 777). HM Stationery Office: London, UK.
MoD (2005b). The BG PSO Combat Estimate. MoD: Warminster, UK.
Moray, N. (2004). Où sont les neiges d’antan? In: D. A. Vincenzi, M. Mouloua, and P. A.
Hancock (Eds.), Human Performance, Situation Awareness and Automation: Current
Research and Trends. Lawrence Erlbaum Associates: Mahwah, NJ.
Morgan, G. (1986). Images of Organisation. Sage: London, UK.
NATO (2006). SAS-050 conceptual model version 1.0. Available at: https://pdfs.semantic-
scholar.org/0937/20a588962d1a80a31360e869515b3f76de02.pdf.
Ogden, G. C. (1987). Concept, knowledge and thought. Annual Review of Psychology, 38,
203–227.
Perla, P., Markowitz, M., Nofi, A., Weuve, C., Loughran, J. and Stahl, M. (2000). Gaming and
shared situation awareness. Available at: http://www.thoughtlink.com/publications/
GamingSSA00Abstract.html.
Rasmussen, J., Petjerson, A. M. and Goodstein, L. P. (1994). Cognitive Systems Engineering.
Wiley: New York, NY.
Rogers, Y. (1997). A Brief Introduction to Distributed Cognition. Interact Lab, University of
Sussex: Brighton, UK.
EAST in Military Command and Control 63

Rogers, Y. and Ellis, J. (1994). Distributed cognition: An alternative framework for analys-
ing and explaining collaborative working. Journal of Information Technology, 9 (2),
119–128.
Salas, E., Bowers, C. A. and Cannon-Bowers, J. A. (1995). Military team research: Ten years
of progress. Military Psychology, 7, 55–75.
Salmon, P. M., Stanton, N. A., Walker, G. H., Baber, C., Jenkins, D. P., McMaster, R. and
Young, M. S. (2008a). What really is going on? Review of situation awareness models
for individuals and teams. Theoretical Issues in Ergonomics Science, 9 (4), 297–323.
Salmon, P. M., Stanton, N. A., Walker, G. H., Jenkins, D. P., Baber, C. and McMaster, R. M.
(2008b). Representing situation awareness in collaborative systems: A case study in the
energy distribution domain. Ergonomics, 51 (3), 367–384.
Smith, E. A. (2006). Complexity, Networking, & Effects-Based Approaches to Operations.
CCRP Publication Series.
Smith, K. and Hancock, P. A. (1995). Situation awareness is adaptive, externally directed
consciousness. Human Factors, 37 (1), 137–148.
Stanton, N. A., Salmon, P., Walker, G. H., Baber, C. and Jenkins, D. (2005a). Human Factors
Methods: A Practical Guide for Engineering and Design. Ashgate: Aldershot, UK.
Stanton, N. A., Hedge, A., Brookhuis, K., Salas, E. and Hendrick, H. (2005b). Handbook of
Human Factors and Ergonomics Methods. CRC: Boca Raton, FL.
Stanton, N. A., Salmon, P. M., Walker, G. H. and Jenkins, D. P. (2009). Genotype and
phenotype schemata and their role in distributed situation awareness in collabora-
tive systems. Theoretical Issues in Ergonomics Science, 10 (1), 43–68.
Vicente, K. J. (1999). Cognitive Work Analysis: Toward Safe, Productive, and Healthy
Computer-Based Work. Lawrence Erlbaum Associates: Mahwah, NJ.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M. and Green, D. (2006a).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics, 49, 1345–1369.
Walker, G. H., Stanton, N. A., Gibson, H., Baber, C., Young, M. and Green, D. (2006b).
Analysing the role of communications technology in C4i scenarios: A distributed
cognition approach. Journal of Intelligent Systems. Special Issue: Information and
Communication Technology, 15 (1–4), 299–328.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. M. and Jenkins,
D. P. (2010). From ethnography to the EAST method: A tractable approach for rep-
resenting distributed cognition in air traffic control. Ergonomics: Special Issue on
Aviation Safety, 53 (4), 184–197.
Watts, L. A. and Monk, A. F. (2000). Reasoning about tasks, activities and technology to
support collaboration. In: J. Annett and N. Stanton (Eds.), Task Analysis (pp. 55–78).
Taylor and Francis: London, UK.
Wilson, J. R. and Corlett, E. N. (Eds.) (1995). Evaluation of Human Work: Practical
Ergonomics Methodology. Taylor and Francis: London, UK.
4 EAST in Energy
Distribution Operations
With Daniel P. Jenkins, Chris Baber
and Richard McMaster

INTRODUCTION
The Event Analysis of Systemic Teamwork (EAST) analysis presented in this
chapter focuses on command and control activities in the energy distribution
domain. Command and control scenarios are characterised by multiple individu-
als and teams working together in pursuit of a common goal (comprising mul-
tiple interacting sub-goals). High levels of communication and coordination are
required, and the onus is often placed on technologies to facilitate this. Various
sub-constructs are also evident within command and control, including planning,
directing, coordinating and control of resources, situation awareness and so on.
When examining command and control, the descriptive constructs of interest can
be distilled down to simply

• why: (the goals of the system, sub-system[s] and actor[s])

• who: (the actors performing the activity, including humans and technologies)
• when: (when activities take place and which actors are associated with them)
• where: (where activities and actors are physically located)
• what: (what activities are undertaken, what knowledge/decisions/processes/
devices are used and what levels of workload are imposed)
• how: (how activities are performed and how actors communicate and col-
laborate to achieve goals).

More than likely, none of the Human Factors methods described in this book
can independently cover all of these constructs. Using an integrated suite of meth-
ods, however, allows scenarios to be analysed exhaustively from the perspectives
described, but more importantly, it allows the effects of constructs on other con-
structs to be considered, for example, how communications influence the way in
which tasks are performed and in turn how the way in which the tasks are being
performed influences the knowledge/decisions/processes being used. There are also
further advantages associated with methods integration, because not only does the
integration of existing methods bring reassurance in terms of a validation history, but
it also enables the same data to be analysed from multiple perspectives. These mul-
tiple perspectives, as well as being inherent in the scenario that is being described

65
66 Systems Thinking in Practice

and measured, also provide a form of internal validity. Assuming that the separate
methods integrate on a theoretical level, then their application to the same data set
offers a form of ‘analysis triangulation’.
EAST (Stanton et al. 2005) provides one such framework of methods that allows
collaborative performance to be comprehensively described and evaluated. Since its
conception, the framework has been applied in many domains, including land and
naval warfare (Stanton et al. 2006), aviation (Stewart et al. 2008), air traffic control
(Walker et al. 2010), railway maintenance (Walker et al. 2006) and the emergency
services (Houghton et al. 2006). EAST is underpinned by the notion that complex
collaborative systems can be meaningfully understood through a network of net-
works approach (see Figure 4.1). Specifically, three networks are considered: task,
social and knowledge networks. Task networks describe the goals and subsequent
tasks being performed within the system. Social networks analyse the organisation
of the system (i.e. communications structure) and the communications taking place
between the actors working in the team. Finally, knowledge networks describe the
information and knowledge (distributed situation awareness) that the different actors
use and share during task performance.
The process begins with the conduct of an observational study (Chapter 2) of the
scenario under analysis. Hierarchical Task Analysis (HTA) (Chapter 3) is then used
to describe the goals, sub-goals and operations involved during the scenario. The
resultant task network identifies the actors involved, what tasks are being performed,
the temporal structure of tasks and the interrelations between tasks. The HTA output
is then used to construct an Operation Sequence Diagram (OSD) (Chapter 3) of the
task, detailing all activities and interactions between actors involved. The social net-
work is embodied by Social Network Analysis (SNA) (Chapter 8), which considers
the associations between agents during the scenario. It is important to note here that
agents may be human or technological, and so SNA caters too for human–machine
interactions. The Critical Decision Method (CDM) (Chapter 4) focuses on the
decision-making processes used during task performance and the information and

TASK
NETWORK

Knowledge required during Operational structure

task performance required

KNOWLEDGE SOCIAL
NETWORK NETWORK
Distribution and
communication of
information underpinning
knowledge

FIGURE 4.1 Network of networks approach.

EAST in Energy Distribution Operations 67

Define
scenario(s)

Data collection methods

Activity Critical
sampling/ Decision
observation Method

Data analysis methods

Hierarchical Social
Task Network
Analysis Analysis

Representational
methods
Operation Social Comms Coordination
Propositional
Task network sequence network Usage Demands
diagram networks Diagram Analysis
diagrams

FIGURE 4.2 Internal structure of EAST framework.

knowledge underpinning decision making. Based on these data, the propositional

network approach (Chapter 6) represents situation awareness during the scenario,
from both the point of view of the overall system and that of the individual agents
performing activities within the system. Thus, the type, structure and distribution
of knowledge throughout the scenario is represented. The overlap between methods
and the constructs they access is explained by the multiple perspectives provided on
issues such as the ‘Who’ and the ‘What’. For example, the HTA deals with ‘what’
tasks and goals, the CDM deals with ‘what’ decisions are required to achieve goals
and the propositional networks deal with ‘what’ knowledge or situation awareness
underpins the tasks being performed and decisions being made. Each is a differ-
ent but complementary perspective on the same descriptive construct, and a differ-
ent but complementary perspective on the same data derived from observation and
interview, which is an example of analysis triangulation. The structure of the EAST
framework is represented in Figure 4.2.

Example Case Study: A Human Factors Analysis

of Civilian Command and Control

The analysis presented in this chapter concerns two scenarios undertaken on a major
UK electrical distribution network. The distribution grid in question consists of 341
geographically dispersed substations in England and Wales, which are used to dis-
tribute electricity to consumers. Power stations (and feeds from continental Europe)
energise the grid, which uses an interconnected network of 400,000 volt (400 Kv),
275 Kv (the super grid network) and 132 Kv overhead lines and towers or cables run-
ning in tunnels to carry electricity from source to substations. The substations are
68 Systems Thinking in Practice

the national distribution company’s interface with regional electricity companies,

which step down the grid’s transmission voltages to 33 Kv, 11 Kv, 400 v and 240 v
for domestic and industrial consumption. Although flexibly manned, in operational
terms, they are remotely manipulated from a central control centre to ensure that the
capacity available in the grid is used in optimal and rational ways and that security
of supply is maintained. Maintenance operations are also coordinated centrally from
another centre, thereby separating operations from safety.

Scenario 1: Switching Operations Scenario (Barking)

Scenario 1 took place at a substation in East London handling voltages and cir-
cuits from 275 Kv down to 33 Kv and a Central Operations Control Room (COCR).
It involved the switching out of three circuits relating to so-called ‘Super Grid
Transformers’ (SGT), which convert incoming transmission voltages of 275 Kv down
to 132 Kv or 33 Kv. Specifically, circuit SGT5 was being switched out for the instal-
lation of a brand-new transformer for a bulk electricity consumer, while SGT1A
and 1B were being switched out for substation maintenance. Associated with such
large pieces of high-voltage apparatus are several control circuits, large overhead
line isolators, remotely operated air blast circuit breakers, other points of isolation,
compressed air equipment and oil cooling apparatus. All of this disparate equipment
had to be handled and made safe by qualified personnel in a highly prescribed man-
ner. In addition, the work had to be centrally pre-planned by electrical engineers to
ensure that other circuits were not affected and that the balance and capacity of the
system was not compromised. Qualified personnel worked on-site to these plans and
liaised with the COCR at key points during this process.

Scenario 2: Maintenance Scenario (Tottenham)

Scenario 2 took place at the COCR and a rural substation site. It involved the switch-
ing out of circuits and overhead lines in order to permit work to commence on pieces
of the control equipment (Current and Voltage Transformers) used to provide read-
ings and inputs into other automatic, on-site current, voltage and phase regulation
devices. In addition, maintenance work was to be carried out on a Line Isolator (the
large mechanical switching device that provides a point of isolation for a specific
overhead line that departed from this substation [A] and terminated at another sub-
station [B]), and major maintenance on a device called an earth switch. There were
three main parties involved in the outage: a party working at Substation B on the
outgoing substation A circuit, a party working at Substation A on the substation B
circuit and an overhead line party working in between the two sites.
In both scenarios, the COCR operator, located a central control room (see Figure
4.3), took on the role of commander, planning the work to be undertaken and dis-
tributing work instructions to the Senior Authorised Persons (SAPs) and Authorised
Persons (APs) located at the substations where the work was to be undertaken. In
addition to overseeing the activities that were analysed for the purposes of this
research, the COCR operator was also involved in other activities being undertaken
elsewhere on the grid and so had other responsibilities and tasks to attend to during
the study. The other agents involved in the scenarios included the Central Command
(CC) Operator and Overhead Line Party (OLP) personnel working on the overhead
EAST in Energy Distribution Operations 69

FIGURE 4.3 Central control room.

lines. The COCR operator communicated with the other agents via landline tele-
phone and mobile phone. The COCR operator also had access to substation dia-
grams, work logs and databases and the internet. The structure of personnel for
Scenario 1 is presented in Figure 4.4. The structure of personnel for Scenario 2 is
presented in Figure 4.5.

METHODOLOGY
Design
The study was an observational study that involved direct observation of the activi-
ties undertaken during the scenarios analysed. Three researchers acted as observers,
with one located in the field at the substation involved and two located at the central
control room.

Participants
This study involved 11 participants who work for the energy distribution organisa-
tion in question. Scenario 1 involved the following four participants: a CC operator, a
COCR operator and a SAP and AP. Scenario 2 involved the following seven partici-
pants: a CC operator, a COCR operator and a SAP, AP and Competent Person (CP)
at one substation, and SAP at another substation and an overhead line party contact.
Due to access restrictions and the nature of the study (observation during real work
activities), it was not possible to collect demographic data for the participants involved.
70 Systems Thinking in Practice

Central
Command Centre

CC
Operator

Central
COCR
Operator
Operations
Control Room

SAP AP

Substation

FIGURE 4.4 Scenario 1 network structure; CC Operator = Central Command Operator;

COCR Operator = Central Operations Control Room Operator; SAP = Senior Authorised
Person; AP = Authorised Person.

Materials
The observers used pen and paper, video and audio recording equipment to collect
data during the observations. All observational and verbal transcripts were tran-
scribed using Microsoft Word. A set of pre-defined interview probes were used for
the CDM interviews, along with a CDM pro-forma to record participants’ responses.
A Dictaphone was also used to record the CDM interviews. Various other software
programs were used to support the analysis presented, including the Agna network
analysis and WESTT software packages and Microsoft Visio.
EAST in Energy Distribution Operations 71

CC Central
Operator Command

Central
Operations
COCR Control Room
Operator OLP
Contact

Overhead Lines

SAP#1 SAP#2

Substation#2

AP#1 CP#1

Substation#1

FIGURE 4.5 Scenario 2 network structure; CC Operator = Central Command Operator;

COCR Operator = Central Operations Control Room Operator; SAP = Senior Authorised
Person; AP = Authorised Person; CP = Competent Person; OLP Contact = Overhead Line
Party Contact.

Procedure
The analyses were based on data collected during live observational study of the two
scenarios. In each scenario, two observers were located at the COCR observing the
COCR operator, and one observer was located in the field with the SAPs/APs at the
substation involved. The analysts located at the COCR observed all of the COCR
operator’s activities and were able to discuss the activities being undertaken and
to query different aspects of the scenarios as they unfolded. The analyst located at
the substations observed the SAP/APs undertaking the work required and was also
able to discuss aspects of the scenario with them. Observational transcripts were
constructed, and audio recordings were used to record the communications between
those involved. The data recorded via observational transcripts included a descrip-
tion of the activity (specific task steps, e.g. ‘issue instructions to SAP at substation’)
performed by each of the agents involved, transcripts of the communications that
72 Systems Thinking in Practice

occurred between agents during the scenarios, the technology used to mediate these
communications, the artefacts used to aid task performance (e.g. tools, computers,
instructions, substation diagrams), time and additional notes relating to the tasks
being performed (e.g. why the task was being performed, what the outcomes were).
CDM interviews were conducted with the key agents involved (the COCR opera-
tor and the SAPs) upon completion of the scenario. This involved decomposing the
scenario into a series of key decision points and administering CDM probes in order
to interrogate the decision-making processes used at each point. For validation pur-
poses, a subject matter expert from the energy distribution company reviewed the
data collected and the subsequent analysis outputs.

RESULTS
Task Networks
HTAs were constructed for each scenario based on standard operating procedures
and direct observation of the scenarios. The HTAs describe each scenario in terms of
a structured hierarchy of goals and sub-goals, along with feedback loops. One useful
way of summarising large and complex HTA outputs is through the construction of
a task network, which provides a summary of the main higher-level goals and tasks
involved and the interaction between them. Task networks for each scenario are pre-
sented in Figure 4.6.

Social Networks
Social network analysis is used to examine individuals or teams who are linked to
each other by communications and to subject that network to mathematical analysis
using Graph Theory (Driskell and Mullen 2004). Social networks were constructed
for each scenario based on the observed communications between agents during task
performance. The structure of each network was then analysed using three network
analysis metrics: density, sociometric status and centrality. Network density represents
the level of interconnectivity of the network in terms of communications links between
agents. Density is expressed as a value between 0 and 1, with 0 representing a network
with no connections between agents and 1 representing a network in which every agent
is connected to every other agent (Kakimoto et al. 2006; cited in Walker et al. 2011).
Sociometric status provides a measure of how ‘busy’ each agent is relative to the total
number of agents within the network under analysis (Houghton et al. 2006). Centrality
is also a metric of the standing of each agent within a network (Houghton et al. 2006),
but here this standing is in terms of its ‘distance’ from all other agents in the network.
A central agent is one that is close to all other agents in the network, and a message
conveyed from that agent to an arbitrarily selected other agent in the network would, on
average, arrive via the least number of relaying hops (Houghton et al. 2006). The social
network analysis for Scenarios 1 and 2 are presented in Figure 4.7.
The social network analysis outputs allow identification of the key communi-
cations ‘hubs’ within each network, as well as a comparison of the communica-
tions network apparent in each scenario. Key agents, specifically those that act as
EAST in Energy Distribution Operations 73

Scenario 1 Scenario 2

Consult operational log Consult operational Perform pre-

log op activities

Prepare instructions Prepare instructions

Deal with other tasks and unplanned events

Send and receive task-related information

Deal with other tasks and unplanned events

Distribute instructions

Send and receive task-related information

Distribute instructions

Conduct switching Perform de-earthing

operations

Report completion of Render equipment operative

operations

Report operations complete

Transfer control of circuit

Perform post-restorative activities

= COCR operator = SAP/AP

FIGURE 4.6 Task models for Scenarios 1 and 2; shading denotes which agents undertake
activities related to each goal.

communications hubs within each network, are identified through examination of

the centrality and sociometric status values. Agents with values above the central-
ity and sociometric status mean values for the overall network are defined as key
agents. For Scenario 1, the key agents are the COCR operator and the SAP/AP at the
substation. For Scenario 2, the COCR operator is the key agent. The network density
values show that the communications network in Scenario 1 was denser than that of
Scenario 2. That is, there are less agents involved in the scenario but a similar num-
ber of communications between them.

Information Networks
Information networks were constructed for each scenario phase (as defined by the
task networks) using the observational transcripts, CDM interview responses and the
HTA of the tasks performed. The information network for Scenario 1 is presented
in Figure 4.8. Within Figure 4.8, each node represents a concept, and the directional
arrows show the relationships between the concepts (e.g. substation ‘has’ location,
display ‘shows’ circuit). The information networks were interrogated to identify the
usage and ownership of concepts by each agent involved as well as the key concepts
74 Systems Thinking in Practice

Scenario 1

CC
Operator Agent Density Centrality S. Status

COCR Operator 2.16 6

1 CC Operator 1.85 3.66
2
0.31
3 3 SAP/AP 2.16 6.3
REC 1 REC Operator 1.85 2
2
Operator

1
COCR
Operator
2
8
1

3
SAP/AP

Scenario 2
Agent Density Centrality S. Status
CC COCR Operator 4.72 6.4
Operator SAP
substation CC Operator 2.6 1.2
3
1 SAP/AP Sub 1 3.25 2.4
4 0.2
4 SAP/AP Sub 2 2.73 2
2
SAP/AP Sub 3 2.73 1.2
1
COCR Overhead line 2,73 0.4
Operator Overhead party
1 line party

8
1
2
2 8
SAP/AP SAP/AP
substation substation
1 2

FIGURE 4.7 Social network analysis outputs for Scenarios 1 and 2.

underpinning situation awareness. The usage of concepts by each agent is presented

in Figures 4.9 (Scenario 1) and 4.10 (Scenario 2). Further, the usage of concepts by
each agent per scenario phase is presented in Table 4.1.
The key concepts underpinning situation awareness were identified through the
use of the centrality and sociometric status (see social network analysis). Key con-
cepts are defined as those that have salience for each scenario phase, salience being
defined as those concepts that act as hubs to other concepts. Those concepts with a
sociometric status and centrality value above the mean values for the overall network
are taken to be key concepts (Figure 4.11).

Additional EAST Analyses

A deeper level of analysis is provided through additional methods from the frame-
work, including Operation Sequence Diagrams (OSDs), Coordination Demands
EAST in Energy Distribution Operations 75

Analysis (CDA) (Burke 2004) and Communications Usage Diagram (CUD) (Watts
and Monk 2000). In the present study, the task, social and knowledge network analy-
ses were supplemented by OSD, CDA and CUD analyses. The purpose of these
additional analyses was to examine levels of workload and teamwork during the
scenarios and also to evaluate the technologies used to mediate communications.
CUD (Watts and Monk 2000) is used within the EAST framework to describe
the communications between agents and evaluate the technologies used to mediate

FIGURE 4.8 Propositional network for Scenario 1.

76 Systems Thinking in Practice

80
70
60
50
40
30
20
10
0
COCR Operator CC Operator Senior Authorised
Authorised Person
Person

FIGURE 4.9 Information element usage for Scenario 1 (overall).

100
90
80
70
60
50
40
30
20
10
0
COCR CC Operator Senior Authorised Overhead
Operator Authorised Person line party
Person

FIGURE 4.10 Information element usage for Scenario 2 (overall).

TABLE 4.1
Information Element Usage by Agents throughout
Each Phase of Scenario 2

Information Element Usage

Scenario Phase
Agent 1 2 3 4 5 6 7 Total
COCR Operator 22 41 24 0 20 30 0 67
SAP 34 36 0 29 18 37 0 84
AP 33 29 0 24 18 3 14 91
CC Operator 26 0 0 0 0 0 0 57
OLP Contact 15 0 0 9 13 0 0 43
EAST in Energy Distribution Operations 77

Scenario One Key Information Scenario Two Key Information

Elements Elements

Check, Open, Check, Open,

Circuits Lock and Isolation Lock and
Caution Caution

Switching
Confirmation Equipment Confirmation
Operations

System State
Locations Locations 275Kv CVTs
Certificate

Switching Work
Reconfiguration Switching Log
Instructions Instructions

Status Requirements Motor Fuses Caution Tape

Switching
Time Job Cards Cabinet
Phone

Break In Readback Lock Out Key Caution Sign

Time Earthing

FIGURE 4.11 Key information elements for Scenarios 1 and 2.

these communications. For the present study, those bottom-level task steps from the
HTAs involving communications between agents were examined. An extract from
the CUD analysis from Scenario 1 is presented in Figure 4.12.
CDA (Burke 2004) is used to derive a rating of the level of coordination exhibited
between team members during collaborative scenarios. Again, using the HTA as its
primary input, teamwork tasks were first identified, following which coordination
between team members on each teamwork task step was rated using a taxonomy
of key teamwork behaviours. An extract of the CDA for Scenario 1 is presented in
Table 4.2.
For Scenario 1, the CDA analysis revealed that 64% of the bottom-level HTA
tasks were teamwork tasks. The mean level of coordination across all teamwork
tasks was 1.57, which represents a medium level of coordination.

SUMMARY
The aim of this chapter was to demonstrate an integrated framework of Human
Factors methods for analysing performance in complex sociotechnical systems.
Although only a summary of the outputs derived were presented, the utility of tak-
ing a framework approach to the analysis of performance in complex sociotechnical
systems is demonstrated. In this case, EAST enables description and examination of
78

Scenario 1
SAP/AP at Comms Analysis of comms
Time COCR Operator CC Operator REC Operator
substation media media
Contact COCR to agree
08:56 SSC in order to release
part of circuit for
substation to control by
COCR + Clear and stable sound
Contact SAP/AP at + Two way comms
+ Allows concurrent viewing
09:08 substation to discuss
outage at Supergrid of site diagrams/desktop
transformer computer screens
+ Confidentiality maintained
Contact SAP/AP at
- Data is not recorded;
substation for information may be lost
09:37 preliminary
discussions about
forthcoming work

SAP and AP engage in

09:54 pre-amble regarding
forthcoming job

Contact SAP at
substation to
confirm isolation
10:13 requirements

FIGURE 4.12 Communications Usage Diagram extract.

Systems Thinking in Practice
TABLE 4.2
CDA Extract

Teamwork = 64% of HTA Tasks Task Work = 36% of HTA Tasks

Mean Coordination Across all Teamwork Tasks = 1.57
Task Agent Task Type Comms SA DM MA Lead Adapt Assert Mean Co–ord
1.1.1. Use phone to contact COCR CC Operator Task work – – – – – – – –
1.1.2. Exchange Ids CC Operator & Team work 3 3 1 1 1 1 1 1.57
COCR Operator
EAST in Energy Distribution Operations

1.1.3. Agree SSC documentation CC Operator & Team work 3 3 3 1 1 1 1 1.86

COCR Operator
1.1.4.1. Agree SSC with CC Operator COCR Operator Team work 3 3 3 1 1 1 1 1.86
1.1.4.2. Agree time with CC Operator COCR Operator Team work 3 3 3 1 1 1 1 1.86
1.1.5.1. Record details onto log sheet COCR Operator Task work – – – – – – – –
1.1.5.2. Enter details into worksafe system COCR Operator Task work – – – – – – – –
1.2.1. Ask for isolators to be opened COCR Operator Team work 3 3 1 2 2 1 1 1.86
remotely
1.2.2. Perform remote isolation CC Operator Task work – – – – – – – –
1.2.3. Check substation diagram on screen COCR Operator Task work – – – – – – – –
1.2.4. Terminate comms. CC Operator & Team work 3 1 1 1 1 1 1 1.29
COCR Operator
79
80 Systems Thinking in Practice

TABLE 4.3
Elements of Command and Control Examined Through EAST

Social
Network Propositional
HTA Analysis Networks CDM CUD CDA OSD
Tasks
Communications
Teamwork
Situation Awareness
Decision Making
Technologies
Workload

the task and social and knowledge networks underpinning complex sociotechnical
system performance, along with a deeper analysis of various specific constructs.
For example, in both scenarios, the methods with the EAST framework supported
examination of the following elements of command and control.
Table 4.3 demonstrates how, when taken individually, the outputs offer insight
into various components of command and control (e.g. situation awareness, team-
work, communications); for example, the propositional networks alone can be used
to make judgements on the levels of situation awareness attained throughout the
scenarios and also to determine what knowledge underpins situation awareness dur-
ing energy distributed maintenance scenarios. Integration of the outputs, however,
provides a much deeper insight. For example, viewing the task, social and knowledge
networks together enables judgement to be made on how the procedures used and
the organisation of the social network supported situation awareness during opera-
tions. In the case of the task networks, the standard operating procedures engendered
high levels of closed loop communication between the agents involved throughout
the scenario, including issuing of work instructions, regular updates on work prog-
ress and reporting of completed operations. In the case of the social networks, in
both scenarios, each of the agents involved had one or more ways of communicating
with all other agents (e.g. landline telephone, mobile phone and emails). Further, the
hierarchical structure of the network meant that the COCR operator, effectively as
network commander, would regularly contact (or be contacted by) agents in the field
in order to gather or provide work progress updates, a process that served to update
situation awareness throughout the activities. Combining the analysis outputs thus
allows examination of not only how situation awareness was developed and main-
tained and what it comprised but also why efficient levels of situation awareness were
achieved.
The EAST framework lends itself to in-depth evaluations of complex sociotechni-
cal system performance, examination of specific constructs within complex socio-
technical systems (e.g. situation awareness, decision making, teamwork), and also
EAST in Energy Distribution Operations 81

system, training, procedure and technology design. Whilst not providing direct rec-
ommendations, the analyses produced are often highly useful in pinpointing specific
issues limiting performance or generating system redesign recommendations.

REFERENCES
Burke, S. C. (2004). Team task analysis. In: N. A. Stanton, A. Hedge, K. Brookhuis, E. Salas
and H. Hendrick (Eds.), Handbook of Human Factors and Ergonomics Methods (pp.
56.1–56.8). CRC Press: Boca Raton, FL.
Driskell, J. E. and Mullen, B. (2004). Social network analysis. In: N. A. Stanton, A. Hedge,
K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors and
Ergonomics Methods (pp. 58.1–58.6). CRC Press: Boca Raton, FL.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49, 1204–1225.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork. Ashgate: Aldershot, UK.
Stanton, N. A., Salmon, P. M., Baber, C., Walker, G. H. and Jenkins, D. (2005). Human
Factors Methods: A Practical Guide for Engineering and Design. Ashgate: Aldershot,
UK.
Stanton, N. A., Salmon, P. M., Rafferty, L. A., Walker, G. H., Baber, C. and Jenkins, D.
(2013). Human Factors Methods: A Practical Guide for Engineering and Design (sec-
ond edition). Ashgate: Aldershot, UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon, P.
M. et al. (2006). Distributed situation awareness in dynamic systems: Theoretical devel-
opment and application of an ergonomics methodology. Ergonomics, 49, 1288–1311.
Stewart, R., Stanton, N. A., Harris, D., Baber, C., Salmon, P. M., Mock, M., Tatlock, K.,
Wells, L. and Kay, A. (2008). Distributed situation awareness in an airborne warn-
ing and control system: Application of novel ergonomics methodology. Cognition
Technology and Work, 10 (3), 221–229.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M. and Green, D. (2006).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics, 49, 1345–1369.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2011). Cognitive compatibility of motorcy-
clists and car drivers. Accident Analysis & Prevention, 43 (3), 878–888.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Jenkins, D. P. and Salmon, P. M. (2010).
From ethnography to the EAST method: A tractable approach for representing distrib-
uted cognition in air traffic control. Ergonomics, 53 (2), 184–197.
Watts, L. A. and Monk, A. F. (2000). Reasoning about tasks, activities and technology to
support collaboration. In: J. Annett and N. Stanton (Eds.), Task Analysis (pp. 55–78).
Taylor and Francis: London, UK.
5 EAST in a Submarine
Control Room

ANALYSING DISTRIBUTED COGNITION

Distributed cognition is characterised by multiple individuals and teams working
together in pursuit of a common goal (comprising multiple interacting sub-goals).
High levels of communication and coordination are required, and there is often an
onus placed on technologies to facilitate this. Hutchins’ (1995) investigation into navi-
gation on ships showed how distributed cognition worked in practice. In the analysis
of a navigation task, the ship’ s crew were trying to determine two things: to identify
the ship’ s current position and to estimate the ship’ s future position. Hutchins noted
that this task was distributed among the crew and artefacts they use. No one indi-
vidual was able to explain the entirety of navigation. Rather, to understand how ship
navigation is performed, one needs to study the interactions within the wider socio-
technical system. Hutchins studied navigation in restricted waters as this is perhaps
the most challenging task and requires up to 10 personnel: the navigator, assistant
to the navigator, navigation plotter, navigation bearing recorder and timer, starboard
pelorus operator (a starboard observer who takes bearing fixes from landmarks), port
pelorus operator (a port observer who takes bearing fixes from landmarks), restricted
manoeuvring helmsman, quartermaster of the watch, restricted manoeuvring helms-
man in after steering and the fathometer operator. His explanation and analysis were
presented in a mixture of transcripts of spoken voice, narrative, diagrams and pho-
tographs. The richness of the interpretation is derived by the reader through mental
simulation of the scenario as they absorb the information presented. No single Human
Factors method has convincingly covered all of this complexity in its entirety (Stanton
et al. 2005) making the modelling of distributed cognition a significant challenge.
Using an integrated suite of methods however, allows scenarios to be analysed from
the perspectives described. More importantly, it allows the effects of one set of con-
structs on other sets of constructs to be considered, for example, how communications
influence the way in which tasks are performed and, in turn, how the way in which
the tasks are being performed influences the information being used (Stanton et al.
2008). There are also advantages associated with methods integration, because not
only does the combination of existing methods bring reassurance in terms of a valida-
tion history, but it also enables the same data to be analysed from multiple perspec-
tives. Assuming that the separate methods integrate on a theoretical level, then their
application to the same data set offers a form of ‘ analysis triangulation’ .
Various approaches have been put forward to analyse complex sociotechnical sys-
tems (Farrington-Darby et al. 2006; Furniss and Blandford 2006; Patrick et al. 2006;
Stanton et al. 2006; Waterson 2009; Jenkins et al. 2011). The aim of the approaches
is both to make the complexity of sociotechnical systems more explicit, so that the

83
84 Systems Thinking in Practice

interactions between sub-system boundaries may be examined, and to reduce the

complexity to a manageable level. This may seem at odds with the non-linear nature
of complex systems and the unpredictable properties that may emerge from them
(Walker et al. 2010a, b; Stanton et al. 2012). Nevertheless, the representations do
enable inspection of the relationships between sub-systems and their boundaries,
even if some of the fidelity is lost (Griffin et al. 2010). The co-evolution of systems
design requires some occasional stock-taking before the next evolution (Stanton
et al. 2012). The notion that problems become apparent at boundaries of layers, lev-
els, structures and functions is not new, as it underpins general systems theory (von
Bertalanffy 1950). Of particular interest here is the boundary between distributed
cognitive sub-systems, as this is where problems may become apparent.
Walker et al. (2010a) proposed Event Analysis of Systemic Teamwork (EAST)
as a method that could be used to show how distributed cognition could be mod-
elled in air traffic control. The analysis showed how distributed cognition for com-
plex systems could be represented by networks, with the distinct advantage being
that the networks enabled both qualitative and quantitative investigations. Walker
et al. argued that the multi-faceted nature of the different networks revealed the
aggregated behaviours that emerge in complex sociotechnical systems. This repre-
sentation was proposed as an alternative to the reductionistic approaches often used
to understand systems, which present systems in their constituent parts but fail to
capture the system as a whole. Walker et al. suggested that the insights gained by
network modelling were superior to the traditional ethnographic narrative that has
previously been used to describe distributed cognition because they present graphi-
cal models of systems. Griffin et al. (2010) go further to show how the EAST method
offers insight into system failure. Again, the cited advantage of the approach was the
non-reductionistic, non-taxonomic method for analysing non-normative behaviour
of systems. Whilst EAST does not employ taxonomies in the analysis, the resultant
network structures may be classified into archetypes. The systemic approach allows
system interactions to be understood in their entirety (Plant and Stanton 2012).
EAST is underpinned by the notion that complex collaborative systems can be
meaningfully understood through a network of networks approach (see Figure 5.1).

FIGURE 5.1 Network of networks approach.

EAST in a Submarine Control Room 85

Specifically, three networks are considered: task, social and information networks.
Task networks describe the relationships between tasks and their sequence and
interdependences. Social networks analyse the organisation of the system (i.e. com-
munications structure) and the communications taking place between the actors
working in the team. Finally, information networks describe the information that
the different actors use and communicate during task performance (i.e. distributed
situation awareness). Each of these approaches have been presented independently
in other papers, such as Farrington-Darby et al.’ s (2006) presentation of task dia-
grams in a study of railway controllers (an example of a task network); Furniss and
Blandford’ s (2006) presentation of communication channels in emergency medi-
cal dispatch teams (an example of a social network); and Sanderson et al.’ s (1989)
analysis of verbal protocols for a process control task (an example of an information
network). What EAST does is bring these three networks together into the same
analysis framework.
The EAST framework lends itself to in-depth evaluations of complex system per-
formance, examination of specific constructs within complex sociotechnical systems
(e.g. situation awareness, decision making, teamwork) and also system, training, pro-
cedure and technology design. Whilst not providing direct recommendations, the
analyses produced are often highly useful in identifying specific issues limiting per-
formance or highlighting areas where system re-design could be beneficial.
This chapter presents a new shortened form of EAST that focuses on the analysis
of collaborative performance (Stanton et al. 2008). This version of EAST does not
rely upon the previously defined constituent methods of Hierarchical Task Analysis,
Critical Decision Methods, Coordination Demand Analysis, Communications Usage
Diagram and Operation Sequence Diagram. Rather, the task, social and information
networks are developed directly from the raw data. It also extends the quantitative and
qualitative network analysis methods to show how network statistics may be applied
to all three networks and how the different networks’ perspectives may be integrated.
Since its conception, the framework has been applied in many domains, including
naval warfare (Stanton et al. 2006), aviation (Stewart et al. 2008), air traffic control
(Walker et al. 2010a), emergency services (Houghton et al. 2006), energy distribution
(Salmon et al. 2008) and railway maintenance (Walker et al. 2006). The approach is
gaining momentum as well as showing its domain independence.

SYSTEM PROPERTIES OF EAST

EAST is a Systems Ergonomics method and as such is able to be applied to all sys-
tems levels: micro (i.e. individual human–machine interaction), meso (i.e. organisa-
tions operating highly automated systems) and macro (i.e. multi-layered networked
system). At each level of the systems analysis, it is possible to construct the triad of
task, social and information networks as each is a distributed cognition system at dif-
ferent units of analysis. Further, the EAST representations could be nested such that
the micro networks are contained with the meso networks which are also contained
in the macro networks. So, whilst the networks might change in terms of the granu-
larity of the analysis, the methodology would remain the same. Wilson (2012) iden-
tified six of the defining characteristics essential to Systems Ergonomics: systems
86 Systems Thinking in Practice

TABLE 5.1
Systems Characteristics of EAST

Characteristic Property of EAST

Systems focus Captures the whole sociotechnical system in the network analysis and does not favour
one system over the other.
Context Analyses system behaviour at work using observed and recorded from data from that
context with input from subject matter experts. System boundaries are defined by
subject matter of interest and may also emerge from the analysis conducted.
Interactions The interacting parts of the system are revealed in the three networks and the
relationships between the networks, as indicated in Figure 5.1. Thus, both
interactions within and between networks can be analysed showing distributed
cognition in terms of task-social, task-informational, social-informational and
task-social-informational interactions.
Holism The networks are analysed as a whole, both quantitatively (using SNA metrics) and
qualitatively (using network archetypes). The networks are also superimposed upon
each other to produced combined networks.
Emergence The emergent properties of the system are revealed through the SNA metrics and the
network archetypes.
Embedding The method itself is embedded in the communications and systems engineering
disciplines, so it offers familiarity to organisations wishing to scrutinise their
sociotechnical systems. It has the benefit of representing the networks in graphical
form as well as supporting metrics for detailed analysis.

Source : Wilson, J. R. Work , 41, 3861– 3868, 2012.

focus, context, interactions, holism, emergence and embedding. Any method that
lays claim to be a Systems Ergonomics, such as EAST, should embody these charac-
teristics. Table 5.1 indicates how EAST encapsulates these properties.
As the contents of Table 5.1 attest, EAST passes the ‘ systems’ test. By conceiv-
ing of systems as sets of interacting networks (i.e. task, social and information net-
works), it has the distinct advantage of enabling network analysis and classification
of network archetypes. This means that the systems analysis methods can be framed
in terms of quantitative and qualitative assessments. Networks do not differentiate
between different types of node (e.g. artefacts and/or people) so that from a model-
ling perspective they are not constrained by existing structures of people and arte-
facts; rather, they are related to the tasks associated with a scenario. It is also possible
to model the temporal aspects of networks by identifying critical moments in the
sequence of activity. To do this, the scenario is divided into task phases allowing
active and non-active elements to be specified and represented.

DATA COLLECTION AND ANALYSIS

EAST was undertaken as an ‘ information audit’ of the relationship between the
sound room and control room on board a submarine. The control and sound rooms
have all the characteristics of distributed cognition problems (as described by
EAST in a Submarine Control Room 87

Hutchins 1995). The purpose was to understand how information flowed around
the system in order to determine what requirements might be included in the next
generation of tactical systems.
The data were collected over a series of observations in a control room and sound
room layout in the Talisman Command Team Trainer. There is a strong communica-
tion link between the Officer of the Watch (OOW), Chief Petty Officer for Tactical
Systems (CHOPS[TS]: sometimes called the OpsO or Operations Officer) and the
Sound Room Controller (SRC, sometimes called the Sonar Controller or CHOPS[S]).
The layout of the control room and sound room is fixed by the equipment layout but has
been optimised over the decades to the current design. The photograph in Figure 5.2
shows a view into the control room with the periscope on the right-hand side.
All of the internal communications were recorded by plugging into the system
and recording on a laptop computer using Audacity (version 1.3.14 software). The
ambient communications were recorded using an external Yoga BM-26D boundary
microphone.
The purpose of this chapter was to see if the networks could be analysed quantita-
tively using Social Network Analysis (SNA) metrics. SNA offers a means of analysing
the network as a whole as well as the behaviour of individual nodes and their inter-
actions. As such, SNA is potentially a very powerful tool for Systems Ergonomics.
Whilst it has traditionally been applied to the analysis of social networks (as implied
by the name of the method) Driskell and Mullen 2005; Houghton et al. 2006), there
is no reason why it cannot be applied to other networks, such as task and information
networks. This is a new application for the method but a potentially useful one. The
method can also be applied to the design of anticipated networks, so that more effec-
tive task, social and information networks can be designed into new systems, which

FIGURE 5.2 View into the control room with periscope on right-hand side.
88 Systems Thinking in Practice

is another new avenue of research for Systems Ergonomics that would enable network
resilience to be explored in a practical manner. The first step in a SNA involves defin-
ing the network that is to be analysed. Once the overall network type is specified,
the tasks, agents or information should be specified. Once the type of network under
analysis has been defined, the scenario(s) within which they will be analysed should
be defined. For the purposes of this chapter, the scenario for return to periscope depth
(RTPD) was considered. Once the network and scenario(s) under analysis are defined
clearly, the data collection phase can begin. There are a number of metrics associated
with the analysis of social networks, depending upon the type of evaluation that is
being performed. The size of the network determines the number of possible relations,
and the number of possible relations grows exponentially with the size of the network.
This defines the network’s complexity. An explanation of the metrics analysed using
AGNA (version 2.1.1 – a software program for computing the SNA metrics) are pro-
vided below. The first set of metrics analyse the individual nodes:

• Emission and reception degree are the number of ties emanating from, and
going to, each agent in the network.
• Eccentricity is defined by the largest number of hops an agent has to make
to get from one side of the network to another.
• The sociometric status of each agent refers to the number of communica-
tions received and emitted, relative to the number of nodes in the network.
• Agent centrality is calculated in order to determine the central or key
agent(s) within the network. There are a number of different centrality cal-
culations that can be made. For example, agent centrality can be calculated
using Bavelas-Leavitt’ s index.
• Closeness is the inverse of the sum of the shortest distances between each
individual and every other person in the network. It reflects the ability to
access information through the ‘ grapevine’ of network members.
• Farness is the index of centrality for each node in the network, computed as
the sum of each node to all other nodes in the network by the shortest path.
• Betweenness is defined by the presence of an agent between two other agents,
which may be able to exert power through its role as an information broker.

The second set of metrics analyse the whole network:

• The density of a network is defined by the number of social relations that

are actually observed and can be represented as some fraction of the total
possible.
• Cohesion is defined as the number of reciprocal connections in the network
divided by the maximum number of possible connections.
• The largest geodesic distance within a network defines its diameter , which
can be thought of as another metric of the network’ s size, that is, the num-
ber of hops to get from one side of the network to the other.

The first step in the analysis was to calculate the statistics for each of the nodes in
the network, of which a range can be produced to represent the metrics of distance
EAST in a Submarine Control Room 89

(i.e. eccentricity), sociometrics (i.e. emission/reception and sociometric status) and

centrality (i.e. centrality, closeness, farness and betweenness). The second step was
to calculate statistics for the whole network (i.e. density, cohesion and diameter).
The final step was to combine the networks for a qualitative assessment. Each of the
networks is presented in turn followed by combinations of the networks.

TASK NETWORK ANALYSIS

The task network was constructed from the main phases of the RTPD activities
involved in returning the submarine to periscope depth. Rather than represent the
phases as a linear process, the task network portrays the relationships between the
tasks that are non-sequential. The decision to RTPD is continually being assessed
and reassessed. If an unexpected contact (i.e. another vessel) appears in the area (or
going towards the area) where the submarine is heading, then the decision to RTPD
will be cancelled and the submarine will head down to the safe depth (below that of
the deepest ship’ s hull, i.e. greater than 30 metres). A depiction of the task network
is shown in Figure 5.3. As Figure 5.3 shows, the focus of the task is on identifying all
of the contacts surrounding the area where the submarine is intending to RTPD. This
focus continues all the way up and when the ‘ look’ (i.e. viewing the surface using the
periscope) is established. At any point in the manoeuvre, the ship will return to a safe
depth if a contact is detected. The decision to RTPD begins with the OOW calling an
outstations briefing to ensure all people and systems on the submarine are in order
and ready for the manoeuvre. This is followed by ballasting (to ensure the submarine
is in trim) and clearing of stern arcs (so that the sound room can check for vessels
behind the submarine). Then the sound room and control room are engaged in the
activities to range all contacts within the local area to find a safe area of sea to RTPD.
If no safe area is found, then the submarine will need to change area and continue
to range contacts. When all contacts have been ranged and an area of sea has been
identified, the OOW reports to the Captain for permission to RTPD. If permission
is forthcoming, the OOW will request final reports from outstations to check that
the people and systems on the submarine are in order and ready for the manoeuvre.
A decision will be made by the OOW whether to conduct the standard routine (where
range and bearing of all contacts are called out as the submarine returns to periscope
depth) or a silent routine (where the submarine returns to periscope depth as stealth-
ily as possible). At periscope depth (called PD in Figure 5.3) the OOW calls ‘ break-
ing’ and conducts two sweeps with the periscope to check that the submarine is clear
from potential collision with contacts and will then range the contacts manually to
update the Submarine Maritime Command System (SMCS) operators. Then the mis-
sion intentions can be carried out. If the submarine is on a potential collision course
with any of the contacts, the OOW will give the order to dive to the safe depth. As
stated previously, the task network helps to identify interdependencies and relation-
ships between tasks.
The network in Figure 5.3 was used to construct an association matrix to indi-
cate the presence or absence of links between tasks. The number of links into and
out of each task are shown in the ‘ reception’ row and ‘ emission’ rows respectively
of Table 5.2. The network statistics (as described in Section 3 on data collection
90 Systems Thinking in Practice

FIGURE 5.3 Task network for RTPD.

and analysis) were then computed using AGNA. Whilst using SNA to analyse task
structures is somewhat new to ergonomics, it can give us some insights into the use
of metrics. Obviously, the task ‘ detect close contacts’ is critical in RTPD, which is
reflected in most of the metrics. Some of the other tasks are less obvious. The two
tasks of ‘ clear stern arcs’ and ‘ range all contacts’ score quite low on Sociometric
Status but quite high on Centrality and Betweenness. The higher the value means the
higher each task is scored on that metric.
The network density was 0.14 (i.e. a low distribution network) and cohesion was
0.02 (i.e. a very low level of reciprocal links – suggesting very strict dependences
in the task network, as shown in the task network diagram) with a diameter of 9
(i.e. nine hops from one side of the network to the other). The network can also be
described as weighted (i.e. non-uniform) and asymmetric (i.e. unbalanced).
TABLE 5.2
Analysis of Task Network
Final Warner Establish
OOW Detect OOW report Conduct Conduct Return to clearances look at
Task Network Safe outstations Clear Range all close report to from standard silent periscope at periscope periscope Conduct
EAST in a Submarine Control Room

Analysis depth briefing stern arcs Ballasting contacts contacts Captain outstations routine routine depth depth depth mission

Reception 2 1 2 2 1 7 1 2 1 1 2 1 2 1
Emission 1 3 2 1 3 1 1 3 2 2 3 2 2 0
Eccentricity 8 7 6 7 5 9 6 6 7 7 7 8 9 0
Sociometric 0.2 0.3 0.3 0.2 0.3 0.6 0.2 0.4 0.2 0.2 0.4 0.2 0.3 0.1
Status
Centrality (B-L) 7.7 7.8 7.9 6.9 7.7 7.5 6.5 7.3 6.4 6.4 6.3 5.7 5.8 10.4
Closeness 0.2 0.3 0.3 0.2 0.4 0.2 0.3 0.3 0.3 0.3 0.3 0.2 0.2 0
Farness 58 46 41 53 36 67 46 39 46 46 49 56 63 0
Betweenness 67 68 70 0 70 65 0 59 17 17 31 0 12 0
91
92 Systems Thinking in Practice

SOCIAL NETWORK ANALYSIS (SNA)

The social network analysed the relationships (e.g. communications and activity)
between the agents during the scenario, that is, the Captain, OOW, Ships Control,
WECDIS, Weapons Engineering Officer (WEO), OpsO, Sound Control and SMCS.
The matrix represents the frequency of communications between each agent in the
network that could be captured through the network communications system and the
ambient communications (it is accepted that not all of the communications in the net-
work could be captured, particularly those outside the network communications sys-
tem). The association matrix for the RTPD scenario is presented in Table 5.3. This
matrix shows whether an agent within the system can be associated with any other
agent, specifically through frequency of communications.
Once the matrix of association is completed, the social network diagram can be
created. The social network depicts each agent in the network and the communica-
tions between them. The communications are represented by directional arrows, and
the frequency of communications is also presented, as shown in Figure 5.4. The
thicker the line, the more communications have occurred, and vice versa.
Table 5.4 shows the metrics for the social network in the RTPD scenario. This
highlights the OOW and Operations Officer (OpsO) as the highest on sociometric
status and betweenness metrics, whereas the SMCS operator is highest on Centrality.
The Captain, who has the most executive power in the ship, is modestly placed on
the sociometric status metric for the RTPD scenario. The Captain’ s betweenness
score hints at his executive power. The higher the value means the higher each task
is scored on that metric.
Network density is equal to the total number of links between the agents in the
network divided by the total number of possible links. Low network density figures
are indicative of a broadly spread network with few links. High density figures indi-
cate a tight network that has many links. In this case the network density was 0.4

TABLE 5.3
Association Matrix for the RTPD Scenario

Sound Ships
From/To Captain OOW OpsO Control SMCS WEO Control Warner WECDIS
Captain 2 1 1 1 2 1 1 1
OOW 9 9 4 4 4 21 4 4
OpsO 3 3 38 5 3 3 8 3
Sound 0 0 28 0 0 0 0 0
Control
SMCS 0 0 0 0 0 0 0 0
WEO 2 0 0 0 0 0 0 0
Ships 0 31 0 0 0 0 0 0
Control
Warner 0 0 4 0 0 0 0 0
WECDIS 0 1 0 0 0 0 0 0
EAST in a Submarine Control Room 93

FIGURE 5.4 Social network diagram for the RTPD scenario.

(i.e. a medium distributed network – which is likely to have good resilience against
failure of the network) and cohesion was 0.2 (i.e. a low level of reciprocal links) with
a diameter of 2 (i.e. two hops from one side of the network to the other) and 29 edges
(i.e. when an in-link to an agent had a corresponding out-link to another agent) for a
network of nine nodes. The network can be described as weighted (i.e. non-uniform)
and asymmetric (i.e. unbalanced). As a baseline study, this characterised the network
for returning the submarine to periscope depth.

INFORMATION NETWORK ANALYSIS

The information networks were created by analysing the communications tran-
scripts. Each ‘ concept’ in the transcript was identified and paired with its nearest
relation (i.e. concepts from the same sentence) to build up a network of information
concepts in the control room related to the activities of returning the submarine to
periscope depth. The networks were then presented to the subject matter experts
(SMEs) for verification. The information may be related temporally or spatially to
agent or tasks or both. In this way, it is possible to build up a picture of who knows
what when and how this relates to task performance. This notion takes the activities
of the sound room and control room into the realm of distributed cognition. In this
94

TABLE 5.4
Social Network Metrics for the Network for the RTPD Scenario

Social Network Analysis Captain OOW OpsO Sound SMCS WEO Ship Cont Warner WECDIS
Reception 14 37 37 43 10 9 25 13 8
Emission 10 54 66 28 0 2 31 4 1
Eccentricity 1 1 1 2 0 2 2 2 2
Sociometric Status 3 11 12 9 1 1 7 2 1
Centrality(B-L) 5.2 5.5 5.5 3.8 7.6 3.8 3.8 3.8 3.8
Closeness 1 1 1 0.5 0 0.5 0.5 0.5 0.5
Farness 8 8 8 15 0 15 15 15 15
Betweenness 7 14 14 0 0 0 0 0 0
Systems Thinking in Practice
EAST in a Submarine Control Room 95

view, cognition is not an individual phenomenon but rather a systemic endeavour

performed by the whole team working together. It is argued that cognition transcends
the boundaries of individual actors and becomes a function that is achieved by coor-
dination between the human and technological agents working within the collabora-
tive system. Systems theory argues that the cognitive system needs to be analysed
as a whole rather than on the basis of its constituent parts. To this end, the entire
transcript is presented in the information network model, as the unit of analysis is
not the individual person but the entire system under investigation. The information
network for returning the submarine to periscope depth is presented in Figure 5.5
(note that the node ‘ cuts’ at the bottom right of Figure 5.5 is duplicated merely as a
convenience to disentangle the lines). By viewing the system as a whole, it does not
matter if humans or technology own this information, just that the right information
is activated and passed to the right agent at the right time. Individual human agents
are not required to know everything, provided that the system has the information in
some form or other.
As Figure 5.5 shows, the information network has 68 nodes. Central nodes in the
network (informally defined on the basis of the centrality in the network) appear to
be ‘ cuts’ , ‘ tracks’ , ‘ contacts’ , ‘ classification’ , ‘ reports’ , ‘ manoeuvre’ , ‘ depth’ and
‘ trim’ . The informational nodes coming off these more central key concept nodes
provide the detail that would pertain to the situational specifics, such as the ‘ bearing’
and ‘ range’ for the ‘ fire control solution’ or the type and nature of the vessel clas-
sification. The purpose of the information network was to identify the type of infor-
mation required to return the submarine to periscope depth. A simplified version of
the network was formed (abstracted from Figure 5.5) for the practical convenience of
presenting the SNA metrics in this chapter, as is shown in Figure 5.6.
The metrics in Table 5.5 suggest three classes of information in the network.
‘ Contacts’ has the highest sociometric status and is probably the most important
piece of information in the control room. Next are two clusters of information. One
refers to information about the activities of the submarine (i.e. depth, manoeuvre
and course) and one refers to the activities of other vessels (i.e. tracks, picture and
reports). The network statistics are useful in highlighting the important underlying
features of the relationships between information elements from the control room.
The higher the value, the higher each task is scored on that metric.
The network density was 0.18 (i.e. a low distribution network) and cohesion was
0.17 (i.e. a very low level of reciprocal links – suggesting very strict dependences in
the information network, as shown in the task network diagram) with a diameter of 7
(i.e. seven hops from one side of the network to the other). The network can also be
described as weighted (i.e. non-uniform) and asymmetric (i.e. unbalanced).

COMBINING NETWORK MODELS

As stated in the introduction to the chapter, the purpose of EAST is not only to con-
sider the task, social and information networks in isolation of each other but also
to consider the relationships between the three networks. These relationships are
explored in the following sections. As proposed earlier, the method presents a dis-
tributed cognition perspective within a multi-person– machine system to show how
96

FIGURE 5.5 Information network for returning the submarine to periscope depth.
Systems Thinking in Practice
EAST in a Submarine Control Room 97

FIGURE 5.6 Simplified information network.

information is used and parsed between people performing tasks. First, the combined
task and social networks are presented, wherein the people are mapped onto the task
network. Next, the information and social networks are presented, wherein the people
are mapped onto the information network. Finally, the three networks are integrated,
wherein the information network is coded by the people and tasks.

Task and Social Networks

The relationships between the task and social networks show which agent is primar-
ily involved in each aspect of the task network. These relationships are presented in
Figure 5.7, in which the task network in Figure 5.3 has been colour coded by social
network to indicate which role is primarily responsible for each task. As Figure 5.7
shows, the role specialists are involved in different tasks throughout the RTPD. The
OOW calls the outstation briefing. The Ships Control Officer (SCO) and Planesman
clear the stern arcs and ballast the submarine. The SRC, Sound Room Operators
(SRO), Tactical Picture Supervisor (also called the Ops Officer, or OpsO) and SMCS
are responsible for ranging contacts and reporting to the OOW. The OOW reports to
the Captain that the submarine is ready to RTPD. The OOW receives final reports
from the outstations. All are involved in the silent or standard routine. The OOW,
SCO and Planesman are responsible for getting the submarine to periscope depth,
whilst the SRC and SRO are listening for close contacts (which if detected would
lead the OOW to order the submarine back to a safe depth). The Warner is respon-
sible for giving clearances at periscope depth. Then the OOW and Periscope Watch
Keeper (PWK) are responsible for establishing the look at periscope depth. The first
98

TABLE 5.5
Analysis of Simplified Information Network

Information
Network Analysis Periscope Trim Depth Manoeuvre Course Steer Report Contacts Classification Merchant Picture Tracks Cuts
Reception 1 1 3 2 3 1 3 3 3 1 3 3 1
Emission 1 1 3 2 3 1 3 4 2 1 3 3 1
Eccentricity 7 7 6 5 4 5 4 5 6 7 5 6 7
Sociometric Status 0.2 0.2 0.5 0.3 0.5 0.2 0.5 0.6 0.4 0.2 0.5 0.5 0.2
Centrality (B-L) 4.9 4.9 6.2 7.5 8.8 6.4 9.1 8.2 6.7 5.2 8.0 6.8 5.3
Closeness 0.2 0.2 0.3 0.3 0.4 0.3 0.4 0.4 0.3 0.2 0.4 0.3 0.2
Farness 52 52 41 34 29 40 28 30 39 50 32 37 48
Betweenness 0 0 42 54 78 0 72 42 23 0 22 23 0
Systems Thinking in Practice
EAST in a Submarine Control Room 99

FIGURE 5.7 Task network coded by social agents.

sweep is to search for close in contacts and the second sweep for contacts at greater
distance. Any contacts on a potential collision course would lead the OOW to order
the submarine back to a safe depth. If the submarine is safe at periscope depth, then
the mission intentions can be carried out, in which all of the specialist roles will be
performing their respective tasks.

Information and Social Networks

The dynamic nature of information communication means it is subject to change
moment by moment in light of changes in the task, environment and interactions
100 Systems Thinking in Practice

(both social and technological). These changes need to be tracked if the phenomena
are to be properly understood. As Figure 5.8 shows, the sound room and control
room really do function as a distributed cognition system, with ownership of infor-
mation flowing around the network. No one individual owns all of the information;
rather, the system works as a functional unit.
The colour coding by the social agents reveals clusters of information that fea-
ture around the main information nodes identified in the Section 5 on ‘ information
networks’ (i.e. cuts, tracks, picture, merchant, classification, contacts, reports, steer,
course, manoeuvre, depth, trim and periscope). These are presented in Table 5.6.
The relationships between the key concepts and the roles in Table 5.6 show the
information exchanges that occur in the course of the submarine returning to peri-
scope depth. These information exchanges are called transactions , as the informa-
tion flows both ways. In a simple exchange, this may be a request, followed by an
information transfer and ending with a confirmation (as revealed in the communica-
tion analyses). The transaction informs both the person requested and the requestor.

Information and Task Networks

The grouping of information networks by the tasks shows how different tasks use,
distribute and share information, as shown in Figure 5.9. The groupings show the links
between information and tasks, such that ‘cuts’ are linked to ‘tracks’, ‘tracks’ are
linked to ‘contacts’, ‘contacts’ are linked to the ‘picture’ and ‘reports’, which are
linked to the ‘course’ (via ‘all positions’ reports), the ‘course’ is linked to ‘manoeuvre’,
which is linked to ‘speed’ and ‘depth’, and ‘depth’ is linked to ‘trim’ and ‘periscope’.

Combined Task, Social and Information Networks

The final combined picture puts the task, social and information networks into one
network, as presented in Figure 5.10. The dependences between tasks cannot be rep-
resented in this view (the reader is referred back to Figures 5.3 and 5.7 for those
views). The integrated network model reveals the multi-modal nature of the work in
the sound and control rooms in returning the submarine to periscope depth, that is,
how people utilise information to conduct tasks.
As Figure 5.10 shows, no one individual possessed all of the information to per-
form the RTPD task. The tasks were distributed among the crew and artefacts they
use. To understand how a submarine is returned to periscope depth, one needs to
study the interactions within the sociotechnical system. In this study of the sound
and control room, at least ten personnel had main roles in performing different
tasks.

SUMMARY AND CONCLUSIONS

In summary, this analysis has reported on the application of the EAST to the analy-
sis of the RTPD task in the control room of the Trafalgar class submarine. EAST
acknowledges that systems are inherently complex, and multiple perspectives on the
problem are required to more fully appreciate the relationships between the social
EAST in a Submarine Control Room

FIGURE 5.8 Information network coded by social agents.

101
102 Systems Thinking in Practice

TABLE 5.6
Analysis of Concepts by Social Agents

Key Concept XO OOW SCO OPSO SMCS SRC SRO PWK

Cuts
Tracks
Picture
Merchant
Classification
Contacts
Reports
Steer
Course
Manoeuvre
Depth
Trim
Periscope

and technical aspects of the system. EAST accepts that systems are intertwined and
analyses the system as a whole rather than on the basis of its constituent parts. The
research has shown that the network models are able to characterise the domain
in different, but complementary, ways. The seven outputs, namely the individual
task networks, social and information networks (and associated metrics), and the
combined networks (i.e. task and social network, information and social network,
information and task network, and task, social and information network) offer a
graphical representation of distributed cognition (e.g. the sound and control room)
from different perspectives. The different perspectives offered by the representations
are an attempt to characterise the activities between the sound and control rooms in
returning the submarine to periscope depth.
This chapter extends the work of Walker et al. (2010a) by applying metrics to all
three task, social and information networks to provide insights into the structural
integrity and the relative contribution of each of the nodes. It has also been shown
that it is possible to construct the networks directly from the observational data.
Walker et al. (2010a) demonstrated the benefits of the network representations over
the traditional ethnographic narratives and pictures (Hutchins 1995), which has been
reinforced in the current chapter. It has been shown that the network models offer a
useful way of considering distributed cognition in systems to reveal the interdepen-
dences between tasks, agents and information. There are some similarities between
the approach taken by EAST and that of Actor Network Theory (ANT) (Engestrom
2000), but EAST represents the networks separately as well as together. Both EAST
and ANT have, at their core, a conceptual triangulation between objects, actors and
events. Both use networks of relationships to graphically display their analysis. Also,
EAST goes further than ANT to apply statistical analysis to the networks as well
as identify network archetypes. Nevertheless, there are important similarities in the
EAST in a Submarine Control Room

FIGURE 5.9 Combined information and task networks.

103
104

FIGURE 5.10 Integrated networks model.

Systems Thinking in Practice
EAST in a Submarine Control Room 105

approach and representation. Arguably, ANT relies more heavily upon the skill of the
analyst to identify themes in the networks, whereas the use of verbal protocol data and
SNA metrics by EAST has reduced this subjectivity. The use of social network sta-
tistics may be one way of examining the potential resilience of networks (Hollnagel
et al. 2006) – which may be particularly useful as metrics of distributed cognition. This
is a new concept and methodology, so further studies are required to test its efficacy.
The representations in this chapter revealed the clustering of tasks, social agents
and information. These clusters show how the constituent parts of the networks have
been bound together, either by chance or design. As the system has developed over
a century of submarine operations, it reflects a high state of evolution, but that does
not mean that the system cannot be improved upon. The representations afford both
quantitative and qualitative structural analysis. The quantitative analyses have been
presented at some length and offer insights into the potential integrity and resilience
of the system. As a method for Resilience Engineering, EAST can be used to assess
the potential weaknesses and points of failure in socio-technical structures. The
qualitative analyses enable the network structures to be classified into archetypes
(e.g. chain, circle, tree, star, mesh, small world and fully connected) (Stanton et al.
2008; Stanton et al. 2012). In these terms, the task network appears to be a hybrid of
a chain and circle archetype, the social network appears to be a hybrid of a star and
circle archetype and the information network appears to be a small world archetype.
This analysis is somewhat speculative at this point, and further research is needed to
understand the relationship between the metrics for network resilience and the arche-
types of network structure. There are some early indications that the small world
networks offer the greatest resilience and efficiency (Stanton et al. 2012).
EAST described the control room in terms of task, social and information net-
works as well as exploring the relationships between those networks. The individual
networks were used to describe the respective relationships between the tasks (such
as the task dependencies and sequences), between social agents (such as sociometric
status of agents based on communications) and information (such as the interdepen-
dences between the concepts discussed). The combined task and social networks
showed which roles were performing the tasks in series and parallel. The combined
information and social networks showed which roles were communicating the infor-
mation concepts. The three integrated networks described how information was used
and communicated by people working together in the pursuit of tasks. Any new
conceptualisation of the command system will need to consider the likely changes
on these sociotechnical networked structures.
EAST was able to characterise the activities in the control room of the Trafalgar
class submarine. Thus, it is capable of being applied to a complex sociotechnical
‘ system of systems’ to present ‘ networks of networks’ . EAST offered complemen-
tary descriptions of the requirements for the RTPD task using the current command
system. The networks show multiple perspectives on the activities in the system,
which is a necessary requirement for sociotechnical analysis. Further analysis
should attempt to characterise future command systems so that the multiple per-
spectives can be compared and ‘ so-what’ questions can be asked as ideas for design
of the social and technical aspects of the system co-evolve (Clegg 2000; Walker
et al. 2010b; Stanton et al. 2012). EAST has the potential to map the task, social
106 Systems Thinking in Practice

and information networks and their interdependencies. Adopting this sociotechnical

systems design approach would help to jointly optimise the whole system rather than
the parts in isolation. To do this requires spending more time in initial modelling
and prototyping, working with end-users and SMEs, with more focus on the social
systems and ways of working (to redress the balance of focus on technical system
development) than is currently the case. In the study reported in this chapter, the
benefit of the EAST method is that it helped users understand what was going on
between the sound and control room and how people share information related to the
tasks they were performing. The ultimate goal of the work will be to model alterna-
tives and provide metrics for choosing one alternative over another.

ACKNOWLEDGEMENTS
The author would like to thank Kevin Bessell of BAE Systems for his assistance with
recording and transcribing the verbal communication data.
This work from the Human Factors Integration Defence Technology Centre (HFI
DTC) was part-funded by the Human Capability Domain of the UK Ministry of
Defence.

REFERENCES
Clegg, C. W. (2000). Sociotechnical principle for system design. Applied Ergonomics , 31,
463– 477.
Driskell, J. E. and Mullen, B. (2005). Social network analysis. In: N. A. Stanton, A. Hedge,
K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors and
Ergonomics Methods . Taylor & Francis: London, UK.
Engestrom, Y. (2000). Activity theory as a framework for analyzing and redesigning work.
Ergonomics , 43 (7), 960– 974.
Farrington-Darby, T., Wilson, J. R., Norris, B. J. and Clarke, T. (2006). A naturalistic study
of railway controllers. Ergonomics , 49 (12– 13), 1370– 1394.
Furniss, D. and Blandford, A. (2006). Understanding emergency medical dispatch in terms of
distributed cognition: A case study. Ergonomics , 49 (12– 13), 1174– 1203.
Griffin, T. G. C., Young, M. S. and Stanton, N. A. (2010). Investigating accident causation
through information network modelling. Ergonomics , 53 (2), 198– 210.
Hollnagel, E., Woods, D. D. and Leveson, N. (2006). Resilience Engineering: Concepts and
Precepts . Ashgate: Aldershot, UK.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics 49 (12– 13), 1204– 1225.
Hutchins, E. (1995). Cognition in the Wild . MIT Press: Cambridge, MA.
Jenkins, D. P., Salmon, P. M., Stanton, N. A., Walker, G. H. and Rafferty, L. A. (2011). What
could they have been thinking? How sociotechnical system design influences cogni-
tion: A case study of the Stockwell shooting. Ergonomics , 54 (2), 103– 119.
McIlroy, R. C. and Stanton, N. A. (2011). Specifying the requirements for requirements speci-
fication: The case for work domain and worker competencies analyses. Theoretical
Issues in Ergonomics Science , 13 (4), 450– 471.
Patrick, J., James, N. and Ahmed, A. (2006). Human processes of control: Tracing the goals
and strategies of control room teams. Ergonomics , 49 (12– 13), 1395– 1414.
EAST in a Submarine Control Room 107

Plant, K. L. and Stanton, N. A. (2012). Why did the pilots shut down the wrong engine?
Explaining errors in context using Schema Theory and the Perceptual Cycle Model.
Safety Science , 50 (2), 300– 315.
Salmon, P. M., Stanton, N. A., Walker, G. H., Jenkins, D. P., Baber, C. and McMaster, R.
(2008). Representing situation awareness in collaborative systems: A case study in the
energy distribution domain. Ergonomics , 51 (3), 367– 384.
Sanderson, P., Verhage, A. G. and Flud, R. B. (1989). State-space and verbal protocol methods
for studying the human operator in process control. Ergonomics , 32 (11), 1343– 1372.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork . Ashgate: Aldershot, UK.
Stanton, N. A., Salmon, P. M., Walker, G. H., Baber, C. and Jenkins, D. P. (2005). Human
Factors Methods: A Practical Guide for Engineering and Design . Ashgate: Aldershot,
UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
P. M. et al. (2006). Distributed situation awareness in dynamic systems: Theoretical
development and application of an ergonomics methodology. Ergonomics 49 (12– 13),
1288– 1311.
Stanton, N. A., Walker, G. H. and Sorensen, L. J. (2012). It’ s a small world after all:
Contrasting hierarchical and edge networks in a simulated intelligence analysis task.
Ergonomics , 55 (3), 265– 281.
Stewart, R., Stanton, N. A., Harris, D., Baber, C., Salmon, P. M., Mock, M., Tatlock, K. and
Wells, L. (2008). Distributed situational awareness in an airborne warning and control
aircraft: Application of a novel ergonomics methodology. Cognition, Technology and
Work , 10 (3), 221– 229.
Vicente, K. J. (1999). Cognitive Work Analysis: Toward Safe, Productive, and Healthy
Computer-Based Work . Lawrence Erlbaum Associates: Mahwah, NJ.
Von Bertalanffy, L. (1950). The theory of open systems in physics and biology. Science , 111,
23– 29.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M. and Green, D. (2006).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics , 49 (12– 13), 1345– 1369.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. M. and Jenkins, D. P.
(2010a). From ethnography to the EAST method: A tractable approach for representing
distributed cognition in air traffic control. Ergonomics , 53 (2) 184– 197.
Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D. P. and Rafferty, L. A. (2010b).
Translating concepts of complexity to the field of ergonomics. Ergonomics , 53 (10),
1175– 1186.
Waterson, P. (2009). A systems ergonomics analysis of the Maidstone and Tunbridge Wells
infection outbreaks. Ergonomics , 52 (10), 1196– 1205.
Wilson, J. R. (2012). Fundamentals of systems ergonomics. Work , 41, 3861– 3868.
6 EAST in Railway
Maintenance
With Huw Gibson and Chris Baber

In order to extract data from complex and diverse scenarios, a methodology called
Event Analysis for Systemic Teamwork (EAST) was developed. With over 90 exist-
ing ergonomics methodologies already available, the approach taken was to inte-
grate the following: a Hierarchical Task Analysis (Annett et al. 1971), a Coordination
Demand Analysis (CDA_ (Burke 2005), a Communications Usage Diagram (CUD)
(Watts and Monk 2000), a Social Network Analysis (Scott 1991; Driskall and
Mullen 2005) and the Critical Decision Method (Klein et al. 1989). The outputs of
these methods provide two summary representations in the form of an enhanced
Operation Sequence Diagram and Propositional Network. These offer multiple and
overlapping perspectives on key descriptive constructs, including who the agents are
in a scenario, when tasks occur, where agents are located, how agents collaborate
and communicate and what information is used and knowledge shared. The appli-
cation of these methods to live data drawn from the UK rail industry demonstrates
how alternate scenarios can be compared on key metrics, how multiple perspectives
on the same data can be taken and what further detailed insights can be extracted.
Applied across a number of scenarios in different civil and military domains, the
ultimate aim of EAST is to provide data to develop generic models of C4i (com-
mand, control, communications, computers and intelligence) activity, and to improve
the design of systems aimed at enhancing this management infrastructure.

INTRODUCTION
C4i is the management infrastructure for any large, complex and dynamic resource
system (Harris and White 1987). An underlying mechanism for organisation and
planning is needed where there are multiple individuals, or teams of individuals,
who need to communicate and coordinate in order to achieve an efficient outcome
to a common goal. Although the notion of command and control is one that might
popularly be thought of as occurring in the military domain, examples of underlying
organisational and planning infrastructures also abound in the civil domain.
C4i embodies people, teams, technology, knowledge and information. It is a
multi-faceted endeavour that can be viewed from a variety of different and concur-
rent perspectives. This is can be challenging not just for those engaged in sustaining
effective performance within a C4i scenario, but also for those engaged in analys-
ing such a scenario. The multiple and concurrent perspectives include the micro-
level descriptions of tasks, actors, and individual cognition, through the meso-level

109
110 Systems Thinking in Practice

descriptions of team performance, through again to macro-level systems analysis.

Add to this the notion that C4i also involves human and non-human elements and
is comprised of activities that are highly related temporally, and it is clear that C4i
research presents a considerable challenge. Smalley (2003) provides a contempo-
rary perspective on these issues and an example of current research in this domain.
Smalley’s (2003) high-level analytical review of command and control systems also
elucidates some of the different challenges inherent in C4i, which Harris and White
(1987) summarise as being simply that

1. there is currently no theory of command and control;

2. there is informational chaos;
3. there is an ongoing technology revolution (Harris and White 1987).

The aim of this chapter is to present an analysis methodology that feeds into the
development of generic models and theories of C4i. In order to do this, the method
seeks to make sense of the informational chaos identified above by the structured and
systematic extraction of data from C4i scenarios. The ultimate aim of this extrac-
tion and modelling process is for the Defence Technology Centre for Human Factors
Integration (DTC HFI) to generate the level of understanding required to deploy
novel forms of technology in ways that contribute optimally towards the effective-
ness of achieving the goals of C4i scenarios. So, in summary, the context of the
proposed descriptive methodology generates the following specific requirements to:

• generate the data necessary to understand key features of C4 scenarios

• use these data to develop theories and generic models
• use this understanding to design C4i systems that accurately meet user
needs.

This chapter describes the development and application of the EAST methodol-
ogy, which was designed to meet these requirements. EAST was developed specifi-
cally to analyse C4i in any domain (DTC HFI 2003).

DEVELOPMENT OF THE EAST METHODOLOGY

At the most basic level, the descriptive constructs in C4i can be distilled down to:

• who: (the agents participating in a C4i scenario are)

• when: (tasks are performed and which agents are associated with their
performance)
• where: (agents are physically located)
• how: (agents collaborate and communicate to achieve scenario aims)
• what: (tasks are agents performing and what knowledge is used and/or
shared).

There are over 90 different methods available to the ergonomist covering virtu-
ally all aspects of system design and description, including the constructs above
EAST in Railway Maintenance 111

(Stanton et al. 2005). It would seem evident based on this that a high degree of
circumspection is required before embarking on the development of yet more. This
is especially so given the multi-faceted, and therefore multi-measurement, approach
that would be required to analyse C4i scenarios. In other words, the task of develop-
ing and then subsequently validating the range of methods needed to access all the
required perspectives on C4i is formidable. Instead, an approach based on method
integration is proposed. This has a number of compelling advantages, because not
only does the integration of existing methods bring reassurance in terms of a vali-
dation history, but it also enables the same data to be analysed from multiple per-
spectives. These multiple perspectives, as well as being inherent in the object that
is being described and measured (C4i scenarios), also provides a form of inter-
nal validity. Assuming that the separate methods integrate on a theoretical level,
then their application to the same data set offers a form of ‘analysis triangulation’.
Therefore, the internal structure of the EAST methodology can be broken down into
three layers comprised of associated methods already established in the literature:

• Layer 1 – data collection methods

• Layer 2 – data analysis methods
• Layer 3 – representational methods

It is important to consider the theoretical issues surrounding the use of methods

applicable to these three levels. This consideration is important not only from the
point of view of validity in terms of the individual method but also in terms of com-
patibility ‘between’ methods.

Layer 1 – Data Collection Methods

Structured data collection methodologies are required to extract meaningful data
from C4i scenarios. Annett (2003) argues that data collection should comprise of
observation and interviews at the very least, and both Annett (2003) and Kieras
(2003) argue for the least intrusive method of observation that circumstances permit.
Therefore, a two-step process of observation and interview is proposed as a means
to collect key information from within live C4i scenarios, where several people are
likely to be working remotely from each other.
Observation itself takes the form of multi-site activity sampling (Stanton et al.
2005). Activity sampling is an aid to unobtrusively recording activity (or actual
behaviour) in the field. This process involves the use of pre-defined categories or
classes of action, which can then be used to create a checklist. The checklist could
be completed at regular time intervals, for example, every 15 seconds (by ticking
the appropriate action in the appropriate column for that observation), or it could
be completed whenever an action occurs (by entering the time against the action).
Activity sampling provides a means of relating the timing of actions to locations
and to the decisions made by specific actors in context. It can be assumed that in a
number of cases there will also be pre-existing material in the form of task analyses
(or similar) that will enable task descriptions to be structured, sampled and even
validated through observation.
112 Systems Thinking in Practice

In basic terms, whilst observational techniques provide information on the

observable inputs and outputs of human information processing, they produce lim-
ited data on the process of decision making. Interviewing people enables the analyst
to capture data on these unobservable processes, particularly if the interviewee is
describing a recent event and how they dealt with it. In recent years, the study of
decision making in real-world situations has received a great deal of attention, and
there is a growing emphasis on the use of interviews to collect such information. The
Critical Decision Method (CDM) (Klein and Armstrong 2005) is a contemporary
example. According to Klein, ‘The CDM is a retrospective interview strategy that
applies a set of cognitive probes to actual non-routine incidents that required expert
judgment or decision making’ (Klein et al. 1989, p. 464). In this approach, the inter-
view proceeds through a series of four stages: briefing and initial recall of incidents,
identifying decision points in a specific incident, probing the decision points and
checking. A slightly modified version of the CDM probes presented in O’Hare et al.
(2000) are adopted within EAST and presented in Table 6.1. These permit elicitation
of information on key decision points as well as non-routine ‘incidents’.
The selection of data collection methods proceeds not just from the theoretical
perspective but also the practical. Given that most C4i scenarios will involve the
simultaneous observation of activities distributed geographically and among indi-
viduals, the proposed two-step process of activity sampling and CDM interview pro-
vides an expedient method of data capture in this context.

Layer 2 – Analysis Methods

Structured analysis methodologies take the data extracted earlier in the data collec-
tion phase and model them in terms of deeper, more fundamental concepts. These
concepts relate to task and social structures, teamworking and mediating communi-
cations technology.

Hierarchical Task Analysis (HTA)

Task analysis is the activity of collecting, analysing and interpreting data on sys-
tem performance (Annett and Stanton 2000; Diaper and Stanton 2004) and is one
of the central underpinning analysis methods within EAST. According to Stanton
(2004), task techniques can be broadly divided into five basic types: hierarchi-
cal lists (e.g. HTA and GOMS), narrative descriptions (e.g. the Crit and Cognitive
Archaeology), flow diagrams (e.g. TAFEI [Task Analysis for Error Identification]
and Trigger Analysis), hierarchical diagrams (e.g. HTA) and tables (e.g. Task-
Centred Walkthrough, HTA, SGT [Sub Goal Template] and TAFEI). Some methods
have multiple representations, such as HTA, which can be viewed as a hierarchical
text list, a hierarchical diagram or in tabular format.
Hierarchical Task Analysis (HTA) is a means of describing a system in terms of
goals and sub-goals, with feedback loops in a nested hierarchy (Annett 2005). Its
enduring popularity and indeed its appropriateness within EAST can be put down to
two key points. First, it is inherently flexible: the approach can be used to describe any
system, even C4i. Second, it can be used for many ends, from person specification,
to training requirements, to error prediction, to team performance assessment and to
EAST in Railway Maintenance 113

TABLE 6.1
CDM Probes

Cognitive Cue Sample Question

Goal specification What were your specific goals at the various decision points?
Goal identification What features were you looking at when you formulated your decision?
How did you know that you needed to make the decision?
How did you know when to make the decision?
Expectancy Were you expecting to make this type of decision during the course of the
event?
Can you describe how this affected your decision-making process?
Conceptual model Are there situations in which your decision would have turned out differently?
Can you describe the nature of these situations and the characteristics that would
have changed the outcome of your decision?
Influence of At any stage, were you uncertain about either the reliability or the relevance of
uncertainty information that you had available?
At any stage, were you uncertain about the appropriateness of the decision?
Information What was the most important piece of information that you used to formulate
integration the decision?
Situation awareness What information did you have available to you when formulating the decision?
Situation Did you use all the information available to you when formulating the decision?
assessment Was there any additional information that you might have used to assist in
formulating the decision?
Options Were there any other alternatives available to you other than the decision that
you made?
Why were these alternatives considered inappropriate?
Decision blocking Was there any stage during the decision-making process in which you found it
difficult to process and integrate the information available?
Can you describe precisely the nature of the situation?
Basis of choice Do you think that you could develop a rule, based on your experience, that
could assist another person to make the same decision successfully?
Do you think that anyone else would be able to use this rule successfully?
Why?/Why not?
Generalisation Were you at any time reminded of previous experiences in which a similar
decision was made?
Were you at any time reminded of previous experiences in which a different
decision was made?

Source: O’Hare et al., Task Analysis, Taylor and Francis, London, UK, 2000, 170–190.

system design. The multiple perspectives available from HTA fit well with the multi-
ple perspectives available from C4i scenarios. HTA has additional pragmatic benefits
because it already underpins a number of subsequent analysis methodologies such as
CDA (Burke 2005) and CUD (Watts and Monk 2000), making it an ideal candidate
for method integration. Task analysts applying HTA are also required to understand
both the ways in which people adapt to their environment and the ways that they adapt
their environment to themselves. Thus, HTA has the further benefit of capturing and
114 Systems Thinking in Practice

specifying the contextual conditions and precursors within C4i scenarios. These are
represented within detailed task descriptions and as ‘plans’ within the hierarchy.
In its application within EAST, the definition of the HTA proceeds with ref-
erence both to what has been observed and to what may have been previously
defined through any pre-existing task analyses. It is possible to meaningfully
integrate these information sources on goal structure to produce a task analysis
that accurately describes what has been observed, is consistent with what has
already been pre-defined and also covers key decision points covered in the CDM
interview.

Coordination Demand Analysis (CDA)

It might be assumed that C4i activity will be dominated by coordination activities,
but this supposition needs to be checked. Individual tasks from the HTA can be
assessed for the type of coordination that is required for successful performance
using a method called CDA (Burke 2005). The method integrates with the HTA,
where the tasks identified are assessed according to multi-dimensional aspects of
teamworking, presented in Table 6.2.
Burke (2005) distinguishes between teamworking tasks (those comprised of the
coordination dimensions in Table 6.2), and task work (individual tasks that are not

TABLE 6.2
Coordination Demand Dimensions

Coordination Dimension Definition

Communication Includes sending, receiving and acknowledging information among crew
members
Situational Awareness (SA) Refers to identifying the source and nature of problems, maintaining an
accurate perception of the aircraft’s location relative to the external
environment and detecting situations that require action
Decision Making (DM) Includes identifying possible solutions to problems, evaluating the
consequences of each alternative, selecting the best alternative and
gathering information needed prior to arriving at a decision.
Mission analysis (MA) Includes monitoring, allocating and coordinating the resources of the
crew and aircraft; prioritising tasks; setting goals and developing plans
to accomplish the goals; creating contingency plans
Leadership Refers to directing activities of others, monitoring and assessing the
performance of crew members, motivating members and
communicating mission requirements.
Adaptability Refers to the ability to alter one’s course of action as necessary, maintain
constructive behaviour under pressure and adapt to internal or external
changes
Assertiveness Refers to the willingness to make decisions, demonstrating initiative and
maintaining one’s position until convinced otherwise by facts
Total coordination Refers to the overall need for interaction and coordination among crew
members
EAST in Railway Maintenance 115

dependent on teamworking). For example, a teamwork task would be dealing with

the issuing of an instruction to another individual, whereas a ‘task work’ task would
be inputting data into the C4i system. It is argued by Burke that existing teamwork-
ing methodologies tackle only a few of the necessary skill aspects; therefore, CDA is
proposed as a comprehensive and systematic descriptor. CDA also integrates readily
within the EAST, where it provides a profile of teamworking skills according to the
tasks identified earlier in the HTA.

Communications Usage Diagram (CUD)

The CUD (Watts and Monk 2000) is another task analysis technique that, in its cur-
rent application, provides an opportunity to systematically critique the communica-
tions technology in use during a scenario. The critique is based upon the task flow.
It identifies actors in specific locations, the communications in use within identified
tasks and an approximate indication of the sequencing of events. The critique of
communications technology currently in use is based upon this. It identifies the posi-
tive and negative points associated with a given communications media within the
current context of use and enables the analyst to propose alternative solutions. The
CUD is useful as a way of drawing out practical, design-related outputs as well as
providing a more theoretical level of description.

Social Network Analysis (SNA)

SNA is a means to present and describe, in a compact and systematic fashion, the net-
work structure underlying individuals or teams who are linked through communica-
tions with each other (Driskell and Mullen 2005). The relationships that are specified
from this analysis can be used to determine what aspects of the network structure
constrain or enhance the performance of agents in the network (Driskell and Mullen
2005). Unlike the other analysis methods within EAST, SNA is not directly based
on the task analysis. However, foundation data on actors and communications are
derived from the HTA and CUD.
SNA can be used to analyse the formal and informal relationships between people
in a network, but there is no reason why they cannot also show technological media-
tion of communication and networks where some of the nodes are non-human. This
is of particular relevance within C4i. In a practical sense, the approach might reveal
sub-optimal networks and bottlenecks in communication. The approach also allows
a complex network to be summarised on just a small number of key metrics and
for these same metrics to permit easy comparison between different C4i scenarios.
Another major advantage that Driskell and Mullen (2005) identify is that this ‘net-
work approach focuses on the relationships among actors embedded in their social
context’ (pp. 58). Again, another important consideration within C4i. Houghton
et al. (2006) deal in more detail with this network perspective.

Layer 3 – Representational Methods

Representational methods take the extracted data modelled in the analysis section
and provide a means of simplifying and presenting these deeper concepts in various
integrated, graphical forms.
116 Systems Thinking in Practice

Operation Sequence Diagram (OSD)

Process charts offer a systematic approach to describing activities. They empha-
sise essential features using a graphical representation that is easy to follow and
understand (Kirwan and Ainsworth 1992). Charting techniques such as this preserve
the ability of preceding methods, such as SNA and HTA, to represent human and
non-human elements of the C4i system (Drury 1990). Charting techniques are also
capable of representing one of the key contextual factors in C4i scenarios, and that is
the temporal structure and interrelations between and among processes.
Several enhancements to the OSD representation have been made to reflect the
outputs of supporting analysis methods. The OSD presents a temporal overview
of tasks, the outputs of the CDA, communications media from the CUD and links
between agents from the SNA. Therefore, the approach summarises a large amount
of supporting analysis in a fashion that is graphical and relatively easy to follow,
whilst also being scalable to suit different ‘sized’ C4i scenarios. This enhanced ver-
sion of OSD is therefore one of the key summary representations within EAST.
Two additional reasons underlie the selection of the OSD. First, OSD has long been
a popular and useful tool in the Human Factors toolkit. Second, many approaches
to systems analysis (such as Unified Modelling Language) make use of sequence
diagrams in their representations, and so it ought to be possible to integrate a Human
Factors OSD with the more technical sequence diagrams used by systems analysts.
In this manner, we propose that the process-based analysis derived from EAST pro-
vides us with a convenient route into Human Factors Integration. This factor is con-
sistent with the DTC HFI’s overall aims.

Propositional Network (PN)

Propositional Networks (PNs) are like semantic networks in that they contain nodes
(with words) and links between nodes. It is argued that the application of basic prop-
ositions and operators enables dictionary-like definitions of concepts to be derived
(Ogden 1987). Stanton et al. (2006) take this basic notion and extend it to offer a
novel way of modelling knowledge in any scenario. Knowledge relates strongly to
SA. A systems view of SA (and indeed an individual view as well) can be understood
as activated knowledge (Bell and Lyon 2000), and therefore propositional networks
offer a novel and effective means of representing this ‘systems level’ view of SA. The
theoretical background to this approach is described in detail in Stanton et al.’s DSA
methodology (2006). However, an opportunity arises within EAST to apply the DSA
methodology in a real sense to C4i scenarios.
The knowledge used in C4i activities is accessed via the CDM, where a sys-
tematic content analysis of the interview transcripts permits ‘knowledge objects’ to
be extracted and subsequently linked. Knowledge objects are analogous to proposi-
tions and can be defined as an ‘entity or phenomenon about which an individual
requires information in order to act effectively’ (Baber 2004). The resultant network
of knowledge objects enables at least four powerful perspectives on SA.

1. Firstly, a major advantage of PAs is that they do not differentiate between

different types of node (e.g. knowledge related to objects, people or ideas),
and therefore, from a design perspective, they do not constrain assessments
EAST in Railway Maintenance 117

to consideration of existing configurations of people and objects but rather

to the required knowledge elements associated with a scenario.
2. Secondly, the network shows the totality of knowledge used in the scenario
at a systems level, regardless of whether agents in the scenario are human
or technical.
3. Thirdly, shared SA can be accessed from the CDM, where agents can be
attributed to knowledge objects within the network.
4. Fourthly, Endsley (1995) states that SA occurs within a ‘volume of time and
space’ (p. 36), and it is possible to illustrate this key temporal aspect of SA
by animating the propositional network in terms of active and non-active
knowledge objects (Figure 6.8) later in this chapter provides an illustration
of the concept). To do this, the scenario is divided into tasks phases accord-
ing to the higher-level goals of the HTA. The CDM relates to these phases
and accesses information on decision making within them, allowing active
and non-active knowledge objects to be specified and represented.

Summary of Methods
Whilst the methods above are tried and tested, the integration of them into EAST
represents a new approach. Each method provides insight into the main descrip-
tive constructs identified above. The HTA identifies the actors in the scenario, the
temporal structure of tasks, where tasks (and associated actors) are taking place and
what tasks are being performed. The HTA is one of the main foundation methods
within EAST and drives the CDA analysis, which provides structured insight into
the general question of how teamworking tasks are performed and what teamwork-
ing skills they require. The CUD is also founded on data from the HTA; it provides
insight into who the actors are, the flow of tasks (i.e. when), where actors are located
geographically, what items are being communicated and how (i.e. the communica-
tions technology used). The SNA extends the analysis of communications by consid-
ering the ‘links’ between actors rather than task flow. These two perspectives on ‘the
who’ are complementary. Table 6.3 is a methods matrix that relates the component
methods to the descriptive constructs identified above. The overlap between methods
and the constructs they access is explained by the multiple perspectives provided on
issues such as ‘who’ and ‘what’. For example, the HTA deals with ‘what’ tasks, the

TABLE 6.3
Methods Matrix Mapping Descriptive C4i
Constructs onto Component Methods of EAST

HTA CDA CUD SNA

Who
When
Where
What
How
118 Systems Thinking in Practice

CDA deals with ‘what’ teamworking skills, and the CUD deals with ‘what’ com-
munications technology is used. Each is a different but complementary perspective
on the same descriptive construct and a different but complementary perspective on
the same data derived from observation and interview, that is, analysis triangulation.

Structure of EAST Methodology

The internal structure of the methodology is illustrated in Figure 6.1. The live observa-
tion provides input into all of the Layer 2 analysis methods. The HTA serves to provide
a definition of tasks and a goal structure for the remaining analysis methods. The
analysis methods are then summarised and represented on an enhanced form of OSD
(the SNA can also be used as a representational method, e.g. Houghton et al. 2006).
The CDM interview data collected live are represented using a propositional network,

FIGURE 6.1 Internal structure of EAST methodology.

EAST in Railway Maintenance 119

thus representing the structure of knowledge at key decision points in the scenario.
The outputs of EAST provide an integrated picture of multiple perspectives; that is, the
component methods are compatible with each other to the extent that they can be sum-
marised within a common representation (e.g. the OSD), but the individual methods,
and combinations of them, are also capable of generating insights at the core of C4i.

PROCEDURE
The high-level procedure for applying EAST can be summarised in Table 6.4 below
as involving nine key steps. Obviously, within these overall headings, a degree of
flexibility can be adopted to suit the particular domain or circumstances.

APPLICATION OF THE METHOD TO LIVE DATA

Three examples of complex and dynamic resource systems are taken from the UK
rail industry. The scenarios serve as a source of live data to demonstrate the capabil-
ity of the method. The data are sourced, in this case, from written transcripts based
on communications between parties in the scenarios and interviews with subject
matter experts.
Broadly speaking, the C4i activities under consideration are those involved in the
setting up of safety systems required when carrying out maintenance of track. Safety
systems are required so that workers on the track do not come into conflict with mov-
ing trains and that trains do not travel over railway infrastructure that is rendered
unsafe by the maintenance work or the requirement for it. The strict procedures
underpinning these systems are specified nationally in the UK railway industry Rule
Book (RSSB 2003).
Railway operations are an example of civilian C4i, where a ‘management infra-
structure’ is required and in place. Maintenance activities on the railway possess all
the essential C4i ingredients, including

• that there is a common goal

• that there are individuals and teams coordinating to reach it
• that those parties are dispersed geographically
• that there are numerous systems, procedures and technology to support
their endeavour

Background
Under normal conditions, a signaller has the key responsibility for controlling train
movements and maintaining safety for an area of railway line. This control occurs
remote from the line at a control centre (a signal box or signalling centre). These can
be located many miles from where activity could be taking place.
During maintenance, another person takes responsibility for an area of the line
(sometimes referred to as a possession) and/or for preventing trains passing over
the possession (measures referred to as protection). These individuals are normally
termed the person in charge of possession (PICOP) or controller of site safety
120

TABLE 6.4
High-Level Procedure for EAST

STAGE Action Timing Output

1 Organise observation Prior to observation Stakeholder involvement;
It is proposed that many of the sessions can also be undertaken in simulators or details of scenario; number of
training centres as an alternative to the field. personnel; key decision roles
2 Conduct HTAs with SMEs Prior to observation Definition of task scenario;
Existing task analysis material would be sourced during this stage and validated by details of task structure;
SMEs. If required, new task analyses would be undertaken with SMEs. activities and actors
3 Define objectives Prior to observation List of agreed objectives; agree
This stage will involve checking that the proposed technique is feasible (e.g. in terms observation posts;
of security, access, privacy etc.). If possible, define ‘ideal’ performance, perhaps in if possible, description of ideal
terms of a list of roles, responsibilities, decisions and timing of events. performance
4 Brief and train observers. Prior to observation Demonstrate sampling strategy;
30–60 minutes. practice
5 Arrive at scenario On-site
On arrival at ‘scenario’, position observers at their posts. Synchronise watches. 5–10 minutes.
6 Observation During observation Multiple activity samples
The incident is sampled using the pro-forma/experimental materials. Duration of incident or n × 20
minute observations.
7 Interview After observation Interview transcript containing
Participants interviewed by observer(s) using CDM methodology. 1 hour per interviewee. knowledge objects
8 Define CDA and CUD outputs with SMEs After observation (can be Validated CDA and CUD data
This involves engaging SMEs to assist in the rating of tasks along teamwork undertaken at any stage after
dimensions and the critique of communications technology in use during the scenario. Step 2 if required)
2 hours.
9 Collation After session Analysis and Representations
Systems Thinking in Practice

The activity samples are collated to produce full EAST analysis. of C4i scenario
EAST in Railway Maintenance 121

FIGURE 6.2 Overall diagram of the various track possession scenarios.

(COSS). Communication and coordination is required to transfer responsibility

between the signaller and PICOP/COSS. The PICOP/COSS also has to communicate
and coordinate with various other personnel, such as those carrying out maintenance
within their areas of control, drivers of trains and on-track plant that may be in the
possession and personnel implementing aspects of the possession (all of which may
also be dispersed over a certain geographical area). The three specific maintenance
scenarios are briefly described below, with Figure 6.2 providing additional clarity on
the general layout and relative geographical positions of personnel.

Scenario 1 – Planned Maintenance Activities

This scenario describes the processes and activities for setting up a possession for a
stretch of track so that planned maintenance can take place. This requires coordina-
tion between multiple parties, including communication between the signaller and
PICOP (so that appropriate ‘protecting’ signals are set to danger) and the provision
of instructions to a ‘competent person’ (CP) to place a form of protection against
122 Systems Thinking in Practice

oncoming trains at the limits of the possession (these take the form of explosive
charges called detonators, which emit a loud noise to alert drivers who may have just
run over them). Additional complexity comes in the form of a number of engineering
work sites within the possession, each of which has an engineering supervisor (ES)
and COSS responsible for setting up and managing each one. The ES will also use
‘competent personnel’ to place marker boards as a form of additional protection at
the ends of the individual worksites.

Scenario 2 – Emergency Engineering Work

When railway personnel are required to carry out unplanned emergency engineering
work on the line, such as when track or infrastructure has been damaged or has sud-
denly degraded, then the passage of trains must be stopped and an emergency pro-
tection procedure called a T2(X) applied. For this procedure, a portion of the railway
that is normally under the control of a signaller working remotely from a signal box
becomes the responsibility of a COSS, who will work on the line. The signaler places
signals at the limits of the dangerous work zone in order to protect track workers
from train movements. Emergency protection can be arranged between the COSS
and signaller following discussion with the Network Rail Area Operations Manager.
Overall, it can be noted in the scenario above that non-emergency engineering work
involves greater advanced planning and protection, whereas in emergency scenarios,
organisation tends to occur ‘on the day’.

Scenario 3 – Ending a Track Possession

When the possession is ended, the ‘set-up’ procedure outlined in Scenario 1 is largely
reversed. First, the ES of a worksite has to check that the worksite can be closed. This
requires agreement between the ES and each COSS within the worksite (there is a
COSS responsible for each piece of work being undertaken within the worksite).
Once this has been checked and the PICOP has been informed, then the ES can
instruct a CP to remove the worksite marker boards. The PICOP is informed when
this is completed and then, when all the worksites within a possession are closed,
the possession itself can be closed. The PICOP can then instruct a CP to remove the
possession protection. The PICOP will inform the signaller that lines are now safe
and clear for trains to run. Control of the line is then passed from the PICOP to the
signaller, and normal running of trains over the lines can resume.

RESULTS AND DISCUSSION

Analysis Methods
Task Networks
The first step in the EAST methodology, subsequent to collecting the data, is to
model the goal structure of the scenario using HTA. The output of this step is repre-
sented in the form of task networks. These are graphical representations of the ‘plan
0’ within the HTA and depict the task structure in terms of how tasks relate to each
other functionally and temporally. In the present case, the highly proceduralised and
rigid nature of the activity is seen as a more or less linear task flow (Figure 6.3).
EAST in Railway Maintenance 123

FIGURE 6.3 Task networks for each scenario.

The following is an extract from the verbal transcripts and shows a typical com-
munication occurring between the signaller (S) and PICOP:

S ‘Wimbledon …’ [answers phone with name of signal box]

S ‘… panel 1’ [signalling panel, and associated geographical area being
worked by the signaller]
PICOP ‘Hello Wimbledon, it’s [name of PICOP] at Waterloo [station]’
S ‘… Yes …’
124 Systems Thinking in Practice

PICOP ‘the blocks [protective measures] now been put out mate on the down
main slow and the up main slow …’ [referring to different lines]
S ‘… right …’
PICOP ‘… and its clear of 15 64 b points and 15 12 b points.’
S ‘It’s all yours, at, er, what we on, 9:53 then.’
PICOP ‘oh, 9:53, cheers mate.’
S ‘… ok …’
PICOP ‘Can I take your name please, I forgot to write it down earlier’
S ‘… [provides name].’
PICOP ‘[name] thanks a lot mate …’
S ‘ok …’
PICOP ‘… bye.’

It is interesting to note the data collection transcripts that highlight that some of
the required and critical steps may be implied (e.g. the PICOP’s closure of communi-
cation is by saying ‘bye’, which is implied by the signaller to mean that there are no
remaining issues or ambiguity and that points 15 64b, and 15 12b are indeed clear)
and that the sequencing of communications is more flexible than the procedures may
suggest (e.g. the PICOP may only enquire in more detail as to the name of the signal-
ler at the stage when it is needed for the completion of documentation, rather than at
the beginning of a call). However, at the level illustrated above, any informality or
flexibility occurs within the confines of a well-defined procedure.

CDA
Based on Figure 6.4, the supposition that C4 is dominated by coordination tasks
appears to be justified. In the three scenarios under analysis, the tasks that fall into

100%

75%
% Total tasks

50%

25%

0%
Emergency Possession Hand back
Team 71.79 67.57 66.04
Task 28.21 32.43 33.96

Team Task

FIGURE 6.4 Results of CDA analysis showing percentage of task/teamwork activities

undertaken within each scenario.
EAST in Railway Maintenance 125

3.00

2.00

1.00

0.00
n
tio

s
es

ing

sis
ica

ip
en

ity
sh
aly
un

s
ar

n
m

bil
er
m

tio
aw

en
ion

ta
ad
m

ina
ion

ap
Co

tiv
ion

Le
cis

rd
Ad

er
iss
at

oo
s
tu

lc
Si

ta
to
n
ea
M

Emergency possession Hand back possession Normal possession

FIGURE 6.5 Results of CDA analysis showing profile of results on each of the coordination
dimensions.

the ‘teamwork’ track, and that require coordination, form between 66% and 72% of
total tasks undertaken.
Figure 6.5 extracts the teamwork tasks for further analysis. The analysis proceeds
according to seven coordination dimensions and one summary total coordination
score (based on the mean of the individual scores). This analysis reveals a broadly
similar pattern of coordination activity within the scenarios; certainly the total coor-
dination figures are comparable, falling within the mid-point of the rating scale. Of
more interest is the pattern of results across the seven individual dimensions, where a
distinctive footprint emerges. Communication, situational awareness (SA) and deci-
sion making are prominent dimensions, and there is also a smaller ‘blip’ for the lead-
ership dimension. Leadership can be taken as a further indication perhaps of some
decision-making activity. It can be further noted that the larger (and more complex)
the scenario, the larger the leadership ‘blip’ is. Therefore, in summary, not only are
the majority of total tasks dominated by coordination activities, but those activities
are also dominated by communications and the creation and maintenance of SA.

CUD
General observations from the CUD analysis are that the communications are entirely
verbal. Given the nature of the scenario, verbal and telephone communications
126 Systems Thinking in Practice

appear to be appropriate at most stages of the interaction. However, possible technol-

ogy options could be helpful in three respects.

1. by removing possible sources of error inherent in verbal communications

2. by removing the cumbersome nature of read-back procedures
3. by alleviating the physical disturbance to other tasks caused by the unsched-
uled and ad-hoc presentation of verbal communications.

Of course, any new approaches would require fuller risk justification and assess-
ment within the wider task context before application. The key point is that the CUD
method provides a systematic way of presenting the existing situation and consid-
ering alternatives to it based on data. Figure 6.6 summarises the communications
technology in use within the scenarios.

SNA
Social Network Analysis is used in the EAST method to represent and summarise
the communication/information links between agents in the scenario. Figure 6.6
presents a graphical representation of the networks derived from each of the sce-
narios and is also annotated with communications information drawn from the CUD
analysis.
A range of mathematical metrics (derived from Graph Theory) can be applied.
The results show that the PICOP, signaller and ESs have the highest levels of socio-
metric status and centrality. These metrics indicate that they are key agents in the
scenarios. The notion of centrality is also borne out when considering ‘betweenness’;
that is, the PICOP and ES fall between pairs of other positions in the network the
most frequently.
Having identified the key agents, it is also possible to view the network as a whole
under the concept of network density. Density is the degree of interconnectivity
between agents or the number of network links used compared to those that are

FIGURE 6.6 Graphical representation of social networks overlain with comms media
drawn from CUD analysis.
EAST in Railway Maintenance 127

TABLE 6.5
Comparison of Network Density between Scenarios

Possession Emergency Possession Hand Back Possession

Density 0.4 0.67 0.33

theoretically available (the maximum being a case where all agents are linked to
each other). In Table 6.5, it can be seen that the emergency possession scenario has
the densest pattern of connectivity, with the remaining two scenarios being broadly
comparable.
Although the metrics allow comparison between networks, intelligent inter-
pretation is required. For example, a network with every agent connected to each
other would permit easy dissemination of information, but it might also be ineffi-
cient. Similarly, having one central node may have advantages for coordination but
offers the potential for an information bottleneck. Houghton et al. (2006) discuss
the implications of several prototypical networks and their comparison to live sce-
nario networks. The main point is that the interpretation and then subsequent com-
parison of networks has to take into account a range of contextual factors. For the
time being, the networks derived from the scenarios above appear to be relatively
well matched to the procedures being undertaken, with a mix of central agents and
interconnectivity.

Representational Methods
Scenario Process Charts (OSD)
Figure 6.7 presents a sample of an enhanced OSD from scenario 1 (the planned
maintenance activities), and highlights how the preceding methods are integrated
with it.
The operations loading is presented in Table 6.6, showing the PICOP, signaller
and CP as the most heavily loaded individuals in the network in terms of tasks. The
operations loading table provides a further level of summarisation in being able to
capture, in a relatively compact manner, the process-based aspects of what is often
a large OSD.

Propositional Networks (PN)

The PN provides an overview, for each scenario, of all the knowledge elements and
their relationships. It also allows the knowledge elements related to a specific phase
in the scenario to be described, as well as the history of previously used knowledge
points. Figure 6.8 displays the network elements related to taking a possession for
emergency maintenance. Shaded cells denote knowledge objects that are currently
active within the task phase. Faded cells indicate previously active knowledge. The
main image in Figure 6.8 is intended to provide information on knowledge objects,
whereas the smaller networks alongside are merely illustrative of changing activation.
128 Systems Thinking in Practice

FIGURE 6.7 Enhanced OSD summary representation.

Despite the emergency procedure scenario being a relatively simple procedur-

alised task, the PN appears complex. It should be recognised that this complexity
does not necessarily reflect complexity in the task itself. For example, some of the
knowledge elements are internalised skills (e.g. participant knowledge of the railway
rule book or a signaller’s knowledge of the current status of the railway) or simple
objects (e.g. telephones) that would not provide load or add to the complexity per-
ceived by a skilled participant. It does, however, demonstrate that a large number of
related objects are used to accomplish relatively simple tasks. In the context of task
redesign, the propositional network provides a prompt that can allow system design-
ers to question the necessity for knowledge objects, the form those objects can take
and how they are communicated. This is a novel perspective. The network could
also be considered as a tool for designing training, which describes the knowledge
elements and relationships that someone undertaking the task must have available.

TABLE 6.6
Task Loading Table

Operations
Agent Operation Receive Decide Transport Total
PICOP 54 30 1 85
Signaller 17 16 1 34
Eng. Supervisor 13 17 1 31
Competent Person 1 17 20 1 38
EAST in Railway Maintenance 129

FIGURE 6.8 Illustration of propositional networks for phases within Scenario 3.

Summary
The summary and individual outputs derived from EAST have been presented. It should
be evident that the data generated from an EAST analysis are extensive, so in Figure 6.9,
an attempt is made to convey an impression of what the total EAST representation looks
like. It is from this representation that a C4i scenario can be surveyed relatively quickly
according to the key constructs of who, when, where, how and what. In turn, areas and
themes that require further detailed insights can be extracted and examined.

CONCLUSIONS
Further Insights
The individual outputs available from EAST present a number of distinct but overlap-
ping perspectives on the railway maintenance scenarios. This structured collection
of data is required for the subsequent generic modelling and C4i theory develop-
ment that is the main purpose of, but is separate from, the EAST method. However,
along with contributing to theoretical data, a number of more practical themes can
be drawn from each EAST analysis. The main practical themes related to this EAST
analysis of railway data have been identified as follows:

1. The HTA and OSD highlight the sequential nature of the tasks involved.
There is relatively limited decision making, and therefore the complexity
comes from a requirement on the personnel involved to time and precisely
sequence their actions.
130 Systems Thinking in Practice

FIGURE 6.9 Summary of application of EAST to live railway data.

2. Pursuant to this, the analysis highlights a routine requirement in each com-

munication to identify the people involved and the purpose of the call. This
is due to the verbal nature of the telephone technology in use.
3. The analysis highlights the key role of communication in this process and
that communication plays a part in, and is dependent on, SA.
4. Finally, the signaller (and other key agents) are required to develop situ-
ational awareness about the requirements of the possession communicated
by various parties physically present at the trackside. They then have to
consider the feasibility of carrying out the requirements in the context of
planned information (where available) and the current status of train move-
ments in the location.

The creation and maintenance of SA based on communications between actors is

a key issue and can be examined further. One strategy to enable this is to extract the
knowledge objects that the signaller requires, and that different actors possess. The
relevant propositional networks and companion CDMs can be used for this purpose.
The sharing of information can then be related to the sequence and timing of actions,
and actors. Where knowledge objects are shared, consideration can also be given to
the means by which this sharing occurs, for example, the communications technol-
ogy that facilitates this sharing. This particular approach enables the analyst to pro-
ceed from a theoretical level to one that is much closer to practical design outputs.
Figure 6.10 below illustrates this example of an approach to a more detailed analysis.
In describing this example, the intention is to convey how the general representa-
tions provided by EAST can be drilled down into to provide very specific insights
from multiple perspectives. These multiple perspectives can be theoretical and
practical.
EAST in Railway Maintenance 131

FIGURE 6.10 Plan for detailed analysis of communications and SA within railway scenarios

Wider Context
In conclusion, the application of the EAST methodology to live railway data illus-
trates the design and descriptive capability of the method. The method summarises
the task structure, network structure and operational context into a form that enables
ready comparisons on key metrics to be drawn between scenarios and for individual
practical themes and insights to be analysed further. The method is also scalable,
and its application by the DTC HFI to a host of military and civil domains is provid-
ing the data necessary to provide input into the development of generic models and
theories of C4i.

ACKNOWLEDGEMENTS
The assistance of the Rail Safety and Standards Board is gratefully acknowledged,
and the data presented here are used with their kind permission.
This work from the Human Factors Integration Defence Technology Centre
(HFI DTC) was part-funded by the Human Sciences Domain of the UK Ministry of
Defence Scientific Research Programme.

REFERENCES
Anderson, J. R. (1980). Cognitive Psychology and its Implications. Freeman: San Francisco,
CA.
Annett, J. (2003). Hierarchical task analysis. In D. Diaper and N. A. Stanton (Eds.), Handbook
of Task Analysis in Human–Computer Interaction (pp. 67–82). Lawrence Erlbaum
Associates: Mahwah, NJ.
Annett, J. (2005). Hierarchical task analysis. In N. A. Stanton et al. (Eds.), Handbook of
Human Factors and Ergonomics Methods (pp. 33.1–33.7). CRC: London, UK.
Annett, J., Duncan, K. D., Stammers, R. B. and Gray, M. (1971). Task Analysis. Her Majesty’s
Stationery Office: London, UK.
Annett, J. and Stanton, N. A. (2000). Task Analysis. Taylor and Francis: London.
Baber, C. (2004). Personal communication.
132 Systems Thinking in Practice

Bell, H. H. and Lyon, D. R. (2000). Using observer ratings to assess situation awareness. In
M. R. Endsley (Ed.), Situation Awareness Analysis and Measurement (pp. 129–146).
Lawrence Erlbaum Associates : Mahwah, NJ.
Burke, S. C. (2005). Team task analysis. In N. A. Stanton et al. (Eds.), Handbook of Human
Factors and Ergonomics Methods (pp. 56.1–56.8). CRC: London, UK.
Collins, A. M. and Loftus, E. F. (1975). A spreading-activation theory of semantic processing.
Psychological Review, 82, 407–428.
Defence Technology Centre for Human Factors Integration (2003). Methodology for DTC
HFI WP1 Field Trials (Report: Draft 2.1.). Baber, C. & Stanton, N. A: London, UK.
Diaper, D. and Stanton, N. A. (2004). Handbook of Task Analysis in Human–Computer
Interaction. Lawrence Erlbaum Associates: Mahwah, NJ.
Driskell, J. E. and Mullen, B. (2005). Social network analysis. In N. A. Stanton et al.
(Eds.), Handbook of Human Factors and Ergonomics Methods (pp. 58.1–58.6). CRC:
London, UK.
Drury, C. G. (1990). Methods for direct observation of performance. In J. R. Wilson and E.
N. Corlett (Eds.), Evaluation of Human Work – A Practical Ergonomics Methodology
(pp. 45–68). Taylor and Francis: London, UK.
Endsley, M. R. (1995). Toward a theory of situation awareness in dynamic systems. Human
Factors, 37(1), 32–64.
Harris, C. J. and White, I. (1987). Advances in Command, Control & Communication
Systems. Peregrinus: London, UK.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P., Stewart, R. and
Walker, G. (2006). Command and control in emergency services operations: A social
network analysis. Ergonomics, 49 (12–13), 1204–1225.
Kieras, D. E. (2003). GOMS models and task analysis. In D. Diaper and N. A. Stanton (Eds.),
Handbook of Task Analysis in Human–Computer Interaction (pp. 83–116). Lawrence
Erlbaum Associates: Mahwah, NJ.
Kirwan, B. and Ainsworth, L. K. (1992). A Guide to Task Analysis. Taylor and Francis:
London, UK.
Klein, G. and Armstrong, A. A. (2005). Critical decision method. In N. A. Stanton et al.
(Eds.), Handbook of Human Factors and Ergonomics Methods (pp. 35.1–35.8). CRC:
London, UK.
Klein, G. A., Calderwood, R. and MacGregor, D. (1989). Critical decision method for eliciting
knowledge. IEEE Transactions on Systems, Man, and Cybernetics, 19(3), 462–472.
Ogden, G. C. (1987). Concept, knowledge and thought. Annual Review of Psychology, 38,
203–227.
O’Hare, D., Wiggins, M., Williams, A. and Wong, W. (2000). Cognitive task analysis for deci-
sion centred design and training. In J. Annett and N. A. Stanton (Eds.), Task Analysis
(pp. 170–190). Taylor and Francis: London, UK.
Quillian, R. (1969). The teachable language comprehender: A simulation program and theory
of language. Communications of the ACM, 12459–12476.
Rail Safety and Standards Board. (2003). Master Rule Book. Newark: Willsons.
Scott, J. (1991). Social Network Analysis: A Handbook. Sage: London, UK.
Smalley, J. (2003). Cognitive factors in the analysis, design, and assessment of command
and control systems. In E. Hollnagel (Ed.), Handbook of Cognitive Task Design (pp.
223–253). Lawrence Erlbaum Associates: Mahwah, NJ.
Stanton, N. A. (2004). The psychology of task analysis today. In D. Diaper and N. A. Stanton
(Eds.), Handbook of Task Analysis in Human–Computer Interaction (pp. 567–584).
Lawrence Erlbaum Associates: Mahwah, NJ.
Stanton, N. A., Baber, C. and Young, M. S. (2005). Observation. In N. A. Stanton et al.
(Eds.), Handbook of Human Factors and Ergonomics Methods (pp. 28.1–28.7). CRC:
London, UK.
EAST in Railway Maintenance 133

Stanton, N., Hedge, A., Brookhuis, K., Salas, E., and Hendrick, H. (2005). Handbook of
Human Factors and Ergonomics Methods. CRC: London, UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
P., Hoyle, G., Walker, G., Young, M. S., Linsell, M., Dymott, R. and Green, D. (2006).
Distributed situation awareness in dynamic systems: Theoretical development and
application of an ergonomics methodology. Ergonomics, 49(12–13), 1288–1311.
Watts, L. A., and Monk, A. F. (2000). Reasoning about tasks, activities and technology to
support collaboration. In J. Annett and N. Stanton (Eds.), Task Analysis (pp. 55–78).
Taylor and Francis: London, UK.
7 EAST at Road
Intersections
With Michael G. Lenné and J. Ashleigh Filtness

STATEMENT OF RELEVANCE
Intersection safety currently represents a key road safety issue worldwide. In line
with recent calls for a systems approach to road safety, this chapter presents a novel
application of a framework of ergonomics methods for studying intersection ‘system’
performance. The findings support development of interventions that consider all
road users as opposed to one group in isolation.

INTRODUCTION
Road transport–related trauma continues to be one of the leading causes of death
and disability throughout the world (World Health Organisation 2009). Although
significant reductions in injuries and fatalities have been made over the last four
decades in most motorised countries (Elvik 2010), a number of complex intractable
issues remain. One of these is collisions between different types of road users at
intersections. In Australia, for example, the majority of urban crashes and a substan-
tial proportion of rural crashes occur at intersections (McLean et al. 2010). Safety at
intersections, therefore, represents a key road safety issue.
The prevalent approach to understanding and preventing collisions at intersec-
tions has tended to be road user centric and component driven, focusing on specific
road user types (e.g. drivers) and/or fixing a component of the problem (e.g. making
motorcycles more conspicuous; Gershen and Shinar 2013). The fundamental limita-
tion of this reductionist approach is that there is a focus on only a component of the
problem (e.g. rider conspicuity). In recent times, researchers have made a strong case
for the ‘systems’ approach when considering road user behaviour and safety (e.g.
Larsson et al. 2010; Salmon et al. 2012). The systems approach has important impli-
cations for how safety outcomes are achieved. Are they achieved through focusing
on one component of the system in isolation (a reductionist approach) or can they,
instead, be achieved by looking at the wider interacting system of human and techni-
cal elements (a systems approach)? In other words, does safety arise from ‘adding’
together interventions performed on individual elements of the system, such as the
road environment and intersection infrastructure, training, road rules and regula-
tions, environmental conditions and individual road users, or does it multiply or
‘emerge’ from these interactions? If it does, then a focus on individual elements will

135
136 Systems Thinking in Practice

fail to capture important road safety interventions that could have disproportionately
large effects.
Unsurprisingly, the systems approach has long been discussed in the context of
road traffic crashes (e.g. Wagenaar and Reason 1990), but despite this, very few
studies go beyond a focus on just one component of the system. Previous studies of
road user behaviour at intersections have largely been road user centric, focusing
on individual road user factors such as driving errors (e.g. Gstalter and Fastenmeier
2010; Sanding 2009) and pedestrian behaviours (e.g. King et al. 2009). Moreover,
the majority of road safety research focusing on vulnerable road users, such as
cyclists and motorcyclists, has been undertaken from the perspective of drivers (e.g.
Hole and Tyrrell 1995; Schepers and Brinker 2011) or engineering (e.g. Highways
Agency 2012). The consequence is that other factors influencing behaviour are left
untouched and so diminish the effectiveness of the reductionist intervention (Salmon
et al. 2012). For this reason, the reductionist, component driven approach to road
safety has been criticised for failing to take into account the inherent complexity in
transportation systems (e.g. Larsson et al. 2010; Salmon et al. 2012). This article is a
response to this criticism.
This article argues that the long-standing problem of collisions between distinct
road users at intersections cannot be solved through a reductionist approach that
does not consider the overall sociotechnical intersection system. That is, consid-
eration of individual road user behaviours in isolation, whilst yielding important
information, will not necessarily lead to appropriate interventions that support the
behaviour of all road users. This creates a challenging methodological problem.
Existing approaches for studying road user behaviour, such as simulation (e.g. de
Winter et al. 2009), microsimulation (e.g. Bell et al. 2012) and questionnaires (e.g.
Reason et al. 1990) typically focus on one component only (e.g. ‘driving’ simula-
tors, ‘driver’ behaviour questionnaire). To go beyond mere lip service to the ‘sys-
tems approach’ requires new approaches that embed individual road users within a
wider joint cognitive system.
The analysis presented in this chapter moves towards systems analyses in road
transport by considering the intersection ‘system’, comprising different road users
(e.g. drivers, riders, cyclists and pedestrians), vehicles and the road environment.
Using data derived from a recent on-road study, the Event Analysis of Systemic
Teamwork framework (EAST) (Stanton et al. 2013; Walker et al. 2006, 2010) was
used to evaluate behaviour at three major signalised intersections. The aims were to
examine the differences in behaviour of different road users within the intersection
system, to identify the factors underpinning conflicts between distinct road users at
intersections and to inform the development of new more holistic intersection inter-
ventions that are appropriate for all road users.

EAST Intersection Case Study

The key benefit of a systems approach is that it offers a universal language for
describing any type of system. Entities and artefacts are represented as nodes and
the relationships between them as links. As such, there is no conceptual reason
EAST at Road Intersections 137

why a systems approach cannot be applied to intersections. Indeed, previous

studies of road user behaviour at intersections suggest that two ‘systemic’ issues
may be particularly problematic: intersection design and infrastructure that does
not meet users’ expectations (e.g. Cornelissen et al. 2013) and incompatibilities
between different road users, such as drivers and motorcyclists (Salmon et al.
2013; Walker et al. 2011). The aim of the study was, through considering the
entire intersection system as the unit of analysis, to investigate these issues fur-
ther by applying the EAST framework to real-world, on-road data. The EAST
framework lends itself to an in-depth evaluation of intersection performance of
this kind. It facilitates examination of specific constructs relevant to the two key
issues mentioned above, such as situation awareness, decision making, team-
work, training and road design. Whilst not providing direct recommendations,
the analyses will

1. help to reduce complexity

2. identify specific issues limiting performance
3. highlight areas where system redesign could be beneficial.

The data were derived from an on-road study that covered a pre-defined 15km
urban route in the south-east suburbs of Melbourne, Australia. The analysis focused
on three major signalised intersections located along the route (see Figure 7.1). Each
intersection required a right-hand turn in order to pass through the intersection and
remain on the study route.

FIGURE 7.1 The three intersections studied (white arrow denotes route though intersection).
138 Systems Thinking in Practice

METHODOLOGY
Design
The work reported in this chapter is based around an on-road study investigating
the influence of road design on the behaviour of different types of road user. The
study used a semi-naturalistic paradigm whereby participants drove an instrumented
vehicle around a pre-defined urban route. Drivers drove an instrumented test vehicle,
whilst motorcyclists and cyclists completed the route using their own motorcycle or
bicycle, which was equipped with video and audio recording equipment. Pedestrians
negotiated the three intersections on foot whilst wearing video recording glasses
and a microphone connected to a Dictaphone. All participants provided concurrent
verbal (think aloud) protocols as they negotiated the route.

Participants
Seventy-eight participants (52 male, 26 female) aged 21−64 years (mean = 35.81, SD
= 13.03) took part in the study. The sample comprised 20 car drivers, 18 motorcy-
clists, 20 cyclists and 20 pedestrians. An overview of the participants in each group
is presented in Table 7.1.
All participants were recruited through a weekly on-line university newsletter
and were compensated for their time. Prior to commencing the study, ethics approval
was formally granted by the Monash Human Ethics Committee.

Materials
A demographic questionnaire was completed using pen and paper. A desktop driving
simulator was used to provide verbal protocol training. A 15 km urban route, located
in the south-eastern suburbs of Melbourne, was used for the on-road study component.
Drivers drove the route in a 2004 Holden Calais sedan equipped to collect various vehi-
cle, driving scene and driver-related data. A Dictaphone was used to record participant’s

TABLE 7.1
Participant Demographics

Hours Using Respective Mode

Road User Group Mean Age (SD) Gender of Transport per Week
Drivers 34.9yrs (12.53) 10 males 11.5 hours
10 females
Cyclists 32.4yrs (10.42) 15 males 6.85 hours
5 females
Motorcyclists 45.5yrs (12.87) 17 males 7 hours
1 female
Pedestrians 30.5yrs 10 males 8.92 hours
(11.86) 10 females
EAST at Road Intersections 139

verbal protocols. Each motorcyclist participant’s motorcycle was fitted with an Oregon
Scientific ATC9K portable camera, which, depending on the motorcycle model, was
fixed either to the handlebars or front headlight assembly. The ATC9K camera records
the visual scene, speed and distance travelled. A microphone was fitted inside each
motorcyclist participant’s helmet and attached to the Dictaphone to record their verbal
protocols. For cyclist participants, the ATC9K portable camera was fitted to their hel-
met and they wore Imging HD video cycling glasses. Pedestrians negotiated the inter-
sections on foot whilst wearing Imging HD video sunglasses and a microphone linked
to a Dictaphone. Motorcyclist, cyclist and pedestrian participants also carried a Garmin
GPS unit to record aspects such as speed, distance travelled, route taken and so on.
All verbal protocols were transcribed using Microsoft Word. During data analysis, the
Leximancer TM content analysis software was used to construct knowledge networks.

Procedure
Individual participants negotiated the intersections on separate occasions. In order
to control for traffic conditions, all trials took place at the same pre-defined times on
weekdays (10 am or 2 pm Monday to Friday). Participants first completed an informed
consent form and demographic questionnaire and were then briefed on the research
and its aims (expressed generally as a study of driver behaviour). Following this,
they were given training in providing verbal protocols, which included instruction
from an experimenter followed by a desktop driving simulator task where they were
asked to complete the drive whilst providing a verbal protocol. The experimenter
monitored the drive and provided feedback to the participant regarding the quality
of their verbal protocol. Participants were then shown the study route and were given
time to memorise it. When comfortable with the verbal protocol technique and route,
participants were taken to their vehicle and asked to prepare themselves for the test.
They were then given a demonstration of the video and audio recording equipment,
which was also set to record at this point. Following this, the experimenter instructed
the participant to begin the study route. For the drivers, an experimenter was located
in the vehicle and provided route directions if necessary. For the motorcyclists and
cyclists, an experimenter followed behind (in a car for the motorcyclists, on a bicycle
for the cyclists), ready to intervene if the participants strayed off route. Pedestrians
were taken by car to the first intersection and instructed to negotiate the intersection
and walk to a set point following the intersection. Once the participant reached this
point, they were picked up by the experimenter and driven to the next intersection.
This process was repeated until all three intersections had been negotiated.
Participants’ verbal protocols were transcribed verbatim using Microsoft Word.
For data reduction purposes, extracts of each verbal transcript for each intersection
were taken from the overall transcripts. The extracts were taken based on the video
data and pre-defined points in the road environment (e.g. beginning and end of inter-
section). The three network representations were constructed as follows:

1. Task networks: The task networks were built based on experimenter

observations, a review of the video data and the verbal transcripts. This
led to the construction of generic task networks for each road user group,
140 Systems Thinking in Practice

incorporating the range of participant behaviours observed within each

group at the three intersections.
2. Social networks: In the present analysis, the social networks represent
each participant’s interaction with the intersection ‘system’. Construction
of the social networks involved reviewing the verbal transcripts and
recording all instances in which the participant in question verbalised an
interaction with part of the intersection system (e.g. ‘I’m just checking
the traffic lights ahead’, ‘I’ve just noticed a car pulling up behind me’,
‘I’m just going to press the pedestrian crossing button’). The frequency
of interactions with different parts of the intersection system were cal-
culated for each road user group and represented using standard social
network diagrams.
3. Situation awareness networks: Situation awareness networks were con-
structed for each participant group at each intersection through analysis
of participants’ verbal transcripts using the Leximancer content analysis
tool. Leximancer uses text representations of natural language to interro-
gate verbal transcripts and identify themes, concepts and the relationships
between them. The software does this by using algorithms linked to an
in-built thesaurus and by focusing on features within the verbal transcripts
such as word proximity, quantity and salience. Leximancer thus automates
the content analysis procedure by processing verbal transcript data through
five stages: conversion of raw text data, concept identification, thesaurus
learning, concept location and mapping (i.e. creation of network). The out-
put is a network representing concepts derived from the verbal transcript
and the relationships between them reflected within the verbalisations. The
software has previously been used for situation awareness network con-
struction and analysis (e.g. Salmon et al. 2013; Walker et al. 2011) and is
important to analyses of this kind since it provides a reliable, repeatable
process for constructing the networks.

Each form of network was analysed using selected network metrics that have pre-
viously been used in EAST analyses. Although predominately used for the analyses
of social networks, Stanton (2014) demonstrated the utility of applying selected net-
work analysis metrics to the task, social and situation awareness networks generated
through EAST analyses. In the present study, the following network analysis metrics
were applied to each network:

1. Network density – Network density represents the level of interconnectivity

of the network in terms of relations between nodes. Density is expressed as
a value between 0 and 1, with 0 representing a network with no connections
between nodes and 1 representing a network in which every node is con-
nected to every other concept (Kakimoto et al. 2006; cited in Walker et al.
2011).
2. Sociometric status (individual nodes). Sociometric status provides a mea-
sure of how ‘busy’ a node is relative to the total number of nodes within
the network under analysis (Houghton et al. 2006). In the present analysis,
EAST at Road Intersections 141

nodes with sociometric status values greater than the mean sociometric sta-
tus value plus one standard deviation are taken to be ‘key’ (i.e. most con-
nected) nodes within each network.

RESULTS
Task Networks
The ‘negotiate intersection’ task networks provide a summary of the main goals and
tasks involved in negotiating signalised intersections for each road user group. The
task networks for each road user group are presented in Figures 7.2 through 7.5.
The task networks highlight the differences, in terms of goals and tasks per-
formed, between the four road user groups when negotiating intersections. As shown
in Figures 7.2 through 7.5, when negotiating intersections, the goals and subsequent
tasks are markedly different across the four road user groups studied. The only task
common across the four road user groups is that of ‘maintaining situation aware-
ness’, and this task in itself entails maintaining awareness of different aspects of
the intersection depending on road user type (e.g. cyclists and motorcyclists require
awareness of debris on the road, whereas drivers do not).
One key issue highlighted by the task networks is the differences in the level of
flexibility afforded to each road user by the intersections studied. Cyclists have a sig-
nificant level of flexibility in that they can proceed through the intersection in a vari-
ety of different ways. Depending on traffic conditions and perceived level of risk, they
can either turn right on the road within the flow of traffic, turn right via the pedestrian
crossings and along the footpath or turn right using a ‘hook’ turn, whereby they pro-
ceed straight on through the intersection, join the traffic queue to the left-hand side

Approach Check intersection and determine

intersection route through intersection

Select and move into appropriate lane

Maintain
situation
awareness Take place in traffic Maintain
Operate (e.g. other Follow traffic queue/at safe
vehicle traffic, own front of traffic queue position on
position, road
lights, speed
limits)
Check lights and wait for green light/turning arrow

Proceed through intersection

FIGURE 7.2 Driver intersection task network.

142 Systems Thinking in Practice

Approach intersection

Check intersection
and traffic and
Maintain Filter to Take place in
determine
situation front of traffic queue Maintain
route through
awareness traffic queue (in appropriate Make self safe
intersection
Operate (e.g. other (in appropriate lane) visible to position on
(e.g. right/ lane)
bicycle traffic, own other road road and
left lane, hook turn,
position, Get onto users avoid surface
service lane, use
lights, road footpath obstacles
pedestrian crossings)
surface) and
cross using
pedestrian
Stop and wait for green light crossings

Proceed forward into

Proceed through intersection other traffic flow to
conduct hook turn

FIGURE 7.3 Cyclist intersection task network.

and then wait for a green light and proceed straight through the intersection (achiev-
ing the originally desired right-hand turn). This flexibility is also not restricted to a
certain point within the intersection; cyclists can change their route through at any
point prior to the intersection and even when passing through the intersection. For
pedestrians, a degree of flexibility is extant through the ability to either cross via the
pedestrian crossing or the road itself. Motorcyclists’ behaviour is more restricted;
however, they still have some flexibility in that they can choose to use the stay in the
normal traffic queue or to filter up between the traffic to the front of the queue. Of the
four road user groups, drivers are the most restricted in terms of behaviour since they
can only select one of two turning lanes and progress through the intersection.
Other important differences across the task networks are present. These include
the task of filtering up through the traffic queue undertaken by both cyclists and
motorcyclists in pursuit of the goal to get to the front of the traffic queue, the

Check intersection and determine route

Approach intersection
through intersection

Maintain Filter to front of traffic queue (in

Take place in traffic queue (in
situation appropriate lane) onto traffic
appropriate lane)
awareness light sensor Maintain safe
Make self
(e.g. other position on
Operate visible to
traffic, own roadand
motorcycle other road
position, avoid surface
users
lights, road obstacles
surface,
speed limits) Wait for green light

Proceed through intersection

FIGURE 7.4 Motorcyclist intersection task network.

EAST at Road Intersections 143

Check intersection and determine route

Approach intersection
through intersection

Activate pedestrian crossing(s)

Maintain
situation
Walk/stop/ Wait for green man
awareness
stand
(e.g. traffic,
lights)

Cross road (without

Cross road using crossing
crossing)

Leave intersection area

FIGURE 7.5 Pedestrian intersection task network.

requirement for cyclists and motorcyclists to undertake activities in order to make

themselves visible to other road users (e.g. wiggling, flashing lights, eye-balling driv-
ers, revving engine) and the requirement for cyclists and motorcyclists to check the
road surface for debris/hazards/obstacles and to avoid debris on the road as they turn
through the intersection.
The task models confirm that there are key differences in the goals being pursued
and tasks being performed across the different road user groups. Moreover, they
reveal differences that raise concerns regarding potential conflicts between distinct
road users. For example, the level of flexibility afforded to cyclists, motorcyclists
and pedestrians ostensibly gives them a degree of unpredictability in and around
intersections. A cyclist in the right-hand turn lane may decide at a late stage to cross
the intersection via the pedestrian crossing and thus manoeuvre across three lanes
of traffic. Depending on how alert drivers are, this unpredictability could lead to
conflicts. The different means of negotiating the intersection employed by different
road users also raise potential conflicts, for example, cyclist and motorcyclist filter-
ing behaviour coupled with drivers changing lanes and following the traffic ahead,
or cyclists negotiating the intersection on the footpath and via pedestrian crossings.
The density and sociometric status calculations for the task networks are pre-
sented in Table 7.2.
Table 7.2 shows that the driver, cyclist and pedestrian task networks had similar
levels of density; however, the motorcyclist task network is the most dense (0.81).
This indicates that, of the four road users, motorcyclists have a more connected set of
tasks when negotiating intersections. The key tasks, as identified through the socio-
metric status metric, demonstrate the importance of ‘operate vehicle’ and ‘maintain
situation awareness’ tasks for all road users but also shows the importance of the
‘make self visible’ and ‘maintain safe position on road’ for cyclists and motorcyclists.
144 Systems Thinking in Practice

TABLE 7.2
Task Network Analysis

Road User Density Sociometric Status Key Tasks

Drivers 0.62 Operate vehicle
Maintain situation awareness
Maintain safe position on road
Cyclists 0.69 Operate bicycle
Maintain situation awareness
Make self visible to other road users
Maintain safe position on road
Motorcyclists 0.81 Operate motorcycle
Maintain situation awareness
Make self visible to other road users
Maintain safe position on road
Pedestrians 0.63 Walk/stop/stand
Maintain situation awareness

Social Networks
The social networks depict each road user group’s interaction with the intersections
studied. For example, the social networks for each road user group at Intersection 1
are presented in Figure 7.6. This shows the total number of interactions with differ-
ent parts of the intersection for each road user group. The thickness of the arrows
represents the frequency of interactions, with the thicker arrows representing a
higher frequency across each road user group.
The social network analyses highlight the fact that intersections are characteristic
of joint cognitive, or distributed cognition, systems. The networks show a high inter-
action between human and non-human agents and also the key role of non-human
agents within the right turn at intersection task. For each form of road user, the
turn right task involves multiple non-human agents, such as the traffic lights, road
and road markings, other vehicles and road signage. The importance of this finding
is that compatibility between human and non-human agents is known to be a key
element of efficient distributed cognition system performance (Salmon et al. 2008;
Stanton et al. 2006; Walker et al. 2010).
Density was calculated for each network, and the sociometric status metric was
used to identify the most connected nodes within each network (outside of the road
users themselves). These network analysis results are presented in Table 7.3.
Table 7.3 demonstrates that the pedestrians were the most connected in terms of
their interactions with the intersection environments. They interacted with the great-
est number of intersection components (e.g. lights, button, traffic), had the highest
number of interactions and achieved the highest sociometric status values across all
three intersections. The drivers, on the other hand, were the least connected of the
four groups, with smaller networks, the lowest number of interactions and the small-
est sociometric status values across the three intersections.
EAST at Road Intersections 145

On- Traffic
coming lights
traffic On- Traffic
coming
traffic lights
Ped Road
26 name
44 crossing (straight
ahead) 21
Traffic
Traffic Lane Parked cars
Speed 2 82 (service
1 lane)
limit 11 Service
4 Road 9 signage Traffic lane
5 lights 2
surface Road 17
2 surface
Traffic 1 Green/ P.
Traffic
lights in Red man Traffic Highway
7 front 9 in
6 front 9
Left turn 11 Ped
arrow crossing Turning
1
4 21 traffic
Pedes- 2
trians Driver 1
Camera
6 Footpath
4
Cyclist

Pedes- 2
6 trians
14

Traffic
behind Traffic
behind

Traffic
On- Traffic
coming lights
traffic
Lights

Traffic 15 Footpath
Parked
180
56 cars 133
Traffic 12 Speed 11
lights 1 limit Bin Road
Road signage
3 surface 2
3
P. Driveway 29 Crossing
Traffic Highway
9 in 8 10
6 front 1
Service
Ped lane
Pedestrian Button
crossing 12 15
2 17
1
Police
Cars
4 P.
Signage 1
(driveway) Highway
Rider 1 Camera

1 4 1 Horizon
10 1 Island Camera
Pedes- Cyclist
trians 10 19 2
1
Peds 2 Gardener
Road
2
Traffic surface
behind Shops
Caryard

FIGURE 7.6 Social networks for Intersection 1.

The social networks also reveal important differences in how the different road
user groups interact with intersection systems. Again, differences across the four
road user groups are apparent. The networks show that the drivers also interacted
less with certain parts of the intersection than did the other road users. For example,
across the three intersections, the most frequent interaction for drivers was observing
the oncoming traffic, traffic lights ahead and the traffic in front. Although cyclists
and motorcyclists had frequent interactions with these parts of the intersections, they
tended to have more interactions with these components, and other interactions were
also prominent, such as checking for traffic behind and observing the traffic gener-
ally (i.e. travelling across the intersection). The low number of driver interactions
focusing on the traffic generally, the traffic behind and adjacent traffic is a concern,
particularly when coupled with the task network finding that the other road user
groups may be unpredictable in and around the intersection.
The other road user groups also had interactions not found in the driver group.
Unsurprisingly, pedestrians had various interactions related to the activity of inter-
acting with the pedestrian crossing infrastructure, such as pressing the ‘button’.
146 Systems Thinking in Practice

TABLE 7.3
Summary of Social Network Analysis for Each Intersection

Number Sociometric
Road User of Number of Status of Nodes with Highest
Group Nodes Interactions Density Road User Sociometric Status Values
Intersection 1
Drivers 12 116 0.083 11.09 Lights, Oncoming traffic,
Traffic (In front), Lane,
Pedestrians, Traffic
(Behind)
Cyclists 13 204 0.08 17.17 Lights, Oncoming traffic,
Traffic (In front), Service
lane, Traffic (Behind)
Motorcyclists 14 208 0.07 11.31 Lights, Traffic (In front),
Oncoming traffic, Traffic
(General), Traffic (Behind),
Pedestrians, Parked cars,
Road, Camera, Cyclist
Pedestrians 20 423 0.05 22.78 Traffic, Lights, Road,
Pedestrians, Service lane
Intersection 2
Drivers 10 127 0.1 14.11 Lights, Traffic (In front),
Traffic (General), Traffic
(Behind), Sign
Cyclists 20 208 0.05 10.95 Lights, Traffic (General),
Service lane, Traffic (In
front), Road
Motorcyclists 13 116 0.08 9.66 Lights, Traffic (In front),
Road, Traffic (General),
Traffic (Behind)
Pedestrians 28 388 0.036 14.37 Traffic (General), Lights,
Road, Intersection, Button
Intersection 3
Drivers 14 152 0.071 11.69 Lights, Traffic (General),
Traffic (In front),
Intersection, Lane
Cyclists 20 243 0.05 12.78 Lights, Traffic (General),
Traffic (Behind), Lane,
Traffic (In front)
Motorcyclists 18 113 0.053 6.28 Lights, Traffic (General),
Lane, Road, Intersection
Pedestrians 30 767 0.033 26.44 Traffic (General), Lights,
Road, Pedestrians,
Intersection
EAST at Road Intersections 147

Interestingly, cyclists also had a number of interactions with the footpath and pedes-
trian crossing infrastructure, which reflects the decision by selected cycling par-
ticipants to cross the intersection using the footpath and pedestrian crossing. Other
notable interactions specific to road user groups were motorcyclists and cyclists
checking the ‘road surface’ for debris and oil, cyclists and motorcyclists monitoring
traffic adjacent to them, motorcyclists checking the traffic lights facing other traffic
(to work out when other traffic would be stopping and going) and cyclists monitoring
pedestrian behaviours in and around the intersection.

Situation Awareness Networks

Leximancer was used to construct overall driver, cyclist, motorcyclist and pedestrian
situation awareness networks for each intersection. For example, Figure 7.7 depicts
each road user group’s situation awareness network, overlaid on one another, for
Intersection 1. Within Figure 7.7, the nodes and links are shaded to depict each road
user group’s situation awareness. Figure 7.7 shows how situation awareness differed
across the distinct road user groups whilst negotiating Intersection 1, both in terms
of the concepts underpinning situation awareness (i.e. nodes in the network) and in
the way in which the concepts were linked together (i.e. links between the nodes in
the network). Moreover, the network demonstrates that, even when the different road
users were using the same concept, they were doing so in conjunction with other dif-
ferent concepts. This pattern is repeated over the other two intersections studied. The
multi-road user situation awareness networks therefore confirm that driver, cyclist,
motorcyclist and pedestrian situation awareness was different when negotiating the
three intersections studied.
A summary of the situation awareness network analysis is presented in Table 7.4.
Table 7.4 highlights some important differences in situation awareness across
the four road user groups. Pedestrian situation awareness networks comprised more
situation awareness concepts at each intersection, whereas the driver networks
were the smallest. The driver situation awareness networks were the most dense at
Intersections 1 and 2, whereas the motorcyclists were most dense at Intersection 3.
The same pattern was found for the mean sociometric status values, which indi-
cates that the driver networks were most connected in terms of links between con-
cepts at Intersections 1 and 2, whereas the motorcyclists were most connected at
Intersection 3.
The key concepts underpinning each road user groups’ situation awareness
also reveal important differences. At Intersection 1, they key concepts are broadly
similar across all four road user groups, with situation awareness predominantly
underpinned by the lights, the lights’ status (red or green) and the traffic. Notable
differences at Intersection 1 include a focus of cyclists on the ‘intersection’ itself and
a focus of motorcyclists on the ‘lane’ within which to progress through the intersec-
tion. At Intersection 2, more differences in key concepts are apparent. Again, the
drivers’ focus is on the lights; however, the cyclists are also focused on the ‘service
lane’, which ‘lane’ to use, ‘turning’ traffic and the ‘road’ itself. The motorcyclists’
key concepts include ‘looking’, ‘behind’, and again the ‘turning’ traffic and the
‘lane’ to use. At Intersection 3, the key concepts are again different. The drivers’ key
148 Systems Thinking in Practice

= Cyclists = Drivers = Motorcyclists = Pedestrians

Checking
Orange
Lane
Service

Busy
Moment
Hand
Round Middle
Fairly
Car

Road
Looking Walking
I’ll
Check Able
Turning
Intersec on

Walk
Looks
Cross
Princes Safe
highway
Cars
Coming Stopping Proceed
Truck Sure

Buon Green Unclear

Traffic
Red

Front
Change Light Arrow

Man

Crossing Lights
Pedestrian
Slowly Wait

Amber

Behind
Straight
Oncoming
Wai ng

Stay Ahead
People

FIGURE 7.7 Intersection 1 situation awareness network showing driver, cyclist, motorcy-
clist and pedestrian situation awareness networks mapped onto each other.

concepts now include distinct areas around the car, such as ‘front’ and ‘behind’, and
also the ‘lanes’. The cyclists’ key concepts also include ‘ahead’ and ‘behind’, ‘wait’
and ‘crossing’. The motorcyclists’ key concepts relate to the lights (i.e. green) and
the ‘lane’ within which to progress through the intersection. Finally, the pedestrians
have a larger set of key concepts at Intersection 3, with key concepts covering the
EAST at Road Intersections 149

TABLE 7.4
Summary of Situation Awareness Network Analysis for Each Intersection

Mean
Road User Number of Sociometric
Group Concepts Density Status Key Concepts
Intersection 1
Drivers 26 0.075 0.151 Green, Oncoming, Traffic
Cyclists 28 0.061 0.122 Coming, Cars, Intersection, Red, Traffic
Motorcyclists 28 0.066 0.132 Lane, Traffic, Cars, Arrow
Pedestrians 29 0.069 0.138 Green, Red
Intersection 2
Drivers 27 0.074 0.148 Lights, Light, Lane
Cyclists 29 0.069 0.138 Cars, Green, Road, Take, Service, Lane,
Turning
Motorcyclists 28 0.071 0.143 Looking, Behind, Green, Cars, Turning,
Lane
Pedestrians 31 0.064 0.129 Lights, Light, Green, Turning, Coming,
Road, Crossing
Intersection 3
Drivers 27 0.074 0.148 Front, Red, Dandenong road, Making,
Behind, Lanes
Cyclists 28 0.071 0.143 Turning, Stay, Behind, Ahead, Wait, Lane,
Lights, Crossing
Motorcyclists 26 0.075 0.151 Green, Lane
Pedestrians 34 0.055 0.110 Road, Cars, Coming, Cross, Wait, Traffic,
Lights, Green, Red, Light,

‘road’, the traffic and its behaviour (e.g. ‘cars’, ‘traffic’, ‘coming’) and the lights and
their status.
The differences in the situation awareness networks were explored further by
examining the ‘common’ concepts (i.e. those found in all road user groups’ situation
awareness networks) along with the concepts unique to each road user group at each
intersection (See Figure 7.8). Figure 7.8 shows that the concepts common across all
four road user groups at the three intersections were mainly related to the cars, traf-
fic, the road, the lights and their status (e.g. green), the intersection and the act of
turning. This reflects a high focus of all road users on cars, the traffic lights and the
intersection environment itself.
Important differences in situation awareness concepts are shown in Figure 7.8.
At Intersections 1 and 2, the cyclist networks include the concepts ‘service’, ‘stay’
and ‘route’, all of which reflect the key decision that they face regarding whether
or not to use the service lane on approach to the intersection and then cross via
the footpath and pedestrian crossing, or to stay on the road and go through the
intersection with the normal traffic flow. Moreover, prior to the intersection,
150 Systems Thinking in Practice

Intersection 1
Common across all road Unique to Unique to Unique to
Unique to pedestrians
user groups drivers cyclists motorcyclists
Walk/ Check/
Car(s) Road Safe Service Stopping
Walking Checking

Light(s) Green I’ll Road name Fairly Button Ahead

Cross/
Intersection Traffic Slowly Moment Man
Crossing

Wait/
Coming Stay Round Able Straight
Waiting

Intersection 2
Common across all road Unique to Unique to Unique to
Unique to pedestrians
user groups drivers cyclists motorcyclists
Lane/
Car(s) Pull Service Hand Turned Anyway
Lanes

Light(s) Green Time Left Hand Button Man

Intersection Sure Ready Bike Seems

Red Straight Stay Gear

Turning Take Line

Route

Intersection 3
Common across all road Unique to Unique to Unique to
Unique to pedestrians
user groups drivers cyclists motorcyclists
Change/ Walk/
Car(s) Coming Hook I’ll Middle
Changing Walking

Check/
Light(s) Green Right hand Doing Hand Able
Checking

Intersection Red Forward Moving Clear Flashing

Look/
Turning Making Gear Button
Looking

Notice Merging Man

Route Assume

Stopping

FIGURE 7.8 Common and unique concepts across road user groups at each intersection.

the cyclists also decide whether they will leave the road and get back into the
service lane once they have passed through the intersection. At Intersection 3,
the cyclists’ network included the ‘hook’ concept, which refers to their decision
regarding whether or not to use a hook turn in order to turn right at the intersec-
tion. Again, this reflects a key decision whereby cyclists try to work out whether
it is safe enough to pass through the intersection on the road within the flow of
traffic or whether they need to perform a hook turn to avoid conflict with other
traffic also turning right.
EAST at Road Intersections 151

For the motorcyclists, the unique concepts relate primarily to the selection of the
left or right-hand lane to negotiate the intersection (e.g. ‘hand’, ‘left hand’, ‘merg-
ing’), the motorcycle itself (e.g. ‘bike’, ‘gear’) and the ‘line’ that they should take
through the intersection. The ‘stopping’ concept refers not only to motorcyclists’
own braking behaviour but also to them checking that other traffic approaching from
behind is stopping when the traffic lights are on red. For the pedestrians, the unique
concepts were primarily related to the physical acts of walking (e.g. ‘walk/walking’)
and crossing the road (e.g. ‘cross/crossing’) and also the crossing infrastructure (e.g.
‘button’, ‘green man’). Interestingly, only the pedestrian networks included the con-
cepts ‘check/checking’ and ‘look/looking’, which indicate that the other road users
placed less emphasis on checking other traffic and the road environment when nego-
tiating the intersections.

DISCUSSION
The aim of this article was to demonstrate how systems analysis methods such EAST
can be used to better understand the complex behaviour of intersection systems. In
doing so, we wished to examine the differences in behaviour across distinct road
users and to identify factors that potentially create conflicts between them. The study
aimed to demonstrate that the EAST method provides a means to reduce complex-
ity, identify specific issues limiting performance and highlight areas where system
redesign could be beneficial.

Reducing Complexity
The analysis demonstrates that, even when faced with the same intersection and right-
hand turn task, the goals and tasks, interactions and knowledge used by distinct road
users is different. This might not be particularly ground-breaking; however, identi-
fying a structured methodology and associated metrics that are capable of revealing
these differences using on-road, real-world data is a substantive contribution to the
current state of the art. Further, the framework moves towards the much-heralded use
of systems analysis techniques rather than reductionist, individual road user techniques
for road transport analyses. Notably, although a systems approach has been called for
by many (e.g. Larsson et al. 2010), to date there has been a dearth of appropriate meth-
ods proposed for supporting systems analyses in road transport (Salmon et al. 2012).
The EAST analysis also confirms the notion that road transport environments
represent joint cognitive systems in which cognition is distributed across both human
and non-human agents and is achieved through coordination between agents. This
not only confirms the requirement for systems analysis techniques in road transport
applications but also highlights the pressing need for studies of distributed cognition
in the road transport context. In addition, this suggests that a distributed situation
awareness approach (e.g. Salmon et al. 2008) in which the distribution of awareness
across the intersection system is focused on, rather than the awareness of one road
user group alone (e.g. drivers in motorcycle conspicuity studies), is required. Given
the rapid advances in the capability of in-vehicle and roadside technologies and
their likely increased role in road systems (e.g. intelligent infrastructure, car-to-car
152 Systems Thinking in Practice

communications), distributed cognition and distributed situation awareness studies

in support of future road system design efforts are recommended.

Performance Limitations
If tasks, interactions and knowledge are different across road users, and inter-
sections are joint cognitive systems, then the key to intersection safety is ensur-
ing that the differences are compatible and that road user activities and cognition
connect together. This raises two lines of inquiry for the present study findings.
First, do the differences revealed by EAST create conflicts between road users,
or are they compatible? And second, does intersection system design (e.g. road
environment design, training, road rules and regulations) support these differences
in a way that ensures compatibility between distinct road users? It is in synthesis-
ing the tripartite EAST perspective to answer these questions that the framework
becomes most powerful. Unfortunately, examination of the networks produced
indicates a level of disharmony that suggests compatibility is under threat in some
instances; in some circumstances, intersections may behave more like disjoint cog-
nitive systems. For example, the networks demonstrate that the limited latitude for
behaviour for drivers at intersections is in fact in conflict with the highly flexible
and thus unpredictable behaviour of cyclists, motorcyclists and pedestrians. Here,
flexibility, an often-sought-for characteristic in complex sociotechnical systems,
coupled with poor intersection design, is creating collisions between distinct road
users at intersections.
In the case of cyclists, for example, the combined networks clearly demon-
strate their high level of flexibility since they can proceed through the intersec-
tion in multiple ways (i.e. hook turn, in normal flow of traffic, filtering, on the
footpath and via pedestrian crossings). The networks also show that the result of
this is that cyclists’ behaviour and situation awareness are focused on working out
which way through is appropriate, and that their behaviour is generally unpre-
dictable since their chosen route through can change on a moment-by-moment
basis. Examination of the driver social and situation awareness networks, however,
shows that they may not be expecting cyclists, and indeed motorcyclists, to make
major manoeuvres in close proximity to the intersection. Consequently, they do
not behave or think in a manner that is compatible with other cyclists and motor-
cyclists. The driver situation awareness networks embodied a high focus on cars,
the traffic lights and the act of turning right and an absence of ‘check/checking’,
‘look/looking’, and left- or right-‘hand’ side concepts. The driver social networks
showed a limited interaction in which the oncoming traffic, traffic lights and traf-
fic directly in front of the vehicle represent the majority of their interaction with
the intersection system. Drivers also had only limited interactions with the areas
adjacent to and behind their own vehicle.
This ‘flexibility problem’, as we will call it, is likely brought about by the fact
that it represents flexibility not designed into the system per se but flexibility
derived from emergence. For example, cyclists’ range of possible behaviours at
intersections is brought about by the lack of support for, and high risk associated
with, turning right in the normal flow of traffic. Working within the constraints of
EAST at Road Intersections 153

intersection systems, cyclists and motorcyclists have come up with new ways of
taking a safer path through the intersection. The design of the three intersections
is therefore both creating this flexibility and then limiting compatibility between
the cyclist and the driver. There is currently no signage warning drivers of the pres-
ence of cyclists or motorcyclists or the likelihood that they will be manoeuvring
in and around the intersection. Also, where present, the cycling lanes preceding
the intersections end a significant distance before the intersection itself, and there
are no dedicated cycle or motorcycle lanes proceeding through the intersections.
In addition, the road rules prohibit cyclists from cycling on footpaths and motor-
cyclists from filtering up the traffic queue. The effect of this is that drivers and
pedestrians have limited experience of the prohibited behaviours and often may
not be expecting them.

System Redesign
One of the aims of the analysis presented was to identify opportunities to create
interventions designed to support safer interactions between distinct road users at
intersections. The findings suggest that there are various avenues that can be pur-
sued; however, the key is that all road users and non-human agents be considered,
as opposed to one group in isolation. Intersection design should strive to create joint
cognitive intersection systems.
To solve the flexibility problem, it would appear that the key is not only to ensure
that different road users have some degree of flexibility, but also to ensure that the
way in which the intersection is designed supports this flexibility and ensures that
the resulting behaviours are compatible. For example, it is appropriate to maintain
the level of flexibility for cyclists; however, other measures should be taken to sup-
port it. For example, avenues for making other road users (pedestrians, drivers) more
expectant of cyclists and their range of behaviours will be fruitful. The use of cross-
mode training has previously been raised as a way of developing anticipatory schema
that support perception of other distinct road users (e.g. Magazzu et al. 2006; Walker
et al. 2011). Cross-mode training incorporating both cyclist and motorcyclist training
for drivers is likely to increase their expectancy levels regarding cyclist and motor-
cyclist behaviours at intersections. There is also a clear role for road design, with
dedicated cycling lanes (on the road and on the footpath) and signage warning drivers
of the presence of cyclists and the likelihood that they will make major manoeuvres
in close proximity to the intersection.
As an exploratory study, this research did have some limitations. First, road
users negotiated the same intersections under similar traffic conditions but not at the
same time as one another (i.e. they were not negotiating the intersections together
at the same time). Analysis of intersection system performance when road users
are interacting together at the same intersection at the same time would provide
more valid data on the potential conflicts between them. Second, the social networks
were based on verbal protocols describing interactions rather than eye-tracking data
showing actual interactions with objects in the road environment. Future studies in
which eye-tracking data are used to construct the social networks are planned by
these authors.
154 Systems Thinking in Practice

CONCLUSION
It is concluded that systems analysis methods, such as EAST, provide a useful avenue
for understanding and eradicating road safety issues. From the analysis presented,
it is concluded that there was a significant level of disharmony in the intersections
studied. Although intersections are characteristic of joint cognitive systems, the
intersections studied often behaved like disjointed cognitive systems. Differences
were found in the way in which components (or road users) act, think and interact
with the environment; however, it appears that intersections are designed from the
viewpoint that all road users need to act and think in the same manner. Moreover, the
analysis suggests that the way in which the intersections are currently designed may
be restricting compatibility between different road user groups. It is argued that a
failure to consider the overall intersection system during intersection design, includ-
ing distinct road users’ physical and cognitive behaviours and their differences, is
facilitating the problem of intersection collisions. Moreover, a failure to consider
these differences in future will render intersection collisions an intractable problem.
Only a systems-based approach will be able to make substantive and sustainable
inroads into accident reduction.

REFERENCES
Bell, M. C., Galatioto, F., Giuffrè, T. and Tesoriere, G. (2012). Novel application of red-light
runner proneness theory within traffic microsimulation to an actual signal junction.
Accident Analysis and Prevention, 46, 26–36.
Cornelissen, M., Salmon, P. M., McClure, R. and Stanton, N. A. (2013). Using Cognitive
Work Analysis and the Strategies Analysis Diagram to understand variability in road
user behaviour at intersections. Ergonomics, 56 (5), 764–780.
de Winter, J. C. F., de Groot, S., Mulder, M., Wieringa, P. A., Dankelman, J. and Mulder, J. A.
(2009). Relationships between driving simulator performance and driving test results.
Ergonomics, 52 (2), 137–153.
Elvik, R. (2010). Why some road safety problems are more difficult to solve than others.
Accident Analysis and Prevention, 42, 1089–1096.
Farrington-Darby, T., Wilson, J. R., Norris, B. J. and Clarke, T. (2006). A naturalistic study
of railway controllers. Ergonomics, 49, 1370–1394.
Furniss, D. and Blandford, A. (2006). Understanding emergency medical dispatch in terms of
distributed cognition: A case study. Ergonomics 49 (12–13), 1174–1203.
Gershon, P. and Shinar, D. (2013). Increasing motorcycles’ attention and search conspicu-
ity by using Alternating-Blinking Lights System (ABLS). Accident Analysis and
Prevention, 50, 801–810.
Gstalter, H. and Fastenmeier, W. (2010). Reliability of drivers in urban intersections. Accident
Analysis and Prevention, 42 (1), 225–234.
Highways Agency (2012). Design Manual for Roads and Bridges. Highways Agency: London,
UK.
Hole, G. J. and Tyrrell, L. (1995). The influence of perceptual ‘set’ on the detection of motor-
cyclists using daytime headlights. Ergonomics, 38 (7), 1326–1341.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49, 1204–1225.
EAST at Road Intersections 155

King, M. J., Soole, D. and Ghafourian, A. (2009). Illegal pedestrian crossing at signalised
intersections: Incidence and relative risk. Accident Analysis & Prevention, 41 (3),
485–490.
Larsson, P., Dekker, S. W. A. and Tingvall, C. (2010). The need for a systems theory approach
to road safety. Safety Science, 48 (9), 1167–1174.
Magazzu, D., Comelli, M., and Marinoni, A. (2006). “Are Car Drivers Holding A Motorcycle
License Less Responsible for Motorcycle–Car Crash Occurrence? A Non-Parametric
Approach.” Accident Analysis & Prevention 38 (2): 365–370.
McLean, J., Croft, P., Elazar, N. and Roper, P. (2010). Safe Intersection Approach Treatments
and Safer Speeds Through Intersections: Final Report, Phase 1, AP–R363/10.
Austroads: Sydney, Australia.
Reason, J., Manstead, A., Stradling, S., Baxter, J. and Campbell, K. (1990). Errors and viola-
tions on the roads: a real distinction?. Ergonomics, 33, 1315–1332.
Salmon, P. M., McClure, R. and Stanton, N. A. (2012). Road transport in drift? Applying con-
temporary systems thinking to road safety. Safety Science, 50 (9), 1829–1838.
Salmon, P. M. Young, K. L. and Cornelissen, M. (2013). Compatible cognition amongst road
users: The compatibility of driver, motorcyclist, and cyclist situation awareness. Safety
Science, 56, 6–17.
Salmon, P. M., Stanton, N. A., Walker, G. H., Jenkins, D. P., Baber, C. and McMaster, R.
(2008). Representing situation awareness in collaborative systems: A case study in the
energy distribution domain. Ergonomics, 51 (3), 367–384.
Sandin, J. (2009). An analysis of common patterns in aggregated causation charts from inter-
section crashes. Accident Analysis & Prevention, 41 (3), 624–632.
Sanderson, P., Verhage, A. G. and Flud, R. B. (1989). State-space and verbal protocol methods
for studying the human operator in process control. Ergonomics, 32 (11), 1343–1372.
Scheper, P. A. and den Brinker, B. B. (2011). What do cyclists need to see to avoid single-
bicycle crashes? Ergonomics, 54 (4), 315–327.
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57:3, 403–418.
Stanton, N. A., Salmon, P. M., Rafferty, L., Walker, G., Baber, C., and Jenkins, D. P. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design. Second
Edition. Ashgate: Aldershot, UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
et al. (2006). Distributed situation awareness in dynamic systems: theoretical develop-
ment and application of an ergonomics methodology. Ergonomics, Vol. 49, pp. 221–229.
Stewart, R. J., Stanton, N. A., Harris, D., Baber, C., Salmon, P. M., Mock, M., Tatlock, K.,
Wells, L., and Kay, A. (2008). Distributed situational awareness in an airborne warn-
ing and control aircraft: Application of a novel ergonomics methodology. Cognition,
Technology and Work, 10 (3), 221–229.
Wagenaar, W. A. and Reason, J. T. (1990). Types and tokens in road accident causation.
Ergonomics, 33 (10–11), 1365–1375.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2011). Cognitive compatibility of motorcy-
clists and car drivers. Accident Analysis & Prevention, 43 (3), 878–888.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M., and Green, D. (2006).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics, 49, 1345–1369.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Jenkins, D. P. and Salmon, P. M. (2010).
From Ethnography to the EAST method: A tractable approach for representing distrib-
uted cognition in air traffic control. Ergonomics, 53 (2), 184–197.
World Health Organisation (2009). Global status report on road safety: Time for action.
World Health Organization: Geneva, Switzerland. Available at: www.who.int/
violence_injury_prevention/road_safety_status/2009.
8 EAST in Elite Women’s
Cycling Teams
With Clare Dallat and Amanda Clacy

INTRODUCTION
Situation awareness remains one the most ubiquitous of all ergonomics concepts,
with applications in an ever-expanding list of domains (Stanton et al. 2015). One area
in particular that has been receiving increasing attention is elite sport, with recent
studies considering situation awareness among Australian Rules Football umpires
(Neville et al. 2016), handball teams (De Keukelaere et al. 2013) and elite ham-
mer-throwing and rowing athletes and coaches (Macquet and Stanton 2013). Whilst
applications have included testing of theory and methods, the primary function has
been to inform the development of coaching strategies and interventions designed to
optimize performance (e.g. Macquet and Stanton 2014).
Despite exhibiting many features that are characteristic of the domains in which
ergonomics practitioners work, elite cycling has not yet been studied from this per-
spective. One interesting feature is the critical role that situation awareness appar-
ently plays at the individual rider level, at the cycling team level and at the overall
cycling system level. To date, however, the focus of research in elite cycling has
largely been on the development of quantitative models to describe the behaviour of
cycling pelotons (e.g. Trenchard 2015; Trenchard et al. 2014). What rider, team and
system situation awareness comprises, and how it can be optimised, remained unex-
plored. Such knowledge could potentially provide teams with a performance edge.
Much has been written about the contention surrounding situation awareness and
the appropriate theoretical and methodological approach to adopt in ergonomics
studies (see Salmon et al. 2008 and Stanton et al. 2017). In a recent review of situation
awareness models and their utility for studying and optimising elite sport systems,
Neville and Salmon (2016) identified Stanton et al.’s Distributed Situation Awareness
(DSA) model as highly suitable for sporting applications. Prior to this, Salmon et al.
(2010) had noted that the Event Analysis of Systemic Teamwork (EAST) framework
offers a powerful framework for understanding elite sports systems. Salmon et al.
(2010) concluded from test applications that the EAST framework was highly suited
to describing and analysing sports system performance and to supporting the devel-
opment of interventions designed to improve sports performance.
This chapter brings together both of these arguments, presenting an EAST analy-
sis of DSA and teamwork in an elite women’s cycling team during two road race
stages. The study was exploratory in nature and aimed to describe and understand
DSA and teamwork to assist the team in optimising performance in future events.
157
158 Systems Thinking in Practice

A secondary aim of the analysis was to further examine the utility of applying both
DSA and EAST in elite sports. In presenting this study here, the intention is to
encourage further applications of EAST in the sporting context.
Elite Women’s Cycling
The analysis presented in this chapter focuses on two elite women’s road races under-
taken as part of the 2016 Australian Subaru National Road Series (NRS). NRS road
races typically range in length from 50 kilometres (km) to 130 km and involve teams
of four to five riders. Each team has a so-called ‘protected’ or ‘general classification’
(selected based on overall NRS points standing), with the remaining riders acting in
a ‘domestique’* or support capacity.
The team’s goal is to ensure that their protected rider finishes each stage of the
race in the best position and time possible. Accordingly, the domestiques employ
various tactics to assist their protected rider, such as allowing the protected rider
to ride behind them (‘on their wheel’) to reduce their physical workload or ‘leading
them out’ to initiate a sprint finish.
The riders from each team form a peloton of cyclists that negotiate the race route
together. The peloton has been formally defined as ‘a group of cyclists that are coupled
together through the mutual energy benefits of drafting, whereby cyclists follow oth-
ers in zones of reduced air resistance’ (Trenchard et al. 2014, p. 92). Throughout the
race, riders may attempt to ‘attack’ and break away from the peloton, and other teams’
domestiques may attempt to chase them down and bring them back to the peloton or
let them go until a later point in the race. Alternatively, the peloton may stay together,
resulting in a ‘bunch sprint’ finish. Throughout the race, slower riders drop out of the
peloton and often work together to attempt to re-join the peloton. Likewise, domes-
tiques who have fulfilled their role of putting the protected rider in a position to finish
the race often drop out of the peloton due to fatigue. An additional feature of the NRS
women’s road races are shorter sprint and Queen of the Mountain (QOM) sub-sections,
whereby bonus points are offered for the top three cyclists within each sub-race.
A convoy of vehicles comprising the Commissaires; race officials, medical sup-
port, neutral spares, the media and team support vehicles follows behind the peloton.
The Commissaires controls the race and following convoy, communicating with the
cyclists via loudspeaker and the vehicles in the convoy via UHF radio or in person by
driving alongside them. A medical support vehicle follows behind the Commissaires
to provide assistance in the event of a crash or injury. Each team then has their own
specific support vehicle, typically containing a Director Sportif (DS), a mechanic
(in some cases this may be the DS), spare parts (e.g. wheels), bicycle tools and food
and drink for the riders. Each support vehicle is permitted to provide mechanical or
nutritional support to team riders and can offer tactical guidance and encouragement
through verbal exchanges. The positioning of the support vehicles in the peloton
convoy is based on the order of riders in the race (e.g. the team with a rider in first
place is given the first support vehicle spot within the convoy).
Together the peloton and convoy form a complex, highly dynamic system compris-
ing multiple human (riders, DS, mechanic) and non-human agents (bicycles, vehicles,

* In French, domestique translates as ‘servant’

EAST in Elite Women’s Cycling Teams 159

bicycle computers), which exhibits emergent properties, non-linear interactions, mul-

tiple control and feedback loops, loose and tight coupling and rapid decision making.
As such, it is amenable to ergonomics inquiry whilst offering relevant examples of
sociotechnical system behaviour that could inform theoretical, methodological and
practical contributions in other safety critical domains. It is also characteristic of pre-
cisely the kind of system that EAST was developed to describe and analyse.

Event Analysis of Systemic Teamwork

The EAST framework provides an integrated suite of ergonomics methods for ana-
lysing the behaviour of complex sociotechnical systems. Heavily underpinned by
network theory and analysis, EAST takes the premise that system performance
can be meaningfully described via a ‘network of networks’ approach, and it uses
three interlinked network-based representations to describe and analyse behaviour.
Task networks are used to describe the goals and subsequent tasks being performed
within a system (i.e. which agents, both human and non-human, do what). Social net-
works are used to analyse the structure of the system and the communications taking
place between agents (i.e. who/what interacts and communicates with whom/what).
Situation awareness networks show how information and knowledge is distributed
across different agents within the system (i.e. who/what knows what at different
points in time). By illustrating the relationships between task, social and situation
awareness networks and then by interrogating these networks, an in-depth under-
standing of system behaviour is achieved.
Salmon et al. (2010) first discussed the potential utility of using EAST to analyse
sports performance, presenting an EAST analysis of a fell running race that exam-
ined runner goals, situation awareness, decision making and workload. Whilst oth-
ers have since used components of EAST to examine elite sports performance (e.g.
Neville et al. 2016), the three forms of EAST network are yet to be used in conjunc-
tion with one another to analyse elite sports performance.

EAST Analysis of Elite Women’s Cycling

The study was a naturalistic study whereby the research team collected data from
an elite women’s road racing team during two Australian NRS events from the 2016
NRS season. Both events were elite cycling road race events incorporating a series of
stages, including a time trial, road races and a criterium race. The analysis presented
in this chapter focuses on one of the races from the first NRS event: stage 1 of the
Battle on the Border, which took place in the Tweed Coast region, New South Wales.
The full analysis will be included in the conference presentation.

METHODS
Participants
The participants were members of an elite women’s Australian NRS cycling team.
For the present analysis, the sample included five riders, one DS and one mechanic.
160 Systems Thinking in Practice

Due to the naturalistic nature of the study, it was not possible to gather complete
participant demographic data; however, all were experienced in elite cycling and had
raced throughout the 2016 NRS season. The study was granted ethics approval by
the University of the Sunshine Coast’s Human Ethics Committee.

Materials
The research team observed the race from within the cycling team’s support vehicle.
Go ProTM cameras were used to record the races, and a Dictaphone was used to record
the verbal communications occurring within the team support car. Dictaphones were
also used to record team race planning meetings, the post-race Critical Decision
Method (CDM) (Klein et al. 1989) interviews and post-race team debriefs. The inter-
viewers used a pro-forma containing the CDM interview probes. The interview tran-
scripts were transcribed using Microsoft WordTM. For data representation, the task
and social networks were created using Microsoft VisioTM.

Procedure
The research team provided the cycling team with an overview of the study, its aims
and the data collection methodology prior to the first race of each event. The research
team travelled to each race with the cycling team and observed the warm up and pre-
race discussions and planning activities. The Go Pro cameras and Dictaphone were
placed within the support vehicle and set to record shortly before the start of each race.
Once the races began, the research team observed the peloton from the support vehicle
and made hand-written notes regarding the team’s performance, tactics and the inter-
actions between the DS, the Commissaire and riders. During low workload periods of
the race, the researchers were also able to discuss the team’s performance with the DS,
including tactics, planned activities and factors influencing rider performance.
Following the race, the research team attended the riders’ post-race debrief and
planning session for the next race, recording both using a Dictaphone.
At the conclusion of the planning and debrief sessions, the research team con-
ducted one-on-one CDM interviews with each rider regarding the earlier race. The
interviewers used a set of CDM probes adapted for the cycling race context and
recorded each interview using a Dictaphone. Following the interview, each rider was
asked to complete a social network analysis questionnaire to show who and what
they interacted with during the race. This involved presenting them with a social
network diagram showing all agents (riders, other riders, DS, Commissaire, cycling
computer) and asking them to rate their level of communication with each during the
race on a scale of low (<10 communications), medium (between 10 and 20 commu-
nications) and high (>20 communications).
The video and audio footage were transcribed verbatim using Microsoft Word.
The three EAST network representations were constructed as follows:

1. Task network. The task network was built based on the research team’s
observations, a review of the video data and the interview transcripts. A
draft version of the task network was reviewed and refined by the cycling
team during the Battle on the Border event.
EAST in Elite Women’s Cycling Teams 161

2. Social networks. Construction of the social network involved transforming

the social network diagrams obtained following the races into a social net-
work diagram detailing the number of interactions between the cyclists and
other agents.
3. Situation awareness networks. Situation awareness networks were con-
structed through content analysis of the riders’ interview transcripts.
Specifically, the riders’ responses to the CDM questions ‘What were your
specific goals for the race?’ ‘What information were you using throughout
the race?’ and ‘What was the most important piece of information?’ and
riders’ initial description of the race content were analysed by one analyst.
Keywords and the relationships between them were extracted and used to
construct a situation awareness network comprising concepts and the rela-
tionships between them. For example, from the sentence, ‘I was checking
my position in the peloton’, the related concepts ‘Checking’, ‘Position’ and
‘Peloton’ were extracted. Once the situation awareness network was con-
structed, concepts were shaded based on their presence in each of the rider’s
interview transcripts.

RESULTS
Task Network
The task network is presented in Figure 8.1. Within Figure 8.1, the circular nodes
represent distinct tasks undertaken immediately prior to and during the road races.
The linkages between the nodes represent relationships between tasks; for example,
the ‘attack’ and ‘support protected rider’ tasks are closely linked, since the attacks
are made by domestique riders in order to either help the protected rider break away
from the peloton or to tire out opposing team riders.
The coloured rings around each node represent the agents in the cycling system
who undertake each task. For example, the ‘attack’ task is undertaken by the rid-
ers only, whereas the ‘monitor other riders’ task is undertaken by the riders, the
Commissaire and the DS.
Two features of the task network should be noted. First, there are a range of
tasks involved, including both individual tasks that are unique to specific agents and
shared tasks that are common across agents. For example, the task of communicat-
ing with other riders is undertaken by team riders within the peloton (e.g. asking for
assistance to initiate a break), by the DS in the support vehicle (e.g. communicating
tactics with team riders) and by the Commissaire (e.g. informing other riders in the
peloton and the DSs in the support vehicles of the time gap between a breakaway
rider and the peloton). Second, the relationships between each task can be used to
identify those tasks that are integral to team success. In this case, the task network
indicates that the tasks of ‘race planning and tactics’, ‘establish and maintain appro-
priate position in the peloton’, ‘monitor other riders’, ‘communicate with other rid-
ers’ and ‘monitor speed, cadence, power and distance’ are integral to team success
as they are the most connected to other tasks in the network. It is notable that these
tasks are also highly interrelated.
162 Systems Thinking in Practice

Warm up
Start race
Race planning
and tactics
Maintain
appropriate Monitor
position in speed/
convoy cadence/
distance etc

Attack Bridge across

Provide Establish/
Control convoy maintain
mechanical/
appropriate
nutritional
position in
support peloton
Monitor other
Communicate Bring back the riders
with break
commissaire Request
mechanical/
nutritional
support

Support Form break

protected rider QOM/Sprint
Update convoy
competition

Communicate Communicate
Update riders Sprint finish
with team with other
mates riders

Organise lead
out

= Riders = Commissaries = Director sportif

FIGURE 8.1 Generic task network for women’s cycling road racing.

Social Network
The social network for the Battle on the Border stage 1 road race is presented in
Figure 8.2. Within Figure 8.2, the circular nodes represent agents (e.g. riders, sup-
port vehicle, bicycle computers), and the arrows between them represent the extent to
which the agents communicated with one another during the race.
The social network shows the structure of communications throughout the road race.
In this case, the protected rider and domestique rider 2 are the most connected in terms of
incoming and outgoing communications. In addition, both the protected rider and domes-
tique rider 2 have a greater frequency of communications with other agents, both having
five connections that were rated as ‘high’ on the frequency of communications scale.

Situation Awareness Network

The situation awareness network for the Battle on the Border stage 1 road race is pre-
sented in Figure 8.3. Within the network, the nodes represent pieces of information,
and the lines linking the nodes represent relationships between the information (e.g.
‘Computer’ displays ‘Speed’).
The network shows that certain pieces of information were critical during the
race. For example, the nodes ‘Peloton’, ‘Position’ and ‘Riders’ are the most connected
nodes within the network, indicating that they were integral to situation awareness
and decision making during the race. This included the riders self-monitoring their
EAST in Elite Women’s Cycling Teams 163

FIGURE 8.2 Social network for Battle on the Border road race stage 1.

position in the peloton, as well as the position of their teammates, and relating this
information to tactics and energy levels. The network also shows a focus on the
protected riders from other teams in terms of what they were doing and their posi-
tion in the peloton throughout the race. The constant monitoring of own and other
riders’ energy levels was reported by all riders as important throughout the race and
is shown through the nodes related to the ‘energy’ node. Nodes related to the team’s
race plan were also prominent. Finally, it is interesting to see safety-related nodes
in the network, including the nodes ‘crashes’, ‘weather’, ‘conditions’ and ‘pot holes’.

CONCLUSION
What Does the Analysis Tell Us About DSA and
Teamwork in Elite Women’s Cycling Teams?
The analysis highlights the high requirement for teamwork during elite women’s rac-
ing. Salas et al.’s (2005) ‘Big Five’ model of teamwork identifies five key teamwork
164 Systems Thinking in Practice

Points
To go
Level

Computer QOM

Sprint
Distances
Energy
Descent Conser-
vation
Speed Main
Handle group
Chase
bars
Power
Use
Time Peloton
gap Win Breaks
Other
groups
Tactics Crashes
Attacks Front
Protected
rider
Goals
Plan Course
Protect
Own Position
Support team
vehicle Behind

Riders Weather
Conditions
Form
Skill Experience
GC level
riders Bike
Pot holes
Other
Tyres
Fatigue teams

Training
Working
Contenders together
Brakes

FIGURE 8.3 Situation awareness network for Battle on the Border road race stage 1.

behaviours (leadership, mutual performance monitoring, back-up behaviour, adapt-

ability, team orientation) along with three key supporting mechanisms (shared
mental models, mutual trust, communication). Throughout the three networks, it is
possible to identify clear examples of key teamwork behaviours as outlined in the
‘Big Five’ model (see Table 8.1).
In terms of DSA, the networks demonstrate that the overall peloton possesses
DSA that comprises the situation awareness of individual riders, riding teams, the
peloton as a collective and the convoy and its elements (e.g. the Commissaire, race
officials, medical support, neutral spares, the media and team support vehicles fol-
low behind the peloton). In addition, the analysis shows that situation awareness is
held by both human and non-human agents. For example, the situation awareness
network shows the importance of non-human agents, such as the bike-mounted
computer and annotated handlebars. Here, the DS annotates the bicycle handlebars
to provide critical race information to the riders, such as when the sprint section
of the race will occur. Riders then used their race computers in conjunction with
EAST in Elite Women’s Cycling Teams 165

TABLE 8.1
Examples of ‘Big Five’ Teamwork Behaviours

‘Big Five’ Model

Behaviour Example Source
Leadership The tactical direction provided by both riders within the Task and situation
team and the DS awareness network
Mutual performance Riders need to monitor their own teammates in terms of Task and situation
monitoring position, fatigue, etc. awareness network
Back-up behaviour The ability of riders to respond to the needs of other Task and situation
riders in the team, e.g. chasing down a breakaway for awareness network
the protected rider
Adaptability The need for the riders to respond to other teams’ Task and situation
behaviour, e.g. breaks from the peloton, attacks awareness network
Team orientation The fact that all riders in the team are working together Task and situation
to ensure that the protected rider wins the race (team awareness network
goal overriding individual goals)
Shared mental The use of race planning to ensure that the teams have a Task and situation
models shared understanding of the plan and of the team’s awareness network
goals
Mutual trust The belief among the riders that team members will Task and situation
perform their roles, e.g. the protected rider’s awareness network
knowledge that her teammates will sacrifice their own
overall position and protect her until the sprint finish
Communication Multiple communications occurring continuously Social and task
between all riders network

the information written on the handlebars to be aware of when to attack or initiate

a sprint.
The importance of continual transactions of situation awareness within the
peloton was also emphasised, as is the fact that the transactions can be both verbal
(e.g. talking to a teammate) and non-verbal (monitoring another rider’s behaviour
to assess their energy level). Notably, these transactions include intra-team trans-
actions (exchanges of awareness between teammates) and inter-team transactions
(exchanges of awareness between riders from different teams).

What Are the Implications for Optimising Performance

in Elite Women’s Cycling Teams?

Although the analysis presented was exploratory in nature, it is possible to identify

avenues that could be used to optimise race performance. The importance of pre-
race planning and the implementation of the plan itself during the race was empha-
sised throughout the present analysis, with race planning identified as a key node
within the task network and information related to the race plan identified as a key
node within the situation awareness network. This suggests that robust, in-depth and
166 Systems Thinking in Practice

detailed planning processes are likely to yield improvements in team performance,

and that communication of the plan is critical, both before and throughout the race.
Further, the importance of devising a race plan that is agile and adaptable and con-
tains a series of well-defined contingency plans is apparent. This will enable teams
to respond appropriately and rapidly in the face of variable tactics from opposing
teams.
The importance of transactions in awareness was also highlighted; it may be that
educating cyclists on forms of non-verbal transaction may provide ways in which
teammates can communicate without giving away tactics to other teams. Further,
behaviours designed to mask key information relating to opposition team’s situation
awareness may be useful to create situation awareness decrements, for example, dis-
guising high levels of fatigue and sending out spurious information regarding race
tactics. Finally, teaching cyclists ways in which to identify other teams’ tactics and
energy levels will be useful and serve to optimise team situation awareness.
A final interesting feature of the analysis is the importance of communication
within the team, even when team riders may be spread throughout the peloton
without the ability to communicate directly (radio communications are not used in
Australian elite women’s cycling). This suggests that future work may be required to
investigate strategies for rapid and effective communications during races.

What Are the Implications for Future EAST Applications in Elite Sport?
The analysis presented provides some evidence of the utility of applying EAST in the
elite sports context. EAST provided new perspectives on elite cycling, outlining the
key tasks, interactions and information required during elite women’s cycling road
races. Further applications of the EAST framework are encouraged, both in cycling
and in other forms of elite team sports such as football, rugby league and union,
Australian Rules football, American football, Formula 1 racing and baseball. In
addition, formal comparisons between EAST and other existing analysis approaches
are recommended.

REFERENCES
De Keukelaere, C., Kermarrec, G., Bossard, C., Pasco, D. and De Loor, P. (2013). Formes,
contenus et évolution du partage au sein d’une équipe de sport de haut niveau. Le travail
humain, 76 (3), 227–255.
Klein, G. A., Calderwood, R., and MacGregor, D. (1989). Critical decision method for elicit-
ing knowledge. IEEE Transactions on Systems, Man and Cybernetics, 19 (3), 462–472.
Macquet, A. C. and Stanton, N. A. (2014). Do the coach and athlete have the same ‘Picture’
of the situation? Distributed situation awareness in an elite sport context. Applied
Ergonomics, 45 (3), 724–733.
Neville, T. and Salmon, P. M. (2016). Never blame the umpire? A review of situation awareness
models and methods for examining the performance of officials in sport. Ergonomics,
59 (7), 962–975.
Neville, T., Salmon, P. M., Read, G. J. M. and Kollinitas, A. (2016). Play on or call a foul?
Testing and extending DSA theory through sports officiating. TIES, 17 (1), 80–103.
Salas, E., Sims, D. E. and Burke, C. S. (2005). Is there a ‘Big Five’ in teamwork? Small Group
Research, 36 (5), 555–599.
EAST in Elite Women’s Cycling Teams 167

Salmon, P. M., Stanton, N. A., Gibbon, A. C., Jenkins, D. P. and Walker, G. H. (2010). Human
Factors Methods and Sports Science: A Practical Guide. Taylor & Francis: Boca
Raton, FL.
Salmon, P. M, Stanton, N. A., Walker, G. H., Baber, C., Jenkins, D. P. and McMaster, R.
(2008). What really is going on? Review of situation awareness models for individuals
and teams. Theoretical Issues in Ergonomics Science, 9 (4), 297–323.
Stanton, N. A., Salmon, P. M., Rafferty, L., Walker, G. H., Jenkins, D. P. and Baber, C. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design. 2nd Edition,
Ashgate: Aldershot, UK.
Stanton, N. A., Salmon, P. M., Walker, G. H. (2015). Let the reader decide. A paradigm shift
for situation awareness in socio-technical systems. Journal of Cognitive Engineering
and Decision making, 9:1, 45–50.
Stanton, N. A., Salmon, P. M., Walker, G. H., Salas, E., Hancock, P. (2017). State-of-the-
science: Situation Awareness Models. Ergonomics. 60(4): 449–466.
Trenchard, H. (2015). The peloton superorganism and protocooperative behavior. Applied
Mathematics and Computation, 270, 179–192.
Trenchard, H., Richardson, A., Ratamero, E. and Perc, M. (2014). Collective behavior and the
identification of phases in bicycle pelotons. Physica A: Statistical Mechanics and its
Applications, 405, 92–103.
9 EAST in Automated
Driving Systems
With Victoria A. Banks

INTRODUCTION
There have been rapid developments in automotive automation over the past decade
whereby partial and fully automated vehicles are reaching a point where widespread
deployment is feasible (National Highway Traffic Safety Administration; NHTSA
2016). It is expected that by the end of 2016, NHTSA will publish best-practice
guidelines on establishing principles of safe operation for vehicles operating at Level
4 ‘full’ automation. Fully autonomous cars promise to deliver abundant socioeco-
nomic advantages (Casner et al. 2016), including improvements to traffic flow and
mobility and significant improvements to road safety. With 90% of accidents being
attributable to driver error (Smiley and Brookhuis 1987; Stanton and Salmon 2009),
vehicle manufacturers may have good reason to remove drivers from active con-
trol. There are, of course, other benefits of ‘driverless’ vehicles, one being to give
vehicle occupants time to engage in other tasks not related to driving (Fagnant and
Kockelman 2015). This is seen as one of the main drivers for market implementation
to improve comfort and convenience. According to the Department for Transport
(2015), UK drivers spend, on average, 235 hours a year behind the wheel. This
equates to approximately six working weeks where the driver has no spare capacity
to engage in other tasks. The advent of fully automated vehicles could therefore com-
pletely transform our experience of driving and provide the driver with additional
productive time (making the journey similar to taking public transport – without the
drawbacks).
The human factors issues pertaining to vehicle automation have been speculated
about since the 1970s (Sheridan 1970). Simply removing the driver from the control-
feedback loop and eliminating their responsibility over safe vehicle operation does
not, however, render human factors completely redundant. Instead, vehicles oper-
ating at increased levels of autonomy with ‘self-driving’ capabilities open up new
avenues of investigation. Some of these avenues are explored within this chapter.

Levels of Automation and the Role of the Driver

Automation of the driving task brings with it a shift in the role and responsibilities of
the human driver. Within the literature, Kaber and Endsley (2004) captured this idea
eloquently by explaining that the role of the driver shifts from one of active operation
to more of a passive monitor as the level of automation increases. Within industrial

169
170 Systems Thinking in Practice

practice, the role and responsibilities of the driver are alluded to, but not explicitly
described, in automation taxonomies. The Society for Automotive Engineers (SAE
2015), NHTSA (2013) and BASt Expert Group (Gasser 2014) have all developed
their own versions of automated driving taxonomies. These are often used inter-
changeably within the literature, which can lead to confusion about what the driver
can and cannot do under different levels of automation within the driving system.
SAE went some way in trying to standardise these descriptions (SAE J3016). Even
so, in any instance whereby a take-over request is issued or indeed possible, the
driver is expected to be able to regain control of the vehicle and resume their tradi-
tional driving role. As long as the driver remains in the control-feedback loop, which
would be the case for any system whereby control transitions may be made between
the automated system and driver, an acknowledgement of their role and how they
can be supported back into their traditional driving role remains an important area
of investigation (Eriksson and Stanton 2017). Thus, we cannot assume that at higher
levels of automation, the driver or human element will no longer be required. This is
because any potential for a take-over request to be issued implies that (1) the vehicle
does not offer full ‘self-driving’ functionality and (2) the driver should be able to
regain control of the vehicle as may be required by the situation and/or operational
limits of the system.
We also cannot overlook the fact that individuals may even want to resume con-
trol from the vehicle at some point. Thus, increasing levels of autonomy in driving
do not eliminate all of the human factors issues that are typically associated with
lower levels, for example, Levels 2 and 3. We already know from the literature
that automation within the driving task can lead to decreased situation awareness
(Stanton and Young 2005; Stanton et al. 2011), erratic changes to driver workload
(Stanton et al. 1997; Young and Stanton 2002, 2004; de Winter et al. 2014, 2016),
skill degradation (Stanton and Marsden 1996) and issues relating to trust (Walker
et al. 2016), overreliance and complacency (Stanton 2015). It seems likely that
some, if not all, of these will remain enduring challenges for systems designers as
long as the driver remains within the control-feedback loop to some extent. This
means that strategies for transferring control back to the driver after prolonged
periods of autonomous driving must be carefully designed to ensure that they have
appropriateness levels of situation awareness and are deemed capable of regaining
control of the vehicle. The entire spectrum of driver responsibilities and workload
should be considered in delivering an effective ‘hand-over’ between the driver and
autonomous vehicle.
With this in mind, it is becoming increasingly important to acknowledge the role
of the driver within an automated driving system. There are many lessons that can
be taken from the field of aviation, as we increasingly see the role of the driver
becoming analogous to the role of a pilot (Stanton and Marsden 1996). In aviation,
Hutchins (1995) described two roles in which the pilot can serve; Pilot Flying (PF)
and Pilot Not Flying (PNF). Whilst the PF is responsible for overall control of the
plane, the PNF is responsible for communicating with Air Traffic Control and air-
craft systems as well as completing all of the checklists that are required during each
phase of flight. Thus, the burden of responsibility simply changes rather than being
reduced. In recognition of the changing responsibilities of the PNF, the Federal
EAST in Automated Driving Systems 171

FIGURE 9.1 Driver mode transition network.

Aviation Administration (FAA 2003) altered the terminology of PNF to that of Pilot
Monitor (PM). In the same way, a Driver Driving (DD) would be responsible for
overall control of the vehicle and represents the traditional role of the driver, whilst a
Driver Monitor (DM) would assume a similar role to that of PM. The latter role is to
monitor the behaviour of the vehicle and automated sub-systems to ensure that safe
and normal driving practice is maintained. However, the introduction of automa-
tion into the driving task does not necessarily mean that the driver will assume the
role of DM. We already know that the perception of increased reliability can leave
drivers vulnerable to becoming complacent and over-reliant on automated function-
ality (Parasuraman et al. 1993; Lee and See 2004). Without active vehicle control,
a DM, for example, could become vulnerable to the onset of boredom or fatigue
(e.g. Molloy and Parasuraman 1996; Stanton et al. 1997; Young and Stanton 2002).
Thus, DMs may unintentionally drift in and out of the Driver Not Driving (DND)
role. This becomes particularly problematic in instances whereby the automated sys-
tems are not able to adequately resolve a scenario without human intervention. If
the driver should be in the role of DM (i.e. during partial automation) but is in fact
behaving in a manner more akin to the role of DND, active control of the vehicle may
be transferred to a DND (rather than a DM who is prepared to resume the role of DD)
who may fail to respond appropriately due to a sudden increase in driver workload,
reduced situation awareness (Dozza 2012) or startle (Sarter et al. 1997) (Figure 9.1).
We therefore need to better understand the impact that automation implementation
has upon the role of the driver. One way to do this is to adopt the theoretical under-
pinnings of Distributed Cognition (DCOG), which acknowledges that both human
and non-human agents can work together in pursuit of a common goal (Hutchins
1995). In driving, this essentially reflects the relationship between the human driver
and automated sub-systems that become increasingly capable of performing the tra-
ditional roles of the driver (Banks et al. 2014). The DCOG paradigm provides the
necessary foundations and methods to explore how the role of the driver changes
within the complex sociotechnical systems in which they are involved (Walker et al.
2010; Walker et al. 2015).

METHOD
This chapter explores the changing role of the driver within automated driving sys-
tems using the Event Analysis of Systemic Teamwork (EAST) Stanton et al. 2006,
2013; Walker et al. 2006, 2010) framework. EAST is a descriptive method and
172 Systems Thinking in Practice

proposes that the performance of a system can be described using three interlinked
network representations: task, social and information (Walker et al. 2006, 2010).
Whilst it was originally developed for understanding command and control activi-
ties, it has since been applied to transportation domains including aviation (Walker
et al. 2010; Stanton and Harvey 2017), rail (Walker et al. 2006), driving (Banks
and Stanton 2016) and maritime transport (Stanton et al. 2006; Stanton 2014; Baber
et al. 2013).
Task networks are used to provide a summary of system goals (Salmon et al. 2014)
and offer a description of the sequences and interdependencies that exist between
individual sub-tasks that must be completed to attain these goals. Social networks
are used to analyse the structure of the system in terms of the communications that
take place between different system ‘agents’. Finally, information networks show the
information that is used and communicated by system agents during a task (Walker
et al. 2010). They therefore detail aspects of communication that underpin the com-
pletion of a task and the relationships between informational nodes. This chapter
uses the representations afforded by EAST to explore the differences between the
DD and DM roles using network metrics. These networks can then be subjected to
quantitative analysis using the Applied Graphic and Network Analyses (AGNA) tool
(AGNA version 2.1; Benta 2005). AGNA is a platform-independent freeware appli-
cation that can be used to analyse social networks. Nodes within the network can be
analysed individually to assess agent centrality/prominence or as a whole. In driving
research, network metrics can be used to identify key agents, key tasks and key infor-
mational elements required to complete the task. The following metrics have been
chosen due to their previous application to the driving task (see Banks and Stanton
2016; Salmon et al. 2009; Walker et al. 2011).
Density represents the level of interconnectivity between system agents. It is
expressed as a value between 0 and 1, where 0 represents a network that has no con-
nections between system agents and 1 indicates that the network is fully connected
(Kakimoto et al. 2006). It is calculated using the formula

Network density = 2e /n ( n − 1)

where:
e is the number of links in the network
n is the number of information elements within the network (Walker et al. 2012)

Diameter is used to analyse the connections and pathways between different sys-
tem agents within the network (Walker et al. 2011). Denser networks (i.e. the route
through the network is shorter and more direct) have smaller values. It is calculated
using the following formula:

Diameter = max uy d ( ni , n j )

Cohesion represents the number of reciprocal connections divided by the total num-
ber of possible connections (Stanton 2014).
EAST in Automated Driving Systems 173

Sociometric status is also of interest because it gives an indication of agent promi-

nence as a communicator. Agents with high sociometric values are highly connected
(Salmon et al. 2012). It is calculated using the formula

∑
1 g
Sociometric Status =
( g − 1) j =1
( x ji , xij )
where:
g is the total number of nodes in the network
i and j are individual nodes and are the edge values from node i to node j
(Salmon et al. 2012)

RESULTS
Task Networks
The driving task itself is said to be made up of approximately 2500 subtasks (see
Hierarchical Task Analysis of Driving, HTAoD; Walker et al. 2015). These are
divided into pre-drive, basic, operational, tactical, strategic and post-drive tasks.
Whilst a DD would be involved in all of these tasks, the implementation of auto-
mation into the driving task would see the role of the human operator reduced. For
this reason, the analysis focuses upon some of the basic, and strategic sub-tasks of
driving, as it is in these categories that the changing role of the driver is most vis-
ible. Table 9.1 shows the terminologies, taken directly from the HTAoD (Walker
et al. 2015) and categorised into task type, that were chosen to construct the task
networks presented in this chapter. Notably, both networks should be considered as
continuous processes that are performed as long as the vehicle is in operation. Figure
9.2a reflects the linkages between these sub-tasks of driving from the perspective
of a DD. A DD would be physically be responsible for completing all of the basic
sub-tasks of driving as well as being solely responsible for completing the strategic
sub-tasks. However, as automation is introduced into the driving task, the task net-
work evolves. Figure 9.2b represents the task network from the perspective of a DM.
Whilst the tasks within the network remain largely the same, the emphasis upon

TABLE 9.1
Basic and Strategic Sub-Tasks of Driving

Basic Tasks Strategic Tasks

Perform steering actions Perform surveillance
Control vehicle speed Perform navigation
Undertake directional control Comply with rules
Negotiate bends Respond to environmental conditions
Negotiate gradients

Source: Walker, G. H., et al., Human Factors in Automotive

Engineering and Technology, Ashgate, Aldershot, UK, 2015.
174 Systems Thinking in Practice

(a)

(b)

FIGURE 9.2 Task networks for (a) the role of DD and (b) the role of DM.
EAST in Automated Driving Systems 175

TABLE 9.2
Description of DD and DM Task Networks

DD DM
Number of nodes 9 10
Number of edges 144 160

physical input is removed. Instead, a DM would be responsible for monitoring the

completion of basic tasks as well as strategic tasks. There is therefore no removal of
tasks within the DM task network; the burden of responsibility simply changes from
a physical input to a monitoring input.
Analysis of the whole networks, given their all-connected nature (Leavitt 1951),
render more sophisticated network metrics redundant as they would elicit the same
score. For this reason, only a basic description of the networks is presented in Table
9.2. It shows that whilst one additional task is added to the DM task network (‘Prepare
to take back control’), in comparison to the DD network, the number of edges increases
by 26. This suggests that the level of task demand actually increases in a system
whereby the driver adopts the role of DM. This is supported by empirical literature
on mental workload, showing greater demand for semi-automated driving (Stanton
et al. 2007; de Winter et al. 2014) Therefore, we must acknowledge that the DM role is
likely to be more mentally demanding than the DD role – by way of an analogy, moni-
toring another driver (particularly a learner driver where there is a real possibility of
having to intervene) can be more demanding than driving oneself. This is essentially
what the DM role consists of – monitoring the behaviour of the automated system,
monitoring the road environment and anticipating the behaviour of other road users
as well as being prepared to regain control of the vehicle if necessary.

Social Networks
In order to further our understanding of how the role of the driver affects network
dynamism, high-level social networks were constructed. The authors identified five
system agents relating to the DD network (Figure 9.3a), whilst eight system agents
were identified for a DM social network (Figure 9.3b). The three additional agents
within the DM network represent nodes associated with automated functionality.
Banks and Stanton (2016) identified the Longitudinal and Lateral Controllers as sep-
arate system agents, given their differing capabilities. The same viewpoint has been
taken in the social network representation shown in Figure 9.3b.
Taken at face value, it is clear that the social network associated with the role of
DM is more complex than that of DD. This is likely to be attributed to the increased
communication and coordination that is required within the system network to main-
tain the goal of safe and normal driving practice following automation implementa-
tion. Thus, the transition between DD and DM appears to bring with it a change in
overall system dynamism.
176 Systems Thinking in Practice

(a) (b)

FIGURE 9.3 Social network diagrams for (a) the role of DD and (b) the role of DM.

These social network representations were transformed into association matrices

to enable further analysis using network metrics. Both networks can be defined as
binary (i.e. they can be represented by a zero-one matrix) and non-symmetric (i.e.
directed). The results of the social network analysis using network density, diameter
and cohesion calculation are shown in Table 9.3. The results show that the DM social
network is denser than the DD network. This is unsurprising, considering that the
DM role is only made possible by the introduction of automation into the driving
system. This brings with it the creation of new links and interactions between system
agents. The analysis also revealed that the DM social network is more cohesive than
the DD social network. No difference was found in network diameter between the
DD and DM social networks.
In addition, sociometric status was calculated for both DD and DM social networks
(see Table 9.4). The identification of key agents (denoted by asterisks) is based upon the
rule that any value above the mean sociometric status value reflects network dominance
(Salmon et al. 2009). Unsurprisingly, the DD role is considered to be the most dominant

TABLE 9.3
Basic Description of the DD and DM Social Networks

DD DM
Number of nodes 5 8
Number of edges 10 37
Network density 0.50 0.70
Network diameter 2.00 2.00
Network cohesion 0.30 0.54
EAST in Automated Driving Systems 177

TABLE 9.4
Sociometric Status for the Roles of DD and DM

Node DD DM
Driver 1.75 a 1.57a
Vehicle 1.00 1.00
Vehicle displays 0.75 1.00
Road users 0.75 1.14
Road environment 0.75 1.29
Automated system – 2.00a
Longitudinal controller – 1.14
Lateral controller – 1.00
Mean 1.00 1.32
a Key system agents based upon the rule that any value above the
mean sociometric status value reflects dominance (From Salmon,
P. M. et al., Distributed Situation Awareness: Advances in Theory,
Measurement and Application to Teamwork, Ashgate, Aldershot,
UK, 2009.)

agent in a system whereby full responsibility of the driving task lies with the human
operator. In an automated system where the driver adopts the role of DM, both the
DM and automated system become the most prominent system agents (see Table 9.4).
Notably, the automated system ranks highest in sociometric status, which is unsurpris-
ing given its role within the system of vehicle operation. Even so, this representation
demonstrates that as long as the driver adheres to their responsibilities of monitoring
system behaviour, they maintain a central role within the control-feedback loop.
Of course, ensuring that drivers actually adhere to their changing responsibili-
ties is important, because if a driver transitions to a DND, problems may arise. It is
difficult to predict and understand the behaviour of a DND because they could be
engaged in any task of their choosing. For example, let’s assume that a DM becomes
engaged in a non-driving-related secondary task that keeps their eyes averted from
the road environment for approximately 26 seconds (Eriksson and Stanton 2017).
This essentially would signal the breaking of links between the driver, vehicle dis-
plays, road users and road environment within Figure 9.3b. A ‘broken-links’ analysis
can be used to demonstrate how momentary lapses in efficient monitoring could
impact upon overall network dynamism. This approach essentially represents com-
munication breakdowns between system agents (Stanton and Harvey 2017) simi-
lar to that of removing nodes from the network altogether (e.g. Baber et al. 2013).
‘Broken links’ have typically been used by EAST analysts to analyse accidents post
event to explore vulnerabilities within the system (e.g. Griffin et al. 2010; Rafferty
et al. 2012). In this instance, reciprocal relationships occurring between the DM and
other social agents, shown in Figure 9.3b, are rescinded. The consequences of this
are clearly shown in the results of a social network analysis (see Tables 9.5 and
9.6). As is to be expected, network density and cohesion is rapidly reduced. The
178 Systems Thinking in Practice

TABLE 9.5
Results of a ‘Broken-Links’ Analysis Whereby Links Are
Broken Between the DM, Vehicle Displays, Road Users and
Environment

DM DM (Broken Links Indicative of

(Complete Network) DND Mode Transition)
Number of nodes 8 8
Number of edges 37 30
Network density 0.70 0.54
Network diameter 2.00 2.00
Network cohesion 0.54 0.39

TABLE 9.6
Sociometric Status in a ‘Broken Links’ Network

prominence of the driver within this network also reduces rapidly (see Table 9.6).
This places further pressure on systems designers to ensure that their systems are
safe, reliable and failsafe, especially given the prominence of the automated system
and vehicle under these circumstances. Providing that remaining system agents have
the information required to function effectively and are able to use this information
to perform appropriately in any given context, safety should be maintained (Salmon
et al. 2012) even if the DM transitions to a DND for short periods. There is evidence
that individual agents are able to compensate for each other, enabling the system to
maintain safe operation (Stanton et al. 2006; Stanton 2016). For example, if a driver
fails to respond to system warnings, some autonomous systems can bring the vehi-
cle to a ‘safe stop’ without any human input. Similarly, an Autonomous Emergency
Brake system operates in the background and will only activate if activation thresh-
olds are met. However, this is only possible if system agents have all of the informa-
tion required. This implicates the ‘information network’.
EAST in Automated Driving Systems 179

Information Networks
According to Walker et al. (2010), information networks show the information
that is used by and communicated by agents during a task. Their work specifi-
cally explored systemic situation awareness of air traffic control using observa-
tional notes, interview and live audio feeds. The following information networks
were created by analysing driver verbalisations collected as part of another study
by Banks and Stanton (2015), which utilised the Critical Decision Method (Klein
and Armstrong 2005) to elicit information relating to driver decision making. In
total, 48 verbal transcripts were analysed to construct the DD and DM informa-
tion networks. The DD information network was constructed using evidence from
verbal transcripts collected during manual driving (i.e. no automation), whilst the
DM information network was constructed using evidence from verbal transcripts
collected during a drive in which the basic and strategic sub-tasks of driving
(Table 9.1) were automated.
Each ‘informational node’ within the transcripts was identified and paired with
its closest relation (i.e. nodes from the same sentence). This strategy was used
to build up a network of information concepts relating to the driving task. The
network representations were presented to subject matter experts for verification.
Figure 9.4 presents a schematic overview of the DD information network, includ-
ing 42 nodes and 46 connections. Sociometric status was calculated to identify key
concepts. The results indicate that for a DD, the following information nodes were
most prominent:

1. Traffic type: Includes vehicles, pedestrians and public transport or services.

2. Traffic: Considers the properties of road users, such as speed and route.
3. Infrastructure: Considers physical aspects such as road type, capacity and
lane markings.

Approaching

Traffic priority
Capacity Exiting Positive
Cyclists Roadworks instruction Warnings
Pedestrians Location
Junctions
Signage
Direction
Lane position Information
Motorcyclists Junction type
Traffic
Traffic type Infrastructure management Speed limits
systems
Traffic Lane markings
Public transport Road type Traffic lights
Vehicle Speed
or services

Impact on
others Colour
Route Traffic
behaviour Time
demands
Other
Risk assesment Location of
Hazard type hazard

Previous
Location Environmental
experience conditions Weather

Severity
Visibility

FIGURE 9.4 Schematic overview of the DD information network.

180 Systems Thinking in Practice

4. Risk assessment: Includes hazard type, previous experience and environ-

mental conditions.
5. Signage: Considers the meaning and information presented on road signs
including speed, warnings and instruction.

The informational nodes coming off these key nodes within Figure 9.4 provide greater
detail pertaining to specific situations and contexts within driving (Stanton 2013).
In contrast, the DM information network, presented in Figure 9.5, includes 53
nodes and 61 connections. Sociometric status was once again calculated and revealed
that in addition to the key nodes outlined above for a DD, a DM must also consider
the following:

1. Autonomous driving features: Consider system limits and system-con-

trolled responses.
2. System mode: Considers the state of the system, either active or inactive.
3. Feedback: Includes the modalities in which information is fed back to the
driver.

Thus, despite the intention of autonomous driving features being to reduce the burden
of responsibility placed upon the driver, it actually increases the amount of information
that a driver must consider in order to maintain safe driving practice. In order to further
understand how this additional information requirement impacts upon network dyna-
mism, the information networks as presented in Figures 9.4 and 9.5 were further anal-
ysed using the AGNA software tool. The results of this analysis are shown in Table 9.7.

Approaching

Traffic priority
Capacity Exiting Positive
Cyclists Roadworks instruction Warnings

Pedestrians Location
Junctions
Signage
Direction
Lane position Information
Motorcyclists Junction type
Traffic
Traffic type Infrastructure management Speed limits
systems
Traffic Lane markings
Public transport Roadtype Traffic lights
Vehicles Speed
or services

Impact on
others Colour
Route Traffic
behaviour Time demands
Other
Risk assesment Location of
System hazard
controlled Hazard type
System limits response
Previous
Automated Location Environmental
experience Weather
driving features conditions
Severity

Take over Visibility

requests System mode
Inactive

Feedback Visual Active

Haptic
Auditory

FIGURE 9.5 Schematic overview of the DM information network (grey boxes indicate
‘new’ nodes within the information network as a result of automation implementation).
EAST in Automated Driving Systems 181

TABLE 9.7
Information Network Metrics for the
Roles of DD and DM

Metric DD DM
Number of nodes 42 53
Number of edges 84 112
Network density 0.05 0.04
Network diameter 6.00 7.00
Network cohesion 0.04 0.03

DISCUSSION
This chapter shows how the application of EAST, and the representations afforded it,
can be used to analyse the changing role of the driver. The application of quantitative
network metrics to analyse these networks enables us to see how the changing role of
the driver impacts upon overall system structure. Importantly, these representations
are intended to provide a foundation to the discussion surrounding the role of the
driver within automated driving systems. The findings do go a long way to explain
why the empirical research shows high levels of mental workload with driving auto-
mation that requires constant monitoring (Stanton et al. 1997; Young and Stanton
2004; de Winter et al. 2014). It is a paradox that this level of automation actually
results in more, rather than less, work for the driver (Stanton 2015).
From a theoretical modelling perspective, the DND role is particularly problem-
atic. Driver disengagement from the primary driving task is a serious concern. The
literature is cluttered with instances whereby engagement in secondary tasks can
lead to performance decrements (Eriksson and Stanton 2017). For example, Merat
et al. (2014) found that if an automated system disengaged at the point where the
driver’s attention was diverted away from the road centre, resumption of manual con-
trol was erratic for up to 40 seconds after the transfer of control. The main concern
for the DND role is that it is not possible to construct a task, social or information
network for the DND role. This is because drivers are free to participate in any task
of their choosing. This makes it difficult to predict and understand the behaviour of a
DND. A ‘broken-links’ analysis, however, points to a dramatic shift in system struc-
ture when a DM transitions to a DND role. Whilst some may argue that the DM role
becomes redundant at higher levels of autonomy (e.g. Gasser 2014), the authors cau-
tion that this is not strictly the case. The Department for Transport’s report (2015),
for example, recognizes that some Level 4 systems may still offer a full set of con-
trols that enable manual driving. This means that at some point, the DND could be
required to regain control of the vehicle, whether this be due to a ‘forced’ transfer of
control due to some form of mechanical failure (e.g. sensor failure), through choice
(e.g. the driver may want to abort or change the destination of travel or they may sim-
ply want to be in control) or simply because autonomous driving features only oper-
ate in some driving modes at Level 4. For this reason, the DND will need to adopt
182 Systems Thinking in Practice

the role of DM during the exchange of control between them and the autonomous
vehicle. The success of this transition of control will, however, be based upon a num-
ber of interacting psychological constructs including situation awareness, workload,
trust and skill (Stanton and Young 2000; Heikoop et al. 2016). If vehicle manufactur-
ers are to handle this transition effectively, a greater understanding of how drivers
appraise and make use of higher-level autonomy is needed (Richards and Stedmon
2016). Thus, whilst less emphasis is placed upon the driver as the level of automation
increases, vehicle manufacturers still need to think about ways in which the driver
can be supported if and when they choose to regain control of the vehicle, especially
during early versions of highly automated driving systems. This is because the DD,
DM and DND are closely related and likely to be adopted interchangeably through-
out the duration of a drive, especially during the intermediate phases of automation.
Whilst the DND role represents the aspiration of many OEMs, there are no such
systems that exist on the market today that allow this to happen. If a driver does find
themselves in the role of DND, they should be supported back into the role of DM to
ensure that the overall goal of the system network can be appropriately maintained.
For contemporary vehicle automation systems, the role of DND could be seen as
a form of automation misuse (Parasuraman and Riley 1997) given the functional
limits of automated architecture. Much more research is needed to further popu-
late these network representations for the varying role of the driver to validate the
networks proposed in this chapter. The hypothetical representations presented in
this chapter do, however, provide an avenue for discussion as they represent how an
‘ideal’ network may function. Realistically, however, the authors acknowledge that
prolonged exposure to high levels of automation (and driver inactivity) could result in
issues surrounding boredom or fatigue (e.g. Stanton et al. 1997; Young and Stanton
2002). Strategies to support the role and maintenance of the DM role are therefore
important avenues for further research. We already know from the literature that the
level and type of automation can have a direct effect on driver engagement (Stanton
et al. 1997; Merat et al. 2014), changes to driver workload (Stanton et al. 1997; Young
and Stanton 2002; de Winter et al. 2014), situation awareness (Stanton et al. 2011;
Dozza 2012) and decision making (Banks and Stanton 2015).

CONCLUSIONS
The qualitative models of DD and DM support and explain the findings from empir-
ical studies on driver workload changes with automated systems. The increase in
workload associated with monitoring automated systems may be explained by the
fact that there are simply more tasks to undertake. These qualitative insights are
supplemented with the quantitative network statistics, showing the importance
of the role change when driving with an automated system (i.e. from the human
driver to the computer driver). Modelling future systems is an important step for
Ergonomics as a discipline, as it allows us to anticipate the likely behaviour of
future technologies and the role for humans. In this way, we can compare alter-
native designs to identify which are likely to have the better outcomes. Future
research should seek to extend and validate the models presented in this chap-
ter using empirical data generated from observational studies exploring driver
EAST in Automated Driving Systems 183

behaviour at varying levels of automation. In particular, prolonged exposure to

high levels of autonomy may elicit DND behaviours that enable us to properly
understand the complexities associated with bringing a driver in such a role back
into the control-feedback loop.

ACKNOWLEDGEMENTS
Professor Neville A. Stanton was funded by the European Marie Curie International
Training Network project on the Human Factors of Automated Driving
(PITN-GA-2013-605817) and the EPSRC/JLR TASCC project (Human Interaction:
Designing Autonomy in Vehicles – HI:DAVe – Grant number: EP/N011899/1).

REFERENCES
Baber, C., Stanton, N. A., Atkinson, J., McMaster, R. and Houghton, R. J. (2013). Using
social network analysis and agent-based modelling to explore information flow using
common operational pictures for maritime search and rescue operations. Ergonomics,
56 (6), 889–905.
Banks, V. A. and Stanton, N. A. (2015). Contrasting models of driver behaviour in emer-
gencies using retrospective verbalisations and network analysis. Ergonomics, 58 (8),
1337–1346.
Banks, V.A. and Stanton, N.A. (2016). Driver-centred vehicle automation: Using network
analysis for agent-based modelling of the driver in highly automated driving systems.
Ergonomics, DOI:10.1080/00140139.2016.1146344.
Banks, V. A., Stanton, N. A. and Harvey, C. (2014). Sub-systems on the road to vehicle auto-
mation: Hands and feet free but not ‘mind’ free driving. Safety Science, 62, 505–514.
Benta, M. (2005) Studying communication networks with AGNA 2.1. Cognition Brain
Behaviour, 9, 567–574.
Casner, S. M., Hutchins, E. L. and Norman, D. (2016). The challenges of partially automated
driving. Communications of the ACM, 59 (5), 70–77.
Department for Transport. (2015). The pathway to driverless cars: Summary report and action
plan. Available at: https://www.gov.uk/government/uploads/system/uploads/attach-
ment_data/file/401562/pathway-driverless-cars-summary.pdf [Accessed 19.09.2016].
de Winter, J. C. F., Happee, R., Martens, M. H. and Stanton, N. A. (2014). Effects of adaptive
cruise control and highly automated driving on workload and situation awareness: A
review of the empirical evidence. Transportation Research Part F: Traffic Psychology
and Behaviour, 27, 196–217.
de Winter, J. C. F., Stanton, N. A., Price, J. S. and Mistry, H. (2016). The effects of driving
with different levels of unreliable automation on self-reported workload and secondary
task performance. International Journal of Vehicle Design, 70 (4), 297–324.
Dozza, M. (2012). What factors influence drivers’ response time for evasive maneuvers in real
traffic? Accident Analysis & Prevention, 58, 299–308.
Endsley, M. R. and Kaber, D. B. (1999). Level of automation effects on performance, situa-
tion awareness and workload in a dynamic control task. Ergonomics, 42 (3), 462–492.
Eriksson, A. and Stanton, N. A. (2017). Take-over time in highly automated vehicles:
Transitions to and from manual control. Human Factors, DOI:10.1177/001872081
6685832.
Fagnant, D. J. and Kockelman, K. (2015). Preparing a nation for autonomous vehicles:
Opportunities, barriers and policy recommendations for capitalising on self-driven
vehicles. Transportation Research Part A, 77, 167–181.
184 Systems Thinking in Practice

Federal Aviation Administration. (2003). Standard operating procedures for flight deck crew
members. AC No: 120-71A, 27.2.03.
Gasser, T. (2014). Vehicle automation: Definitions, legal aspects, research needs. UNECE
Workshop: German Federal Highway Research Institute: Towards a new transportation
culture: Technology innovations for safe, efficient and sustainable mobility. Brussels,
17–18 November 2014.
Heikoop, D. D., de Winter, J. C. F., van Arem, B. and Stanton, N. A. (2016). Psychological
constructs in driving automation: A consensus model and critical comment on con-
struct proliferation. Theoretical Issues in Ergonomics Science, 17 (3), 284–303.
Hutchins, E. (1995). How a cockpit remembers its speed. Cognitive Science, 19 (3), 265–288.
Kaber, D. B. and Endsley, M. R. (2004). The effects of level of automation and adaptive auto-
mation on human performance, situation awareness and workload in a dynamic control
task. Theoretical Issues in Ergonomics Science, 5 (2), 113–153.
Kakimoto, T., Kamei, Y., Ohira, M. and Matsumoto, K. (2006). Social network analysis on
communications for knowledge collaboration in OSS communities. In: Proceedings of
The 2nd International Workshop on Supporting Knowledge Collaboration in Software
Development (KCSD’06), pp. 35–41 Tokyo, Japan.
Leavitt, H. J. (1951). Some effects of certain communication patterns on group performance.
Journal of Abnormal and Social Psychology, 46, 38–50.
Lee, J. D. and See, K. A. (2004). Trust in automation: Designing for appropriate reliance.
Human Factors, 46, 50–80.
Merat, N., Jamson, A. H., Lai, F. C., Daly, M. and Carsten, O. M. (2014). Transition to
manual: Driver behaviour when resuming control from a highly automated vehicle.
Transportation Research Part F: Traffic Psychology and Behaviour, 27 (B), 274–282.
National Highway Traffic Safety Administration. (2013). Preliminary Statement of Policy
Concerning Automated Vehicles System. Washington, DC.
National Highway Traffic Safety Administration. (2016). Update to ‘Preliminary Statement
of Policy Concerning Automated Vehicles System’. Available at: http://www.nhtsa.gov/
Research/Crash+Avoidance/Automated+Vehicles [Accessed 19.09.2016].
Nowakowski, C., Shladover, S. E. and Tan, H. S. (2015). Heavy vehicle automation: Human
factors lessons learned. Procedia Manufacturing, 3, 2945–2952.
Parasuraman, R., Molloy, R. and Singh, I. L. (1993). Performance consequences of automa-
tion-induced ‘complacency’. The International Journal of Aviation Psychology, 3, 1–23.
Parasuraman, R. and Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse.
Human Factors: The Journal of the Human Factors and Ergonomics Society, 39 (2),
230–253.
Richards, D. and Stedmon, A. (2016). To delegate or not to delegate: A review of control
frameworks for autonomous cars. Applied Ergonomics, 53 (B), 383–388.
Salmon, P. M., Stanton, N. A. and Young, K. L. (2012). Situation awareness on the road:
Review, theoretical and methodological issues, and future directions. Theoretical
Issues in Ergonomics Science, 13 (4), 472–492.
Salmon, P. M., Stanton, N. A., Walker, G. H. and Jenkins, D. P. (2009). Distributed Situation
Awareness: Advances in Theory, Measurement and Application to Teamwork. Ashgate:
Aldershot, UK.
Salmon, P. M., Walker, G. H. and Stanton, N. A. (2016). Pilot error versus sociotechnical sys-
tems failure: A distributed situation awareness analysis of Air France 447. Theoretical
Issues in Ergonomics Science, 17 (1), 64–79.
Sarter, N. B., Woods, D. D. and Billings, C. E. (1997). Automation surprises. Handbook of
Human Factors and Ergonomics, 2, 1926–1943.
Sheridan, T. B. (1970). Big brother as driver: New demands and problems for the man at the
wheel. Human Factors: The Journal of the Human Factors and Ergonomics Society,
12, 95–101.
EAST in Automated Driving Systems 185

Smiley, A. and Brookhuis, K. A. (1987). Alcohol, drugs and traffic safety. In: J.A. Rothengatter
and R. A. de Bruin (Eds.), Road Users and Traffic Safety (pp. 83–105).Van Gorcum:
Assen, The Netherlands.
Smith, B. W. (2013). SAE levels of driving automation. Available at: http://cyberlaw.stanford.
edu.loda [Accessed 11.04.2016].
Society for Automotive Engineers (2015). Taxonomy and definitions for terms related to on-
road motor vehicle automated driving systems. Available at: http://standards.sae.org/
j3016_201401/. [Accessed 12.10.2015].
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57 (3), 403–418.
Stanton, N. A. (2015). Responses to autonomous vehicles. Ingenia, 62, 9.
Stanton, N. A. (2016). Distributed situation awareness. Theoretical Issues in Ergonomics
Science, 17 (1), 1–7.
Stanton, N. A., Dunoyer, A. and Leatherland, A. (2011). Detection of new in-path targets by
drivers using stop and go adaptive cruise control. Applied Ergonomics, 42 (4), 592–601.
Stanton, N. A. and Marsden, P. (1996). From fly-by-wire to drive-by-wire: Safety implications
of automation in vehicles. Safety Science, 24 (1), 35–49.
Stanton, N. A. and Harvey, C. (2017). Beyond human error taxonomies in assessment of risk
in sociotechnical systems: A new paradigm with the EAST ‘broken-links’ approach.
Ergonomics, 60 (2) 221–233.
Stanton, N. A. and Salmon, P. M. (2009). Human error taxonomies applied to driving: A
generic driver error taxonomy and its implications for intelligent transport systems.
Safety Science, 47 (2), 227–237.
Stanton, N. A., Salmon, P. M, Rafferty, L., Walker, G. H., Baber, C. and Jenkins, D. P. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design. Ashgate:
Aldershot, UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
P., Hoyle, G., Walker, G., Young, M. S., Linsell, M., Dymott, R. and Green, D. (2006).
Distributed situation awareness in dynamic systems: Theoretical development and
application of an ergonomics methodology. Ergonomics, 49 (12–13), 1288–1311.
Stanton, N. A. and Young, M. (2000). A proposed psychological model of driving automation.
Theoretical Issues in Ergonomics Science, 1 (4), 315–331.
Stanton, N. A. and Young, M. (2005). Driver behaviour with adaptive cruise control.
Ergonomics, 48, 1294–1313.
Stanton, N. A., Young, M. and McCaulder, B. (1997). Drive-by-wire: The case of mental
workload and the ability of the driver to reclaim control. Safety Science, 27 (2–3),
149–159.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. and Green, D. (2006). Event
analysis of systematic teamwork (EAST): A novel integration of ergonomic methods to
analyse C4i activity. Ergonomics, 49 (12–13), 1345–1369.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. and Jenkins, D.
(2010). From ethnography to the EAST method: A tractable approach for representing
distributed cognition in Air Traffic Control. Ergonomics, 53 (2), 184–197.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2011). Cognitive compatibility of motorcy-
clists and car drivers. Accident Analysis & Prevention, 43, 878–888.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2015). Human Factors in Automotive
Engineering and Technology. Ashgate: Aldershot, UK.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2016). Trust in vehicle technology.
International Journal of Vehicle Design, 70 (2), 157–182.
Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D. P., Monnan, S. and Handy, S. (2012).
Communications and cohesion: a comparison between two command and control para-
digms, Theoretical Issues in Ergonomics Science, 13(5), 508–527.
186 Systems Thinking in Practice

Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D. P. and Rafferty, L. A. (2010).
Translating concepts of complexity to the field of ergonomics. Ergonomics, 53 (10),
1175–1186.
Young, M. S. and Stanton, N. A. (2002). Malleable attentional resources theory: A new expla-
nation for the effects of mental underload on performance. Human Factors, 44 (3),
365–375.
Young, M. S. and Stanton, N. A. (2004). Taking the load off: Investigations of how adaptive
cruise control affects mental workload. Ergonomics, 47 (8), 1014–1035.
10 EAST in Future Road
Transportation Systems
With Victoria A. Banks, Gary
Burnett and Setia Hermawati

INTRODUCTION
Over the past two decades, there have been major developments in the integration of
wireless and autonomous technologies in the road transportation network (Talebpour
and Mahmassani 2016). Automated vehicles in particular are quickly becoming an
engineering reality (Stanton 2015), and whilst much research has primarily focused
upon driver–automation interaction (e.g. Banks et al. 2014; Zeeb et al. 2015; Louw
and Merat 2017), many issues remain. Some of these issues relate to how automation
can be regulated, legislated and standardised, but more importantly, we do not fully
understand how automation will impact overall road system behaviour. For example,
Atkins Mobility (2016) speculate that if automation brings about improvements to
road safety, we may see a future wherein crash barriers are no longer necessary
and roadway signs become redundant as information can be shared using Vehicle-
to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication streams and
other location services (e.g. Global Positioning Systems [GPS]). V2V and V2I are
made possible through both the use of Dedicated Short Range Communication
(DSRC) sensors and wireless network architectures such as 5G. In vehicles, DSRC
represents on-board sensor units whilst external roadside units represent the means
to achieve V2I communication. This wireless connectivity would enable intra-vehi-
cle communication as well as real-time communication with traffic management
systems. Essentially, camera and radar-based technologies enable the vehicle to ‘see’
(e.g. vision systems that process video data and fuse with microwave radar data),
whilst technologies such as DSRC enable the car to ‘talk’ (i.e. transmit data to other
vehicles and infrastructure) and ‘listen’ (i.e. receive data from other vehicles and
infrastructure).
The concept of Connected and Autonomous Vehicles (CAV) is not new, with
research and innovation dating back to the early 1990s (De La Fortelle 2005). The
Science and Technology Select Committee (2017) cite numerous benefits associated
with CAV, including increased accessibility and mobility and improvements to road
safety and congestion. KPMG (2015) hypothesise that by 2030, all new vehicles sold
within the United Kingdom will be ‘fully connected’. It is clear, then, that future
transportation systems will be reliant upon the exchange of information between both
human and non-human entities to ensure effective system functioning. This type of

187
188 Systems Thinking in Practice

communicative behaviour is the essence of Distributed Cognition (DCOG) Hutchins

1995), whereby interactions take place between humans, resources and materials
across space and time(Hollan et al. 2000; Hutchins 1995). DCOG is related to the
theory of ‘transactional memory’, whereby individuals (both human and non-human
entities) tend to rely upon others to remember things for them (Stanton et al. 2015).
Thus, DCOG is characterised by multiple system ‘agents’ that work together in order
to achieve a common goal (Hutchins 1995). ‘Agents’ in this sense can receive, hold
and share information with one another in order to pursue a common goal (Clark
1995). This implies the need for communication and coordination to exist between
them (Christoffersen and Woods 2002; Stanton 2014; Eriksson and Stanton 2017). Of
course, there are many challenges associated with Vehicle-to-Anything (V2X) com-
munication, including, but not limited to, issues relating to the size of the network
(both in terms of geography and availability). This means that in some instances, the
exchange of information between system agents (both human and non-human) may
be delayed, inaccurate or incomplete. DCOG provides a theoretical framework that
can be used in the effective design of new communication and interactive technolo-
gies to support the relationship that exists between human and non-human agents by
highlighting areas of potential weakness (Hollan et al. 2000). To date, DCOG has
only been used to consider small sociotechnical systems in isolation, such as an air-
line cockpit (Hutchins 1995) and command teams (Stanton 2014). In this chapter, we
argue that DCOG provides not only the theoretical foundation but also the methods
that can be used to explore complex sociotechnical systems at a macro level.
The road transportation system is a good example of a macro-level sociotechnical
system. This is because it can dynamically configure itself to ensure that multiple
sub-systems acting within it (e.g. Traffic Management Centres [TMC] work alongside
External Agencies [EA]) can operate simultaneously to achieve various functions.
Thus, the network is based upon a large number of complex interactions and interde-
pendencies between multiple system agents at a number of levels (Salmon et al. 2014).
These include system agents within the road environment (RE) (e.g. drivers, pedes-
trians and vehicles), TMC (e.g. traffic management and CCTV operators) and EA
(e.g. radio stations and emergency services). Whilst these categories of system agent
are typically analysed independently, this chapter applies the principles underpinning
DCOG to all of the agents and agencies involved within the road transportation net-
work. Given the uncertainty as to how the transportation network will be affected by
CAV functionality, this chapter provides a comparison between non-CAV and CAV
networks to explore how network dynamism may change as a result of increased con-
nectivity. This comparison is important because it provides initial insights into how
system agents will react to, and interact with, intelligent transportation systems.

METHOD
DCOG in complex sociotechnical systems can be further explored and understood
using the Event Analysis of Systemic Teamwork framework (EAST) Stanton et al.
2008). EAST is a descriptive method that proposes that a system can be described
using three interlinked network representations: task, social and information (Walker
et al. 2006, 2010). Task networks provide analysts with a means to show the processes
EAST in Future Road Transportation Systems 189

involved in attaining network goals (Salmon et al. 2014). They can provide a descrip-
tion of the sequences and interdependencies that exist between individual sub-tasks
that must be completed to attain these goals. Social networks are used to analyse
the structure of the system in terms of the communications that take place between
different system ‘agents’. Finally, information networks show the information that
is used by, and communicated by, system agents during a task (Stanton et al. 2008).
Information networks detail aspects of communication that underpin the completion
of a task as well as the relationships that exist between these different informational
nodes. EAST has been used to focus upon specific tasks within varied domains,
including aviation (Sorensen et al. 2011), rail (Walker et al. 2006), driving (Banks
and Stanton 2016) and maritime transport, (Stanton et al. 2006, 2017a, 2017b; Stanton
2014; Baber et al. 2013), providing meso-level representations of DCOG (Grote et al.
2014). However, this chapter goes further by using the representations afforded by
EAST (Stanton et al. 2008) to explore DCOG at a macro level (Grote et al. 2014).
EAST makes it possible to provide an overview of how different agents and agencies
within the road transportation network can function simultaneously within a shared
space (i.e. the road network). The networks can then be subjected to quantitative
analysis using the Applied Graphic and Network Analysis (AGNA) tool (AGNA,
version 2.1; Benta 2005). AGNA is a platform-independent freeware application that
can be used to analyse task, social and information networks. Nodes within each
network can either be analysed individually to assess agent centrality/prominence
or as a whole to give an overall impression of system complexity. Network metrics
can be used to identify key agents, tasks and informational elements within system
operation. Within driving research, the following network metrics have been applied
to analyse EAST representations:
Density represents the level of interconnectivity between system agents. It is
expressed as a value between 0 and 1, where 0 represents a network that has no con-
nections and 1 indicates that the network is fully connected (Kakimoto et al. 2006).
It is calculated using the formula

Network density = 2e / n(n − 1)

where:
e is the total number of links within the network
n is the number of nodes within the network (Walker et al. 2012)

Diameter is used to analyse the connections and pathways that exist between
nodes within the network (Walker et al. 2011). Denser networks (i.e. the route
through the network is shorter and more direct) have smaller values. It is calculated
using the formula

Diameter = max uy d ( ni , n j )

where d(ni,nj) is the ‘largest number of [agents] which must be traversed in order to
travel from one [agent] to another when paths which backtrack, detour, or loop are
190 Systems Thinking in Practice

excluded from consideration’ (i.e. maxuy,) (Weisstein 2008; Harary 1994). Cohesion
represents the number of reciprocal connections divided by the total number of pos-
sible connections (Stanton 2014).
Finally, sociometric status provides an indication of agent prominence (Houghton
et al. 2006; Salmon et al. 2014). Key agents (i.e. those most prominent within the
network) have higher sociometric values (Salmon et al. 2012). It is calculated using
the formula

∑
1 g
Sociometric Status = ( x ji , xij )
( g − 1) j =1

where:
g is the total number of nodes in the network
i and j are individual nodes
xji are the number of communications between node j and node i
xij are the number of communications between node i and node j (Salmon et al.
2012; Houghton et al. 2006)

RESULTS
Identification of System Agents
For the purposes of this analysis, a total of 21 system agents were identified from
previous work conducted by Price (2016) and Banks and Stanton (2016). Their work
specifically sought to identify system agents involved in Traffic Management opera-
tions (e.g. Price 2016) and within automated driving environments (e.g. Banks and
Stanton 2016). The 21 agents broadly span three operational categories: RE, TMC
and EA (see Table 10.1 for complete list and descriptions). These agents represent the
main human and non-human entities that can be found within the road transportation
network.

Task Networks
On the basis of the list presented in Table 10.1, it is possible to consider the types of
tasks in which these system agents engage and how they may be related. This makes
it possible to construct a high-level task network for the entire road transportation
system involving all 21 agents. Walker et al. (2006) suggest that task networks can
show how sub-tasks may relate to other sub-tasks based upon their functional or
temporal properties.
The task network for the road transportation system, shown in Figure 10.1,
should be viewed as a continuous process to reflect the notion that the system always
remains active (i.e. 24 hours a day, 7 days a week). This network representation does
not consider any CAV or intelligent functionality. The task network portrays the rela-
tionships that exist between tasks in a non-sequential manner. Whilst not exhaustive,
this representation includes some of the main tasks (n=19) associated with RE, TMC
and EA agents (and the interconnections that may exist between them).
EAST in Future Road Transportation Systems 191

TABLE 10.1
System Agents Involved in the Road Transportation System

Operational
Category Subcategory Agent Description
Road Environment Drivers (DD / Host Driver The categories of individuals
(RE) – this DM)* Host Passenger occupying vehicles
category Other Drivers Role of driver dependent upon
represents all Other Passengers level of automation (DD, DM)
agents that are Vehicles (Connected Host Vehicle The categories of traffic using
present within and Autonomous Other Vehicles (or potentially using) the road
the road Vehicles; CAV) Services/Goods network
environment Vehicles Addition of intelligent sensors
Emergency Vehicles enable CAV function
External Roadside Traffic Monitoring E.g. CCTV cameras and
Equipment (ERE) Equipment induction loops
Traffic Management E.g. traffic lights and Variable
Equipment Message Signs (VMSs)
Vulnerable Road Users E.g. cyclists and pedestrians
(VRU)
Traffic Management Centre (TMC) – this TMC Operator Responsible for managing
category represents all agents that have traffic
direct access to information relating to Closed-Circuit Television Controls the TMC’s CCTV
the overall traffic situation (CCTV) Applications cameras
Urban Traffic Collects data relating to road
Management Control environment (e.g. vehicle
(UTMC) Applications counts)
Police Closed-Circuit Monitor CCTV for crime,
Television (CCTV) assisting police operations
Personnel
External Agencies (EA) – this category Radio Stations Distribute information to traffic
represents all agents that both share and and other agents
receive information relating to traffic Information Providers Provide additional information
situations (e.g. Met Office, Highways
England)
Other Transport Control Includes other road TMCs as
Centres well as public transport
control centres (e.g. Bus)
Emergency Services Manage emergency service
Control Centres operations
Traffic Data Distribution Dissemination of information
Services to traffic and third parties

Note: DD and DM reflect the ‘Driver Driving’ and ‘Driver Monitoring’ roles identified by Banks, V. A.
and Stanton, N. A., Theoretical Issues in Ergonomics Science, in press.
192 Systems Thinking in Practice

FIGURE 10.1 Task network showing high-level tasks essential for effective functioning of
the road transportation network.

Analysis of the task network presented above was conducted using AGNA (Benta
2005). The AGNA analysis indicates that the network can be described as binary (i.e.
it can be represented by a zero-one matrix) and symmetric (i.e. it is non-directed). In
total, the network has 68 connections (i.e. the number of in-out connections between
nodes). The network has a diameter of 6 (i.e. six hops from one side of the network to
the other) and a density score of 0.2 (i.e. a low distribution network).
However, in order to compare and contrast how non-CAV and CAV task networks
may differ, a combined networks approach was adopted (e.g. Stanton 2014). A com-
bined task and social network approach enables us to see ‘who’ is involved in com-
pleting tasks within the transportation network in parallel to keep the task network
functioning. In terms of DCOG, this approach provides a means to explore the allo-
cation of system function among social agents spanning the RE, TMC and EA cat-
egories (outlined in Table 10.1). For the purposes of discussion, the authors assume
that the same basic task network exists regardless of whether intelligent features
exist within it. Thus, Figure 10.2 represents the task network (shown in Figure 10.1)
coded by social agents for a non-CAV network, whilst Figure 10.3 represents the task
network for a CAV network – note that subcategories of RE have been added.
Figure 10.2 essentially demonstrates that 14 out of 20 tasks are completed in iso-
lation by a single agent category (i.e. RE, TMC or EA). In practical terms, this means
that there is a clear division between tasks associated with different agent categories.
Importantly, the role of the driver reflected within Figure 10.2 represents that of
‘Driver Driving’ (DD) (Banks and Stanton 2018). The DD role essentially represents
the idea that the driver is responsible for completing all of the physical and cognitive
tasks associated with driving (Walker et al. 2015) and is thus responsible for control-
ling the vehicle, avoiding hazards, navigating routes, obeying laws, monitoring traf-
fic and maintaining safety in the non-CAV network portrayed in Figure 10.2.
EAST in Future Road Transportation Systems 193

FIGURE 10.2 Task network for a non-CAV transportation network coded by social agents.

In contrast, Figure 10.3 represents the task network (coded by social agents) for
a CAV network. Here we see many more tasks becoming shared (i.e. completed
in parallel by two or more agents) in comparison to the non-CAV network. Tasks
completed in isolation appear to have halved as a result of CAV implementation.
This indicates that the divisions between the RE, TMC and EA categories are less
rigid, opening up possibilities to create new communication channels in an effort
to improve information sharing and network cohesiveness. ‘Task sharing’ is par-
ticularly more pronounced within the RE category, as CAV vehicles are capable of
completing the same tasks as a traditional DD. To acknowledge the changing role
of the driver within a CAV network, an additional driver role has been added to this
network representation. A Driver Monitor (DM) reflects the ‘intended’ supervisory
role of the driver during the intermediate phases of automation (Banks and Stanton
2018). A DM would need to continue monitoring the behaviour of the vehicle and
automated sub-systems despite the transfer of physical control and aspects of deci-
sion making to CAV vehicles. This is to ensure that safe and normal driving prac-
tice is maintained. Note that the role of DD remains in the representation to reflect
the potential for mixed traffic driving scenarios (e.g. both automated and manually
driven vehicles may occupy the road), but it also demonstrates further the utility of
CAV to perform all of the tasks typically associated with a DD.

Social Networks
To better understand the structure of communications that may occur between
agents within RE, TMC and EA, social networks were constructed to demonstrate
194 Systems Thinking in Practice

FIGURE 10.3 Task network for CAV coded by social agents.

the directional flow of communication between system agents for both non-CAV
and CAV networks. Social networks essentially provide a means to assess the rela-
tionship (i.e. in terms of communication and activity) between different system
agents (Stanton 2014). ‘Communication’ between system agents can range from
being obvious (e.g. gestures between drivers such as flashing lights, radio broad-
casts sharing traffic updates, vehicle displaying information on internal interface to
alert driver to changes in system status, driver manipulating controls within vehicle
to alter its behaviour) to being ‘invisible’ (e.g. Urban Traffic Management Control
[UTMC] data can be used to change behaviour of traffic management equip-
ment). Simplified versions of these networks were then constructed and validated
by Subject Matter Experts (SMEs) for ease of interpretation and are presented in
Figures 10.4 and 10.5. Four SMEs from a leading automotive manufacturer and
members of the Human Factors Research Group at the University of Nottingham
were involved in this process, with a collective 18 years’ experience within the driv-
ing automation field.
Figure 10.4 presents a simplified social network of a non-CAV transportation
network. Here, we see that whilst links exist between the RE, TMC and EA, com-
munications typically remain within operational categories. In contrast, Figure 10.5
presents a social network for future CAV networks and shows that the links that
exist between different operational categories become stronger as new links are
formed. This is because wireless connectivity between vehicles and infrastructure
enables both intra-vehicle, inter-vehicle and real-time communication with traffic
management systems. CAV functionality therefore binds the social network more
EAST in Future Road Transportation Systems 195

Information Radio
Drivers
providers stations

External Roadside
equipment

Traffic data
Vehicles
distribution

TMC UTMC
Operator applications
Emergency vehicles
only
Services/
Goods
only Emergency Police CCTV Engineers
Other transport
services control personnel
control centres
centre

CCTV
applications

Road Environment External agencies Traffic management centre

FIGURE 10.4 Simplified social network representation of macro-level communications

within a non-CAV network.

closely together, bridging the ‘communication gap’ between different operational

categories.
Further analysis of these networks was performed using AGNA (Benta 2005)
to demonstrate that even simplified social network models of macro-level systems
can be used to identify differences between agent communications in non-CAV and
CAV transportation systems. Table 10.2 presents the findings of this analysis and
shows that whilst the networks have the same number of nodes (i.e. number of sys-
tem agents), the CAV network incorporates additional connections (i.e. demonstrat-
ing the increased communication that exists within the network). The CAV network
yields a smaller diameter, which is perhaps attributable to the more direct nature of
interactions within it. With fewer hops required to get from one side of the network
to the other, together with greater network density, CAV implementation does indeed
appear to improve network efficiency.
The calculation of sociometric status was used to assess agent prominence within the
networks. To enable us to see more clearly the shift in agent prominence occurring as a
result of CAV implementation, only agents holding different sociometric status values
between the social networks are highlighted in Table 10.3. Agents within the RE had the
largest changes in prominence. Both ‘Vehicle’ and ‘External Roadside Equipment’ cat-
egories also gain prominence, which is attributable to the introduction of V2V and V2I
sensor units. These changes also influence dynamic shifts in other areas of the transpor-
tation network, as, for example, UTMC becomes even more central to TMC operation.
In addition, CAV implementation also strengthens the link between RE and EA agents.
196 Systems Thinking in Practice

Information
Radio stations Drivers
providers

External Roadside
Equipment*

Traffic data
Vehicles*
distribution

TMC UTMC
Operator applications

Emergency Police CCTV Engineers

Other transport
services control personnel
control centres
centre

CCTV
applications

Road Environment External agencies Traffic management centre

FIGURE 10.5 Simplified social network representation of macro-level communications

within future CAV networks. Bold lines represent new and strengthened links between agents
as a result of CAV implementation.

TABLE 10.2
Contrasting Network Metrics for Non-CAV and
CAV Social Networks

Social Network Analysis Non-CAV CAV

Number of nodes (i.e. system agents) 13 13
Number of connections 43 45
Diameter 5.0 4.0
Density 0.28 0.37

Information Networks
The final representational medium within the EAST framework is information net-
works (Stanton et al. 2008, 2013; Walker et al. 2006, 2010). Information networks
detail aspects of communication that underpin the foundations of the system and
how they are linked. The process of constructing information networks typically
follows the interrogation of operator verbalisations. These may be captured using
observational notes, interviews and live audio feeds recorded during task comple-
tion (e.g. Walker et al. 2010). Information nodes are identified and then paired with
their closest relation. For example, nodes within the same sentence will be linked
EAST in Future Road Transportation Systems 197

TABLE 10.3
Contrasting Sociometric Status for Agents Involved in Non-CAV and CAV
Networks

Note: Highlighted Agents Yield Different Sociometric Status Values as a Result of CAV Implementation.
Asterisks Identify ‘Key’ Agents.

(e.g. Stanton 2014; Walker et al. 2010). This strategy quickly builds a network of
information concepts (Banks and Stanton 2018). According to Stanton (2014), infor-
mation may be related temporally or spatially to different agents or tasks. The infor-
mation network presented in Figure 10.6 was constructed based upon this strategy
and combines driver-orientated knowledge (see Banks and Stanton 2018) with TMC-
orientated knowledge (see Price 2016) to represent a non-CAV transportation net-
work. This representation contains 53 nodes and 119 connections and demonstrates
that in order to ensure the effective functioning of the road transportation network, a
plethora of information must be captured, processed and distributed between all of the
agents involved. In contrast, the information network representing the operation of a
CAV network has 18 additional nodes and a further 66 connections (see Figure 10.7).
These additional nodes essentially represent the information that can be generated
from the implementation of CAV into the network.
The information networks presented in Figures 10.6 and 10.7 were subjected to
further analysis using AGNA (Benta 2005), which showed that whilst network dyna-
mism does not significantly change at a global level (see Table 10.4), the CAV net-
work grows rapidly, even at a high level of analysis.
Underlying differences between non-CAV and CAV information networks is
important because we need to ensure that agents within the network are equipped
198 Systems Thinking in Practice

FIGURE 10.6 Information network for non-CAV networks (note that ‘key’ nodes high-
lighted in grey are based upon the calculation of sociometric status – see Table 10.5).

with ‘key’ knowledge about its functioning. Sociometric status was calculated to
assist in the identification of key informational nodes within the non-CAV and CAV
networks above. Data were ranked in descending order to enable the authors to high-
light key informational nodes within both non-CAV and CAV information networks.
The results are shown in Table 10.5 and appear to confirm that any node with four

FIGURE 10.7 Hypothetical information network for CAV networks (note that ‘key’ nodes
highlighted in grey are based upon the calculation of sociometric status – see Table 10.5).
EAST in Future Road Transportation Systems 199

TABLE 10.4
Comparison of Network Metrics Relating to the
Information Networks for Non-CAV and CAV
Networks

Information Network Analysis Non-CAV CAV

Number of nodes 53 71
Number of connections 119 185
Diameter 8 8
Density 0.04 0.04

TABLE 10.5
Key Informational Nodes Based Upon Rank Sociometric
Status Scores

Non-CAV CAV
1 Strategy Connective functionality
2 Risk assessment Strategy
3 Traffic Risk assessment
4 Signage Traffic
5 Infrastructure Signage
6 Traffic type Infrastructure
7 Hazard type Traffic type
8 Time of occurrence System mode
9 Traffic management Feedback
10 Complexity Hazard type

or more connections should be considered to play a central role in the knowledge

base that underpins effective functioning of the system (Banks and Stanton 2018).
Table 10.5 shows that key terminologies within non-CAV and CAV networks remain
largely the same, although there is greater emphasis upon ‘connectivity’, particularly
in relation to system mode and feedback in CAV networks.

DISCUSSION
This chapter presents a practical method for analysing and evaluating DCOG at
a macro level. Unlike previous applications that have tended to explore DCOG in
meso-level systems, the authors of this chapter have taken a much ‘broader’ systems
view. Rather than isolating a sub-system operating within a much larger sociotech-
nical system, the authors have taken the road transportation system as their ‘unit of
analysis’ (Grote et al. 2014). This in turn provides a macro-level account of DCOG
200 Systems Thinking in Practice

that has enabled us to explore the dependencies and interdependencies that exist
between human and non-human agents across different operational domains for both
non-CAV and CAV networks.
Inspired by Hutchins (1995), the authors have shown how cognitive processes
can be distributed socially across system agents and internal and external tasks and
through time (acknowledging that earlier processes may transform the nature of later
processing) using EAST. Although EAST has been primarily used to focus on meso-
level activities within varied domains (e.g. Baber et al. 2013; Banks and Stanton
2016; Stanton et al. 2006; Stanton 2014; Walker et al. 2006), this chapter shows that
network representations can also provide a means to explore macro-level DCOG
within much larger sociotechnical systems. Rather than focusing upon an individual
component of a system (for example, the RE), the authors have considered other
dependencies and interdependencies spanning multiple operational environments
(i.e. the inclusion of TMC and EA).
EAST was used to describe non-CAV and CAV networks in terms of task, social
and information networks to help improve our knowledge of how these systems may
work in practice. Whilst the network representations presented in this chapter are by
no means exhaustive, they do provide a basis for discussion into (a) the agents that
play a role in our road transportation system, (b) how these agents are connected to
one another, (c) the types of communication or information that are shared within the
network and (d) how this may change as a result of CAV implementation. Regardless,
any new conceptualisation based upon future transportation systems will need to
consider the changes in these sociotechnical networks.
The combined task and social network approach using the EAST framework can
provide insight into DCOG within the transportation network by identifying ‘who’
is doing ‘what’. It demonstrates that even though the underlying task network for
the transportation system remains the same, a connected and autonomous system
can lead to multiple system agents completing the same tasks in parallel. As long as
CAV architecture complements existing infrastructures and frameworks that uphold
non-CAV networks, tasks spanning the RE, TMC and EA categories could all ben-
efit from CAV implementation. Specifically, CAV should increase the efficiency
and reliability of network operations for all social agents involved (Talebpour and
Mahmassani 2016). However, there are a number of issues that must be considered
in order to make CAV networks a success. For example, there remain issues relating
to cybersecurity, compatibility of data sharing between different cloud computing
platforms and ‘car hacking’ (e.g. Checkoway et al. 2011; Frost and Sullivan 2014).
With non-human agents set to become the most prominent agents within a CAV net-
work, designers must ensure that cloud computing platforms are secure, reliable and
efficient. They must be capable of holding, processing, selecting and re-transmitting
relevant information to relevant social agents. A failure within the network could,
after all, have far-reaching implications in terms of RE, TMC and EA operations.
For example, a loss of network coverage could mean information is shared intermit-
tently, which will affect the reliability and efficiency of the network. It is therefore
important to consider the differences between CAV and non-CAV networks, because
until the network is fully connected (across space and time), transitions between
CAV and non-CAV operation are still likely to occur.
EAST in Future Road Transportation Systems 201

Quantitative analysis of networks using network metrics enables systematic assess-

ment of existing systems (i.e. non-CAV) and future systems (i.e. CAV). The network
analysis presented in this chapter is rudimentary but arguably provides insight into
the distributed nature of system function, which has not previously been explored.
Network metrics have been used to quantitatively analyse the underlying differences
between non-CAV and CAV networks. Sociometric status, in particular, has been use-
ful in identifying key agents or nodes within the networks as well as highlighting how
the implementation of CAV can change the prominence of individual agents or nodes.
These methods and tools provide a means to explore possible strengths and weaknesses
within system function. In the future, EAST may be used to provide further insight into
potential network resilience issues via a ‘broken-links’ approach (Stanton and Harvey
2017). After all, it is likely that failure in one part of the CAV network is likely to have
a substantial knock-on effect in others, especially if human agents become accustomed
to CAV functionality. Perceptions of competence and expectancy can all impact upon
the way in which agents interact on the road (e.g. Rudin-Brown and Parker 2004).

CONCLUSIONS
Vehicle automation and the connected services that they offer appear set to revolutio-
nise the way in which people interact with and behave within the road transportation
network. Essentially, CAV points to a future whereby most, if not all, of the poten-
tial communication links between vehicles and infrastructure within the RE may be
connected. Thus, rather than CAV eliminating ergonomics and human factors issues
from the transportation network entirely, it could actually increase the importance of
them as it highlights the types of activities human and non-human agents will engage
in. Whilst networked simulators exist, enabling multiple drivers to interact in a single
virtual environment, research outputs are limited to a single operational environment
(e.g. Sawyer and Hancock 2012). Thus, it has only been possible to explore how
intelligent transport systems, such as CAV, will impact individual agents involved
within the road environment. EAST goes further than this and provides a framework
that can enable researchers to visualise the impact of CAV on a much larger scale.
Adopting sociotechnical systems design in this way could offer a means to improve
the performance of the whole system rather than individual components in isolation
(Stanton 2014). Modelling system interaction between system agents can provide
an extensive qualitative overview of DCOG within large-scale network operations.

REFERENCES
Atkins Mobility. (2016). Connected & autonomous vehicles: Introducing the future of mobil-
ity. Available at http://www.atkinsglobal.co.uk. [Accessed 10.01.2017].
Baber, C., Stanton, N. A., Atkinson, J., McMaster, R. and Houghton, R. J. (2013). Using
social network analysis and agent-based modelling to explore information flow using
common operational pictures for maritime search and rescue operations. Ergonomics,
56(6), 889–905.
Banks, V. A. and Stanton, N. A. (2016). Driver-centred vehicle automation: Using network
analysis for agent-based modelling of the driver in highly automated driving systems.
Ergonomics, 59(11), 1442–1452.
202 Systems Thinking in Practice

Banks, V. A. and Stanton, N. A. (2018). Analysis of driver roles: Modelling the changing
role of the driver in automated driving systems using EAST. Theoretical Issues in
Ergonomics Science, (In press).
Banks, V. A., Stanton, N. A. and Harvey, C. (2014). Sub-systems on the road to vehicle auto-
mation: Hands and feet free but not ‘mind’ free driving. Safety Science, 62, 505–514.
Benta, M. (2005). Studying communication networks with AGNA 2.1. Cognition Brain
Behaviour, 9, 567–574.
Checkoway, S., McCoy, D., Kantor, B., Anderson, D., Shacham, H., Savage, S., Koscher, K.,
Czeskis, A., Roesner, F. and Kohno, T. (2011). Comprehensive experimental analy-
ses of automotive attack surfaces. Proceedings of USENIX Security Symposium. San
Francisco, CA.
Christoffersen, K. and Woods, D. D. (2002). How to make automated systems team players.
Advances in Human Performance and Cognitive Engineering Research, 2, 1–12.
De La Fortelle, A. (2005). Cybercars: Past, present and future of the technology. 12th Word
Congress on Intelligent Transport Systems. San Francisco, CA.
Eriksson, A. and Stanton, N. A. (2017). Takeover time in highly automated vehicles:
Noncritical transitions to and from manual control. Human Factors, 59(4), 689–705.
DOI:10.1177/0018720816685832.
European Road Transport Research Advisory Council (ERTRAC). (2015). ERTRAC auto-
mated driving roadmap, July 2015. [Accessed 12.13.2016].
Frost and Sullivan. (2014). Cybersecurity in the automotive industry. 30 October. https://
www.ertrac.org/uploads/documentsearch/id38/ERTRAC_Automated-Driving-2015.
pdf [Accessed 01.09.2017].
Grote, G., Stanton, N. A. and Weyer, J. (2014). Beyond human-centred automation.
Ergonomics, 57(3), 289–294.
Harary, F. (1994). Graph Theory. Addison-Wesley: Reading, MA.
Hollan, J., Hutchins, E. and Kirsh, D. (2000). Distributed cognition: Toward a new foundation
for human–computer interaction research. ACM Transactions on Computer Human
Interaction, 7, 174–196.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49, 1204–1225.
Hutchins, E. (1995). How a cockpit remembers its speeds. Cognitive Science, 19(3), 265–288.
KPMG. (2015). Connected and autonomous vehicles – The UK economic opportunity. Available
at: https://www.smmt.co.uk/wp-content/uploads/sites/2/CRT036586F-Connected-
and-Autonomous-Vehicles-%E2%80%93-The-UK-Economic-Opportu.1.pdf.
Lavrinc, D. (2014). This is how bad self-driving cars suck in the rain. Available at: http://
jalopnik.com/this-is-how-bad-self-driving-cars-suck-in-the-rain-.
Louw, T. and Merat, N. (2017). Are you in the loop? Using gaze dispersion to understand
driver visual attention during vehicle automation. Transportation Research Part C:
Emerging Technologies, 76, 35–50.
Price, J. (2016). Human factors in the design of traffic management systems. University of
Southampton, Faculty of Engineering and the Environment, London. EngD thesis.
Rudin-Brown, C. M. and Parker, H. A. (2004). Behavioural adaptation to adaptive cruise con-
trol (ACC): Implications for preventative strategies. Transportation Research Part F,
7(2), 59–76.
Salmon, P. M., Lenné, M. G., Walker, G. H., Stanton, N. A. and Filtness, A. (2014). Using the
event analysis of systemic teamwork (EAST) to explore the conflicts between different
road user groups when making right hand turns at urban intersections. Ergonomics, 57
(11), 1628–1642.
EAST in Future Road Transportation Systems 203

Salmon, P. M., Stanton, N. A. and Young K. L. (2012). Situation awareness on the road:
review, theoretical and methodological issues, and future directions. Theoretical Issues
in Ergonomics Science, 13(4), 472–492.
Sawyer, B. D. and Hancock, P. A. (2012). Development of a linked simulation network to
evaluate intelligent transportation system vehicle to vehicle solutions. Proceedings of
the Human Factors and Ergonomics Society Annual Meeting, 56(1), 2316–2320.
Science and Technology Select Committee. (2017). Connected and autonomous vehicles: The
future? Available at: https://www.publications.parliament.uk/pa/ld201617/ldselect/
ldsctech/115/115.pdf.
Society for Automotive Engineers. (2016). Taxonomy and definitions for terms related to on-
road motor vehicle automated driving systems. http://standards.sae.org/j3016_201401/
[Accessed 12.12.2016].
Sorensen, L. J., Stanton, N. A. and Banks, A. P. (2011). Back to SA school: Contrasting three
approaches to situation awareness in the cockpit. Theoretical Issues in Ergonomics
Science, 12(6), 451–471.
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57(3), 403–418.
Stanton, N. A. (2015). Responses to autonomous vehicles. Ingenia, 62, 9.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork. Ashgate: Aldershot, UK.
Stanton, N. A. and Harvey, C. (2017). Beyond human error taxonomies in assessment of risk
in sociotechnical systems: A new paradigm with the EAST ‘broken-links’ approach.
Ergonomics, 60(2), 221–233.
Stanton, N. A. and Roberts, A. P. J. (2017a). Examining task, social and information networks
in submarine command and control. IEEE Transactions on Human-Machine Systems,
DOI: 10.1109/THMS.2017.2720659.
Stanton, N. A., Roberts, A. P. J. and Fay, D. T. (2017b). Up periscope: Understanding sub-
marine command and control teamwork during a simulated return to periscope depth.
Cognition, Technology and Work, 19 (2–3), 399–417.
Stanton, N. A., Salmon, P. M., Rafferty, L., Walker, G. H., Baber, C. and Jenkins, D. P. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design. Ashgate:
Aldershot, UK.
Stanton, N. A., Salmon, P. M. and Walker, G. H. (2015). Let the reader decide: A para-
digm shift for situation awareness in sociotechnical systems. Journal of Cognitive
Engineering and Decision Making, 9 (1), 44–50.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
P. M. et al. (2006). Distributed situation awareness in dynamic systems: Theoretical
development and application of an ergonomics methodology. Ergonomics, 49 (12–13),
1288–1311.
Talebpour, A. and Mahmassani, H. S. (2016). Influence of connected and autonomous vehicles
on traffic flow stability and throughput. Transportation Research Part C: Emerging
Technologies, 71, 143–163.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. and Green, D. (2006). Event
analysis of systematic teamwork (EAST): A novel integration of ergonomic methods to
analyse C4i activity. Ergonomics, 49 (12–13), 1345–1369.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. and Jenkins, D.
(2010). From ethnography to the EAST method: A tractable approach for representing
distributed cognition in Air Traffic Control. Ergonomics, 53 (2): 184–197.
Walker, G. H., Stanton, N. A. and Salmon, P. M. (2015). Human Factors in Automotive
Engineering and Technology. Ashgate: Aldershot, UK.
204 Systems Thinking in Practice

Walker, G. H. Stanton, N. A. and Salmon, P. M. (2011). Cognitive compatibility of motorcy-

clists and car drivers. Accident Analysis & Prevention, 43(3), 878–888.
Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D. P., Monnan, S. and Handy, S. (2012).
Communications and cohesion: a comparison between two command and control para-
digms, Theoretical Issues in Ergonomics Science, 13(5), 508–527.
Weisstein, E. W. (2008). Graph diameter. From MathWorld – A Wolfram Web Resource.
Available at: http://mathworld.wolfram.com/GraphDiameter.html. [Accessed 6 October
2017].
Zeeb, K., Buchner, A. and Schrauf, M. (2015). What determines the take-over time? An inte-
grated model approach of driver take-over after automated driving. Accident Analysis
& Prevention, 78, 212–221.
Section III
Future Developments in EAST
11 Adding a Control Network to
STAMPING on EAST

EAST to Examine the Safety

Controls in the Railway Level
Crossing System Lifecycle
With Gemma J. M. Read, Natassia
Goode, Eryn Grant, Clare Dallat, Tony
Carden and Anjum Naweed

INTRODUCTION
The popularity of systems thinking is now such that various systems ergonomics
methods are being applied across the transport domains, including rail (Salmon et al.
2016b; Stefanova et al. 2015), road (Parnell et al. 2017; Salmon et al. 2016a), aviation
(Allison et al. 2017) and maritime transport (Lee et al. 2017). In the case of rail trans-
port specifically, these methods have been applied to a diverse set of issues, including
railway level crossings (RLXs) (Salmon et al. 2016b; Stefanova et al. 2015), ticketing
system design (Read et al. 2015), accident analysis (Chen et al. 2015; Ouyang et al.
2010; Salmon et al. 2013), signals passed at danger (Madigan et al. 2016) and train
cabin design (Jansson et al. 2006). Given the increasing complexity of rail transport
systems, it is likely that systems ergonomics methods will increasingly be applied for
analysis and design in rail systems worldwide.
Although all of these methods share a common alignment with systems theory,
most have differing theoretical underpinnings, giving rise to their own unique
analysis approach. Systems Theoretic Accident Model and Process (STAMP), for
example, is based on control theory and seeks to identify what control structure is
in place in a given system and, in the event of an accident, how the adaptive control
function within the system failed. Event Analysis of Systemic Teamwork (EAST),
on the other hand, focuses on the relationship between tasks, agents and informa-
tion and uses networks to identify issues that impact performance (e.g. information
bottlenecks, overloaded agents). Further differences can be found when considering
the focus of other systems ergonomics methods such as Hierarchical Task Analysis
(goals), Cognitive Work Analysis (CWA) (constraints) and Functional Resonance
Analysis Method (FRAM) (performance variability and resonance). As a result, each
systems ergonomics method brings with it a unique set of strengths and weaknesses,
207
208 Systems Thinking in Practice

and there has been much discussion on the potential benefits of converging different
methods together (Dallat et al. in press; Houghton et al. 2015; Naweed 2014; Salmon
et al. 2010, 2011; Stanton et al. 2013, 2016; Patriarca et al. 2017).
Researchers and practitioners are therefore beginning to explore the utility of
integrating systems ergonomics methods. Patriarca et al. (2017) combined the
abstraction hierarchy from CWA with FRAM, suggesting that using the two in
an integrated manner enables additional analyses of intra-agent and intra-system-
level interactions. Houghton et al. (2015) combined social network analysis (SNA)
and CWA to examine military planning tasks, concluding that using the two
approaches together enables a richer understanding of system structure. Stanton
et al. (2016) integrated CWA and SNA to compare different cockpit crewing con-
figurations and their likely impact on safety on future flight decks. CWA was used
to define system constraints, whereas SNA metrics were used as a measure of
system resilience. Given that modern-day systems are stretching the capabilities
of our methodological toolkit (Salmon et al. 2017; Walker et al. 2017), and the fact
that many methods do not fully align with systems thinking, it is our opinion that
the integration of systems ergonomics methods represents an important area of
investigation for our discipline. This builds on a long history and tradition of meth-
ods integration in ergonomics research and practice (e.g. Kirwan 1992; Stanton
et al. 2005, 2013).
In the case of EAST, it is apparent that adding further forms of network may
enhance the utility of the framework. Candidate additional EAST networks that the
authors have discussed include a network of safety controls, a network of unexpected
emergent behaviours and a risk network. This chapter investigates the utility of add-
ing a safety controls network based on principles from the STAMP method (Leveson
2004). To test this, we undertook a case study in which EAST and an additional
control network were used to examine safety management during the design and
operation of RLXs systems in Victoria, Australia. The four networks were used to
examine the system and its safety controls with a view to identifying recommenda-
tions for strengthening safety management at RLXs.

RESEARCH CONTEXT: RAILWAY LEVEL CROSSINGS

Collisions between trains and vehicles at RLXs remain a persistent road and rail
safety issue worldwide. In Australia, there were 97 fatalities resulting from colli-
sions between trains and vehicles at RLXs between 2000 and 2009 and 92 colli-
sions between trains and pedestrians at RLXs between 2002 and 2012 (ATSB 2012;
Independent Transport Safety Regulator 2011). Whilst Australia is used here as an
example, the problem is of a similar magnitude in the United States of America and
Europe (Evans 2011; Federal Railroad Administration 2013).
On top of the personal and social impacts of RLX collisions, the financial bur-
den is also significant. In 2010, the annual cost of RLX incidents in Australia was
estimated at just over one hundred million dollars, taking into account human and
property damage costs and other costs such as emergency service attendance, delays,
investigation and insurance (Tooth and Balmford 2010). Given the combined per-
sonal, societal and economic burden created by RLX collisions, they continue to
STAMPING on EAST 209

represent an area of strategic importance for road and rail safety authorities (Read
et al. 2017).
Analyses of fatal collisions provide evidence that RLX collisions have a complex
web of interrelated causes spanning overall road and rail systems. In short, they are
a systems problem (Read et al. 2017; Salmon et al. 2016b; Stefanova et al. 2015). This
has important implications for the methods used to understand and improve RLX
safety. Describing and understanding RLX systems requires analysis methods that
consider the behaviour of overall road and rail systems (Read et al. 2013; Wilson and
Norris 2014). Key elements requiring analysis include

• the tasks that are undertaken across the system as part of RLX design and
operation
• the diverse set of agents that reside within the RLX system and undertake
design and operation tasks
• the interactions between agents (human and non-human) when undertak-
ing these tasks
• the information required for successful completion of these tasks
• the factors influencing behaviour across the system, including safety con-
trols, environmental factors and financial and production pressures

It is apparent that no one method can completely fulfil these requirements. EAST,
for example, considers tasks, agents and information but does not explicitly consider
safety controls and their impact on behaviour. STAMP, on the other hand, consid-
ers controls but does not detail the tasks and information required when enacting
controls. It is our contention, then, that a richer analysis of sociotechnical systems
behaviour can be achieved by integrating different systems ergonomics methods.
The study was undertaken to identify the range of factors influencing behav-
iour and safety in RLX systems and to identify opportunities for introducing new
interventions design to improve safety. To achieve this, the EAST framework and
STAMP control structure methods were integrated and used together to examine a
RLX system ‘lifecycle’, including all activities involved in the design, implementa-
tion, operation and removal (for grade separation) of RLXs in Victoria, Australia.
A short description of how EAST and the STAMP control structure principles were
integrated is given below.

PART 1: INTEGRATING EAST AND STAMP

STAMP and Control Theory
STAMP (Leveson 2004), originally developed as an accident analysis methodol-
ogy, is underpinned by systems and control theory. The method takes the view that
accidents result from the inadequate control or enforcement of safety constraints –
when disturbances, failures or dysfunctional interactions between components are
not handled by existing control mechanisms (Leveson 2004). In the RLX context,
for example, the model might suggest that collisions involving vehicles and trains
at RLXs occur when controls such as flashing lights and warning bells, education
210 Systems Thinking in Practice

campaigns and road rules and regulations fail to prevent drivers from attempting to
traverse the crossing when a train is present. STAMP therefore views safety as an
issue of control and one that is managed through a control structure that has the goal
of enforcing constraints on actors across the system.
It is worth noting that a broad view of control is adopted when using STAMP.
Leveson (2004) describes various forms of control, including managerial, organ-
isational, physical, operational and manufacturing-based controls; that is, system
behaviour is controlled not only by engineered systems and direct intervention but
also by policies, procedures, shared values and other aspects of the surrounding
organisational and social culture. The first phase of STAMP involves building a
control structure to describe the system under analysis and the control relationships
that exist between actors and organisations related to both system design and sys-
tem operation. The control structure model views systems as comprising interrelated
components that maintain a state of dynamic equilibrium through feedback loops
of control and information (Leveson 2004). Accordingly, control structure models
incorporate a series of hierarchical system levels and describe the actors and organ-
isations that reside at each level. Control and feedback loops are included to show
what control mechanisms are enacted down the hierarchy and what information
about the status of the system is sent back up the hierarchy.
A generic control structure model is presented in Figure 11.1 (Leveson 2004).
The left-hand side of Figure 11.2 shows a generic control structure for system devel-
opment, whereas the right-hand side shows a generic control structure for system
System Development System Operations

Congress and Legislatures Congress and Legislatures

Govt reports Govt reports

Lobbying Lobbying
Legislation Legislation
Hearings/Open meetings Hearings/Open meetings
Accidents Accidents

Govt regulatory agencies, Industry Govt regulatory agencies, Industry

and user associations, Unions, and user associations, Unions,
Insurance Companies, Courts Insurance Companies, Courts
Regulations Regulations
Standards Certification info Standards Certification info
Certification Change reports Certification Change reports
Legal penalties Whistleblowers Legal penalties Whistleblowers
Case law Accidents and incidents Case law Accidents and incidents

Company management Company

Safety policy Status reports Safety policy

Standards Risk assessments Standards Operations reports
Resources Incident reports Resources

Project management Operations management

Hazard analyses
Safety-related changes
Progress reports
Hazard analyses Change requests
Safety standards Work instructions Problem reports
Progress reports
Audit reports

Operating assumptions
Design, Documentation Operating procedures
Operating process

Human controller
Safety reports
Safety constraints Test reports
Standards Hazard analyses
Test requirements Review results
Automated
Manufacturing controller
management
Implementation and assurance
Safety reports Revised procedures Actuator(s) Sensor(s)
Work
Audits
procedures Work logs Software revisions
Inspections Hazard analyses Hardware replacement
Documentation Physical
Design rationale process

Manufacturing
Maintenance and Problem reports
evolution Incidents
Performance audits

FIGURE 11.1 STAMP generic control structure model.

STAMPING on EAST 211

CONTROL NETWORK

TASK
NETWORK

INFORMATION SOCIAL
NETWORK NETWORK

FIGURE 11.2 EAST network of network approach with added control network.

operation. The arrows flowing down the hierarchy represent control relationships (or
reference channels, Leveson 2004), and the arrows flowing up the hierarchy repre-
sent feedback loops (or measuring channels, Leveson 2004). In relation to the exam-
ple described above, the rail level crossing infrastructure would enact the control
of ‘warnings’ on road users. In turn, information regarding collisions, near misses,
violations of the warnings and road rules and maintenance requirements for the
warning infrastructure represent forms of feedback that would be passed back up the
system. These feedback loops enable road and rail safety authorities and policy mak-
ers to assess the status of controls and whether they are having the desired impact.
The control structure has been employed as a modelling tool to describe systems
and the control and feedback loops involved in safety management (Salmon et al.
2016a). Whilst the usefulness of the control structure has been reported, there are
some notable limitations: (1) it describes control and feedback loops only and does
not examine the relationships between controls; (2) it is difficult to represent control
and feedback loops that exist between actors and/or organisations within levels of the
212 Systems Thinking in Practice

hierarchy; and (3) it is often difficult to place actors and organisations across the five
levels provided (Salmon et al. 2016a).

Integrating EAST and STAMP: A Network of Controls

Integration of the two methods was achieved by adding a fourth ‘control network’ to
the EAST framework. This is based on the notion that it is important to understand
both the relationships between different controls (i.e. design standards dictate the
nature of the warnings provided at the RLX) and the relationships between controls,
tasks, agents and information (i.e. enacting a particular control requires certain tasks
to be undertaken by certain agents using specific information). The new four-network
EAST approach is represented in Figure 11.2. It can be seen that conceptually in a
given system, the control network sits around the task, social and information triad
of networks, with controls acting to constrain the interactions within and between
these original networks. Figure 11.3 provides a representation of the relationships
between controls, tasks, agents and information.

Information B Information C

Agent A Agent B

Task A

Control A Control B

= Controls imposed on tasks, agents and information

= Controls imposed on interactions

FIGURE 11.3 Interaction of controls, tasks, agents and information.

STAMPING on EAST 213

For example, the task of negotiating a RLX safely is a task undertaken by both
road vehicle drivers (Agent A) and train drivers (Agent B). When negotiating the
RLX, both use various pieces of information such as the status of the warning
devices, the speed and location of the train and the behaviour of other users such
as pedestrians. Road vehicle drivers and train drivers are bound by controls such as
the warnings themselves, rules and regulations and standard operating procedures.
Likewise, the nature of information they are using is influenced by controls such as
design standards (e.g. timing of warnings). It is important to note that controls are
enacted on tasks, agents and information and also on the interactions between them.
For example, various controls at the RLX (e.g. boom gates, flashing lights, train
horn, road rules and regulations) are designed to manage the interactions between
drivers, their vehicle and train drivers and the train.

PART 2: APPLYING THE INTEGRATED METHOD

TO RLX SAFETY MANAGEMENT
Methodology
Data Inputs
Seven analysts with significant experience in applying either EAST or STAMP in a
range of domains (e.g. defence, road and rail transport, aviation, maritime transport)
were involved in conducting the analysis. The analysts had extensive experience with
and background information on rail safety (see Naweed et al. 2015; Salmon et al. 2016b;
Stanton and Walker 2011) and, more specifically, the Victorian RLX system, as they
had recently completed a four-year research programme involving various analyses of
different RLXs as well as the wider RLX sociotechnical system (see Read et al. 2017;
Salmon et al. 2013, 2016b). This background knowledge and information was also
supplemented by an additional documentation review (e.g. road and rail safety strategy
and policy documents, crash investigation reports, relevant academic literature).

EAST Network Development

Initially task, social and information networks were developed in a workshop setting
involving four analysts. Two of the analysts had extensive experience in applying
EAST in a range of transportation settings (Salmon et al. 2014; Stanton 2014; Stanton
et al. 2016; Walker et al. 2006, 2010). The remaining two analysts were human fac-
tors researchers with significant experience of applying systems ergonomics meth-
ods in a range of safety critical domains (e.g. Newnam and Goode 2015; Read et al.
2016; Salmon et al. 2014). One of these analysts also had significant experience in the
area of rail safety (having worked for 9 years in rail safety regulation and on various
rail safety research projects).
Prior to the workshop, participants were provided with the following materials to
support network construction:

1. Hierarchical Task Analysis (HTA) of the RLX system lifecycle. A HTA

(Stanton 2006) of the RLX system lifecycle was developed based on the data
described above. The HTA describes the goals, sub-goals and operations
214 Systems Thinking in Practice

required when designing, implementing, operating and removing RLXs in

Victoria, Australia.
2. ActorMap of rail level crossing stakeholders. An ActorMap was developed
to identify the stakeholders involved in the RLX system lifecycle (Read et al.
2017). The resulting ActorMap showed which stakeholders currently share
the responsibility for RLX system lifecycle of RLXs in Victoria, Australia.
The stakeholders were placed across a hierarchy of systems levels in line
with Rasmussen’s risk management framework (Rasmussen 1997).

The task, social and information networks were developed based on a process of
first identifying nodes, then identifying relationships between the nodes and finally
by reviewing the nodes and relationships for internal consistency. The analysis rules
surrounding what constituted a node and a relationship for each network representa-
tion are presented in Table 11.1.
Following the workshop, a draft control network was constructed by two analysts
based on a review of Salmon et al.’s (2016a) in-depth analysis of rail level crossing
systems; Salmon et al.’s (2016b) road safety control structure; and a STAMP control
structure model of rail safety under development by the authors as part of a separate
project. Development of the control network followed the process above whereby
nodes (controls) were first identified followed by the relationships between them (see
Table 11.1 for examples). The draft control structure was subsequently reviewed by
the other analysts from the original workshop.

Network Analysis
Each network was analysed using quantitative metrics that have previously been
used to interrogate EAST networks (e.g. Salmon et al. 2014; Stanton 2014). In the
present study, the following metrics were applied to each network:

1. Network density (overall network) – Network density represents the level of

interconnectivity of the network in terms of relations between nodes. Density
is expressed as a value between 0 and 1, with 0 representing a network with
no connections between nodes and 1 representing a network in which every
node is connected to every other concept (Kakimoto et al. 2006; cited in
Walker et al. 2011). Higher density values are indicative of a well-connected
network in which tasks, agents, information and controls are tightly coupled.
2. Sociometric status (individual nodes). Sociometric status provides a mea-
sure of how ‘busy’ a node is relative to the total number of nodes within
the network under analysis (Houghton et al. 2006). In the present analysis,
nodes with sociometric status values greater than the mean sociometric sta-
tus value plus one standard deviation are taken to be ‘key’ (i.e. the most
connected) nodes within each network. These nodes represent either key
tasks, agents, pieces of information or controls. For example, in the case
of the control network, the node with the highest sociometric status is the
control that is the most interrelated with other controls.
3. Centrality (individual nodes). Centrality is used to examine the standing
of a node within a network based on its geodesic distance from all other
STAMPING on EAST 215

TABLE 11.1
Analysis Rules Regarding Relationships between Nodes in EAST Networks in
the Context of RLX Design, Implementation, Operation or Removal

Network Nodes Relationships Examples

Task Represent high-level tasks Represent instances where The nodes ‘Select site for
network that are required during the the conduct of one upgrade’ and ‘Announce
design, implementation, high-level grouping of upgrade programme’ are
operation and removal of tasks (i.e. task network linked because the
RLXs. High-level tasks node) influences, is upgrade programme
were extracted from the undertaken in combination announcement cannot be
sub-ordinate goals level of with, or is dependent on, made until sites have
the HTA another group of tasks been selected
Social Represent human, Represent instances where The nodes ‘RLX
network technological or agents within the social warnings and barriers’
organisational agents who network interact with one and ‘Road users’ are
undertake one or more of another during the linked as the warnings
the tasks involved in the design, implementation, and barriers
design, implementation, operation or removal of communicate with road
operation and removal of RLXs users to inform them that
RLXs (as identified in the a train is approaching
HTA and task network)
Information Represent grouped Represent instances where The nodes ‘Traffic’ and
network categories of information information influences ‘Separation’ are linked
that is required by agents other information or is as road vehicles and
when undertaking RLX the used in combination with trains require separation
design, implementation, other information in the from one another (i.e.
operation and removal network during the design, Traffic ‘requires’
tasks (as identified in the implementation, operation Separation)
task and social network) or removal of RLXs
Control Represent controls that are Represents instances The nodes ‘Design
network enacted on the tasks, where controls are either standards’, ‘Boom gates’
agents, information and used together or where and ‘Flashing lights’ are
interactions during RLX one control influences the linked as the design
design, implementation, nature and/or enactment standard dictates the
operation and removal. A of another control characteristics of both
broad view of controls is warnings in terms of
adopted to incorporate both form, presentation and
engineering, managerial timing
and social controls

nodes in the network (Houghton et al. 2006). Central nodes represent those
that are closer to the other nodes in the network as, for example, informa-
tion passed from one to another node in the network would travel through
less nodes. Houghton et al. (2006) point out that well-connected nodes can
still achieve low centrality values as they may be on the periphery of the
network. For example, in the case of the control network nodes, higher
216 Systems Thinking in Practice

centrality status values are those that are closest to all other controls in the
network as they have direct rather than indirect links with them.

RESULTS
Task Network
The RLX system lifecycle task network is presented in Figure 11.4. The outcomes for
the task network analysis are presented in Table 11.2. For the sociometric status and
centrality analysis, Table 11.2 includes the values for the nodes that scored above the
mean sociometric status and centrality values for the network. Nodes that achieved
values above the mean + standard deviation for sociometric status and centrality are
shaded grey.
As shown in Figure 11.4, the RLX system lifecycle is underpinned by a network
of 15 high-level tasks. Analysis of the task network in Table 11.2 reveals a network
density of 0.2, which is indicative of a relatively loosely connected network (i.e. tasks
are not tightly coupled). The sociometric status and centrality analyses reveal that
‘Risk management’ and ‘Monitor performance’ are the key tasks within the network.

Social Network
The RLX system lifecycle social network is presented in Figure 11.5. The outcomes
for the SNA are presented in Table 11.3. Table 11.3 follows the same convention as
Table 11.2 for the sociometric status and centrality values.
As shown in Figure 11.5, a diverse set of 27 agents undertake tasks throughout the
RLX system lifecycle and operation. These agents range from RLX components (e.g.
barriers, flashing lights), RLX users (e.g. drivers, pedestrians, road vehicles, trains),
infrastructure owners and rail operators to the regulator, government departments,
unions, courts and the media. Analysis of the social network shown in Table 11.3
reveals a network density of 0.18, which is again indicative of a relatively loosely
connected network in which agents are not tightly coupled. The sociometric sta-
tus and centrality analysis reveals that ‘Government departments’, ‘Rail operators’,
‘Rail infrastructure managers’, ‘Regulators’ (sociometric status) and ‘Train drivers’
(Centrality only) are key agents in the RLX system lifecycle.

Information Network
The RLX system lifecycle information network is presented in Figure 11.6. The out-
comes for the information network analysis are presented in Table 11.4. Table 11.4
follows the same convention as Table 11.2 for the sociometric status and centrality
values.
As shown in Figure 11.6, 28 information nodes are required throughout the RLX
system lifecycle. Analysis of the information network in Table 11.5 reveals a network
density of 0.09, which is indicative of a very loosely connected network. The socio-
metric status analysis reveals that ‘Risk level’ and ‘Design concepts’ are the key nodes
STAMPING on EAST 217

= Key node (both sociometric status and centrality)

= Key node (sociometric status only)

= Key node (centrality only)

Grade
Close RLX
separate

Check Performance
compliance with monitoring
legislation

Operate RLX

Select site for Risk

upgrade management

Design
Announce Determine
upgrade Manage
upgrade infrastructure
and issue infrastructure
programme required
tender

Identify
and consult Commission
with RLX and
user construct RLX
community Continuous
improvement of
standards

Budget
allocated to RLX
program

FIGURE 11.4 RLX system lifecycle task network.

218 Systems Thinking in Practice

TABLE 11.2
Task Network Analysis Outputs

Key Nodes Sociometric Status Key Nodes Centrality

Network Metrics (Mean = 0.4, SD = 0.22) (Mean = 7.81, SD = 1.56)
Nodes = 15 Risk management = 1.0 Risk management = 10.88
Edges = 42 Monitor performance = 0.79 Monitor performance = 9.98
Density = 0.2 Operate RLX = 0.5 Operate RLX = 9.23
Check compliance with legislation Close RLX = 8.96
= 0.43
Select site for upgrade = 0.43 Check compliance with legislation = 8.6
Manage infrastructure = 0.43 Select site for upgrade = 8.83
Manage infrastructure = 8.12

Note: Cells Shaded Grey Represent Nodes with Values above the Mean + Standard Deviation.

= Key node (both sociometric status and centrality)

= Key node (sociometric status only)

= Key node (centrality only)

Govt depts Legislators

Investigators
Regulators
and auditors

Standard
Courts
setting bodies Other Rail Ops
Local Govts Media
staff

Rail
infrastructure Rail operators
managers Advocacy
Police
groups

Signallers Unions

Commercial
Trackworkers
companies

Construction
Train control and
systems Maintenance
contractors
RLX warnings Professional
and barriers drivers

Train Road users

Train drivers

Pedestrians Road vehicles

Local
community

FIGURE 11.5 Social network for RLX system lifecycle.

STAMPING on EAST 219

TABLE 11.3
Social Network Analysis Outputs

Key Nodes Sociometric Status Key Nodes Centrality

Network Metrics (Mean = 0.37, SD = 0.17) (Mean = 14.36, SD = 2.48)
Nodes = 27 Government depts. = 0.81 Government depts. = 17.39
Edges = 138 Rail operators = 0.74 Rail operators = 17.39
Density = 0.18 Rail infrastructure managers = 0.66 Rail infrastructure managers = 16.36
Regulators = 0.52 Train drivers = 16.21
Advocacy groups = 0.48 Investigators and auditors = 15.6
Road users = 0.48 Advocacy groups = 15.45
Investigators and auditors = 0.44 Regulators = 15.2
RLX warnings and barriers = 0.44 Media = 14.39
Train drivers = 0.44 Unions = 14.39
Media = 0.44 Road users = 14.39
Unions = 0.41 Police = 14.39
Professional drivers = 0.41 Signallers = 14.39
Legislators = 0.37
Courts = 0.37

Note: Cells Shaded Grey Represent Nodes with Values above the Mean + Standard Deviation.

within the information network. The key nodes according to the centrality analysis are
‘Budget’, ‘Demographics’, ‘Mitigations’, ‘Speed’ and ‘Surrounding land use’.

Control Network
The RLX system lifecycle control network is presented in Figure 11.7. The outcomes
for the control network analysis are presented in Table 11.5. Table 11.5 follows the
same convention as Table 11.2 for the sociometric status and centrality values.
As shown in Figure 11.7, 30 controls are present that constrain activities across
the RLX system lifecycle. Analysis of the control network in Table 11.5 reveals a
network density of 0.11, which is indicative of a relatively loosely connected net-
work. The sociometric status analysis reveals that ‘Audits and Inspections’, ‘Design
standards’ and ‘Road rules’ are the key controls within the control network. Key
controls according to the centrality analysis are ‘Insurance premiums’, ‘Timetables’
and ‘Train protection devices’. Other nodes that achieved sociometric status and cen-
trality scores above the mean include controls relating to the RLX itself (e.g. ‘Train’,
‘Vehicle’, ‘Signals’); safety strategy (e.g. ‘Policy and strategy’, ‘Targets’); enforcement
(e.g. ‘Enforcement’, ‘Legal Penalties’); and education (e.g. ‘Education’, ‘Initiatives’).

DISCUSSION
This chapter presents the findings from a case study in which a control network
was added to EAST to enable EAST to incorporate a control network analysis.
220 Systems Thinking in Practice

Polices and
Hazards
procedures

Programme of Direcon Past

works experience

Risk level System state

Design
Speed
standards

Warnings and
Disrupons Prompts Site condions
Incidents and
Near misses Compliance

Surrounding
Demographics
land use
Stakeholder Legislave
preferences reqs
Separaon

Traffic Competence
Design
Tender reqs
concepts

Evidence Final RLX Budget

design

Complaints and
Migaons
feedback

= Key node (both sociometric status and centrality)

= Key node (sociometric status only)

= Key node (centrality only)

FIGURE 11.6 Information network for RLX system lifecycle.

STAMPING on EAST 221

TABLE 11.4
Information Network Analysis Outputs

Key Nodes Sociometric Status Key Nodes Centrality

Network Metrics (Mean = 0.17, SD = 0.1) (Mean = 16.05, SD = 6.04)
Nodes = 28 Risk level = 0.55 Budget = 27.36
Edges = 65 Design concepts = 0.37 Demographics = 26.85
Density = 0.09 Tender requirements = 0.3 Mitigations = 26.35
Past experience = 0.26 Speed = 23.72
Incidents and near misses = 0.22 Surrounding land use = 22.23
Mitigations = 0.22 Final RLX design = 21.89
Separation = 0.19 Stakeholder preferences = 21.89
Compliance = 0.19 System state = 21.56
Warnings and prompts = 0.19 Site conditions = 20.62
Policies and procedures = 0.19 Direction = 20.62
Design standards = 0.19 Disruptions = 19.49
Site conditions = 0.19 Complaints and feedback = 17.35
Stakeholder preferences = 0.19
Legislative requirements = 0.19

Note: Cells Shaded Grey Represent Nodes with Values above the Mean + Standard Deviation.

TABLE 11.5
Control Network Analysis Outputs

Key Nodes Sociometric Status Key Nodes Centrality

Network Metrics (Mean = 0.22, SD = 0.13) (Mean = 18.70, SD = 14.74)
Nodes = 30 Audits and inspections = 0.66 Insurance premiums = 89.8
Edges = 97 Design standards = 0.59 Timetables = 34.53
Density = 0.11 Road rules = 0.37 Train protection devices = 37.41
Train = 0.34 Signals = 29.93
Initiatives = 0.31 Education = 23.63
Vehicle = 0.28
Policy and strategy = 0.28
Targets = 0.28
Legal penalties = 0.24
Enforcement = 0.24
Education = 0.24

Note: Cells Shaded Grey Represent Nodes with Values above the Mean + Standard Deviation.

The intention was to extend EAST’s utility for analysing the behaviour of complex
sociotechnical systems. The addition of a control network analysis component to
EAST enables analysts to identify the system of controls present within a par-
ticular system and to examine how different controls are related to one another.
222 Systems Thinking in Practice

= Key node (both sociometric status and centrality)

= Key node (sociometric status only)

= Key node (centrality only)

Policy/
Finances/
Strategy Targets
funding
Franchise
agreements Design
standards
Education
Initiatives

Social
norms
Road rules

Safe
working
rules Legal
penalties
e.g. fines
Boom
gates

Enforcement

Flashing
lights

SOPs

Road
markings

Accreditation,
Registration
and Licencing
Signage

Audits and
Inspections
Road

Training
Vehicle

Insurance
premiums
Fences

Timetables
Track

Signals
Horn Train
Train protection
devices

FIGURE 11.7 Control network for RLX system lifecycle.

Integration with the task, social and information networks also allows analysts
to examine the relationship between controls and tasks, agents and information
as well as the extent to which controls are likely to be effective. Finally, network
metrics can be used to compare the relevant importance of controls. This provides
an additional capability over and above the STAMP control structure, and it is
anticipated that it will useful for safety critical systems that rely heavily on the use
of controls to manage risk.
An important contribution of the analysis presented was the use of EAST to
describe and analyse the overall sociotechnical system – in this case, the RLX
STAMPING on EAST 223

system design lifecycle. This is a key capability of the framework; however, few
EAST analyses have sought to examine overall sociotechnical systems. The analysis
presented in this chapter demonstrates that it is possible to use EAST to analyse the
behaviour of overall systems and that a richer analysis is produced in doing so. It is
hoped that readers of this book will be encouraged to apply EAST to analyse entire
sociotechnical systems in other areas.

Implications for Railway Level Crossing Safety Management

Examination of the four networks reveals some important implications for RLX
safety management. The task network identified risk management and perfor-
mance monitoring as critical tasks. Further, the information network identified
‘risk level’ as one of the key information requirements, whereas the control net-
work identified audits and inspections as one of the key safety control measures.
Assessing the risks at different RLXs and subsequently implementing appropri-
ate risk management strategies therefore ostensibly represent key activities within
the RLX system lifecycle. Despite this, recent Australian studies have suggested
that elements of these tasks may be sub-optimal, including a lack of incident and
near miss reporting and analysis mechanisms (Salmon et al. 2016) and inadequate
risk assessment tools (Salmon et al. 2013). The findings suggest that road and rail
stakeholders should attempt to strengthen their risk management and performance
monitoring activities through the development of more comprehensive risk man-
agement processes (e.g. incorporating human factors data) and performance moni-
toring controls (e.g. incident and near miss reporting systems, regular behavioural
assessments and increasing data collection mechanisms at RLXs such as closed-
circuit television cameras).
All four of the networks had relatively low density scores, indicating that the
RLX system lifecycle as a whole is loosely coupled. This intuitively makes sense,
as the analysis describes activities, decisions and actions that occur across various
temporal and spatial planes. However, it may be that attempting to increase the cou-
pling of tasks, agents, information and controls in some cases will prove beneficial
for safety management. For example, in the social network, standards-setting bod-
ies and local governments were not well connected with other agents. Ostensibly,
there would be benefits in better integrating both in the RLX system lifecycle. This
would ensure that design standards are flexible and appropriate and are updated
based on feedback from other agents regarding performance throughout the RLX
system lifecycle. This suggestion is confirmed through the control network in which
design standards was identified as one of the key controls in terms of impact on
other controls.

CONCLUSION
This chapter has demonstrated an extended version of the EAST framework that
enables it to analyse risk controls in addition to task, social and information net-
works. This extension enhances the analytical and explanatory power of the EAST
framework, and the analysis presented provides a rich understanding of the RLX
224 Systems Thinking in Practice

system lifecycle and RLX safety management activities. Potential improvements

around design standards, the coupling of agents and organisations, risk management
and performance monitoring and the clarification of roles and responsibilities across
the wider system were identified. It is hoped that researchers and practitioners con-
duct further applications of the modified EAST framework across the safety critical
domains.

REFERENCES
Allison, C., Revell, K. M. A., Sears, R. and Stanton, N. A. (2017). Systems Theoretic Accident
Model and Process (STAMP) safety modelling applied to an aircraft rapid decompres-
sion event. Safety Science, vol. 98, October 2017, pp. 159–166.
Australian Transport Safety Bureau. (2012). Australian Rail Safety Occurrence Data: 1
July 2002 to 30 June 2012 (ATSB Transport Safety Report RR-2012-010). Canberra,
Australia: ATSB.
Chen, L., Zhao, Y. and Zhao, T. (2015). An AcciMap analysis on the China-Yong-Wen rail-
way accident. In: P. W. Tse et al. (Eds.), Engineering Asset Management-Systems,
Professional Practices and Certification. New York City, NY: Springer International
Publishing, 1247–1253.
Dallat, C., Salmon, P. M. and Goode, N. (in press). Risky systems versus risky people: To
what extent do risk assessment methods consider the systems approach to accident
causation? A review of the literature. Safety Science, accepted for publication March
2017.
Evans, A. W. (2011). Fatal accidents at railway level crossings in Great Britain 1946–2009.
Accident Analysis & Prevention, 43(5), 1837–1845.
Federal Railroad Administration. (2013). Railroad Safety Statistics 2012 Preliminary Annual
Report. Department of Transportation, Washington, D.C.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49, 1204–1225.
Houghton, R. J., Baber, C., Stanton, N. A., Jenkins, D. P., Revell, K. (2015). Combining net-
work analysis with Cognitive Work Analysis: insights into social organisational and
cooperation analysis. Ergonomics, 58:3, 434–449.
Independent Transport Safety Regulator. (2011). Level crossing accidents in Australia. ITSR,
Sydney. Available at: http://www.onrsr.com.au/__data/assets/pdf_file/0020/2963/Transport-
safety-bulletin-Issue-2-Level-crossing-accidents-in-Australia-August-20112.PDF.
Accessed 10/02/15.
Jansson, A., Olsson, E. and Erlandsson, M. (2006). Bridging the gap between analysis and
design: Improving existing driver interfaces with tools from the framework of cognitive
work analysis. Cognition, Technology & Work, 8, 41–49.
Karsh, B. T., Waterson, P. and Holden, R. J. (2014). Crossing levels in systems ergonomics:
A framework to support ‘mesoergonomic’ inquiry. Applied Ergonomics, 45(1), 45–54.
Kirwan, B. (1992). Human error identification in human reliability assessment: Part 2:
detailed comparison of techniques. Applied Ergonomics, 23, 371–381.
Lee, S., Bo Moh, Y., Tabibzadeh, M. and Meshkati, N. (2017). Applying the AcciMap
methodology to investigate the tragic Sewol Ferry accident in South Korea. Applied
Ergonomics, 59(B), 517–525.
Leveson, N. G. (2004). A new accident model for engineering safer systems. Safety Science,
42(4), 237–270.
STAMPING on EAST 225

Madigan, R., Golightly, D. and Madders, R. (2016). Application of Human Factors Analysis
and Classification System (HFACS) to UK rail safety of the line incidents. Accident
Analysis & Prevention, 97, 122–131.
Naweed, A. (2014). Investigations into the skills of modern and traditional train driving.
Applied Ergonomics, 45(3), 462–470.
Naweed, A., Rainbird, S. and Dance, C. (2015). Are you fit to continue? Approaching rail
systems thinking at the cusp of safety and the apex of performance. Safety Science,
76, 101–110.
Newnam, S., and Goode, N. (2015). Do not blame the driver: A systems analysis of the causes
of road freight crashes. Accident Analysis & Prevention, 76, 141–151.
Office of the National Rail Safety Regulator. (2016). Rail Safety Report 2015–2016. Office of
the National Rail Safety Regulator: Adelaide, Australia.
Ouyang, M., Hong, L., Yu, M-H. and Fei, Q. (2010). STAMP-based analysis on the railway
accident and accident spreading: Taking the China–Jiaoji railway accident for example.
Safety Science, 48(5), 544–555.
Parnell, K., Stanton, N. A. and Plant, K. L. (2017). What’s the law got to do with it? Legislation
regarding in-vehicle technology use and its impact on driver distraction. Accident
Analysis and Prevention, 100, 1–14.
Patriarca, R., Bergström, J. and Di Gravio, G. (2017). Defining the functional resonance anal-
ysis space: Combining abstraction hierarchy and FRAM. Reliability Engineering &
System Safety, 165, 34–46.
Rasmussen, J. (1997). Risk management in a dynamic society: A modelling problem. Safety
Science, 27(2/3), 183–213.
Read, G., Salmon, P. M. and Lenné, M. G. (2013). Sounding the warning bells: The need for a
systems approach to rail level crossing safety. Applied Ergonomics, 44, 764–774.
Read, G., Salmon, P. M. and Lenné, M. G. (2016). When paradigms collide at road and rail
interface. Ergonomics, 59(9), 1135–1157.
Read, G., Salmon, P. M, Lenné, M. G. and Jenkins, D. P. (2015). Designing a ticket to ride
with the Cognitive Work Analysis Design Toolkit. Ergonomics, 58(8), 1266–1286.
doi:10.1080/00140139.2015.1013576.
Read., G. J. M., Beanland, V., Lenné, M. G., Stanton, N. A. and Salmon, P. M. (2017).
Integrating Human Factors Methods and Systems Thinking for Transport Analysis and
Design. CRC Press: Boca Raton, FL.
Salmon, P. M., Lenné, M. G., Mulvihill, C., Young, K., Cornelissen, M., Walker, G. H. and
Stanton, N. A. (2016b). More than meets the eye: Using cognitive work analysis to
identify design requirements for safer rail level crossing systems. Applied Ergonomics,
53(Part B), 312–322.
Salmon, P. M., Lenné, M. G., Walker, G. H., Stanton, N. A. and Filtness, A. (2014). Using the
Event Analysis of Systemic Teamwork (EAST) to explore conflicts between different
road user groups when making right hand turns at urban intersections. Ergonomics,
57(11), 1628–1642.
Salmon, P. M., Read, G., Stanton, N.A . and Lenné, M. G. (2013). The Crash at Kerang:
Investigating systemic and psychological factors leading to unintentional non-compli-
ance at rail level crossings. Accident Analysis and Prevention, 50, 1278–1288.
Salmon, P. M., Read, G. J. M. and Stevens, N. (2016). Who is in control of road safety?
A STAMP control structure analysis of the road transport system in Queensland,
Australia. Accident Analysis and Prevention, 96, 140–151.
Salmon, P. M., Stanton, N. A., Jenkins, D. P. and Walker, G. H. (2010). Hierarchical task
analysis versus cognitive work analysis: Comparison of theory, methodology, and con-
tribution to system design. Theoretical Issues in Ergonomics Science, 11(6), 504–531.
226 Systems Thinking in Practice

Salmon, P. M., Walker, G. H., Read, G. J. M., Goode, N. and Stanton, N. A. (2017). Fitting
methods to paradigms: Are ergonomics methods fit for systems thinking? Ergonomics,
60(2), 194–205.
Stanton, N. A. (2006). Hierarchical task analysis: Developments, applications, and exten-
sions. Applied Ergonomics, 37(1), 55–79.
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57(3), 403–418.
Stanton, N. A., Harris, D. and Starr, A. (2016). The future flight deck: Modelling dual, single
and distributed crewing options. Applied Ergonomics, 53, 331–342.
Stanton, N. A., Salmon, P. M., Rafferty, L., Walker, G., Baber, C. and Jenkins, D. P. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design. Second
Edition. Ashgate: Aldershot, UK.
Stanton, N. A., Salmon, P. M., Walker, G. H., Baber, C. and Jenkins, D. (2005) Human
Factors Methods: A Practical Guide for Engineering and Design. First Edition.
Ashgate: Aldershot, UK.
Stanton, N. A. and Walker, G. H. (2011). Exploring the psychological factors involved in the
Ladbroke Grove rail accident. Accident Analysis & Prevention, 43(3), 1117–1127.
Stefanova, T., Burkhardt, J.-M., Filtness, A., Wullems, C., Rakotonirainy, A. and Delhomme, P.
(2015). Systems-based approach to investigate unsafe pedestrian behaviour at level
crossings. Accident Analysis & Prevention, 81, 167–186.
Tooth, R. and Balmford, M. (2010). Railway Level Crossing Incident Costing Model. LECG
for the Rail Industry Safety and Standards Board: Canberra, Australia.
Vicente, K. J. (1999). Cognitive Work Analysis: Toward Safe, Productive, and Healthy
Computer-based Work. Lawrence Erlbaum Associates: Mahwah, NJ.
Vivekanandan Dhukaram, A. and Baber, C. (2015). Modelling elderly cardiac patients deci-
sion making using Cognitive Work Analysis: Identifying requirements for patient deci-
sion aids. International Journal of Medical Informatics, 84(6), 430–443.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M. and Green, D. (2006).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics, 49, 1345–1369.
Walker, G. H., Stanton, N. A., Salmon, P. M. (2011). Cognitive compatibility of motorcyclists
and car drivers. Accident Analysis & Prevention, 43 (3): May 2011, pp. 878–888.
Walker, G. H., Salmon, P. M., Bedinger, M., Stanton, N. A. (2017). Quantum ergonomics:
Shifting the paradigm of the systems agenda. Ergonomics, 60(2), 157–166.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Jenkins, D. P. and Salmon, P. M. (2010).
From ethnography to the EAST method: A tractable approach for representing distrib-
uted cognition in air traffic control. Ergonomics, 53(2), 184–197.
Wilson, J. R., Norris, B. J. (2005). Rail human factors: Past, present and future. Applied
Ergonomics, 36(6), 649–660.
12 The EAST ‘ Broken-
Links’ Approach
Assessing Risk in
Sociotechnical Systems
With Catherine Harvey

INTRODUCTION
The term Sociotechnical Systems (STS) is used to refer to the interaction between
humans and machines, from the small and simple to the large and highly complex
(Walker et al. 2008; Walker et al. 2010b; Read et al. 2015). These sub-systems oper-
ate and are managed as independently functioning (autonomous) entities with their
own goals, but they must collaborate with other sub-systems to achieve the higher
goals of the STS (Dul et al. 2012; Wilson 2012). A key characteristic is that these
goals can only be achieved by the STS and not by individual sub-systems function-
ing in isolation (Rasmussen 1997; von Bertalanffy 1950). STS present unique chal-
lenges for safety management and risk assessment (Rasmussen 1997; Alexander and
Kelly 2013; Flach et al. 2015; Waterson et al. 2015). Traditional approaches to risk
assessment, such as THERP (Technique for Human Error Rate Prediction) Swain and
Guttmann 1983), TRACEr (Technique for the Retrospective and Predictive Analysis
of Cognitive Errors) Shorrock and Kirwan 2000) and SHERPA (Systematic Human
Error Reduction and Prediction Approach) Embrey 1986), are typically reductionistic
in nature (Stanton et al. 2013), focusing on individual tasks and technologies rather
than the system as a whole (Stanton 2006; Stanton and Stevenage 1998; Stanton et al.
2009; Waterson et al. 2015). These methods use error taxonomies to identify risk, but
recent research has suggested that the term human error is obsolete (Dekker 2014). In
its place, the term human performance variability has been proposed, which includes
both normative and non-normative performance. This latter approach emphasises the
broad spectrum of human behaviour rather than a dichotomy and, therefore, a need
to build resilient systems (Hollnagel et al. 2006). The Systemic Accident Analysis
(SAA) approach treats systems as whole entities with complex, non-linear networks
(Underwood and Waterson 2013). A number of SAA methods were assessed for
their potential for prospective risk analysis within STS in a previous study (Stanton
et al. 2012). Some system methods incorporate error taxonomies, such as CREAM
(Cognitive Reliability and Error Analysis Method) Hollnagel 1998), HFACS (Human
Factors Analysis and Classification System) Shappell and Wiegmann 2001) and

227
228 Systems Thinking in Practice

STPA (System-Theoretic Process Analysis) (Leveson 2012), which, given the recent
shift away from the term human error is something of a conundrum. Rather than
considering risks in systems to be the result of error, the approach taken in this chap-
ter is to propose risks as the failure to communicate information via social and task
networks. This type of failure may be seen in several major incidents. For example,
in the MS Herald of Free Enterprise accident (1987), the state of the bow doors was
not communicated to the ship’ s bridge (Noyes and Stanton 1997), so the ship left
harbour and subsequently capsized. In the Kegworth air disaster (1989), the aircraft
failed to communicate which engine was on fire, leading to the pilots shutting down
the wrong engine (Griffin et al. 2010; Plant and Stanton 2012). In the Ladbroke Grove
rail incident (1999), the signals failed to communicate to the train driver that the sec-
tion of the rail network was protected (Moray et al. 2016). Rather than stopping, the
driver actually increased his speed as he passed the red signal, leading to a collision
with an oncoming high-speed intercity train (Stanton and Walker 2011). So, rather
than conceive these behaviours as errors, we have reconceived them as the failure to
communicate information in the system. This is a new paradigm for risk assessment
that incorporates the value of a holistic perspective for appreciating the relationships
between the various sub-systems and the network diagrams for visualising important
aspects of STS, such as the constraints on communications (see Flach et al. 2015). In
order to analyse the information communications in systems, the Event Analysis of
Systemic Teamwork (EAST) method was selected. EAST takes a different approach
to the error taxonomic methods by modelling and analysing STS-level interactions.
In a previous study, Stanton (2014) analysed communications between various actors
within a submarine control room; in contrast, this case study analyses a retrospec-
tive account of actions within a Royal Navy training activity and is conducted at the
macro level (Grote et al. 2014). The aim of this chapter is to extend the EAST network-
level analysis to include risk prediction by ‘ breaking’ links within networks.
The EAST method was first proposed by Stanton et al. (2005, 2013) and fur-
ther elaborated by Stanton et al. (2008) for modelling distributed cognition in STS.
The method represents distributed cognition in networks, which enables both quali-
tative and quantitative investigations to be performed (Stanton 2014). One of the
main advantages of EAST is its aim to capture the whole system, as opposed to
reductionist methods, which split a system into constituent parts for analysis (Walker
et al. 2010a). It is therefore considered in this study to be a suitable technique for
representing a STS and potential non-normative behaviours. The analysis describes
a system as three different types of network:

• Social : Representing the agents (human, technical and organisational)

within a system and communications between them
• Task : Representing the activities performed by the system and the relation-
ships between them
• Information : Representing the information that is used and communicated
within a system and links between different information types

The social, task and information networks are developed individually and then
combined to create a complete social-task-information network diagram, showing
The EAST ‘Broken-Links’ Approach 229

all the links and information flows (i.e. distributed cognition) within a network of
networks. EAST has been applied in many domains, including aviation (Stewart
et al. 2008, Walker et al. 2010a, Stanton et al. 2016), the military (Stanton et al. 2006,
Stanton 2014), road (Salmon et al. 2014) and rail (Walker et al. 2006) environments
and the emergency services (Houghton et al. 2006, Baber et al. 2013). The aim of
this work is to extend the EAST method to consider risk in systems via a case study
and to provide an initial STS method evaluation criteria presented by Harvey and
Stanton (2014). The premise of the risk assessment is that STS failures are predomi-
nately caused by the failure to communicate information between agents and tasks.
This will be studied within the context of the following case study.

CASE STUDY OF HAWK MISSILE SIMULATION TRAINING

Operation of the Hawk jet to simulate missile attacks against surface ships by the UK
Royal Navy was selected as the case study. This activity is viewed as a STS because
it comprises many interconnected sub-systems, which are themselves complex. The
context for this study is illustrated in an AcciMap (Rasmussen 1997; Jenkins et al.
2010) in Figure 12.1. The AcciMap places different subsystems within the STS at
different levels and shows the links in communication and decision making between
the sub-systems. Each node in the AcciMap is labelled (a, b, c, etc.) to correspond
with the description in the text in the following paragraphs. The year in which each
event occurred is also included, where applicable, to give an indication of timescale.
The Royal Navy uses the Hawk jet to simulate air attacks on ships during sea
training of ships’ gunners and radar operators (event ‘ a’ in Figure 12.1). The Hawks

FIGURE 12.1 AcciMap showing sub-systems within the Hawk jet STS [Please note that the
labels do not indicate a timeline; rather, they are added for clarity of the description below].
230 Systems Thinking in Practice

are used to simulate enemy aircraft attacks and high-speed skimming missiles fired
against ships (Royal Navy 2012). In order to perform these simulation activities, the
Hawk must be flown at a low height above sea level (b); however, the Hawk is not
equipped with a Radar Altimeter (Rad-Alt), which provides a highly accurate mea-
sure of the altitude of the aircraft above the sea. This makes flying the Hawk accu-
rately at very low levels extremely difficult, and a high level of expertise is required
to perform this safely. Prompted by events over a number of years, there have been
some significant changes to the method for assessing the safety of the Hawk STS.
In 2000, a Hawk jet was involved in a sea-strike incident as a consequence of
very low-level flight (c). Although there was no resulting loss of life, this incident
prompted a decision by the Royal Navy to increase the minimum allowable flying
height above sea level for the Hawk (d).
As part of its Strategic Defence and Security Review, HM Government (2010)
took the decision to retire the Harrier jet from service in October 2010 (e). As a
result of this, a number of Royal Navy pilots who would have flown the Harrier were
diverted into the Hawk program (f). Traditionally, the Hawk has been flown by civil-
ian pilots under contract to the Royal Navy (2012); these pilots have extensive mili-
tary experience in fast jets, which includes low-level flight supported by a Rad-Alt.
This experience provided mitigation against the Risk to Life (RtL) for the Hawk air
attack simulation task; however, the cohort of military pilots did not have this same
level of experience, and the RtL had to be reassessed in light of this (g).
In 2006, RAF Nimrod XV230 suffered a catastrophic explosion after a mid-air
refuelling procedure (h); this caused the deaths of all 12 crew members plus two
mission specialists and the total loss of the aircraft (i). The government requested
a comprehensive review into the airworthiness and safe operation of the Nimrod
(j), which was delivered by Haddon-Cave (2009). The report described the develop-
ment of the safety case for the Nimrod as ‘ a story of incompetence, complacency,
and cynicism’ (p. 161) and concluded that it was undermined by the widespread
assumption that the Nimrod was safe because it had been flown successfully for the
preceding 30 years (Haddon-Cave 2009). The report also identified organisational
changes in the years prior to the Nimrod accident as having significant influence;
these included a shift in organisational culture toward business and financial targets
‘ at the expense of functional values such as safety and airworthiness’ (Haddon-Cave
2009, p. 355). As a consequence of the findings, Haddon-Cave (2009) recommended
the establishment of an independent Military Aviation Authority (MAA) to properly
assess RtL and shape future safety culture (k). Further recommendations included
the need for strong leadership , a greater focus on people to deliver ‘ high standards
of safety and airworthiness’ (Haddon-Cave 2009, p. 355) and increased simplicity
of rules and regulations. The tragic consequences of the Nimrod accident, along
with the recommendations of the Haddon-Cave report, effected a culture change
within military aviation; this resulted in a decision to assign individual accountabil-
ity for RtL assessments to ‘ Duty Holders’ (DH), whereas previously, responsibility
for risk had been held at the organisation level (l). The newly established MAA pro-
duced guidelines for the assessment of RtL in the form of the Defence Aviation
Hazard-Risk Matrix (MAA 2011), which supports the classification of single risks
according to their estimated severity (catastrophic, critical, major, minor) and
The EAST ‘Broken-Links’ Approach 231

likelihood (frequent, occasional, remote, improbable). The resulting risk level deter-
mines at which level of DH the risk is held.
The organisational changes brought about by the events described above (i.e.
influx of junior pilots) prompted reassessment of the RtL for the Hawk air attack sim-
ulation activity. The goal of safety management in the UK military is to reduce risk to
a level which is As Low As Reasonably Practicable (ALARP); this is reached when
‘ the cost of further reduction is grossly disproportionate to the benefits of risk reduc-
tion’ (Ministry of Defence 2007). The RtL for all Hawk operations is frequently reas-
sessed, and the shift in pilot experience levels, as described above, prompted changes
to the RtL for the Hawk air attack simulation activity. In order to reduce this RtL to a
level that was ALARP, a decision was taken by the Royal Navy DH with subject mat-
ter expert (SME) advice to further increase the minimum height above sea level (m).
A potential consequence of this decision is the degradation of Royal Navy surface
fleet training against very live low-level targets, as the Hawk can no longer accu-
rately simulate sea-skimming missile attacks on surface ships (n). These events
have changed the nature of Hawk operations within the UK Ministry of Defence
(MoD(2007)) (o).
Potential risks to the safe operation of the missile simulation activity are, in part,
assessed according to the MAA’s Regulatory Articles (MAA 2011). This assessment is
based on the principle that risks can be tolerated provided they are reduced to ALARP.
The MAA regulatory policy outlined its approach to the management of RtL:

Aviation DHs [Duty Holders] are bound to reduce the RtL within their AoR [Area of
Responsibility] to at least tolerable and ALARP; the application of effective and coher-
ent risk management processes will be fundamental to achieving this.
(MAA 2011, p. 18).

Regulatory Article (RA) 1210 – Management of Operating Risk (Risk to Life) –

defined risk as:

a measure of exposure to possible loss [combining] severity of loss (how bad) and the
likelihood of suffering that loss (how often).
(MAA 2011, p. 1).

The MAA suggested that risks can be identified via a number of different meth-
ods including previous occurrences, checklists, HAZOPS (Hazard and Operability),
zonal hazard (safety) analyses and error trend monitoring. Previous work has showed
that these techniques are likely to be inadequate for the analysis of STS (Stanton
et al. 2012). RA 1210 specifically encourages the use of fault trees as accident mod-
els ‘ to assist understanding of the interrelationship between risks and to support the
prioritisation of effort to maximise safety benefit’ (MAA 2010, p. 6). This technique,
along with other traditional error and risk prediction methods, does not account for
the interactions of distributed actors within a STS (Salmon et al. 2011b). Furthermore,
there is also no clear method outlined by the MAA for structuring risk identification;
for example, the recommendation is that a combination of these methods should be
used with the aim of identifying all credible risks, but there is no way of knowing
232 Systems Thinking in Practice

when all possible credible risks have been defined and therefore how many methods
to use and when to stop applying them.
The Hawk RtL case study was identified through interviews with a subject matter
expert (SME) as part of this project. The analysts were provided with a high-level
overview of the case study in an initial interview with the SME. This was followed
up by a second, in-depth interview about the case study with the SME, conducted
by two analysts. This resulted in a detailed account of the Hawk-Frigate STS, which
was supplemented by extra information from official documentation including MAA
(2010, 2011) guidelines, the official report into the Nimrod accident (Haddon-Cave
2009) and Royal Navy safety assessment guidance (Royal Navy 2012). The EAST
method (Stanton 2014) was used to develop the three network diagrams, based upon
the analysis of all case study information, in an iterative process that involved the
SME providing feedback during development.

ANALYSIS OF NETWORKS
Social Network Analysis (SNA) metrics provide quantitative measures that represent
the structures and relations between nodes in the EAST networks (Baber et al. 2013;
Driskell and Mullen 2005; Walker et al. 2009). The SNA metrics describe individual
nodes (including reception, emission, eccentricity, sociometric status, centrality, close-
ness, farness and betweenness). The SNA metrics applied in the current study, along
with their descriptions, are presented in Table 12.1. Analysis software, AGNA version
2.1 (Benta 2005), was used to calculate the SNA metrics. For each EAST network, key

TABLE 12.1
SNA Metrics, along with their Descriptions

Safety Constraint Description

Node-level Emission The number of edges (links to other nodes)
metrics Reception originating at that node
Sociometric status The number of edges incident to that node
Bavelas-Leavitt (B-L) centrality Number of communications received and emitted
Eccentricity relative to the number of nodes in the network
Closeness Degree of connectivity to other nodes in the network
Standard closeness Length of the longest geodesic path originating in
Farness that node (a geodesic path is the path between two
Betweenness given nodes that has the shortest possible length)
Inverse of the sum of the geodesic distances from
that node to all the other nodes, that is, the extent
to which a node is close to all other nodes
Closeness multiplied by (g-1), where g is the
number of nodes in the network
Sum of the geodesic distances from that node to all
other nodes
Frequency with which a node falls between pairs of
other nodes in the network
The EAST ‘Broken-Links’ Approach 233

nodes were identified according to sociometric status. Sociometric status was selected
to define key nodes because it identifies the prominence of an individual node’s com-
munications with the rest of the network, which influences the whole network’s perfor-
mance (Stanton 2014). In a STS, all of the nodes will have complex safety management
rules and behaviours; however, as the ‘key’ nodes have the largest number of connec-
tions to the rest of the network, these nodes will have the highest degree of influence
over the behaviour of the entire STS. Sociometric status key nodes are defined as nodes
that have a higher sociometric status score than the sum of the mean sociometric status
score plus the standard deviation sociometric status score for all nodes in the network.
SNA metrics were calculated for the EAST networks created for this case study; key
agents for sociometric status are indicated in the social, task and information network
diagrams below.

RESULTS
The step-by-step application of the shortened version of EAST is described in detail
in the following sections. This is accompanied by the outputs of the method along
with interpretation of the results. The first stage in EAST was the identification of all
social, task and information nodes within the Hawk missile simulation case study,
based on the SME’ s account of activities, which informed the analysts’ knowledge
of the case study. The nodes were arranged in social, task and information networks
and links drawn between related nodes. Related nodes were those between which
some information was transferred. As well as providing a visual representation of
a STS, the EAST network diagrams can be analysed to produce quantitative SNA
metrics.

Social Network
Seven social ‘ agents’ and their connections were identified from the Hawk RtL case
study with the SME; these are shown in Figure 12.2. The social network was con-
structed by first identifying the main agents that are in the system, then by examin-
ing the interdependencies between those agents. The SME agreed that the social
network was a reasonable representation of the main agents and their relationships.
The ‘ edges’ , or links, between the agents show where information is transferred
and the direction of transfer. There are 19 edges in total; in some cases, information
transfer is reciprocal, but in others it only goes in one direction between two agents.
The ‘ pilot’ node was identified as the key agent according to sociometric status
(1.33), with the highest number of links to and from other nodes in the network; in
fact, the pilot receives and/or emits information from/to all of the other agents in the
social network.
The pilot had the highest betweenness score (10.0) as it is located on the paths
between a number of other agent pairs. The pilot also had the highest score for recep-
tion (6), highlighting a high degree of connectivity to other nodes in the network and
indicating that the pilot’ s actions and communications are integral to the function-
ing of the STS. The high farness score for the regulator (10) indicates that this agent
is located farthest from most other nodes, and this is supported by the information
234 Systems Thinking in Practice

FIGURE 12.2 Social nodes and their links within the STS.

in the case study, which showed that the regulator really only communicates with
the DH and possibly the pilots but has no contact with the frigate or crew. This is
because the regulator in this case is the MAA, which does not have direct control
over the Navy’ s surface ship operations. The sea scored highest for emission (4) and
lowest for reception (0) as it does not receive information from any other nodes but is
used for feedback only. In this sense, the sea can be regarded as a ‘ passive’ agent, as
it cannot respond to feedback; the social agents can only respond to it.

Task Network
Ten tasks nodes and their connections were identified from the Hawk RtL case study
with the SME; these are shown in Figure 12.3. The task network was constructed by
first identifying the main tasks that are performed by the system, then by examin-
ing the interdependencies between the tasks. The SME agreed that the task network
was a reasonable representation of the system. There are 12 edges in total, and in all
cases, the transfer is uni-directional.
In the task network, key nodes were identified as ‘ safe control of aircraft to simu-
late missile’ and ‘ issuing of RtL document’ , which had sociometric status scores of
.56 and .44 respectively. The task network contains more nodes but fewer edges than
the social network, indicating that there are fewer communications between tasks.
The EAST ‘Broken-Links’ Approach 235

FIGURE 12.3 Task nodes and their links within the STS.

Cohesion is zero because there are no mutual, or bidirectional, links between nodes.
The highest score for betweenness was for ‘ safe control of aircraft to simulate mis-
sile’ (45), demonstrating that this task is integral in the STS as it is located between a
high number of other task nodes. This is unsurprising as this task can be considered
to be the main objective of the STS configuration investigated in this case study. This
task also scored highest on emission (2), reception (3) and B-L centrality (6.26), as
well as sociometric status (.56), showing a high level of connectivity to other nodes.

Information Network
EAST identified 25 information nodes, and their connections were based the Hawk
RtL case study and further knowledge of the STS from the SME; these are shown
in Figure 12.4. There are 50 edges in total; however, in this case, the links are not
directional.
236 Systems Thinking in Practice

FIGURE 12.4 Information nodes and their links within the STS.

Six information nodes were identified as key nodes according to sociometric sta-
tus: missile simulation, (absence of) Rad-Alt, altitude, boundaries, safety and RtL.
Safety had the highest betweenness (315.8), standard closeness (.53) and B-L cen-
trality scores (18.51), and this reflects the importance of this in the case study; the
aim within the STS is to achieve a safe solution for missile simulation. Density and
cohesion were relatively low, that is, compared with the social network, as the edges
between nodes were single and non-directional.
The EAST ‘Broken-Links’ Approach 237

BROKEN-LINKS ANALYSIS
Studies of networks have discussed the effects of removing one or more nodes from a
network on the resilience of that network to systemic failures and the resulting desta-
bilisation (Baber et al. 2013; Houghton et al. 2008; Stanton 2014). This has been used
to explore the resulting influence on network structure, rather than as a method for
predicting specific risks. Previously, the network diagrams in EAST have been used
to provide a visual representation of a system to further the users’ understanding
of distributed cognition (Stanton 2014). In this study, however, the EAST network
analysis was extended to identify and examine possible risks by ‘ breaking’ the links
between the various nodes, in a similar approach to the removal of nodes, to explore
system effects. ‘ Broken links’ represent failures in communication and information
transfer between nodes in the networks and these failures can then be used to make
predictions about the possible risks within the STS. Previously, broken links have
only been investigated by EAST analysts when looking retrospectively at accidents
to identify underlying causes. Griffin et al. (2010) demonstrated that the broken link
between the Engine Vibration Indicator and the pilots in the cockpit was a causal
factor in their failure to shut down the correct engine in the Kegworth accident. If
this information had been communicated more effectively, it could have helped to
prevent the crash. Similarly, the EAST method has been adapted to analyse incidents
of fratricide (Rafferty et al. 2012), although this has been conducted as retrospec-
tive and concurrent, rather than predictive, analyses. The broken-links analysis was
performed for the Hawk missile simulation case study on the social and task net-
works shown in Figures 12.2 and 12.3 respectively. The information network was
not subject to the broken-links analysis because broken links between information
nodes were not considered to represent risks as they are caused by a failure in either
the social or task networks. In other words, information does not fail in isolation;
it is the failure to use or communicate the information correctly, and in all cases,
this can be attributed to social nodes, task nodes or both. For the social and task
networks, each link was identified and documented in a table. The combined EAST
networks diagram (see Figure 12.5) shows the information network tagged with the
social networks nodes (to show who owns each information node in the network)
and grouped by the task network nodes (to show which task each information node
belongs to). Details on construction of the combined network have been reported by
Stanton (2014). This combined network was used to identify what information (from
the information network) should be communicated from the origin node to the des-
tination node in the task and social networks and therefore what information would
not be communicated if the links between the nodes in the task and social networks
were removed.
Figure 12.5 also shows the combined information-task-social network as a single
depiction of the entire STS. This shows the overlaps between the three networks, in
other words, which information is being communicated by which agents in which
tasks and how these nodes are interlinked.
In order to conduct the broken-links analysis, the social and task networks were
compared to the combined information-task-social network in turn. For example,
there is a reciprocal relationship between the duty holder and the pilot (in the social
238 Systems Thinking in Practice

FIGURE 12.5 Combined information-task-social network for the Hawk case study (shading
represents the different tasks being undertaken).

network shown in Figure 12.2), and the duty holder and the pilot share the nodes of
boundaries, RtL, risk likelihood, risk severity, procedures, document, responsibility
and safety (in the combined information-task-social network shown in Figure 12.5).
The risk assessment procedure requires that the relationship between the duty holder
and pilot be interrogated to see what would happen if each information element was
not transmitted, as shown in Table 12.2. The pilot was identified as having the high-
est Sociometric Status in the analysis presented in Figure 12.2, so was chosen for the
The EAST ‘Broken-Links’ Approach 239

TABLE 12.2
Extract from Broken-links Analysis for EAST Social Network
From (Agent) To (Agent) Information Resulting Risk Mitigation Strategy
not
Communicated
Duty Holder Pilot Boundaries Pilots are not aware of Boundaries for risk
the boundaries for reporting must be made
flight operations and clear to pilots as part of
for the identification the RtL process.
and reporting of risks
within this.
Duty Holder Pilot RtL Pilots are not made Results and
aware of the results and consequences of the
consequences of the RtL assessment process
RtL assessment process must be effectively
after it is conducted at communicated to pilots.
DH level.
Duty Holder Pilot Risk likelihood Pilots are not made Risks identified as
aware of risks having a high
assessed that their likelihood of
likelihood of occurrence must be
occurrence. reported to pilots.
Duty Holder Pilot Risk severity Pilots are not made Risks deemed as having
aware of risks a high severity of
assessed and their impact must be
severity of impact. reported to pilots.
Duty Holder Pilot Procedures Pilots are not aware of Pilots must be provided
how the RtL process is with clear procedures
conducted at DH level describing the
and of procedures for assessment of RtL at
reporting incidents to DH level and the
the DH. reporting of risks to DH.
Duty Holder Pilot Document Pilots are not provided Pilots must be provided
with documentation with documentation
covering the RtL covering the RtL
process and its results. process and its results.
Duty Holder Pilot Responsibility Pilots are not aware of The responsibilities of
the DH’ s or their own both the pilot and DH
responsibilities for for safety must be
safety. clearly defined and
understood by pilots.
Duty Holder Pilot Safety Pilots do not receive The safety of operations,
information about the as assessed during the
safety of operations RtL process, must be
based on the RtL reported to the pilots.
assessment process.
(Continued)
240 Systems Thinking in Practice

TABLE 12.2 (CONTINUED)

Extract from Broken-links Analysis for EAST Social Network
From To Information Resulting Risk Mitigation Strategy
(Agent) (Agent) not
Communicated
Pilot Duty RtL The DH does not receive Pilots must clearly report
Holder information about new all relevant risks to the
risks identified by the DH.
pilots.
Pilot Duty Risk likelihood The DH does not Pilots must report their
Holder receive information estimate of the
about the likelihood of likelihood of
new risks identified by occurrence of all
the pilots. relevant risks.
Pilot Duty Risk severity The DH does not receive Pilots must report their
Holder information about the estimate of the severity
severity of new risks of impact of all relevant
identified by the pilots. risks.
Pilot Duty Incident (near The DH does not receive Pilots must clearly report
Holder miss) information about all relevant incidents
incidents (or near that occur during Hawk
misses) that occur operations to the DH.
during Hawk operations.
Pilot Duty Experience The DH cannot learn Pilots must clearly report
Holder from the pilots’ their experience levels to
experience of Hawk the DH. Pilots must
operations and the report their assessment
risks encountered. of risks and any
consequent assumptions,
based on this experience.
Pilot Duty Skill The DH cannot learn Pilots must clearly report
Holder from the pilots’ skill their skill levels to the
in Hawk operations. DH. Pilots must report
their assessment of risks
and any consequent
assumptions, based on
this skill level.
Pilot Duty Safety The DH does not Pilots must report their
Holder receive information estimates of the safety
about the safety of impact of any risks
Hawk operations. identified in Hawk
operations to the DH.
Pilot Duty Culture The DH is not aware of Pilots must consider and
Holder the culture of safety report the estimated
among the Hawk influence of safety
pilots. culture on the risks to
Hawk operations.
The EAST ‘Broken-Links’ Approach 241

illustration of the broken-links analysis in Table 12.2. Although the pilot is linked to
all other agents in the social network, for the purpose of illustration, just their recip-
rocal relationship with the duty holder is presented in Table 12.2.
Table 12.2 shows the risks resulting from the failure to pass relevant information
between duty holder and pilot and vice versa. Anecdotal evidence from our SME
suggests that there is variability in what individual pilots will chose to report back
to the duty holder, as they have different interpretations of what they consider to
be a risk and near miss. This shows that there is at least some face validity for the
approach we have proposed.
In the similar manner to the social-information broken-links analysis shown in
Table 12.2, there is a task-information broken-links analysis in Table 12.3. From the
task network, there is a uni-directional relationship between the ‘ Issuing of RtL
document’ and the ‘ Revision of safe flying altitude’ (in the task network shown
in Figure 12.3), and they overlap in the combined information-task-social network
(shown in Figure 12.5). The risk assessment procedure requires that the relationship
between the ‘ Issuing of RtL document’ and the ‘ Revision of safe flying altitude’ be
interrogated to see what would happen if each information element was not transmit-
ted, as shown in Table 12.3. The ‘ Issuing of RtL document’ was chosen as it has the
highest Sociometric Status in the analysis presented in Figure 12.3. The ‘ Safe con-
trol of aircraft to simulate missile’ was chosen for the same reason and is paired with
‘ Navy training’ for the purposes of offering an illustrative example of the method
in Figure 12.3.
Examination of the analysis in Tables 12.2 and 12.3 offers a systematic approach
for examining a system of operation in a holistic manner. For example, increasing
the safe flying attitude (see Figure 12.1) has led to the altitude profile of the Hawk not
matching that of the low-flying missile (see Table 12.3). This has meant that reducing
the risk for the Hawk pilots could have a negative effect on training of the crew on
the frigate, ultimately increasing their risk. So, whilst the top of Table 12.3 is about
improving the safety of the pilot – by increasing altitude, for example – the bottom of
table three shows that this could reduce the safety of the Navy frigate crew as they do
not receive realistic training. The benefit of systems approaches is that the emergent
properties become more readily apparent.

DISCUSSION
This work aimed to explore the use of a modified version of EAST (network model-
ling and broken-links analysis, see Stanton 2014) in a case study of a Royal Navy
training activity. First, the findings of this study are discussed in terms of criteria
that were identified as essential for methods designed to analyse the human com-
ponent of STS (Stanton et al. 2012; Harvey and Stanton 2014). This enabled com-
parisons to be made between the method and the current RtL procedure used in the
Hawk missile simulation case study. Second, the modifications and extensions to
EAST are discussed with reference to use of the method as an assessment of poten-
tial risks within a STS.
Aviation accidents, as with most accidents in STS, usually occur due to a conjunc-
tion of factors (Hodgson et al. 2013; Jenkins et al. 2010), and it is therefore essential
242 Systems Thinking in Practice

that analysis methods are able to explore all of these factors by taking an integrated
and holistic approach (Ramos et al. 2012; Salmon et al. 2011b). EAST specifically
enables the exploration of the social, task and information components of the STS,
allowing a high-level model of the STS to be created (visual diagrams) and analysed
(social network metrics). This visual component is likely to help analysts and other
stakeholders to understand the interactions within networks (Flach et al. 2015); this
is an advantage over many other methods such as HAZOP and Fault Tree analysis,
as well as the MAA’ s RtL/HRM (Hazard-Risk Matrix) approach. Baber et al. (2013)
argued that it is sensible to speak of a ‘ useful’ (rather than ‘ complete’ ) network, as
there will always be a possibility that some connections have been left out due to not
being observed, reported and/or documented. This is certainly applicable to the net-
works generated by EAST, as it is impossible to know whether an analysis has been
exhaustive, and it is therefore safest to assume that it has not. It is also particularly
true in this case as the analysis was performed on an SME’ s reports of activities
within the STS rather than communications between STS actors (as in Stanton 2014).
A consequence of this approach is a lack of richness of information, although if the
main contribution of EAST lies in its ability to visually represent an STS, then this
may not be a significant issue. EAST includes the calculation of SNA metrics, which
provide the analyst with quantitative values to represent various characteristics of
the networks. In this way, the analysis encompasses all elements of a STS and pro-
vides the analyst with an understanding of the structure of a system as a whole and
the relationships between individual system components. These metrics can provide
potential insight into the resilience of the networks (Stanton et al. 2016).
The inclusion of particular agents in an accident model is dependent on the infor-
mation put into the analysis and therefore on the analysts and SMEs involved. This
also is true for more traditional HAZOP and error identification methods and the
current RtL assessment process, as well as EAST. However, because HAZOP and
RtL assessment essentially focus on a list of potential errors, there is no formal pro-
cedure for identifying the decision makers involved. In contrast, EAST enabled a
visual representation of the decision-making agents (Flach et al. 2015) and their
relationships with other nodes in a STS to be constructed, thereby encompassing the
identification of decision makers into the analysis process. This can allow analysts to
understand where responsibility for risks resides within the STS and so target miti-
gation strategies appropriately (Lundberg et al. 2010). This case study showed that
EAST provided a useful visual representation of relationships between the various
components of the STS. EAST examines the links between nodes and so is focused
on communications, and therefore on the consequences of an action at a node, rather
than its causes (Rafferty et al. 2012; Walker et al. 2010a).
In this case study, the analysts used a modified version of EAST, concentrat-
ing on the social, information and task networks (Stanton 2014). Guidance is pro-
vided on structuring a model of the STS under investigation, and there are numerous
examples of previous EAST models (e.g. Griffin et al. 2010; Rafferty et al. 2012;
Walker et al. 2010a; Stanton 2014) in the literature. The ‘ broken-link’ process is very
straightforward indeed and would be a useful addition to the current RtL assessment
process (Haddon-Cave 2009). The current guidance from the MAA states that risks
should be identified from a number of sources including HAZOP, error data and
The EAST ‘Broken-Links’ Approach 243

TABLE 12.3
Extract from Broken-links Analysis for EAST Task Network
From (Task) To (Task) Information not Resulting Risk Mitigation Strategy
Communicated
Issuing of RtL Revision Document The information The RtL document must
document of safe contained in the be used by regulators to
flying RtL document does inform changes to
altitude not trigger a regulations and safety
revision of safe guidance where
flying altitude. appropriate.
Issuing of RtL Revision RtL The outcome of the The outcomes of RtL
document of safe RtL process assessment must be used
flying outlined in the RtL by regulators to inform
altitude document does not changes to regulations
trigger a revision of and safety guidelines
safe flying altitude. where appropriate.
Issuing of RtL Revision Risk likelihood The outcome of the The outcome of the Risk
document of safe Risk likelihood likelihood assessment,
flying assessment, conducted as part of the
altitude conducted as part of RtL process and outlined
the RtL process and in the RtL document,
outlined in the RtL must be used to inform
document, does not changes to regulations
trigger a revision of and safety guidelines
safe flying altitude. where appropriate.
Issuing of RtL Revision Risk severity The outcome of the The outcome of the Risk
document of safe Risk severity severity assessment,
flying assessment, conducted as part of the
altitude conducted as part of RtL process and outlined
the RtL process and in the RtL document,
outlined in the RtL must be used to inform
document, does not changes to regulations
trigger a revision of and safety guidelines
safe flying altitude. where appropriate
Issuing of RtL Revision Safety The safety The safety implications of
document of safe implications of the RtL assessment must be
flying RtL process, used by regulators to
altitude outlined in the RtL inform changes to
document, do not regulations and safety
trigger a revision of guidelines where
safe flying altitude. appropriate.
Issuing of RtL Revision Responsibility Responsibility for Responsibility for changes
document of safe the revision of safe to regulations and safety
flying flying altitude is not guidelines based on RtL
altitude outlined in the RtL assessment must be
document. clearly assigned and
accepted.
(Continued)
244 Systems Thinking in Practice

TABLE 12.3 (CONTINUED)

Extract from Broken-links Analysis for EAST Task Network
From (Task) To (Task) Information not Resulting Risk Mitigation Strategy
Communicated
Safe control Navy Missile The overall control The operation of the
of aircraft to training simulation of the Hawk does Hawk must aid Navy
simulate not adequately training for missile
missile simulate missile attack situations.
attack on the frigate
to aid with training.
Safe control Navy Speed The speed profile of The speed of the Hawk
of aircraft to training the Hawk does not during missile simulation
simulate adequately simulate must be sufficiently
missile missile attack on realistic to aid Navy
the frigate to aid training for missile
with training. attack situations.
Safe control Navy Altitude The altitude profile The altitude of the Hawk
of aircraft to training of the Hawk does during missile simulation
simulate not adequately must be sufficiently
missile simulate missile realistic to aid Navy
attack on the frigate training for missile
to aid with training. attack situations.
Safe control Navy Track The track of the The track of the Hawk
of aircraft to training Hawk does not during missile simulation
simulate adequately simulate must be sufficiently
missile missile attack on realistic to aid Navy
the frigate to aid training for missile
with training. attack situations.

experience of previous events; however, there are no explicit instructions on how

many of these methods to use and when to stop this analysis. This means that the RtL
assessment may proceed without a comprehensive list of potential risks. It appears
that EAST could be a useful model for ensuring that this does not happen; however,
it is important to note that provision of guidance may not be sufficient for successful
application of STS methods. The training requirements of these methods can often
be high for practitioners, with many citing a lack of time and difficulty accessing
new information as barriers to STS analysis (Underwood and Waterson 2013).
Stanton et al. (2012, 2014) previously suggested that EAST could be suitable for
prospective analysis of STS risks; however, these studies only demonstrated the util-
ity of methods for retrospective analysis (Salmon et al. 2011b; Waterson et al. 2015).
In this study, EAST has been applied to a STS that is currently in operation in order
to investigate the ability of methods to model the future state of a STS. The Hawk
missile simulation STS has already experienced and been impacted by incidents (e.g.
Hawk sea strike) and accidents (e.g. Nimrod), but this analysis focused on the predic-
tion of a future state given the changes in the STS, such as the alteration in safe flying
The EAST ‘Broken-Links’ Approach 245

altitude for the Hawk and the effects of this on missile simulation for the frigate and
crew. Having said this, the emphasis with EAST is not on predicting accidents per se;
rather, it is about creating a comprehensive model of the links and information flows
within the STS and by doing so making the analysts aware of potential breakdowns
and failures that may occur in the future. This means that the success of EAST for
prospective analysis is dependent on the participation in the assessment process of
those who will be impacted by these failures and those that can apply the appropriate
mitigation strategies.
In summary, this study used a modified version of EAST, following the exam-
ples in Stanton (2014). In this case, only the network analysis phases of EAST were
applied (followed by the new paradigm of the broken-links analysis, which has not
been previously reported) because the preliminary stages of EAST were negated
by having already collected and represented the data via interviews with an SME.
Furthermore, some of the EAST methods require communications data, which were
lacking in this particular case study as the information came from a SME’ s account
of the STS. Compared to Stanton’ s (2014) analysis of the operations within a sub-
marine control room, the current study analysed activity at a macro level (Grote
et al. 2014), using an SME’ s account of activities within the STS rather than a tran-
script of direct communications between STS actors. Recording and transcribing
communications within a working system in real time is difficult, time-consuming
and potentially disruptive to the STS under investigation. The approach presented
in the current chapter would be easier for personnel within the STS itself to apply
to support their own safety management and risk prediction activities, as it relies
on a macro-level account of actions and relationships with a STS. This also allows
these personnel to create a systems view of the STS of interest, which, as previ-
ously discussed, offers benefits over traditional taxonomic techniques. The absence
of communications data obviously means that the analysis lacks detail and a rich-
ness of information that comes from speech data. This also meant that frequency of
communications could not be represented in the same way as Stanton (2014). So, in
effect, the network diagrams in this case study offer a basic visual representation of
a STS that could aid understanding of the relationships between agents, tasks and
information as well as their combination. Of greater importance is the extension to
EAST presented in this chapter: the broken-links analysis. In order to identify poten-
tial risk in the STS, these links between nodes in the networks had to be examined in
more detail; this was accomplished in the broken-links analysis. In this phase, each
link between the task and social nodes was ‘ broken’ to illustrate the effect of a com-
munication breakdown between nodes. In this case study, 19 social links and 12 task
links were broken and assessed against numerous information nodes, resulting in
the identification of 137 risks in total. These breakdowns would result in a failure in
information transfer, so each broken link was analysed against the information nodes
to identify potential risks. This extension to the EAST method provides a structured
method for identifying all of the risks within a STS. The broken links can be listed
in table form, along with ‘ to’ and ‘ from’ information detailing the origin and des-
tination nodes between which information is transferred. The broken-links analysis
is concluded by developing mitigation strategies for each of the identified risks in a
similar way to traditional human-error analysis methods.
246 Systems Thinking in Practice

CONCLUSIONS
This chapter presented a case study of the extension to the EAST method applied
in the analysis of a STS, specifically Hawk missile simulation intended to aid with
the training of Navy crew. The approach models the STS in two phases. In the first
phase, the system is modelled as three connected networks (social, information and
task), and the second phase, the social and task network links are systematically
broken to reveal what risks are introduced by the failure to communicate informa-
tion. The broken-links approach is a substantial and novel innovation over traditional
human-error taxonomic approaches to assessing risks in systems. The approach is
based on the premise that most, if not all, accidents and near misses are caused, at
least in part, by the failure to communicate information between agents and tasks.
By enabling the generation of a system model, EAST ensures that all of the compo-
nents of interest within a STS have at been identified, and this should lead to a more
comprehensive analysis of potential risks. The extension to EAST offers a holistic,
structured and systematic approach to the identification of information communica-
tion failures in task and social networks. The EAST network broken-links approach
is a new paradigm for risk assessment in systems. The approach can be applied to
any STS in any domain where an EAST model has been constructed. Future work
should explore the risks associated with multiple communication failures occurring
simultaneously as well as considering the degree of resilience offered by different
system network models.

ACKNOWLEDGEMENTS
The authors would like to thank Wing Commander Neil Bing (Bingo) of Air Cap
SO1 Lightning, RAF High Wycombe, for his account of the Hawk RtL case study
and his very valuable insights into the challenges faced by this complex sociotechni-
cal system. This work was part-funded by the Defence Human Capability Science
and Technology Centre (DHCSTC) grant reference TIN 2.002.

REFERENCES
Alexander, R. and Kelly, T. (2013). Supporting systems of systems hazard analysis using
multi-agent simulation. Safety Science , 51, 302– 318.
Baber, C., Stanton, N. A., Atkinson, J., McMaster, R. and Houghton, R. J. (2013). Using
social network analysis and agent-based modelling to explore information flow using
common operational pictures for maritime search and rescue operations. Ergonomics ,
56, 889– 905.
Benta, M. I. (2005). Studying communication networks with AGNA 2.1. Cognition, Brain,
Behaviour , 9, 567– 574.
Dekker, S. (2014). The Field Guide to Understanding ‘Human Error’ . Ashgate: Aldershot, UK.
Driskell, J. E. and Mullen, B. (2005). Social network analysis. In: N. A. Stanton, A. Hedge,
K. Brookhuis, E. Salas and H. Hendrick (Eds.), Handbook of Human Factors and
Ergonomics Methods . CRC Press: Boca Raton, FL.
Dul, J., Bruder, R., Buckle, P., Carayon, P., Falzon, P., Marras, W. S., Wilson, J. R. and Van
Der Doelen, B. (2012). A strategy for human factors/ergonomics: Developing the disci-
pline and profession. Ergonomics , 55, 377– 395.
The EAST ‘Broken-Links’ Approach 247

Embrey, D. E. (1986). SHERPA: A systematic human error reduction and prediction approach.
Paper presented at the International Meeting on Advances in Nuclear Power Systems,
Knoxville, TN.
Flach, J. M., Carroll, J. S., Dainoff, M. J. and Hamilton, W. I. (2015). Striving for safety:
Communicating and deciding in sociotechnical systems. Ergonomics , 58(4), 615– 634.
Griffin, T. G. C., Young, M. S. and Stanton, N. A. (2010). Investigating accident causation
through information network modelling. Ergonomics , 53, 198– 210.
Grote, G., Weyer, J. and Stanton, N. A. (2014). Beyond human-centred automation – Concepts
for human– machine interaction in multi-layered networks. Ergonomics , 57(3), 289– 294.
Haddon-Cave, C. (2009). The Nimrod Review: An Independent Review into Broader Issues
Surrounding the Loss of the RAF Nimrod MR2 Aircraft XV230 in Afghanistan in 2006 .
HM Stationery Office: London, UK.
Harvey, C. and Stanton, N. A. (2014). Safety in System-of-Systems: ten key challenges. Safety
Science , 70, 358– 366.
H. M Government. (2010). Securing Britain in an age of uncertainty: the strategic defence
and security review . London, UK: The Stationery Office.
Hodgson, A., Siemieniuch, C. E. and Hubbard, E.-M. (2013). Culture and the safety of
complex automated sociotechnical systems. IEEE Transactions on Human-Machine
Systems , 43, 608– 619.
Hollnagel, E. (1998). Cognitive Reliability and Error Analysis Method . Elsevier Science:
Oxford, UK.
Hollnagel, E., Woods, D. D. and Leveson, N. C. (Eds.) (2006). Resilience Engineering:
Concepts and Precepts . Ashgate: Aldershot, UK.
Houghton, R. J., Baber, C., Cowton, M., Walker, G. H. and Stanton, N. A. (2008). WESTT
(workload, error, situational awareness, time and teamwork): An analytical pro-
totyping system for command and control. Cognition, Technology and Work , 10,
199– 207.
Houghton, R. J., Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics , 49, 1204– 1225.
Jenkins, D. P., Stanton, N. A., Salmon, P. M. and Walker, G. H. (2010). A systemic approach
to accident analysis: A case study of the Stockwell shooting. Ergonomics , 53, 1– 17.
Leveson, N. G. (2012). Engineering a Safer World: Systems Thinking Applied to Safety . MIT
Press: Cambridge, MA.
Lundberg, J., Rollenhagen, C. and Hollnagel, E. (2010). What you find is not always what you
fix – How other aspects than causes of accidents decide recommendations for remedial
actions. Accident Analysis and Prevention , 42, 2132– 2139.
MAA (2010). RA 1210: Management of operating risk (risk to life). In: M.A. AUTHORITY
(Ed.) RA 1210 . Military Aviation Authority: London, UK.
MAA (2011). Regulatory Instruction MAA RI/02/11 (DG) – Air Safety: Risk Management .
Military Aviation Authority: London, UK.
Ministry of Defence (MoD). (2007). Defence standard 00-56, issue 4, part 2. London, UK:
MoD.
Moray, N., Groeger, J. and Stanton, N. A. (2017). Quantitative modelling in cognitive ergo-
nomics: Predicting signals passed at danger. Ergonomics , 60(2), 206– 220.
Noyes, J. and Stanton, N. A. (1997). Engineering psychology: Contribution to system safety.
Computing & Control Engineering Journal , 8(3), 107– 112.
Plant, K. L. and Stanton, N. A. (2012). Why did the pilots shut down the wrong engine?
Explaining errors in context using Schema Theory and the Perceptual Cycle Model.
Safety Science , 50(2), 300– 315.
Rafferty, L., Stanton, N. A. and Walker, G. H. (2012). The Human Factors of Fratricide .
Ashgate: Surrey, UK.
248 Systems Thinking in Practice

Ramos, A. L., Ferreira, J. V. and Barcelό , J. (2012). Model-based systems engineering: An

emerging approach for modern systems. IEEE Transactions of Systems, Man and
Cybernetics: Part C – Applications and Reviews , 42, 101– 111.
Rasmussen, J. (1997). Risk management in a dynamic society: A modelling problem. Safety
Science , 27, 183– 213.
Read, G. J. M., Salmon, P. M., Lenné , M. G. and Stanton, N. A. (2015). Designing socio-
technical systems with cognitive work analysis: Putting theory back into practice.
Ergonomics , 58, 822– 851.
Royal Navy. (2012). Fleet requirements air direction unit (FRADU) . Royal Navy. Available
at: http://www.royalnavy.mod.uk/sitecore/content/home/the-fleet/air-stations/rnas-cul-
drose/fleet-requirements-air-direction-unit-fradu (Accessed 11 January 2013).
Salmon, P. M., Cornelissen, M. and Trotter, M. J. (2011a). Systems-based accident analysis meth-
ods: A comparison of AcciMap, HFACS, and STAMP. Safety Science, 50, 1158–1170.
Salmon, P. M., Lenné, M. G., Walker, G. H., Stanton, N. A. and Filtness, A. (2014). Using the event
analysis of Systemic Teamwork (EAST) to explore conflicts between different road user
groups when making right hand turns at urban intersections. Ergonomics, 57, 1628–1642.
Salmon, P. M., Stanton, N. A., Lenné , M., Jenkins, D. P., Rafferty, L. and Walker, G. H.
(2011b). Human Factors Methods and Accident Analysis . Ashgate: Surrey, UK.
Shappell, S. and Wiegmann, D. (2001). Applying reason: The human factors analysis and
classification system. Human Factors and Aerospace Safety , 1, 59– 86.
Shorrock, S. T. and Kirwan, B. (2000). Development and application of a human error identi-
fication tool for air traffic control. Applied Ergonomics , 33, 319– 336.
Stanton, N. A. (2006). Hierarchical task analysis: Developments, applications, and exten-
sions. Applied Ergonomics , 37, 55– 79.
Stanton, N. A. (2014). Representing distribution cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics , 57, 403– 418.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systematic Teamwork . Ashgate: Surrey, UK.
Stanton, N. A., Harris, D. and Starr, A. (2016). The future flight deck: Modelling dual, single
and distributed crewing options. Applied Ergonomics , 53, 331– 342.
Stanton, N. A., Rafferty, L. A. and Blane, A. (2012). Human factors analysis of accidents in
systems of systems. Journal of Battlefield Technology , 15, 23– 30.
Stanton, N. A., Salmon, P., Harris, D., Marshall, A., Demagalski, J., Young, M. S., Waldmann,
T. and Dekker, S. W. A. (2009). Predicting pilot error: Testing a new methodology and a
multi-methods and analysts approach. Applied Ergonomics , 40(3), 464– 471.
Stanton, N. A., Salmon, P. M., Rafferty, L. A., Walker, G. H., Baber, C. and Jenkins, D. P.
(2013). Human Factors Methods: A Practical Guide for Engineering and Design,
Second Edition. Ashgate : Aldershot, UK.
Stanton, N. A., Salmon, P. M., Walker, G. H., Baber, C. and Jenkins, D. P. (2005). Human
Factors Methods: A Practical Guide for Engineering and Design, First Edition .
Ashgate: Aldershot, UK.
Stanton, N. A. and Stevenage, S. (1998). Learning to predict human error: Issues of reliability,
validity and acceptability. Ergonomics , 41(11), 1737– 1756.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon, P.
et al. (2006). Distributed situation awareness in dynamic systems: Theoretical devel-
opment and application of an ergonomics methodology. Ergonomics , 49, 1288– 1311.
Stanton, N. A. and Walker, G. H. (2011). Exploring the psychological factors involved in the
Ladbroke Grove rail accident. Accident Analysis & Prevention , 43(3), 1117– 1127.
Stewart, R., Stanton, N. A., Harris, D., Baber, C., Salmon, P., Mock, M., Tatlock, K., Wells,
L. and Kay, A. (2008). Distributed situation awareness in an airborne warning and
control system: Application of novel ergonomics methodology. Cognition, Technology
and Work , 10, 221– 229.
The EAST ‘Broken-Links’ Approach 249

Swain, A. D. and Guttmann, H. E. (1983). A Handbook of Human Reliability Analysis

with Emphasis on Nuclear Power Plant Applications , NUREG CR-1278 . USNRC:
Washington, DC.
Underwood, P. and Waterson, P. (2013). System accident analysis: Examining the gap between
research and practice. Accident Analysis and Prevention , 55, 154– 164.
von Bertalanffy, L. (1950). An outline of general system theory. British Journal for the
Philosophy of Science , 1, 134– 165.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. and Green, D. (2006). Event
analysis of systematic teamwork (EAST): A novel integration of ergonomics methods
to analyse C4i activity. Ergonomics , 49, 1345– 1369.
Walker, G. H., Stanton, N. A., Baber, C., Wells, L., Gibson, H., Salmon, P. M. and Jenkins, D.
(2010a). From ethnography to the EAST method: A tractable approach for representing
distributed cognition in air traffic control. Ergonomics , 53, 184– 197.
Walker, G. H., Stanton, N. A., Sal mon, P.M. and Jenkins, D. P. (2008). A review of socio-
technical systems theory: A classic concept for command and control paradigms.
Theoretical Issues in Ergonomics Science , 9, 479– 499.
Walker, G. H., Stanton, N. A., Salmon, P. M. and Jenkins, D. P. (2009). Command and
Control: The Sociotechnical Perspective . Ashgate: Aldershot, UK.
Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D. P. and Rafferty, L. A. (2010b).
Translating concepts of complexity to the field of ergonomics. Ergonomics, 53, 1175–1186.
Waterson, P., Robertson, M. M., Cooke, N. J., Militello, L., Roth, E. and Stanton, N. A.
(2015). Defining the methodological challenges and opportunities for an effective sci-
ence of sociotechnical systems and safety. Ergonomics , 58(4), 565– 599.
Wilson, J. R. (2012). Fundamentals of systems ergonomics. Work , 14, 3861– 3868.
13 Modelling Future
From CWA to SNA

Flight Decks
With Don Harris and Alison Starr

INTRODUCTION
The trend in flight deck design over the past half century has been one of progressive
‘de-crewing’. Fifty years ago, it was not uncommon for there to be five crew on the flight
deck of a civil airliner (two Pilots, Flight Engineer, Navigator and Radio Operator).
Today, just two pilots accomplish the same tasks once undertaken by five. Many func-
tions are now wholly or partially automated. Consequently, the role of the pilot has
changed from one of being a ‘flyer’ to one of being a systems/flight deck manager.
Airline personnel costs vary between about 11% of operating costs to nearly 25%,
depending upon aircraft type, sector length and how much activity is outsourced
(Ryanair 2009; easyJet plc 2013). Crew costs for smaller commercial aircraft can be
between 15% and 35% of the aircraft direct operating costs (Alcock 2004). Annual
accounts from a typical low-cost operator suggest that even for a larger airliner, the
crew represent nearly 19% of operating costs (excluding fuel and propulsion – easy-
Jet plc 2013). The scope to make significant cost savings with the current common
configuration for aircraft (cylindrical fuselage with wings, rudder and tail plane) is
now limited. This configuration is approaching the end of its development potential.
Alternative configurations such as the blended wing body concept, which offer con-
siderable structural and aerodynamic advantages, have met with limited enthusiasm
from potential passengers. Problems of ensuring a safe and efficient means of pas-
senger evacuation have also been identified (Galea et al. 2011). This configuration
also needs a great deal of development in other areas (such as flight control sys-
tems and structural testing) before it will be suitable for service entry. Reducing the
number of crew on the flight deck to just a single pilot will produce significant cost
savings, especially in smaller commercial aircraft operated on shorter and ‘thinner’
(lower-demand) routes.
Some manufacturers (e.g. Embraer) are already developing the technology for a
single-crew aircraft, as are avionics suppliers (e.g. Honeywell – see Keinrath et al.
2010). The approaches being adopted in these instances centre upon the devel-
opment of sophisticated airborne technology to assist the pilot (e.g. Intelligent
Knowledge-Based Systems and adaptive automation). This approach is also being
adopted in other research programmes looking at flight deck automation and

251
252 Systems Thinking in Practice

crewing, for example, the development of an Electronic Standby Pilot (ESP) as

part of the Advanced Cockpit for the Reduction of Stress and Workload (ACROSS)
project (see http://www.across-fp7.eu/). The medium-term objectives of the
ACROSS project are concerned with reducing the number of flight deck crew in
the cruise phase in long-haul flights to permit crew to rest and help prevent fatigue.
It is anticipated that the same technology will aid in the case of partial (or even
full) flight crew incapacitation. The longer-term objectives of the project are to
form the basis for potential single-crew operations.
A similar approach has been adopted by other researchers in the past, par-
ticularly in the military domain, but with only mixed success (e.g. the COGnitive
cockPIT – COGPIT programme – Bonner et al. 2000; Taylor et al. 2000; and
the Cockpit Assistant Military Aircraft – CAMA programme – Schulte and Stütz
2001; Stütz and Schulte 2001). CASSY (the Cockpit ASsistant System) was a
civil aircraft version of the latter developed by the same team (see Onken 1994;
Onken 1997). The Cognitive Adaptive Man–Machine Interface (CAMMI) proj-
ect (Keinrath et al. 2010) also makes use of extensive Artificial Intelligence (AI)
software in its approach to adaptive automation. A later requirements analysis
for developing concepts for single-pilot operations was also predicated upon the
notion of incorporating extensive pilot automated assistance on the flight deck,
particularly synthetic vision systems, data linking and direct voice input/output
systems (Deutch and Pew 2005). The main arguments for the use of two mem-
bers of flight deck crew centre around issues concerned with pilot workload (spe-
cifically instances of workload peaks), the reduction of flight crew error and pilot
incapacitation. Many of the assumptions are either questionable or are becoming
out-dated as discussed in the following three paragraphs.
From the perspective of the person in command of any aircraft, there is a work-
load ‘cost’ associated with the management of crew on the flight deck. The require-
ment to coordinate crew, cooperate and communicate on the flight deck itself has
workload associated with it. Doubling the number of crew does not half the work-
load. Furthermore, modern flight decks are already certificated so that they can be
operated by a single member of flight deck crew (see FAR/CS 25.1523). Automated
flight deck systems have already considerably reduced pilot workload (Weiner and
Curry 1980; Harris 2003).
While the second crew member may distribute the workload around the flight
deck somewhat, it can also be argued that they actually introduce an error mode.
Poor CRM (Crew Resource Management) has been implicated as a contributory
factor in nearly 23% of all fatal commercial jet aircraft accidents (CAA] 2008).
The effectiveness of the second pilot as an ‘error checker’ is also questionable.
Omission of action or inappropriate action was implicated in 39% of accidents, and
an incorrect application of procedures or a deliberate non-adherence to procedures
was implicated in a further 13% (CAA 2008). Becoming ‘low and slow’ (a failure
to cross-monitor the flying pilot) was a factor in 12% of accidents. As a cross-check
on the position of the aircraft, the PM’s (Pilot Monitoring) effectiveness would
also seem to be questionable, as a lack of positional awareness was identified as
a causal factor in 27% of cases (CAA 2008). This is quite a crude analysis, how-
ever. It is acknowledged that what these data do not show is in how many cases
From CWA to SNA 253

the second pilot trapped an error made by the other pilot and avoided an accident;
this is unknown and unknowable. However, observational data obtained from rou-
tine flights reported that 47.2% of errors committed by Captains involved inten-
tional non-compliance with Standard Operating Procedures (SOPs) or regulations;
38.5% were unintentional procedural non-compliance (Thomas 2003). Thomas
also reports that in observations of line operations, crews did not demonstrate
effective error detection, with more than half of all errors remaining undetected by
one or both of the flight crew. As a result, it can be argued that removing one of the
pilots actually reduces the scope for accidents occurring as a result of miscommu-
nication or misunderstanding between the pilots, and that removing the PM does
not double the workload on the flight deck.
Perhaps the greatest concern for the development of a single-crew aircraft is that
associated with pilot death, incapacitation or impairment. However, such instances
are very rare. A study of in-flight medical incapacitations in US airline pilots
between 1993 and 1998 found only 39 instances of incapacitation and 11 instances
of impairment (DeJohn et al. 2004). The rate of in-flight medical events (encompass-
ing both types) was 0.058 per 100,000 flight hours. The probability that one of these
events would subsequently result in an accident was calculated to be 0.04. DeJohn
et al. (2004) observed that the safety of the flight was seriously impacted in only
seven cases and resulted in two non-fatal accidents. A later study of UK commercial
pilots by Evans and Radcliffe (2012) suggest that the annual in-flight incapacitation
rate was 0.25%; however, this study is seriously flawed in than it was not weighted
by flight hour and the rate is expressed as a percentage of all UK registered pilots
(irrespective of flight hours accumulated by each, per year).
It is argued that with the judicious use of existing equipment, there are no
major reasons why a single-pilot-operated commercial aircraft is not feasible in
the very near future using existing technology. Military aviation has flown com-
plex, high-performance single-crew aircraft for many years, and Unmanned Air
Vehicles (UAVs) are now commonplace. UAV technology has matured, and such
aircraft are now regularly being used for national border and port security, home-
land surveillance, scientific data collection and telecommunications services
(Harris 2007). Airworthiness standards for their design and operation in civil
airspace are being developed on both sides of the Atlantic (e.g. UK Civil Aviation
Authority 2010 – Unmanned Aircraft System Operations in UK Airspace –
Guidance [CAP 722]). Several UAVs are now the size of a small aircraft, with
performance similar (or exceeding) that of a conventional aeroplane. It is worth
noting that UAVs still have a designated ‘Pilot in Command’ International Civil
Aviation Organisation Annex 2, ‘Rules of the Air’, states that the ‘Responsibility
of the pilot-in-command’ of an aircraft shall, whether manipulating the controls
or not, be responsible for the operation of the aircraft in accordance with the
rules of the air, except that the pilot-in-command may depart from these rules in
circumstances that render such departure absolutely necessary in the interests of
safety’.
The greatest obstacle to the operation of civilian, single-pilot aircraft is not the
technology per se. Rather, the barriers are combining the ground and airborne tech-
nologies, designing the user interfaces and developing new concepts of operations to
254 Systems Thinking in Practice

make such an aeroplane safe and useable in a wide range of normal and n on-normal
operating situations (when flown by a typical commercial pilot). That is to say that
the Human Factors requirements are the prime driver in this case, not the tech-
nology. The concept evaluated in this chapter is based upon an alternative design
approach to that of utilising a large amount of on-board, complex computing (e.g.
that using agent-based software) first described by Harris (2007). The concept uses
a sociotechnical systems-based design philosophy utilising a great deal of currently
existing technology. In this case, the control and crewing of the aircraft is distrib-
uted in real time across both the aircraft’s flight deck and ground stations (see also
Stanton et al. 2014). The second pilot is not replaced by on-board AI or Intelligent
Knowledge-Based Systems, which would be both difficult to develop and challeng-
ing to certificate; they are merely displaced.

DESIGN APPROACH
The proposed approach regards a future single-crew aircraft as just one part of a
wider operating system, a radical change from the operation of current generation
airliners. The initial high-level design architecture proposed for operating the single-
crew aircraft consists of several discrete elements (Stanton et al. 2014):

• The aircraft itself (including pilot)

• Ground-based components including
• ‘Second pilot’ support station/office
• Real-time engineering support
• Navigation/flight planning support
• System ‘Mirror’.

This is the initial envisaged instantiation of the system upon which to base a cog-
nitive work analyses in order to evaluate its efficacy and to identify required technol-
ogy development paths. The design of this single-crew aircraft operations system is
intended to form the basis of incremental developments, incorporating subsequent
airborne pilot-support technology as it is developed. This will allow the further rapid
development of the concept.

Aircraft Component
In many ways, the aircraft component will be little different to current types. The
requirements are that aircraft should be able to function in all types of airspace
without any special Air Traffic Control/Air Traffic Management (ATC/ATM)
procedures and should be able to be flown by regular Airline Transport Pilots
Licence (ATPL) qualified professional pilots without extraordinary training. It
should exhibit an equivalent level of safety to fourth-generation modern airliners.
Initially, such an aircraft will be optimised for shorter-range, low-cost operations
(including cargo operations) and for ‘thinner’ routes where cost of operation is a
critical factor. Emphasis will be placed upon reduction of workload and error by
simplification of operation.
From CWA to SNA 255

Ground-Based Component
The ground-based component’s primary functions are to support the pilot (e.g. in
navigation, system management, ATC/ATM support or fault diagnosis), not neces-
sarily to duplicate their skills and functions. Suitably skilled personnel on the ground
may simultaneously support several aircraft (resulting in economy of scale and reduc-
tion in duplication of effort). Furthermore, these personnel need not necessarily have
the same skill set as a pilot; rather, they could be specialists in navigation, commu-
nications or avionics systems. In the case of the single-crew aircraft, control from a
ground station would only be required in the advent of pilot incapacitation, although
assistance from the ground may be required in other circumstances, such as high
workload (e.g. take-off/landing) or abnormal (e.g. re-routing/bad weather/emergency)
situations. This approach would allow ground-based operators to undertake many of
the key roles of the second pilot but to do so for several aircraft simultaneously. This
emphasises that the design of a single-crew aircraft is not simply concerned with the
re-design of the flight deck. It requires a change in the overall operating philosophy.
‘Second Pilot’ Support Station/Office The primary functions of the ground-
based ‘second-pilot’ station/office is to provide real-time support for the pilot on
the aircraft as required and to assist in flight planning and pre-flight preparation.
If the hierarchy of tasks when flying an aircraft can be conceptualised as ‘Aviate’,
‘Navigate’, ‘Communicate’ and ‘Manage’, whenever possible the ‘Aviate’ component
will remain on the flight deck itself. However, if necessary, it will be possible to
operate the aircraft remotely from this position. It is envisaged that this support sta-
tion will be able to support a number of aircraft.

The function of this aspect of the ground-support element is four-fold:

• Prior to flight, the ground-based support function provides a cross-check of

flight planning data.
• When pilot workload is high, if required, the operator in the ground-based
workstation may assume many of the duties of the second pilot normally on
board the aircraft.
• During non-normal, abnormal or emergency situations, the operator in the
ground-based workstation will also assume the duties normally undertaken
by the second pilot; however, this will also be in conjunction with other
support (provided as required) by engineering support and navigation/flight
planning support elements.
• In the event of pilot incapacitation, the aircraft can be ‘flown’ from the ground
station (in the same manner as a UAV). It is envisaged that any direct opera-
tion from this position will be undertaken via the aircraft’s automation; direct
operation via the primary flight controls will be a last resort.

The main focus of ground-based support from this operator station will be on
aircraft configuration management and short-term navigation and communication
support to the pilot. Emphasis will be placed upon decision-aiding aspects of pilot
support.
256 Systems Thinking in Practice

This station will be crewed by pilots qualified on the single-crew aircraft, who
‘rotate’ through the post in between actually flying the aircraft. This should enhance
the shared Situation Awareness between ground and air components.
Real-Time Engineering Support: The single-crew aircraft will routinely pass
system operation information to the ground for automated health monitoring. This
is already done routinely for engine data but may be expanded to encompass all
critical aircraft systems in real time. System information will be monitored for sig-
nificant deviations from normal on the ground. In the event of a non-normal situ-
ation being detected, ground-based automated systems and engineering staff will
evaluate the system failure and either provide advice to the pilot or re-configure the
system remotely. Emphasis will be upon stabilising the situation and evaluating con-
sequences rather than attempting to rectify the problem in flight.
In the event of a serious system failure, the implications for the continuation of the
flight will be evaluated. If it is necessary to perform an immediate descent or diver-
sion, the appropriate navigation/flight planning facility will be notified.
Engineering support will be provided by engineers qualified on the aircraft and trained
to provide remote real-time support in the advent of an airborne system malfunction.
Navigation/Flight Planning Support: This is an extension of the existing airline
flight planning functions to encompass real-time, in-flight planning/re-planning
facilities. This facility will be expanded to encompass an ‘aircraft-centric’ point of
view and will be able to access the on-board flight management computers to up-link
new routing information. Support (long and short term) will be provided by flight
planning specialists.

System ‘Mirror’
The system mirror will be an independent, ground-based (software) representation of
the aircraft system states (in particular the Flight Management System [FMS], autopilot
system and autothrust systems and the general configuration of the aircraft – flaps, slats,
etc.). Ground-based system elements (pilot support station, flight/navigation panning
and engineering support) will be able to interact with the ground-based system mirror
without directly affecting aircraft systems (if required). In the advent of a datalink fail-
ure, the system mirror will contain the last known configuration of the aircraft systems
and will be able to update aircraft systems, if required. During normal operations, the
automation mirror will normally be ‘transparent’ to all operators in the system.

MODELLING AND ANALYSIS OF SYSTEM CONFIGURATIONS

Complex sociotechnical systems (such as flight operations) are made up of numerous
interacting parts, both human and non-human, operating in a dynamic, ambiguous
and safety–critical domain (Harris and Stanton 2010). The complexity embodied in
these systems presents significant challenges for modelling and analysis and requires
Systems Ergonomics methods for the effective design of future work systems (Wilson
and Carayon 2014). Cognitive Work Analysis (CWA) is a structured framework specif-
ically developed for considering the development and analysis of such complex socio-
technical systems (Rasmussen et al. 1994; Vicente 1999; Jenkins et al. 2009). The
From CWA to SNA 257

framework leads the analyst to consider the environment within which the task takes
place and the effect of the imposed constraints on the way work can be conducted. The
framework guides the analyst through the process of answering the question of why
the system exists and what activities can be conducted within the domain, as well as
how these activities can be achieved and who can perform them, also identifying the
competencies required. Recent case studies have been conducted in the assessment of
risk associated with nuclear decommissioning (Walker et al. 2014) and command team
activities in control of a submarine (Stanton and Bessell 2014).
The analyses in this chapter commence with a generic CWA concerning the
operation of various options for system configurations in a number of operating sce-
narios. These analyses are supplemented by a Social Network Analysis (SNA) which
provides an indication of the resilience of the operational networks in each case
(Diskell and Mullen 2004; Baber et al. 2013). This approach has previously been
used to examine the characteristics of terror cells (Kenney et al. 2013) and command
networks (Walker et al. 2009). These analyses can be used to examine options for the
proposed distributed system architecture for the operation of a single-crew aircraft.
They also help to define more specific aspects of new on-board automation require-
ments for the single-crew aircraft and the role of the Ground Support Station/Office.
Analyses commence with a ‘conventional’ baseline aircraft (Airbus A320 and
variants) being flown by two crew, followed by an analysis of the same aircraft being
operated by a single pilot. Note that it is a certification requirement that all air-
craft must be capable of being safely operated by a single crew member (FAR/CS
25.1523 – Minimum Flight Crew). These initial analyses were followed by a com-
parative analysis of four different configuration options for a single-crew-operated
aircraft being operated as part of a distributed system. Four potential versions of
single-pilot flight operations are presented for comparison (as shown in Table 13.1),
that is, a single-pilot aircraft (A); a single-pilot aircraft with an additional pilot at a
ground station who can be called upon at times of need (C); a single-pilot aircraft
with an automation mirror on the ground that cross-checks the inputs and outputs
independently of the aircraft automation (B); and a single-pilot aircraft with an addi-
tional pilot at a ground station and also with an independent automation mirror (D).

TABLE 13.1
Comparison of Four Potential Versions of Single-Pilot Flight Operations

Four Versions of the Future Pilot in Ground Station

Automation NO YES
Mirror NO Single-pilot aircraft Single-pilot aircraft with additional pilot on the
(Option A) ground
(Option C)
YES Single-pilot aircraft Single-pilot aircraft with additional automation
with additional mirror on the ground and additional pilot on
automation mirror the ground
on the ground (Option D)
(Option B)
258 Systems Thinking in Practice

Material for undertaking the CWA and SNA was drawn from a number of sources,
including operations manuals and SOPs (Airbus Industrie n.d.). These were comple-
mented by a structured de-brief of an experienced qualified Test Pilot who was also
type rated on a number of Airbus types (and is a training Captain for a major airline),
who also helped to devise the various operating scenarios. The CWA material will
be presented first as the formative design of the future system alternatives. Much
has been made of CWA as a formative approach (Vicente 1999; Jenkins et al. 2009;
Naikar 2013) but little is seen in practice. CWA served as the basis for identifying the
functional loading on agents in the system as well as providing the data for the SNA.

Cognitive Work Analysis (CWA)

In the first phase of CWA, the Abstraction Hierarchy (AH) is used to model the work
domain as follows.

Work Domain Analysis (WDA)

WDA is the most commonly used component within CWA (McIlroy and Stanton
2011), and it identifies the constraints imposed on workers’ behaviour by the pur-
posive and physical context, or problem space, in which workers operate (Naikar
2006a). WDA is conducted at the functional rather than behavioural level; it is used
to define the environment within which the activity is conducted. WDA identifies a
fundamental set of constraints on the actions of any system component, thus pro-
viding a solid foundation for subsequent phases of the development of the aircraft
(McIlroy and Stanton 2011). The abstraction component of the diagram models the
same system at a number of levels; at the highest level, the overall functional purpose
of the system is considered, while at the lowest level, the individual components
within the system are described. Generally, five levels of abstraction are used:

• Functional Purposes (the purposes of the work system and the external con-
straints on its operation)
• Values and Priority Measures (the criteria that the work system uses for
measuring its progress towards the functional purposes)
• Purpose-Related Functions (the general functions of the work system that
are necessary for achieving the functional purposes)
• Object-Related Processes (the functional capabilities and limitations of physi-
cal objects in the work system that enable the purpose-related functions)
• Physical Objects (the physical objects in the work system that afford the
object-related processes).

Functional Purposes
The AH was created in both a top-down and a bottom-up fashion. First, the overall
functional purpose of flight operations was specified. This is the reason why the sys-
tem exists. These purposes are independent of time; they exist for as long as the system
exists. In this case, it is to ‘transport people and cargo safely from A to B’ and ‘increase
From CWA to SNA 259

FIGURE 13.1 Work domain analysis showing the five levels and the means-ends-links.

shareholder dividend’ (i.e. ensure that the aircraft was economic to operate). These
purposes were represented as two nodes at the top of the WDA in Figure 13.1.

Values and Priority Measures

The next stage of the construction was to consider the level below the functional
purposes, the values and priority measures, as shown in Figure 13.1. Here, further
constraints on the system were more explicitly listed; these were measures for deter-
mining how well the intended system achieved its functional purposes. In this case,
they were identified as ‘maximise comfort of passengers’ (in order to enhance the
customer experience), ‘minimise risk’ (to keep the system as safe as possible), ‘maxi-
mise aircraft utilisation’ (to avoid storage costs and enhance revenue) and ‘reduce
operating costs’ (to avoid wastage). The way that the future aircraft system is config-
ured to meet these needs is likely to be heavily contextually dependent.

Physical Objects
As previously stated, a mixture of a top-down and a bottom-up approach has been
employed in this analysis. Moving down to the lowest level of the hierarchy, the
physical objects within the system were listed. The boundaries of this analysis lim-
ited this list to the systems of direct relevance to the operation of a future single-
crew aircraft rather than every single object. The boundaries of the analysis indicate
the levels of fidelity applied here. In an attempt to keep the analysis manageable,
the boundary has omitted individual system elements. Whilst it would have been
possible to decompose many of the listed objects into their component parts and
describe their affordances more concisely, this was outside the remit of this initial
260 Systems Thinking in Practice

evaluation. A list of the physical objects analysed can be found at the base of Figure
13.1. This list comprises primary flight display, navigation display, engine indication,
anti-ice system, environmental conditioning system, electrical systems, fuel system,
hydraulics, auxiliary power unit, flight management system, nav/comms (navigation/
communication), throttle, rudder pedals, high-lift devices, landing gear, brakes, fire-
extinguishing system and external lighting.

Object-Related Processes
The second level from the bottom of the AH, the object-related processes, captures
the processes that are conducted by the physical objects to perform purpose-related
functions. Most importantly, they capture the affordances of the physical objects
independently of their purpose. For example, movement of the aircraft on the ground
is afforded by the combination of the throttle, landing gear (nose wheel steering) and
brakes. A complete list of the object-related processes can be found at the level above
the base level of Figure 13.1 along with the links indicating to which object they
relate. To aid readability, this list comprises attitude of aircraft, airspeed, altitude
of aircraft, rate of climb or descent; Instrument Landing System (ILS) indications,
current heading of aircraft, flight plan, collision avoidance, terrain display, weather
alerting, ground speed, engine thrust; weight and balance of aircraft; engine health;
preventing icing; maintaining breathable atmosphere, maintaining thermal comfort,
powering aircraft avionics, fuel management, management of hydraulics, communi-
cation management, performance/cost index, movement of aircraft on the ground,
safety of cargo and conspicuity of aircraft.

Purpose-Related Functions
In the middle of the AH, the purpose-related functions are listed. These functions
have the ability to influence one or more of the values and priority measures. They
link the purpose-independent processes with the object-independent functions. They
are listed as ‘aviate’ (i.e. keeping the aircraft airborne); ‘communicate’ (i.e. staying
in touch with ATC/ATM and company operations as well as communicating with
crew and passengers); ‘navigate’ (i.e. plan, change and check the route for the aircraft
as required); ‘manage’ (i.e. manage the avionics system; and ‘warn’ (i.e. indicate to
aircrew and company operations when systems are outside their tolerances).
The use of the means-ends-links and the utility of the AH can be described with
the example shown in Figure 13.1. Figure 13.1 shows the node ‘aviate’ in the pur-
pose-related functions level. Following the links out of the top of this node answers
the question ‘why is this needed?’ – in this case, to ‘maximise comfort of passen-
gers’ and ‘maximise aircraft utilisation’. Following the links down from the ‘aviate’,
it is possible to answer the question ‘how can this be achieved?’ – in this case, by
‘attitude of aircraft’, ‘airspeed’, ‘altitude of aircraft’, ‘rate of climb’, ILS indications’
and ‘current heading/track of aircraft’.
The diagram does not prescribe a particular arrangement for providing this func-
tionality; rather, it lists all of the components that can affect it. In this case, there is
redundancy in the system.
From CWA to SNA 261

Control Task Analysis (ConTA)

The second phase of the CWA framework, Control Task Analysis (ConTA), allows
the requirements associated with known, recurring classes of situations to be
identified. Naikar et al. (2006a, 2006b) have developed the Contextual Activity
Template (CAT) (see Figure 13.2) for use in this phase of the CWA. This template
is one way of representing activity in work systems that are characterised by both
work situations and work functions. Work situations can be decomposed based on
recurring schedules or specific locations. Rasmussen et al. (1994) describe work
functions as being activity characterised by its content, independent of its temporal
or spatial characteristics. These functions can often be informed by the abstrac-
tion hierarchy. Rasmussen et al. (1994) recommend that the analyst decompose on
either work functions or work situations; however, Naikar et al. (2006a, 2006b) plot
these on two axes so that their relationship can be investigated, allowing the repre-
sentation of activity in work systems that are characterised by both work situations
and work functions. Typically, the work situations (in this case the phases of flight
as defined by the International Civil Aviation Organization [ICAO]) are shown

FIGURE 13.2 CAT for object-related processes.

262 Systems Thinking in Practice

along the horizontal axis, and the work functions (associated with controlling and
managing the aircraft) are shown along the vertical axis of the CAT. The circles
indicate the work functions with the bars showing the extent of the table in which
the activity typically occurs. The dotted boxes around each circle indicate all of
the work situations in which a work function can occur (as opposed to must occur),
thus capturing the constraints of the system. Figure 13.2 shows the CAT for the
purpose of related functions derived from the abstraction hierarchy (see Figure
13.1). The situations follow the main phases of flight operations, from the aircraft
standing at a gate to landing on a runway (assuming that taxiing and standing will
be similar for both take-off and landing), along with the addition of an emergency
descent.
Looking at Figure 13.2, perhaps the most salient feature is that in an aircraft dur-
ing flight operations, some function constraints are notably contextually influenced
(e.g. rate of climb or descent, terrain display, ground speed, and movement of aircraft
on the ground) whereas others are not (e.g. engine thrust, engine health, powering
aircraft avionics, management of hydraulics and communication systems). To read
Figure 13.2, note that an empty cell means that the function (vertical axis) cannot be
performed in that situation (horizontal axis); a dashed box means that the function
can be performed in the situation, and a ball-and-whisker means that the function is
typically performed in the situation.

Social Organisational Cooperation Analysis – Contextual

Activity Template (SOCA-CAT)
In the next phase of the analysis, roles of personnel are allocated to the functions
across the situation. This phase is called Social Organisation and Cooperation
Analysis – Contextual Activity Template (SOCA-CAT). A list of the key roles and
their related coding can be seen in Figure 13.3. One of the roles is non-human,
denoted as Aircraft Automation in Figure 13.3 (the light speckled coding).
As Figure 13.3 shows, many of the functions in the baseline aircraft are per-
formed either exclusively by, or with the assistance of, automation. There is a clear
demarcation between the roles of the pilot flying (red coding) and the pilot monitor-
ing (green coding). The pilot monitoring is oversight of the pilot flying (red coding),
automation (blue coding) and ATC/ATM (pink coding). It is striking that each func-
tion tends to be assigned to the same set of roles throughout all situations (where that
function applies). Of the 25 functions identified, 10 are typically the responsibility of
just a single agent, whereas 15 are joint responsibility (most being the responsibility
of two agents). Some 240 of the 400 cells (functions against situations) have been
assigned to agent roles in SOCA-CAT.
SOCA-CAT has provided the data for building the social networks. This method
has previously been proven by Baber et al. (2013) when developing operational con-
cepts for search and rescue missions. In essence, every within-cell and between-cell
relationship becomes a link between agents in the network (except for the emergency
descent). In this way, it is possible to step from a formative allocation of function to
a formative SNA. This is a useful extension of the CWA framework, as it enables
quantitative as well as qualitative exploration of future operational concepts.
From CWA to SNA 263

FIGURE 13.3 Social organisation and SOCA-CAT for current flight operations.

Social Network Analysis

SNA is used to analyse and represent the relationships between agents and artefacts
within a network. A social network is defined as a set or team of agents that possess
relationships with one another (Driskell and Mullen 2004). In this case, the social
network extends beyond the aircraft to other agencies such as air traffic control,
264 Systems Thinking in Practice

fuellers, dispatch and so on. With changes in automated support on the ground and in
the air, coupled with increases in air-traffic density, aircraft operational concepts are
constantly changing. Direct routing and self-assured separation (‘free flight’) have
vastly expanded the Joint Cognitive System boundaries around airliner operations
well beyond the flight deck (Hollnagel 2007). SNA is based upon the notion that
the relationship between agents within a social network has a significant effect upon
the actions performed and also the performance achieved by the network. SNA uses
both graphical and mathematical procedures to represent social networks. Typically,
centrality measures are calculated for each agent (e.g. degree, betweenness and
closeness), and the overall network density is calculated. This allows the identifica-
tion of the key agents within the network and also the classification of the network
structure. The technique has previously been used for the analysis of networks in
a number of areas, such as command and control (Stanton et al. 2008), search and
rescue operations (Baber et al. 2013) and submarine command teams (Stanton 2014).
All three studies have shown those people (agents) who are most important to the
success of the team, based on communication frequencies. There are a number of
metrics associated with the analysis of whole networks, depending upon the type
of evaluation that is being performed. The size of the network determines the num-
ber of possible relations, and the number of possible relations grows exponentially
with the size of the network. This defines the network’s complexity. The latter is
expressed in the form of social relations that are actually observed, represented as
some fraction of the total possible.
A major advantage of networks is that they do not differentiate between different
types of node (e.g. artefacts and/or people), so that from a modelling perspective,
they are not constrained by existing structures but rather help to define the tasks
allocation associated with a particular scenario (Baber et al. 2013; Stanton 2014).
It is also possible to model the temporal aspects of networks by identifying critical
moments in the sequence of activity. To do this, the scenario is divided into task
phases, allowing active and non-active elements to be specified and represented.
The first step in a SNA involves defining the network that is to be analysed. Once
the overall aircraft operation network is specified, the people and/or artefacts need to
be defined. In the case of the development of the single-crew aircraft, a number of dif-
ferent networks were identified for analysis over a number of different flight scenarios.
In this case, the agents identified were the pilot flying, the pilot monitoring, ATC/
ATM, dispatch, flight planning, engineering, ground handling and the fuellers. For
comparison, a description of current operations of two crew flying the baseline aircraft
is provided (see Table 13.2), with which the envisaged single crew commercial aircraft
options are contrasted. The initial analyses pertain to aircraft operations as a whole,
not to any particular normal, non-normal or emergency situation. The matrix repre-
sents the frequency of associations between each agent in the network. This matrix
shows whether or not an agent within the current aircraft system of operation can
be associated with any other agent in the proposed architecture for the single-crew
aircraft, specifically through frequency of communications. As an example, the asso-
ciation matrix for the generic aircraft operational scenario is presented in Table 13.2.
An example social network diagram for the takeoff task is presented in Figure
13.4. Comparison with the network archetypes shows that the network is mesh-like,
From CWA to SNA 265

TABLE 13.2
Association Matrix for Baseline Aircraft Generic Operational Scenario

Ground
handling
Pilot
monitoring

Dispatch

Aircra
automaon

ATC/ATM

Flight
planning

Fuelers

Pilot Flying

FIGURE 13.4 Network diagram for baseline aircraft generic operational scenario.

with dispatch and flight planning on the left-hand side; pilot monitoring, ATC/ATM
and pilot flying in the middle; and ground handling, aircraft automation and fuel-
lers on the right-hand side. Mesh structures are considered to be generally robust
(Stanton et al. 2008).
266 Systems Thinking in Practice

TABLE 13.3
Social Network Analysis Statistics

ATC
Nodal statistics PF PM* AA ATM D FP GH F
Emission/Reception 115 202 102 86 7 8 8 8
Eccentricity 1 1 2 2 2 2 2 2
Sociometric Status 25.55 44.88 22.66 19.11 1.55 1.77 1.77 1.77
Centrality (B-L) 5.28 5.28 4.11 4.11 3.70 3.70 3.36 3.36
Closeness 1.28 1.28 1.00 1.00 0.90 0.90 0.82 0.82
Farness 7 7 9 9 10 10 11 11
Betweenness 7.33 7.33 2.00 1.33 0 0 0 0

* Bold highlighting of possible role to be displaced to a ground station

Counting the number of collaborative (i.e. within-cell) and cooperative (i.e.

between-cells) links gives rise to the data presented in Table 13.2 (which assumes
a symmetrical network, with every link reciprocated, such as with a read-back or
acknowledgement of each and every communication). This network forms the basis
for the numerical analysis using SNA statistics (see Table 13.3). These statistics can
either look at the network as a whole (e.g. the density of a network is defined by the
number of social relations that are actually observed and can be represented as some
fraction of the total possible) or analyse the individual nodes, e.g.:

• Emission and reception degree are the number of ties emanating from, and
going to, each agent in the network.
• Eccentricity is defined by the largest number of hops an agent has to make
to get from one side of the network to another.
• The sociometric status of each agent refers to the number of communica-
tions received and emitted, relative to the number of nodes in the network.
• Agent centrality is calculated in order to determine the central or key
agent(s) within the network. There are a number of different centrality cal-
culations that can be made; for example, agent centrality can be calculated
using Bavelas-Leavitt’s index.
• Closeness is the inverse of the sum of the shortest distances between each
individual and every other person in the network. It reflects the ability to
access information through the ‘grapevine’ of network members.
• Farness is the index of centrality for each node in the network, computed as
the sum of each node to all other nodes in the network by the shortest path.
• Betweenness is defined by the presence of an agent between two other
agents, which may be able to exert power through its role as an information
broker.

The same principles were used to develop the four networks associated with
each of the single-pilot options, each representing one of the four visions for future
From CWA to SNA 267

flight operations. Again, the networks appeared mesh-like, with fewer connections
in the network in Option A (single-pilot aircraft) and more connections in Option D
(a single-pilot aircraft with an additional pilot at a ground station and also with an
independent automation mirror).
More robust networks generally have more connections, but that does not neces-
sarily mean that everything should be connected to everything else. Unnecessary
connections in a network can lead to greater coordination problems and could also
mean that irrelevant information is being communicated (Rafferty et al. 2012;
Sorensen and Stanton 2013). Analysis of the SOCA-CATs shows some differences
in the functional loading of the single-pilot configurations compared to the current
version of flight operations, as revealed in Table 13.4.
As Table 13.4 shows, removing the pilot monitoring from the network dramati-
cally increases the functional loading on the remaining pilot on the flight deck as
well as reducing the density of the network (where 0 means no connections, and
1 means everything is connected to everything else). The single pilot with a pilot
on the ground (Options C and D) has the closest resemblance to current operations
(although the pilot on the ground would have responsibility for monitoring several
aircraft in different phases of flight). This comparison is for both the loading of
functions and the network density. The most robust network appears to be option D,
with the single pilot with the pilot on the ground and the automation mirror (which
replicates the aircraft automation, using the same inputs and comparing with the
results of the aircraft automation – only reporting when a discrepancy is found), as
this network has the highest density (Walker et al. 2009)
The SOCA-CAT cooperation and collaboration interactions were compiled
for comparative analysis. As before, the analysis assumes that the networks were

TABLE 13.4
Functional Loading for the Generic Scenario

Current
Function Loading/Options System Option A Option B Option C Option D
Pilot Flying (PF) 90 217 217 90 90
Pilot Monitoring (PM) or Pilot on 160 – – 160 160
the Ground (PG)
Aircraft Automation (AA) 143 143 143 143 143
System Mirror (SM) – – 143 – 143
Air Traffic Control (ATC) and Air 51 51 51 51 51
Traffic Management (ATM)
Dispatch (D) 2 2 2 2 2
Flight Planning (FP) 3 3 3 3 3
Ground Handling (GH) 4 4 4 4 4
Fuelers (F) 3 3 3 3 3
Network Density* 0.422 0.266 0.377 0.422 0.555

* Network density is calculated as a fraction of total possible connections in the network.

268 Systems Thinking in Practice

symmetrical – that each input is met with a corresponding output. For human–
human interaction, this would mean that a request is met with a response or that
an instruction is met with confirmation (this could be verbal or non-verbal, such as
moving a lever that is seen by the person issuing the instruction). In human–machine
interaction, this would mean that depressing a switch or typing an instruction into
the automation is met with some form of output, such as a change in the display sta-
tus or direct voice output (Table 13.5).
The analysis of the networks across the various metrics confirm that the two
networks on the right-hand side of the figures (i.e. the single pilot with additional
pilot on the ground – top right – Option C) and single pilot with additional pilot
on the ground and automation mirror on the ground (bottom right – Option D) are
the most robust versions of the network as there is less dependency on the single
role at the centre of the network (i.e. the pilot flying). This is revealed in several
ways. First, the eccentricity of the nodes in the network (i.e. the largest number of
hops an agent has to make to get from one side of the network to the other) where
the pilot flying and the pilot on the ground have only one hop in the version on
the right-hand side of the figure, making them all connected in the terms of the
network archetypes (as shown by the number of ties). The pilot on the ground has
the highest nodal degree and sociometric status on the right-hand side of the figure,
which is due in part to the number of ties and in part to the greatest number of
emission and reception connections with other nodes in the network. The pilot fly-
ing and the pilot on the ground are highest on the centrality and closeness metrics
because they are both at the centre of the network and have the shortest distance
to every other node. This suggests that they have high importance in the network,
which is not very surprising (Baber et al. 2013; Harris et al. 2015; Kenney et al.
2013). They are also lowest on farness for the same reasons. On the betweenness
metric, the pilot flying exerts less influence over the network because this influence
is shared with the pilot on the ground. The combination of more links and less
dependency on any one agent makes for a more robust network (Stanton 2014). In
summary, SNA can be used to determine the importance of different agents within
a social network and also to classify the network type. SNA offers a comprehensive
analysis of the network in question. The key agents within the network are speci-
fied, as are the frequency and direction of communications within the network. As
such, it proves useful for comparing the different networks that result from the new
single-pilot concepts.
As a development of method, the relationships between CWA and SNA are not
necessarily obvious but have allowed novel exploration of the differences between
dual, single and distributed crewing options in this chapter. To that extent, we have
further developed the approach reported by Houghton et al. (2006) and Baber et al.
(2013). The SOCA-CAT phase of CWA offers a formative allocation of functions
against situations (phases of flight in the present case). As such, it has enabled us
to consider the possible communications between the agents in the system for all
four of our proposed options against the current dual crewing operations. As safety
is paramount in aviation, we require any future option to be as least as safe as con-
temporary crewing. SNA offers the opportunity to consider the likely effects on the
communication networks from a quantitative perspective.
TABLE 13.5
Social Network Analysis of the Four Future Concepts

ATC ATC

OPTION A PF AA ATM D FP GH F OPTION C PF PG AA ATM D FP GH F

Emission/ 151 102 52 5 6 6 6 Emission/ 115 202 102 86 7 8 8 8
From CWA to SNA

Reception Reception
Eccentricity 1 2 2 2 2 2 2 Eccentricity 1 1 2 2 2 2 2 2
Sociometric 33.55 22.66 11.55 1.11 1.33 1.33 1.33 Sociometric 25.55 44.88 22.66 19.11 1.55 1.77 1.77 1.77
Status Status
Centrality 5.00 3.75 3.75 3.33 3.33 3.00 3.00 Centrality 5.28 5.28 4.11 4.11 3.70 3.70 3.36 3.36
(B-L) (B-L)
Closeness 1.5 1.15 1.12 1.00 1.00 0.90 0.90 Closeness 1.28 1.28 1.00 1.00 0.90 0.90 0.82 0.82
Farness 6 8 8 9 9 10 10 Farness 7 7 9 9 10 10 11 11
Betweenness 13 3 2 0 0 0 0 Betweenness 7.33 7.33 2.00 1.33 0 0 0 0
OPTION B PF AA SM ATC D FP GH F OPTION D PF PM AA SM ATC D FP GH F
ATM ATM
Emission/ 245 204 204 54 5 6 10 8 Emission/ 127 284 204 204 88 7 8 12 10
Reception Reception
Eccentricity 1 2 2 2 2 2 2 2 Eccentricity 1 1 2 2 2 2 2 2 2
Sociometric 54.44 45.33 45.33 12.00 1.11 1.33 2.22 1.77 Sociometric 28.22 63.11 45.33 45.33 19.55 1.55 1.77 2.66 2.22
Status Status
Centrality 5.57 4.33 4.33 4.33 3.54 3.54 3.54 3.54 Centrality 5.87 5.87 4.7 4.7 4.7 3.9 3.9 3.9 3.9
(B-L) (B-L)
Closeness 1.28 1.00 1.00 1.00 0.81 0.81 0.81 0.81 Closeness 1.12 1.12 0.9 0.9 0.9 0.75 0.75 0.75 0.75
Farness 7 9 9 9 11 11 11 11 Farness 8 8 10 10 10 12 12 12 12
Betweenness 14 2 2 4 0 0 0 0 Betweenness 8.16 8.16 1.5 1.5 2.66 0 0 0 0

KEY: single pilot (top left – Option A), single pilot with system mirror on the ground (bottom left – Option B), single pilot with additional pilot on the ground (top
right – Option C) and single pilot with additional pilot on the ground and system mirror on the ground (bottom right – Option D).
269
270 Systems Thinking in Practice

CONCLUSIONS
Previous approaches to developing a single-crew aircraft have focused on the
development of sophisticated airborne technology to assist the pilot, but they have
achieved only mixed success (Harris 2007). However, there are no major reasons
why a single-pilot-operated commercial aircraft is not feasible in the very near future
using existing technology. Military aviation has flown complex, high-performance,
single-crew aircraft for many years, and Uninhabited Air Vehicles (UAVs) are now
commonplace. It is time for these technologies to be spun out further into the com-
mercial domain, where they may be applied to financial advantage. The greatest
obstacle to the operation of civil, single-pilot aircraft is not the technology per se.
It is combining the technology, designing the user interfaces and developing a new
concept of operations to make such an aircraft safe and useable in a wide range of
normal and non-normal operating situations when flown by a typical commercial
pilot. The Human Factors requirements are the prime driver in this case, not the
technology (Harris 2007).
Instead of developing a concept of single-pilot operations based upon equipping
an aircraft with complex automation to aid the pilot, the approach described in this
chapter begins to demonstrate how by using a distributed, air/ground, sociotechnical
system, such an aircraft could be ready for service entry within a decade. Instead
of replacing many of the functions of the second pilot, they are simply displaced
to assistance on the ground (q.v. the operation of UAVs). Furthermore, in many cir-
cumstances, the assistance provided from the ground may be of higher quality and
better targeted than that normally available from the second seat on the flight deck
as a result of being able to draw upon a wider range of engineering and flight plan-
ning resources.
CWA, especially the SOCA-CAT phases, has been useful as a formative system
design approach to aid understanding of the function allocation distribution between
the agents and the functional loading on them. Extending this analysis to understand
communication interactions in SNA has revealed Option D (single pilot with addi-
tional pilot on the ground and system mirror on the ground) to be the most resilient,
even more than the current dual-pilot cockpit. It is debatable whether or not this option
could have prevented the recent tragedy with Germanwings (BBC 2015), which has
revealed a further weakness with current operations. Nevertheless, it is most likely that
reduced crewing options are most likely to be implemented in short-haul cargo opera-
tions before they ever become a possibility for passenger-carrying aircraft.
The baseline descriptions for general operations of a conventional two-crew air-
liner (using CWA and SNA) and the same airliners being operated by just a single
crew member give an overall idea of the level of complexity and resilience in the
current system that need to be achieved in future single-crew operations. There is a
natural split between the PF and PM functions that have evolved over decades. The
main question being addressed here is whether these roles have to be co-located. In
particular, the CWA and SNA describe the functions required irrespective of their
location. The social network statistics produced for the various configurations of
single-crew operations indicate that equivalent (or even enhanced) levels of system
resilience may be achieved.
From CWA to SNA 271

Future research will extend the analysis into contemporary issues such as
trajectory, conflict and system-wide information management. Another important
goal of this research is to validate the SOCA-CAT and SNA models in flight sim-
ulators. This would require developing the scenarios and structures as presented
in the four options and testing against the baseline with pilot crews. This analysis
would also enable experiential feedback from pilots to be gathered to supplement the
performance data, offering an opportunity to understand how alternative crewing
options might affect the duty of care the pilots have for the passengers that they are
carrying. The immediate future of the flight deck is unlikely to change dramatically,
but we can foresee a time within the next 20 years when the distributed crewing
concept could enter short-haul cargo operations. When this is proven to be safe and
effective, it might even enter into service for short-haul passenger aircraft.

REFERENCES
Airbus Industrie (n.d.). A319/A329/A321 Flight Crew Operating Manual (Volume 3: Flight
Operations) – Revision 36. Author: Blagnac, France.
Alcock, C. (2004). New turboprop push tied to rising fuel costs. Aviation International
News, 29 September 2004. Available at HYPERLINK “http://www.ainonline.
com/Publications/era/era_04/era_newturbop18.html” http://www.ainonline.com/
Publications/era/era_04/era_newturbop18.html (Accessed 30 January 2007).
Baber, C., Stanton, N. A., Atkinson, J., McMaster, R. and Houghton, R. J. (2013). Using
social network analysis and agent-based modelling to explore information flow using
common operational pictures for maritime search and rescue operations. Ergonomics,
56, 889–905.
BBC (2015). Germanwings crash latest updates. Available at http://www.bbc.com/news/live/
world-europe-32070570 (Accessed 4 June 2015).
Bonner, M., Taylor, R., Fletcher, K. and Miller, C. (2000). Adaptive automation and decision
aiding in the military fast jet domain. In: D. B. Kaber and M. R. Endsley (Eds.), Human
Performance, Situation Awareness and Automation: User-Centred Design for the New
Millennium (pp. 154–159). Omnipress: Madison, WI.
Civil Aviation Authority (2008). Global Fatal Accident Review 1997–2006 (CAP 776). Civil
Aviation Authority: London, UK.
DeJohn, C. A., Wolbrink, A. M. and Larcher, J. G. (2004). In-Flight Medical Incapacitation
and Impairment of U.S. Airline Pilots: 1993 to 1998 (DOT/FAA/AM-04/16). US
Department of Transportation, Federal Aviation Administration, Office of Aerospace
Medicine: Washington, DC.
Deutch, S. and Pew, R. W. (2005). Single Pilot Commercial Aircraft Operation (BBN Report
No. 8436). BBN Technologies: Cambridge, MA.
Driskell, J. E. and Mullen, B. (2004). Social network analysis. In N. A. Stanton et al. (Eds.),
Handbook of Human Factors and Ergonomics Methods (pp. 58.1–58.6). CRC Press:
London, UK.
easyJet plc (2013). Annual report and accounts 2013. Available at: http://corporate.easyjet.
com/~/media/Files/E/Easyjet-Plc-V2/pdf/investors/result-center-investor/annual-
report-2013.pdf (Accessed 27 June 2014).
European Aviation Safety Agency (2014). Certification Specifications for Large Aeroplanes
(CS-25): Amendment 15. EASA: Cologne. Available at: http://easa.europa.eu/system/
files/dfu/00%20Annex%20to%20ED%20Decision%202014-026-R.pdf.
Evans, S. and Radcliffe, S. A. (2012). The annual incapacitation rate of commercial pilots.
Aviation, Space and Environmental Medicine, 83(1), 42–49.
272 Systems Thinking in Practice

Galea, E. R., Filippidis, L., Wang, Z., Lawrence, P. J. and Ewer, J. (2011). Evacuation analysis
of 1000+ seat Blended Wing Body aircraft configurations: Computer simulations and
full-scale evacuation experiment. In: R. D. Peacock, E. D. Kuligowski and J. D. Averill
(Eds), Proceedings Pedestrian and Evacuation Dynamics 2010: 5th International
Conference (pp. 151–162). Springer: New York, NY.
Harris, D. (2003). The human factors of fully automatic flight. Measurement and Control,
36, 184–187.
Harris, D. (2007). A human-centred design agenda for the development of a single crew oper-
ated commercial aircraft. Aircraft Engineering and Aerospace Technology, 79, 518–526.
Harris, D. and Stanton, N. A. (2010). Aviation as a system of systems. Ergonomics, 53(2),
145–148.
Harris, D., Stanton, N. A. and Starr, A. (2015). Spot the difference: Operational event sequence
diagrams as a formal method for work allocation in the development of single-pilot
operations for commercial aircraft. Ergonomics, 58 (11), 1773–1791.
Hollnagel, E. (2007). Flight decks and free flight: Where are the system boundaries? Applied
Ergonomics, 38, 409–416.
Houghton, R. J, Baber, C., McMaster, R., Stanton, N. A., Salmon, P. M., Stewart, R. and
Walker, G. H. (2006). Command and control in emergency services operations: A
social network analysis. Ergonomics, 49(12–13), 1204–1225.
Jenkins, D. P., Stanton, N. A., Walker, G. H. and Salmon, P. M. (2009). Cognitive Work
Analysis: Coping with Complexity. Ashgate: Aldershot, UK.
Keinrath, C., Vašek, J. and Dorneich, M. (2010). A cognitive adaptive man–machine Interface
for future flight decks. In: A. Droog and M. Heese (Eds), Performance, Safety and Well-
being in Aviation Proceedings of the 29th Conference of the European Association
for Aviation Psychology (20–24 September 2010, Budapest, Hungary). European
Association of Aviation Psychology.
Kenney, M., Horgan, J., Horne, C., Vining, P., Carley, K. M., Bigrigg, M. W., Bloom, M. and
Braddock, K. (2013). Organisational adaptation in an activist network: Social networks,
leadership, and change in al-Muhajiroun. Applied Ergonomics, 44(5), 739–747.
McIlroy, R. C. and Stanton, N. A. (2011). Getting past first base: Going all the way with cogni-
tive work analysis. Applied Ergonomics, 42(2), 358–370.
Naikar, N. (2006a). Beyond interface design: Further applications of cognitive work analysis.
International Journal of Industrial Ergonomics, 36, 423–438.
Naikar, N. (2006b). An examination of the key concepts of the five phases of cognitive work
analysis with examples from a familiar system. Proceedings of the Human Factors and
Ergonomics Society 50th Annual Meeting, 447–451.
Naikar, N. (2013). Work Domain Analysis: Concepts, Guidelines, and Cases. CRC Press:
London, UK.
Onken, R. (1994). Human-centred cockpit design through the knowledge-based cockpit assistant
system CASSY. In: NATO DRG Panel 8 Workshop on Improving Function Allocation for
Integrated System Design. TNO Institute for Perception: Soesterberg, NL.
Onken, R. (1997). Functional development and field test of CASSY – A knowledge based
cockpit assistant system. Knowledge-Based Functions in Aerospace Systems, AGARD
Lecture Series 200, AGARD, Neuilly-sur-Seine, France.
Rafferty, L. A., Stanton, N. A. and Walker, G. H. (2012). Human Factors of Fratricide.
Ashgate: Aldershot, UK.
Rasmussen, J., Pejtersen, A. and Goodstein, L. P. (1994). Cognitive Systems Engineering.
Wiley: New York, NY.
Ryanair (2009). Full Year Results, 2009. Available at: http://www.ryanair.com/doc/inves-
tor/2009/q4_2009_doc.pdf (Accessed 17 August 2010).
From CWA to SNA 273

Schulte, A. and Stütz, P. (2001). Cognitive concepts in mission management for air-to-ground
attack aircraft. In D. Harris (Ed.), Engineering Psychology and Cognitive Ergonomics
(pp. 23–30), Volume 5. Ashgate: Aldershot, UK.
Sorensen, L. J. and Stanton, N. A. (2013). Y is best: How distributed situational awareness is
mediated by organisational structure and correlated with task success. Safety Science,
56, 72–79.
Stanton, N. A., Baber, C. and Harris, D. (2008). Modelling Command and Control: Event
Analysis of Systemic Teamwork. Ashgate: Aldershot.
Stanton, N. A. (2014). Representing distributed cognition in complex systems: How a subma-
rine returns to periscope depth. Ergonomics, 57(3), 403–418.
Stanton, N. A. and Bessell, K. (2014). How a submarine returns to periscope depth: Analysing
complex socio-technical systems using cognitive work analysis. Applied Ergonomics,
45(1), 110–125.
Stanton, N. A., Harris, D. and Starr, A. (2014). Modelling and analysis of single pilot opera-
tions in commercial aviation. Proceedings of HCI Aero 2014. 30 July–1 August 2014,
Santa Clara, CA.
Stütz, P. and Schulte, A. (2001). Evaluation of the Cockpit Assistant Military Aircraft
(CAMA) in flight trials. In D. Harris (Ed.), Engineering Psychology and Cognitive
Ergonomics (pp. 15–22), Volume 5. Ashgate: Aldershot, UK.
Taylor, R. M., Howells, H. and Watson, D. (2000). The cognitive cockpit: Operational require-
ment and technical challenge. In: P. T. McCabe, M. A. Hanson and S. A. Robertson
(Eds.), Contemporary Ergonomics 2000 (pp. 55–59). Taylor and Francis: London, UK.
Thomas, M. J. W. (2003). Improving organisational safety through the integrated evaluation
of operational and training performance: An adaptation of the Line Operations Safety
Audit (LOSA) methodology. Human Factors and Aerospace Safety, 3, 25–45.
Vicente, K. J. (1999). Cognitive Work Analysis: Toward Safe, Productive, and Healthy
Computer-based Work. Lawrence Erlbaum Associates: Mahwah, NJ.
Walker, G. H., Cooper, M., Thompson, P. and Jenkins, D. (2014). Practitioner versus ana-
lyst methods: A nuclear decommissioning case study. Applied Ergonomics, 45(6),
1622–1633.
Walker, G. H., Stanton, N. A., Salmon, P. M., Jenkins, D., Stewart, R. and Wells, L. (2009).
Using an integrated methods approach to analyse the emergent properties of military
command and control. Applied Ergonomics, 40(4), 636–647.
Wiener, E. L. and Curry, R. E. (1980). Flight deck automation: Promises and problems.
Ergonomics, 23, 995–1011.
Wilson, J. R. and Carayon, P. (2014). Systems ergonomics: Looking into the future – Editorial
for special issue on systems ergonomics/human factors. Applied Ergonomics, 45(1),
3–4.
14 Future Directions for
the Event Analysis of
Systemic Teamwork

INTRODUCTION
This book has presented a series of studies in which the Event Analysis of Systemic
Teamwork (EAST) framework has been used to describe, understand, and even fore-
cast the behaviour of complex systems. EAST has shown itself capable of dealing
with highly complex, multi-agent, distributed systems, and in most cases the outputs
were used to identify interventions to enhance performance or safety. Whilst this is
encouraging, however, the systems in which human factors and ergonomics practitio-
ners work are arguably becoming more complex and technology driven (Grote et al.
2013; Dekker et al. 2013; Salmon et al., 2017; Walker et al. 2017; Woods and Dekker
2000). This, coupled with a dramatic shift towards the systems thinking paradigm, is
beginning to stretch the capabilities of many human factors and ergonomics methods.
We are certainly not alone in expressing these concerns. Dekker (2014), Leveson
(2011), Salmon et al. (2011; 2017) and Walker et al. (2017) discuss the increasing
complexities of modern-day systems and the extent to which they are rapidly outpac-
ing the capabilities of our methodological toolkit. Salmon et al. (2017) in particular
outlines five key challenges that human factors and ergonomics methods face. It is
our contention that the EAST framework can respond to these challenges.
The aim of this chapter is to demonstrate this by examining the contemporary ergo-
nomics problem space and the extent to which EAST can meet the challenges posed.
Specifically, we outline Salmon et al.’s five key challenges, and discuss the capacity of
EAST to respond to them. The five challenges include (Salmon et al., 2017):

• normal performance as a cause of accidents (Dekker 2011; Leveson 2004;

Rasmussen 1997);
• accident prediction (Salmon et al. 2014a);
• system migration (Rasmussen 1997);
• systems concepts (Hutchins 1995; Stanton et al. 2006); and
• human factors and ergonomics in design.

NORMAL PERFORMANCE AS A CAUSE OF ACCIDENTS

Significant progress has been made in understanding accident causation in safety crit-
ical systems. Systems models, in particular, are now widely accepted (Leveson 2004;

275
276 Systems Thinking in Practice

Rasmussen 1997), and there are a range of methods that enable accidents to be ana-
lysed from this perspective (e.g. Hollnagel 2012; Svedung and Rasmussen 2002;
Leveson 2004). Accidents are now widely acknowledged to be systems phenomena,
just as safety is (Hollnagel 2004; Dekker 2011). Both safety and accidents, there-
fore, are emergent properties arising from non-linear interactions between multiple
components distributed across a complex web of human and machine agents and
interventions (e.g. Leveson 2004).
This form of thinking, and the evolution of it, brings the methods we currently use
into question. Salmon et al. (2017) point out that, despite the great progress in safety
performance that has been made in most safety-critical sectors since the Second
World War, significant trauma still occurs, and in some areas, progress may be
slowing. Worse still, in some domains, incidents appear to be rising again. Leveson
(2011) agrees, and suggests that one reason is that our methods do not fully uncover
the underlying causes of accidents. Salmon et al. (2017) argue that a major the issue
may be that the evolution in accident causation models is not reflected in current
accident analysis methods.
One of the fundamental advances provided by state-of-the-art accident causa-
tion models relates to the idea that the behaviours underpinning accidents do not
necessarily have to be errors, failures or violations (e.g. Dekker 2011; Leveson
2004; Rasmussen 1997). As Dekker (2011) points out, systems thinking is about
how accidents can happen when no parts are broken. ‘Normal performance’
plays a role too (Perrow 1984). Salmon et al. (2017) outlines two tenets that have
important implications for the methods that we use. First, normal performance
plays a role in accident causation; and second, accidents arise from the very same

FIGURE 14.1 The pattern of global passenger fatalities per 10 million passenger miles on
scheduled commercial air transport since 1993. The graph shows that the precipitous drop in
fatality rates since 1945 has, since 2003, levelled off.
Future Directions for Systems Ergonomics Methods 277

behaviours and processes that create safety. Dekker (2011), for example, argues
that the seeds for failure can be found in ‘normal, day-to-day processes’ (p. 99)
and are often driven by goal conflicts and production pressures. These normal
behaviours include workarounds, improvisations, and adaptations (Dekker 2011)
but may also just be normal work behaviours routinely undertaken to get the job
done. It is only with hindsight and a limited investigation method that these normal
behaviours are treated as failures.
The tenets described above provide an interesting shift in the requirements for
accident analysis methods. Dekker (2014) argues that practitioners should not look
for the known problems that appear in incident reporting data or safety management
systems. Instead, he argues that the focus should be in the places where there are
no problems, in other words, normal work. In addition, the burgeoning concept of
Safety II (Hollnagel et al. 2013) argues that safety management needs to move away
from attempting to ensure that as little as possible goes wrong and towards ensuring
that as much as possible goes right. A key part of this involves understanding perfor-
mance when it went right as well as when it went wrong.
Salmon et al. (2017) discuss how, out of three popular contemporary acci-
dent analysis methods (Accimap, STAMP, HFACS), only the Accimap method
(Rasmussen, 1997) is capable of describing normal performance and of consid-
ering the role of normal performance in accident causation. AcciMap is used to
describe accidents in terms of contributory factors and the relationships between
them. This enables a comprehensive representation of the network of contributory
factors involved. It does this by decomposing systems into six levels, across which
analysts place the decisions and actions that enabled the accident in question to
occur (although the method is flexible in that the number of levels can be adjusted
based on the system in question). Interactions between the decisions and actions are
subsequently mapped onto the diagram to show the relationships between contribu-
tory factors within and across the six levels. A notable feature of AcciMap is that
it does not provide analysts with taxonomies of failure modes; rather, analysts have
the freedom to incorporate any factor deemed to have played a role in the accident
in question.
Both HFACS and STAMP provide taxonomies of error and failure modes that are
used to classify the behaviours involved in accident scenarios, which in turn means
that there is little scope for analysts to include behaviours other than those deemed to
have been failures of some sort. There is no opportunity for analysts to incorporate
normal behaviours in their descriptions of accidents – they have to force-fit events
into one of the error or failure modes provided. A consequence is that the normal
behaviours that contribute to accidents are not picked up during accident analysis
efforts. This may impact accident prevention activities by providing a false sense
of security that nothing else is involved and thus nothing needs modification (apart
from error-producing human operators).
AcciMap, on the other hand, does not use a taxonomy of failure or error modes
and so enables analysts to incorporate normal performance and to show its relation-
ship with other behaviours (Salmon et al., 2017). Despite this, AcciMap descriptions
still tend to incorporate contributory factors that relate only to failures, and few
Accimap analyses actually incorporate normal performance.
278 Systems Thinking in Practice

The EAST framework provides an approach that can be used to integrate analyses
of normal performance and accident causation. As demonstrated in this book, the
EAST framework is used primarily to describe ‘normal performance’ by describ-
ing the task, social and information networks underpinning behaviour. In addi-
tion, EAST has also been used for accident analysis purposes (Griffin et al., 2008;
Rafferty et al., 2012; Salmon et al., 2011; 2016). Since EAST was originally devel-
oped to describe normal performance, and has since been applied for accident analy-
sis purposes, it seems suited to incorporating both normal performance and failures
in descriptions of accidents. It is therefore recommended that further applications of
EAST for accident analysis purposes are undertaken. In particular, within the three
networks a focus should be placed on how normal performance either creates or
interacts with failures. This will allow rich, in-depth descriptions of the role of both
normal performance and failures in accidents.
The conclusion, then, is that there is room for improvement in our accident
analysis methods. Not all state-of-the-art methods are consistent with our current
understanding of accident causation. Further, even for the methods that are, such as
AcciMap, it is questionable whether they are being used in a manner that is consis-
tent with contemporary models of accident causation. This represents a key issue
for human factors and ergonomics researchers and practitioners and for safety sci-
ence generally. On the one hand, there is now a widespread understanding that the
role of normal performance in accidents is apparent and needs to be understood
(Dekker 2011, 2014; Leveson 2011; Rasmussen 1997). On the other hand, accident
analysis efforts, regardless of domain, do not seem to be dealing particularly well
with this feature. This means our understanding of accidents may be incomplete.
The EAST framework described in this book provides can be used to respond to
this challenge.

ACCIDENT PREDICTION
Forecasting accidents before they occur has been labelled the final frontier for human
factors and ergonomics (Moray 2008; Salmon et al. 2011; Stanton and Stammers
2008). Although there have been various attempts at developing accident prediction
models (e.g. Deublein et al. 2013), most are statistical models that are unable to iden-
tify and describe how behaviours across overall sociotechnical systems might com-
bine to create failure scenarios. Other predictive methods are available, such as those
that can be used to predict the ‘human errors’ that lead to accidents (see Stanton et al.
2013). Indeed, some of these methods have been shown to achieve acceptable levels
of reliability and validity (e.g. Stanton et al. 2009). The problem is that they predict
what is likely the last behaviour in a long and complex network of interacting and
emergent behaviours occurring across various parts of the system. They predict con-
sequences, not causes, and they do not identify the network of contributory factors
that might co-occur to create accidents. Accident prevention efforts are better served
by looking at the interactions that occur before the human operator makes the error.
In short, it is the entire accident scenario, including interacting factors and emergent
behaviours, that are important for understanding how to prevent accidents.
Future Directions for Systems Ergonomics Methods 279

A Systems Approach to Prediction

A key requirement, then, is a prediction method that is underpinned by systems
thinking. Error prediction methodologies can be thought of as reductionist (although
they are not entirely reductionist as they do focus on human–machine interactions).
Reductionist approaches, those that rely on taking the system apart in order to under-
stand the components and then reassembling the components back into the complete
system (on the tacit assumption that the whole cannot be greater or less than the sum
of its parts), do not allow us to detect the emergent properties associated with the
types of risk issues upon which we wish to make progress (see Walker et al. 2009).
Systems approaches such as EAST do. One means by which they can enable fore-
casting and prediction is to consider the causal texture of the systems environment
and the system’s movement through that environment.
As discussed, the systems approach has become popular in part because of vari-
ous systems analysis methodologies that can, to some extent at least, do this (e.g.
Rasmussen 1997). These methods, for example AcciMap (Rasmussen 1997), are
becoming increasingly popular for accident investigation purposes. A major limita-
tion of these methods is that, so far, they have not been used predictively; organ-
isations are effectively waiting for loss events to occur before they can work on
prevention strategies. The lack of data resulting from improved safety trends com-
bined with greater operational intensity and risk exposure means that, if anything,
loss events are more likely to be large scale and unexpected, meaning that ‘learning
from disasters’ is becoming increasingly dubious from an ethical perspective. The
need for systems-based prediction approaches is discussed extensively in the litera-
ture (e.g. Moray 2008; Salmon et al. 2011; 2017; Stanton and Stammers 2008) but as
yet, a credible approach has yet to emerge.

East and Accident Prediction

Encouragingly, the EAST framework is beginning to demonstrate its utility for pre-
dicting system behaviour (Stanton & Harvey, 2017).
Stanton & Harvey (2017 See chapter 10) first explored the predictive capacity of
EAST by using it to assess the risks associated with Royal Navy low flying hawk
training scenarios designed to simulate a sea skimming missile attack. This involved
first using EAST to describe the task, social and information networks involved, and
then using a breaking links process to identify potential risks. This involved adding
and breaking links and nodes within the networks to explore different system states.
The effect of this make/break-link process is to create ‘short circuits’, ‘long circuits’
or ‘no circuits’, all of which put systems into new configurations (see Figure 14.2).
For example, adding and breaking links within the networks reveals instances where
a human operator may or may not be aware of a particular piece of information,
where a task will or will not be fulfilled by a human operator or piece of technol-
ogy or where a required communication may or may not be made. By these means,
it is possible to model the majority of possible accident pathways in a given system
model under a wide range of different permutations. On the other hand, adding and
breaking links may reveal aspects of normal performance that move towards the
280 Systems Thinking in Practice

FIGURE 14.2 The EAST models sociotechnical systems by combining task, social and
knowledge networks into composite networks that can be systematically degraded into all
possible configurations.

boundary of safe operations (Rasmussen 1997). This systems level focus encourages
a different approach to forecasting resilience: to examine persistent and emergent
patterns that arise even though the boundaries of the system cannot always be fully
known. Further testing of EAST in this capacity is currently being undertaken by
the authors, and it is recommended that further predictive applications of EAST are
undertaken.

MIGRATION TOWARDS SAFETY BOUNDARIES

Rasmussen’s risk management framework (Rasmussen 1997; see Figure 14.3)
is becoming one of the most popular safety and risk management models of our
time (e.g. Salmon et al. 2014c; Underwood and Waterson 2014; Goode et al. 2014).
In his seminal article, Rasmussen outlined the concept of migration based on the
‘Brownian movements’ of gas molecules, describing how organisations shift towards
and away from safety and performance boundaries due to various constraints includ-
ing financial, production and performance pressures. According to Rasmussen, there
is a boundary of economic failure: the financial constraints on a system that influ-
ence behaviour towards greater cost efficiencies. There is also a boundary of unac-
ceptable workload: the pressures experienced by people and equipment in the system
as they try to meet economic and financial objectives. The boundary of economic
failure creates a pressure towards greater efficiency, which works in opposition to
a similar pressure against excessive workload. As systems involve human as well
as technical elements, and because humans are able to adapt situations to suit their
own needs and preferences, these pressures inevitably introduce variations in behav-
iour that are not explicitly designed and can lead to increasingly emergent system
behaviours, both good and bad (Qureshi 2007; Clegg 2000). Over time, this adaptive
Future Directions for Systems Ergonomics Methods 281

Real, invisible, safety boundary

Economic failure
boundary

Adverse events

Unacceptable
workload boundary

Boundary defined by
official work practices

FIGURE 14.3 Rasmussen’s dynamic safety space (Adapted from Rasmussen 1997).

behaviour can cause the system to cross safety boundaries and accidents to happen
(Qureshi 2007; Rasmussen 1997). The key, then, is to detect in advance a) where
those boundaries are and b) where the system is travelling in relation to them.

Mapping Migration
Unfortunately, there are few human factors and ergonomics methods that are capable
of describing where boundaries are situated and whereabouts organisations may be
in relation to them. Further, the ability to dynamically track an organisation’s migra-
tion is not readily supported by current methods. Whilst information on so-called
lagging indicators can provide an indication of proximity to a safety boundary, there
is an absence of human factors and ergonomics methods that use leading indicators
to track organisational performance and safety. Work currently being undertaken by
the authors of this book is exploring the capacity for EAST to be used as a frame-
work for a. identifying leading indicators, and b. tracking organisational behaviour
and ultimately migration towards and away from safety boundaries.

SYSTEMS CONCEPTS
Salmon et al. (2017) argue that many human factors and ergonomics concepts could
be better understood by adopting a systems thinking lens. Citing accident causa-
tion and distributed situation awareness as examples of where a systems thinking
approach has proved successful, Salmon et al. (2017) argue that other concepts such
as mental workload and decision making should be viewed through a systems lens.
They also argue, however, that we do not possess the methods to support this. It is
282 Systems Thinking in Practice

our view that EAST provides an appropriate method with which to view concepts
such as mental workload and decision making through a systems lens.
Mental workload (see Young et al. 2015) is a case in point. Whilst it has predomi-
nantly been thought of as an individual operator concept, it is increasingly being
examined at the team level (Helton et al. 2014), and there is no reason why it cannot
be considered at a systems level. Just as Stanton et al. (2006) describe how situa-
tion awareness is an emergent property of systems and is distributed across opera-
tors, mental workload can be also thought of in this way. Moreover, similar to the
transactions in awareness described by Stanton and colleagues, transactions in men-
tal workload between operators are readily apparent in sociotechnical systems. An
example of this is the workload shedding by air traffic controllers, dividing sectors
up as the air traffic increases (Walker et al. 2010).
There are many methods available to support the assessment of individual operator
workload, including the highly popular NASA TLX (Hart and Staveland 1988) and
many similar subjective rating scales (see Stanton et al. 2013). In addition, there are
other individual focused methods, such as psychophysiological measures. Further,
methodologies that support the assessment of team mental workload are emerging
(Helton et al. 2014), although these are not without their problems. Unfortunately,
methods that can specifically consider mental workload at a systems level do not yet
exist. Such a method, something akin to ‘distributed mental workload’ assessment,
would need to consider the workload of multiple agents, how interactions between
actors shapes each other’s workload, how different levels of workload dynamically
shift throughout task performance and, further, what wider systemic factors con-
strain or facilitate workload. In addition, under a systems view, workload across
system levels and even non-human actors should be considered.
Again, it is our opinion that EAST can be used to provide such an analysis.
Specifically, EAST considers multiple agents and interactions between them (via
the social network), the tasks being undertaken by different agents (via the task
network), and the information load associated with different tasks and agents (via
the information network). Early applications of the EAST framework used the
information network to provide an indirect assessment of operator workload by
calculating the total number of information nodes being used by different opera-
tors during different scenario phases (e.g. Salmon et al., 2008). This provides a
dynamic indication of workload as different tasks are undertaken. As such, EAST
can be used to assess the level of workload imposed on different agents, the dis-
tribution of this workload across the system, and how this workload is exchanged
across agents.

HUMAN FACTORS AND ERGONOMICS METHODS IN DESIGN

Human factors and ergonomics has a key role to play in the design of new tech-
nologies, interfaces, training programmes, procedures and even overall systems.
Despite this, a criticism of many human factors and ergonomics methods is that
they do not directly contribute to design – that is, they are not used by designers to
design (Salmon et al., 2017).
Future Directions for Systems Ergonomics Methods 283

The requirement to shift the emphasis on human factors and ergonomics to the
front end of the design life cycle is well known (Norros 2014; Salmon et al., 2017).
However, it requires human factors and ergonomics methods that can be used by
designers and design teams to design or that can at least be integrated with existing
design approaches. It also requires human factors and ergonomics researchers and
practitioners to take the lead to facilitate design efforts, which would be a paradigm
shift in the way complex systems are engineered.
Whilst a significant shift is required, it is notable that many of our methods are
suited to directly and indirectly informing design. Rather than develop new meth-
ods per se, the pressing requirement seems to be the development of processes that
bridge the gap between analysis and design. Specifically, processes are required
to translate human factors and ergonomics analyses into specific design require-
ments. One such approach that has been developed recently for this purpose is the
Cognitive Work Analysis design toolkit (CWA-DT) (Read et al. 2015; 2017).
The CWA-DT assists CWA users to identify extract insights from CWA analyses
and to use these insights within a participatory design paradigm. It promotes the col-
laborative involvement of experts (i.e. human factors and ergonomics professionals,
designers and engineers), stakeholders (i.e. company representatives, supervisors,
unions) and end-users (i.e. workers or consumers) to solve design problems based on
insights gained through CWA.
Underlying the CWA-DT is both the design philosophy of CWA (i.e. ‘let the
worker finish the design’) and the related sociotechnical systems theory approach,
which aims to design organisations and systems that have the capacity to adapt and
respond to changes and disturbances in the environment (Trist and Bamforth 1951;
Cherns 1976; Clegg 2000; Walker et al. 2009). Consequently, the CWA-DT provides
design tools and methods that encourage consideration of the values underlying the
sociotechnical systems approach (and indeed underpinning human factors ergonom-
ics more generally). These include the notion of humans as assets or adaptive deci-
sion makers rather than error-prone liabilities; of technology being designed as a tool
to assist humans to achieve their goals, rather than implemented because of assumed
efficiency or cost savings; the consideration of the needs and capabilities of a diverse
set of end-users; and of design focussing on promoting the quality of life or wellbe-
ing for end-users. Further, the consideration of sociotechnical design principles, such
as minimal critical specification, boundary management and joint design of social
and technical elements, intends to achieve the design of systems that can operate
within their safety and performance boundaries both on implementation and in an
ongoing fashion through continual monitoring and re-design.
Whilst the approach was originally designed for use with CWA outputs, there
is no reason why it cannot be used in conjunction with the outputs of other systems
human factors and ergonomics methods such as EAST. Indeed, Salmon et al (2018)
recently used the CWA-DT in conjunction with both CWA and EAST as part of an
intersection design process. This involved using insights derived from the EAST
analyses of road user behaviour presented in Chapter 7 as part of a CWA-DT par-
ticipatory design process to design three new intersections. An emphasis was placed
on creating designs that catered to the needs of all forms of end-user, rather than
designing exclusively for drivers and motorised vehicles. The intention was to create
284 Systems Thinking in Practice

FIGURE 14.4 The EAST models sociotechnical systems by combining task, social and
knowledge networks into composite networks that can be systematically degraded into all
possible configurations.

intersection design concepts that, once implemented, would support safe interactions
between drivers, cyclists, motorcyclists and pedestriansby ensuring that all road
users are aware of each other and understand each others’ likely behaviours. The
design brief was to replace one of the intersections from the EAST study presented
in Chapter 6 (See Figure 14.4).
When we evaluated the three designs with drivers, cyclists, motorcyclists and
pedestrians, two of the designs performed best against key criteria: alignment with
sociotechnical systems values, attainment of key intersection functions (such as to
minimise collisions, maximise efficiency, maximise compliance, optimise flexibil-
ity), and user preferences.
The first design is known as the “turning team” design (see Figure 14.5).
It works on the premise that different road users could work effectively as a
team when proceeding through the intersection. To do this the design aims to
make drivers explicitly aware of other forms of road user (to connect the team)
and provides each with a clear and dedicated path through the intersection. The
design aims to clear cyclists from the intersection before allowing motorised
traffic to enter. Other features include a pedestrian crossing path wide enough to
accommodate cyclists who are not comfortable with using the road, motorcyclist
filtering lanes, and phasing of traffic lights based on road user type and direction
of travel.
The second design is the “circular” concept. It explicitly separates motorised and
non-motorised traffic. A circular pathway around the intersection is provided for
pedestrians and cyclists to use. This pathway links with cycle lanes running down
the centre of the road, separated by a kerb from the roadway. On the roadway, this
design provides a separate bus lane and a motorcycle zone at the front of the intersec-
tion to encourage motorcyclists to filter to the front. Finally, the design incorporates
signs warning motorists to be on the lookout for cyclists and for motorcyclists filter-
ing through the traffic from behind (Figure 14.6).
Future Directions for Systems Ergonomics Methods 285

FIGURE 14.5 Turning teams intersection design concept.

CONCLUSIONS
Following on from Salmon et al. (2017), this chapter examined some of the chal-
lenges posed for human factors and ergonomics methods by the contemporary par-
adigm of systems thinking and discussed how the EAST framework can respond
to the challenges posed. Although human factors and ergonomics researchers and
practitioners have a range of methods at their disposal, the popular paradigm of
systems thinking requires analyses that go beyond their capabilities. The discussion
presented in this chapter has suggested that

• accident analysis methods, though high on explanatory power, do not

describe accident causation in a manner that is congruent with contempo-
rary models. Based on previous accident analysis applications, EAST could
be used to describe the role of normal performance in accident causation;
• despite there being a range of appropriate candidate methods, we currently
do not have a method that supports the prediction of accidents. EAST is
286 Systems Thinking in Practice

beginning to show its capacity for predicting system behaviour and can be
used to predict specific failure scenarios;
• the migration of systems towards and away from safety boundaries has not
yet been dealt with by ergonomics methods. EAST could potentially be
used to identify leading indicators and then track the migration of system
behaviour towards and away from system boundaries;
• various ergonomics constructs may be suited to systems level analysis, but
there is a lack of ergonomics-based systems methods to support this. EAST
can be used to describe and assess concepts such as mental workload and
decision making from a systems perspective; and
• despite its critical role in the design process, few ergonomics methods are
actually used by designers to design. The recently developed CWA-DT
(Read et al., 2015; 2017) can be used in conjunction with EAST to create
novel designs that align with core sociotechnical systems theory values and
principles.

FIGURE 14.6 Circular intersection design concept.

Future Directions for Systems Ergonomics Methods 287

In closing, we would like to emphasise the important role that new and emerging
systems thinking-based methods such as EAST have in ensuring that human factors
and ergonomics remain relevant. In a world where complexity is apparently increas-
ing, and problems are becoming more difficult to resolve, methods that have the
capacity of describing systems in a manner that allows us to understand what compo-
nents exist in the system and how they interact together are required. Only by taking
the overall system as the unit of analysis, and by considering all components and how
they interact together to shape behaviour can we begin to understand systems and
their emergent properties. EAST allows human factors and ergonomics researchers
and practitioners to do this. We hope that the reader has found this book useful, and
that further applications of EAST are undertaken. In particular, we encourage appli-
cations in new domains and focussed on problems not previously examined such as
terrorism, cybersecurity, and artificial intelligence.

REFERENCES
Cherns, A. (1976). The principles of sociotechnical design. Human Relations, 29, 783–792.
Clegg, C. W. (2000). Sociotechnical principles for system design. Applied Ergonomics, 31,
463–477.
Dekker, S. (2011). Drift into Failure: From Hunting Broken Components to Understanding
Complex Systems. Ashgate: Aldershot, UK.
Dekker, S. (2014). The Field Guide to Understanding Human Error, Third Edition. Ashgate:
Aldershot, UK.
Dekker, S. W. A., Hancock, P. A. and Wilkin, P. (2013). Ergonomics and sustainability:
Towards an embrace of complexity and emergence. Ergonomics, 56(3), 357–364.
Deublein, M., Schubert, M., Adey, B. T., Köhler, J. and Faber, M. H. (2013). Prediction of
road accidents: A Bayesian hierarchical approach. Accident Analysis & Prevention,
51, 274–291.
Goode, N. Salmon, P. M., Lenné, M. G. and Hillard, P. (2014). Systems thinking applied to
freight handling operations. Accident Analysis and Prevention, 68, 181–191.
Griffin, T. G. C., Young, M. S. and Stanton, N. A. (2008). Investigating accident causation
through information network modelling. Ergonomics, 53(2), 198–210.
Griffin, T. G. C., Young, M. S. and Stanton, N. A. (2010). Investigating accident causation
through information network modelling. Ergonomics, 53(2), 198–210.
Grote, G., Weyer, J. and Stanton, N. A. (2013). Beyond human-centred automation – concepts
for human–machine interaction in multi-layered networks. Ergonomics, 57(3), 289–294.
Helton, W. S, Funke, G. J, and Knott, B. A. (2014). Measuring workload in collaborative con-
texts: Trait versus state perspectives. Human Factors, 56(2), 322–332.
Hollnagel, E. (2012). FRAM: The Functional Resonance Analysis Method: Modelling
Complex Socio-technical Systems. Ashgate: Aldershot, UK.
Hollnagel, E., Leonhardt, J., Licu, T. and Shorrock, S. (2013). From Safety-I to Safety-II:
A white paper. Eurocontrol. Available at: https://www.eurocontrol.int/sites/default/
files/content/documents/nm/safety/safety_whitepaper_sept_2013-web.pdf (Accessed
21 July 2015).
Hutchins, E. (1995). Cognition in the Wild. MIT Press: Cambridge, MA.
Leveson, N. G. (2004). A new accident model for engineering safer systems. Safety Science,
42(4), 237–270.
Leveson, N. G. (2011). Applying systems thinking to analyze and learn from events. Safety
Science, 49, 55–64.
288 Systems Thinking in Practice

Moray, N. (2008). The good, the bad, and the future: On the archaeology of ergonomics.
Human Factors, 50(3), 411–417.
Norros, L. (2014). Developing human factors/ergonomics as a design discipline. Applied
Ergonomics, 45(1), 61–71.
Perrow, C. (1984). Normal Accidents: Living with High-Risk Technologies. Basic Books: New
York, NY.
Qureshi, Z. H. (2007). A review of accident modelling approaches for complex critical socio-
technical systems. Conferences in Research and Practice in Information Technology
Series, Vol. 336.
Rafferty, L. A., Stanton, N. A., (2012). The Human Factors of Fratricide. CRC Press, Boca
Raton.
Rasmussen, J. (1997). Risk management in a dynamic society: A modelling problem. Safety
Science, 27(2/3), 183–213.
Read, G., Salmon, P. M., Lenne, M. G., Jenkins, D. P., (2015). Designing a ticket to ride
with the Cognitive Work Analysis Design Toolkit. Ergonomics. 58:8, 1266–86. doi:
10.1080/00140139.2015.1013576.
Read, G. J. M., Beanland, V., Lenne, M. G., Stanton, N. A., Salmon, P. M., (2017). Integrating
Human Factors methods and systems thinking for transport analysis and design. CRC
Press, Boca Raton, FL.
Reason, J. (1990). Human Error. Cambridge University Press: New York, NY.
Salmon, P. M., Stanton, N. A., Lenné, M., Jenkins, D. P., Rafferty, L. A. and Walker, G.H.
(2011). Human Factors Methods and Accident Analysis. Ashgate: Aldershot.
Salmon, P. M., Lenné, M. G., Read, G., Walker, G. H. and Stanton, N. A. (2014a). Pathways
to failure? Using Work Domain Analysis to predict accidents in complex systems. In:
T. Ahram, W. Karwowski and T. Marek (Eds), Proceedings of the 5th International
Conference on Applied Human Factors and Ergonomics AHFE 2014, Kraków, Poland,
19–23 July 2014.
Salmon, P. M., Goode, N., Lenné, M. G., Cassell, E. and Finch, C. (2014c). Injury causation in
the great outdoors: A systems analysis of led outdoor activity injury incidents. Accident
Analysis and Prevention, 63, 111–120.
Salmon, P. M., Walker, G. H., Stanton, N. A. (2016). Pilot error versus sociotechnical sys-
tems failure? A distributed situation awareness analysis of Air France 447. Theoretical
Issues in Ergonomics Science, 17(1), 64–79.
Salmon, P. M., Walker, G. H., Read, G. J. M., Goode, N. and Stanton, N. A. (2017). Fitting
methods to paradigms: are ergonomics methods fit for systems thinking?. Ergonomics,
60(2), 194–205.
Salmon, P. M., Read, G. J. M., Walker, G. H., Lenne, M. G. and Stanton, N. A. (2018).
Distributed situation awareness in road transport: theory, measurement, and applica-
tion to intersection design. CRC Press, Boca Raton, FL.
Stanton, N. A. and Harvey, C. (2017). Beyond human error taxonomies in assessment of risk
in sociotechnical systems: A new paradigm with the EAST ‘broken-links’ approach.
Ergonomics, 60(2), 221–233.
Stanton, N. A. and Stammers, R. B. (2008). Bartlett and the future of ergonomics. Ergonomics,
51(1), 1–13.
Stanton, N. A., Salmon, P. M., Rafferty, L., Baber, C., Walker, G. H. and Jenkins, D. P. (2013).
Human Factors Methods: A Practical Guide for Engineering and Design, Second
Edition. Ashgate: Aldershot, UK.
Stanton, N. A., Stewart, R., Harris, D., Houghton, R. J., Baber, C., McMaster, R., Salmon,
P. M., Hoyle, G., Walker, G. H., Young, M. S., Linsell, M., Dymott, R. and Green, D.
(2006). Distributed situation awareness in dynamic systems: Theoretical development
and application of an ergonomics methodology. Ergonomics, 49, 1288–1311.
Future Directions for Systems Ergonomics Methods 289

Svedung, I. and Rasmussen, J. (2002). Graphic representation of accident scenarios: Mapping

system structure and the causation of accidents. Safety Science, 40, 397–417.
Trist, E. L. and Bamforth, K. W. (1951). Some social and psychological consequences of the
longwall method of coal-getting: An examination of the psychological situation and
defences of a work group in relation to the social structure and technological content of
the work system. Human Relations, 4, 3–38.
Underwood, P. and Waterson, P. (2013). Examining the gap between research and practice.
Accident Analysis & Prevention, 55, 154–164.
Underwood, P. and Waterson, P. (2014). Systems thinking, the Swiss Cheese Model and acci-
dent analysis: A comparative systemic analysis of the Grayrigg train derailment using
the ATSB, AcciMap and STAMP models. Accident Analysis & Prevention, 68, 75–94.
Walker, G. H., Stanton, N. A., Salmon, P. S. and Jenkins, D. P. (2009). Command and Control:
The Sociotechnical Perspective. Ashgate: Aldershot, UK.
Walker, G. H., Gibson, H., Stanton, N. A., Baber, C., Salmon, P. M. and Green, D. (2006).
Event analysis of systemic teamwork (EAST): A novel integration of ergonomics meth-
ods to analyse C4i activity. Ergonomics, 49, 1345–1369.
Walker, G. H., Stanton, N. A., Ba‑ber, C., Wells, L., Gibson, H., Salmon, P. M. and Jenkins,
D. P. (2010). From ethnography to the EAST method: A tractable approach for repre-
senting distributed cognition in air traffic control. Ergonomics, 53 (2), 184–197.
Walker, G. H., Salmon, P. M., Bedinger, M., Cornelissen, M. and Stanton, N. A. (2017).
Quantum ergonomics: Shifting the paradigm of the systems agenda. Ergonomics,
60(2), 157–166.
Woods, D. D. and Dekker, S. (2000). Anticipating the effects of technological change: A new
era of dynamics for human factors. Theoretical Issues in Ergonomics Science, 1(3),
272–282.
Young, M. S., Brookhuis, K. A., Wickens, C. D. and Hancock, P. A. (2015). State of science:
Mental workload in ergonomics. Ergonomics, 58(1), 1–17.
Index
2016 Australian Subaru National Road Series flight data strip bay, 23
(NRS), 158, 159 operational level activities, 23
2D imagery, 48 radar display, 23
3D plus time, 48 standard routes, 23
4D space, 48 Air traffic management, 32
Airline cockpit, 188
Accident analysis methods, 3 Airspace, 19
AcciMap, 3, 231,281–282 Altitude, 23
Activity sampling, 111 Analysis of networks, 234–235
Activity Stereotypes, 52 ANT, see Actor Network Theory (ANT)
Actor Network Theory (ANT), 102 Applied Graphic and Network Analyses (AGNA)
ActorMap, 4, 214 tool, 88,172, 180, 189
Advanced Cockpit for the Reduction of Stress APs, see Authorised Persons (APs)
and Workload (ACROSS), 254 Army Land Warfare, 39–40
Agent centrality, 268 Artefacts, 3, 20
AGNA, see Applied Graphic and Network As Low As Reasonably Practicable
Analyses (AGNA) tool (ALARP), 233
AI or Intelligent Knowledge-Based Systems, 256 Association matrices, 176
Air Traffic Control (ATC), 4, 19–36 170 ATC, see Air Traffic Control (ATC)
application, 28–35 ATC9K portable camera, 139
analysis of agents, 28–30 Atkins Mobility (2016), 187
control architecture, 31–33 Authorised Persons (APs), 68, 70–71
distributed cognition system, 28–30 Automated driving systems, 169–170, 180
facilitating technology, 31 application of EAST, 172
systemic situational awareness, 33–35 impact of, 170
temporal aspects, 35 information networks, 172, 179–180
beyond ethnography, 21–22 levels of, 169–170
problems, 22 method, 171–173
command, 19 results, 173–180
coordinating, 19 role of driver, 169–171
directing, 19 ‘safe stop’, 178
planning, 19 ‘self-driving’ functionality, 169–170
control, 19 social networks, 172
distributed cognition methodology, 24–25 task networks, 172
descriptive vs. formative methods, 24–25 Autonomous emergency brake system, 178
method integration, 25 Aviation, 4, 172
distributed cognition perspective, 19
distributed situation awareness, 20–21 BAE, see Battlefield Area Evaluation (BAE)
representation, 19–20 Barking, 68
scenarios, 25–26 Basic tasks, 173
applying the method, 26 BASt Expert Group, 170
propositional networks, 28 Battle on the Border road race stage 1, 162
representing distributed cognition, 26 Battlefield Area Evaluation (BAE), 40
social networks, 27 Bavelas-Leavitt’ s index, 88
task networks, 26–27 Betweenness, 88, 268
three counts, 21 ‘Big Five’ model, 163
work setting, 23–24 ‘Broken-links’ approach, 177, 201, 229,
aircraft call signs, 24 224, 239–243 combined information
charts, 23 network, 239
communications, 24 EAST analysts investigations, 239
flight data strip, 23 Hawk missile simulation study, 239

291
292 Index

Kegworth accident, 239 Coordination Demand Analysis (CDA), 25, 42,

represents failures, 239 45–48, 75, 109, 114–115
COS, see Commander and Chief of Staff (COS)
C4i, 109–110 COSS, controller of site safety (COSS)
‘Car hacking’, 200 CP, see Competent Person (CP)
CC, see Central Command (CC) CREAM, see Cognitive Reliability and Error
CDA, see Coordination Demand Analysis (CDA) Analysis Method (CREAM)
CDM, see Critical Decision Method (CDM) Criterium race, 159
CDM Probes, 113 Critical Decision Method (CDM), 25, 6, 42,
CDM pro-forma, 70 56–57, 66, 160, 179
Central Command (CC), 66 CUD, see Communications Usage Diagram (CUD)
Central control room, 69 CWA, see Cognitive Work Analysis (CWA)
Central nodes, 9, 33 Cybersecurity, 200
Central Operations Control Room (COCR), Cycling pelotons, 157
68–70
Centrality, 27–28, 33, 51, 54, 72, 88, 214–216 DCOG, see Distributed Cognition (DCOG)
Chief Petty Officer for Tactical Systems Data collection, 6, 40–41
(CHOPS), 87 concurrent verbal protocols, 6
CHOPS, see Chief Petty Officer for Tactical documentation review, 6
Systems (CHOPS) observations, 6
Civilian Command and Control, 67–69 semistructured, 6
Closeness, 88, 268 structured interviews, 6
COCR, see Central Operations Control Room walkthrough analysis, 6
(COCR) DD information network (DD), 179
Cognitive Adaptive Man-Machine Interface DD, see Driver Driving (DD)
(CAMMI) project, 254 De-crewing, see Single-crew aircraft
Cognitive Reliability and Error Analysis Method Dedicated Short Range Communication (DSRC)
(CREAM), 229 sensors, 187
Cognitive Work Analysis (CWA), 3, 24, 207, 258, Defence Aviation Hazard Risk Matrix, 232
260, 277 Defence Technology Centre for Human Factors
Cognitive Work Analysis Design Toolkit, 3 Integration (DTC HFI), 110
Cohesion, 88, 172–173 Denser networks, 189
Combat estimate, 39–40, 48, 52, 55 Density, 9, 27, 51, 54, 88, 172, 189
Command teams, 4, 188 Department for Transport (2015), 169, 181
Commander and Chief of Staff (COS), 51 Descriptive methods, 24–25
Commissaire, 158, 160 Detonators, 122
Communications media, 48 Diameter, 88, 172, 189–190
advantages, 48 Dictaphone, 70, 138, 160
critique of communications usage, 48 Direction, 23
disadvantages, 48 Director Sportif (DS), 158
network enabled solutions, 48 Distances, 23
verbal communications, 48 Distributed Cognition (DCOG), 23, 44, 171, 188
Communications Modality, 31 Distributed Situation Awareness (DSA), 157
Communications Usage Diagram (CUD), 25, 42, DM information network, 175, 179
46–48, 52, 57, 109, 115 DM, see Driver Monitor (DM)
Compatible SA, 20 DND, see Driver Not Driving (DND)
Competent Person (CP), 69, 71 Domestiques, 158
Complex sociotechnical systems, 3, 4 Driver Driving (DD), 171, 191
Composite network analysis, 4, 8, 219–222 Driver Monitor (DM), 171, 193
Computation, 19–20 Driver Not Driving (DND) role, 171
Connected and Autonomous Vehicles (CAV), 187 Driving research, 172
Control architecture, 31–33, 219, 221 DSA, see Distributed Situation Awareness (DSA)
Control Task Analysis (ConTA), 263–264 DSRC, see Dedicated Short Range
Control, 4 Communication (DSRC) sensors
Control-feedback loop, 170 DTC HFI, see Defence Technology Centre for
Controller of site safety (COSS), 121 Human Factors Integration (DTC HFI)
Convoy of vehicles, 158 ‘Duty Holders’ (DH), 232
Index 293

EA, see External Agencies (EA) domain of application, 5

Earth switch, 68 driving, 189
EAST, see Event Analysis of Systemic Teamwork DSA, 157
(EAST) framework elite women’s cycling, 157–158
Eccentricity, 88, 268 energy distribution operations, 4, 65–81
Electronic Standby Pilot (ESP), 254 ergonomics methods, 159
Elite cycling, 4, 157 evaluation, 4
‘Emergence’, 4, 20, 24 example, 12
Emission and reception degree, 88, 268 flowchart, 10
Energy distribution operations, 4, 65–81 forms of, 159
case study, 67–69 highlights of, 163–165
introduction, 65–69 impact of CAV networks, 194
methodology, 69–72 implications of, 165–166
design, 69 information networks, 189, 237–238
materials, 70–71 land use planning, 5
participants, 69 macro-level, 189
procedure, 71–72 maritime transport, 189
result, 72 meso-level activities, 189
additional east analyses, 74–77 methods, 3–15, 159–161
information networks, 73–74 materials, 160
social networks, 72 participants, 159–160
task networks, 72 procedure, 160
EngGeo, 52 military command and control, 39–61
Engine Vibration Indicator, 239 network metrics, 189
Engineering supervisor (ES), 122 network representations, 172
‘Environmental contributions’, 19, 44 non-CAV networks, 195
Ergonomics methods, 41, 157 non-CAV vs. CAV, 172, 196
accident analysis, 281–283 procedure, 5–6
accident prevention, 283–284 analyse networks, 8–9
in design, 288–289 analysis aims, 5
examining, 284–286 construct composite networks, 8
future directions, 277–278 construct information networks, 8
mapping migration, 287 constructing a task network, 6–7
systems concepts, 287–288 data collection, 6
systems thinking, 279–281 social network analysis (SNA), 8
Error, 25 system under analysis, 6
ES, engineering supervisor (ES) task, 6
Ethnographic techniques, 26 transcribe data, 6
Event Analysis of Systemic Teamwork (EAST) railway maintenance, 109–131, 189
method, 3–15, 22, 24, 26, 88, 157, 159, reliability, 10
171, 230, 277 results, 161–163
‘broken links’ analysis, 177–178 situation awareness network, 162
‘network of networks’ approach, 159 social network, 162
advantages, 9–10 task network, 161
advice, 5–6 road intersections, 135–154
air traffic control, 4, 19–36 SNA metrics, 244
analysis of STS risks, 246 social network, 189, 235–236
applications, 4–5, 10 sports performance, 157
approximate training, 10 structure, 45
automated driving systems, 171–173 submarine control room, 83–106
aviation,189 system behaviour, 159
background, 4–5 systems analysis frameworks, 3
broken-link analysis, 245 systems design methods, 3
complex sociotechnical systems, 3 task networks, 188, 236–237
decision-making agents, 244 teamwork in an elite women’s cycling
design, 4 team, 159
disadvantages, 10 tools, 10–12
294 Index

urban design, 5 nodes relationship, 7

validity, 10 non-CAV and CAV, 197–199
Existing systems, see non-CAV risk assessment, 180
External Agencies (EA), 188, 191, 197 safe solution for missile simulation, 238
schematic overview of DM network, 180–181
Face-to-face communication, 31 signage, 180
Facilitation matrix, 53 SME, 238
Farness, 88, 268 STS, 238
Federal Aviation Administration, 171 systematic overview of DD network, 179, 181
Feedback, 180 traffic, 179
Flight deck manager, see Pilot In-Person Voice, 53
Formative methods, 24–25 Internal structure of EAST, 67
FRAM, see Functional Resonance Analysis Inter-team transactions, 165
Method (FRAM) Interviewers, 160
Functional Resonance Analysis Method Intra-team transactions, 165
(FRAM), 3, 207, 277
Future systems, see CAV Kegworth air disaster (1989), 230
Key agents, 172, 190
G2 Intelligence, 51 Key information objects, 58, 172
Geographical dispersion, 28, 32 Key tasks, 172
Global Positioning Systems (GPS), 139, 187 Knowledge network, 4, 34
Go ProTM cameras, 160 Knowledge Objects (KO), 34
GPS, see Global Positioning Systems (GPS) KO, see Knowledge Objects (KO)
Graph Theory, 27, 72
Group Supervisor level, 32 Ladbroke Grove rail incident (1999), 230
Land Use Planning (LUP), 5, 13
Hawk jet STS, 231 Land warfare, 4, 39–40
Hawk Missile Simulation case study, 231–235 Longitudinal and Lateral Controllers, 175
Hawk RtL case, 235 LUP, see Land Use Planning (LUP)
Hawk-Frigate STS, 234
HAZOP and Fault Tree analysis, 244 MAA’ s Regulatory Articles, 233
HAZOPS (Hazard and Operability), 233 MacroErgonomic Analysis and Design method
HFACS, see Human Factors Analysis and (MEAD), 3, 39, 277
Classification System (HFACS) Management of Operating Risk, 233
HFI, see Human Factors Integration (HFI) MEAD, see MacroErgonomic Analysis and
Hierarchical Task Analysis (HTA), 6, 7, 25, 40, Design method
42, 66–67, 112–114, 207, 213–214, 283 Media and team support vehicles, 158
Hierarchical Task Analysis of Driving Mediation of control, 24
(HTAoD), 173 Medical support, 158
HTA, see Hierarchical Task Analysis (HTA) Microsoft Visio TM, 10, 70, 160
Hubs, 72 Microsoft Word TM, 160
Human Factors Analysis and Classification Military Aviation Authority (MAA), 232
System (HFACS), 67–69, 229, Military command and control, 39–61
281–282 description, 39–41
Human Factors Integration (HFI), 58 army land warfare, 39–40
Human performance variability, 229 combat estimate, 39–40
data collection, 40–41
Information audit, 86 east method, 41–44
Information network, 4, 172, 179–180, 214, descriptive vs. formative methods, 41–42
216–219, 230, 237–238 importance, 41
AGNA, 197 key factors and situational awareness,
analysis task, 13 42–44
Hawk RtL case, 238 method integration, 42
information node, 179–180 theoretical bas is, 44
infrastructure, 179 findings, 44–58
knowledge, 4 Communications Usage Diagram (CUD),
metrics analyses, 8 46–48
Index 295

Coordination Demand Analysis (CDA), OOW, see Officer of the Watch (OOW)
44–46 Operation Sequence Diagrams (OSD), 25, 54–56,
EAST structure, 45 66, 116
Operation Sequence Diagrams (OSD), Operational tasks, 173
54–56 Operator and Overhead Line Party (OLP), 68, 72
Propositional Networks (PN), 56–58 OSD, see Operation Sequence Diagrams (OSD)
Social Network Analysis (SNA), 48–54
scenarios, 39–41 Participant Demographics, 138
Modality Matrix, 53 ‘Passive’ agent, 236
Modelling future flight decks Peloton, 158, 160, 162
airline personnel costs, 253 attack, 158, 161
automated deck systems, 254 bunch sprint, 158
design approach, 256 Periscope Watch Keeper (PWK), 97
aircraft component, 256 Periscope, 87
ground-based component, 257–258 Person in charge of possession (PICOP), 119
system ‘mirror’, 258 PICOP, see person in charge of possession
navigation/flight planning support, 258 (PICOP)
pilot workload, 254–255 Pilot, 253–271
poor CRM, 254 Pilot Flying (PF), 170
real-time engineering support, 258 Pilot Monitor (PM), 171
rules of the air, 255 Pilot node, 235
safe passenger evacuation, 253 Pilot Not Flying (PNF), 170
‘second pilot’ support station/office, 257 Planning Aids, 53
single-crew aircraft, 255 PM, see Pilot Monitor (PM)
system configurations, 258–271 Position, cycling 162
tasks of pilot, 253 Possession, 119
working costs, 254 Post-drive tasks, 173
Monitor other riders, 161 Power stations, 67
Motorcyclists, 142 Practitioners, 4
MS Herald of Free Enterprise accident (1987), 230 Pre-drive tasks, 173
Multi-disciplinary approaches, 5 ‘Principal’ planning staff, 51
Multi-faceted, 4 Propositional networks, 25, 28, 42, 116–117
Protection, 119
National Highway Traffic Safety Administration PWK, see Periscope Watch Keeper (PWK)
(NHTSA), 169
Naturalistic study, 160 Queen of the Mountain (QOM), 158
Naval warfare, 4
Navy training, 243 Race officials, 158
NEC, see Network Enabled Capability (NEC) Radar Altimeter (Rad-Alt), 232
Negotiate intersection, 141 Radar contact, 23
Negotiation architecture, 32 Radio, 23, 53
Network centric, 33 RAF Nimrod XV230, 232
Network density, 214 Railway maintenance, 109–131
Network Enabled Capability (NEC), 46, 54 development of EAST methodology, 110–111
‘Network of networks’ approach, 4, 66 Communications Usage Diagram
Network metrics, 8–9,189 (CUD), 115
Neutral spares, 158 Coordination Demand Analysis
NHTSA, see National Highway Traffic Safety (CDA), 114–115
Administration Hierarchical Task Analysis (HTA), 112–114
Nimrod accident, 234 Layer 1 – Data Collection Methods,
Nodes, 7, 9 111–112
Non-human artefacts, 23, 158–159 Layer 2 – Analysis Methods, 112
Non-verbal transactions, 165 Layer 3 – Representational Methods, 115
Operation Sequence Diagram (OSD), 116
Officer of the Watch (OOW), 87, 97–99 Propositional Network (PN), 116–117
OLP, see Operator and Overhead Line Party Social Network Analysis (SNA), 115
(OLP) summary of methods, 117–118
296 Index

discussion, 122–127 results, 141–151

representational methods, 127–128 cyclist intersection, 142
Propositional Networks (PN), 128 driver intersection, 141
introduction, 109–110 motorcyclist intersection, 142
live data, 119–122 pedestrian intersection, 143
background, 119 situation awareness networks, 147–151
emergency engineering work, 122 social networks, 144–147
ending a track possession, 122 task networks, 141–144
planned maintenance activities, 121–122 Road races, 159
procedure, 119 Road transportation systems, 4, 187–188
results, 122–127 automated vehicles, 187
Rail transport systems, 4, 207–208 camera and radar technology, 188, 191
accident analysis, 208–209 CAV functionality, 188
integrating STAMP and EAST, 209–213 driver–automation interaction, 187
railway level crossings, 208–209 DSRC, 187
RLX collisions, 208 future transportation systems, 187
RLX safety, 209 human factors,187
Railway level crossings (RLX), 208, 214, 224 method, 188–190
Rasmussen’s risk management framework, 286 non-CAV vs. CAV, 192
Reception, 88 regulation of automation, 187
Regulatory Article (RA) 1210, 233 socialtechnical systems, 188
Representational states, 19–21 system agents, 191–194
Researchers, 4, 160 Royal Navy, 231–232, 234
Return to periscope depth (RTPD), 88 RtL/HRM (Hazard-Risk Matrix) approach,
Rider, cycling, 162 233, 244
Risk to Life (RtL), 232 RTPD, see return to periscope depth (RTPD)
RLX, see Railway level crossings (RLX)
RLX Safety management, 213 SA, see situational awareness (SA)
methodology, Safety-related nodes, 163
data inputs, 213 Schematic model, 23
EAST network development, 213–214 Science and Technology Select Committee
network analysis, 214–216 (2017), 187
results, 216–222 SCO, see Ships Control Officer (SCO)
composite network, 219–222 ‘Self-driving’ functionality, 170
control network, 219 Senior Authorised Persons (SAPs), 68–72
information network, 216–219 Senior Watch Assistant level, 32
social network, 216 SGT, see Super Grid Transformers (SGT)
task network, 216 SHERPA, see Systematic Human Error
RLX system lifecycle, 214 Reduction and Prediction Approach
information network, 214 (SHERPA)
performance monitoring, 221 Ships Control Officer (SCO), 97
risk management, 221 Single-crew aircraft, 253, 258
social network, 214 Situation awareness networks, 20–23, 25, 42, 157,
Road Environment (RE), 191, 197 159, 161–162, 164–165, 170
Road intersections, 135–154 SME, see Subject Matter Expert (SME)
EAST intersection case stud y, 136–137 SNA, see Social Network Analysis (SNA)
introduction, 135–136 Social agents, 235
methodology, 138–141 Social Network Analysis (SNA), 4, 8, 25, 27–30,
design, 138 42, 48–54, 66, 87, 115, 160, 172, 214,
materials, 138 216, 230, 235–236, 259, 265–270
network density, 140 activity stereotypes, 52
participants, 138 agents and artefacts, 265
procedure, 139–141 AGNA, 195, 234
situation awareness networks, 140 analysis process, 12, 234
social networks, 140 association matric, 266–268
sociometric status, 140 basic description, 175–176
task networks, 139–140 calculation, 53–54
Index 297

combined analysis process, 14 task network analysis, 89–91

command and control, 51 Super Grid Transformers (SGT), 68
communication, 4, 49–51 Support protected rider, 161
advantages, 49–50 Systems analysis frameworks, 3
disadvantages, 49–50 System agents, 190
defining network, 266 System behaviour, 159
diagram, 13 Systems design methods, 3
elite women’s cycling team, 161–162 System mode, 180
facilitation of network links, 52–53 System of representations, 24
‘free flight’, 266 Systematic Human Error Reduction and
functional loading, 269–270 Prediction Approach (SHERPA), 229
future concepts, 270 Systematic situation awareness, 179
metrics analyses, 8 Systemic Accident Analysis (SAA) approach, 229
network analysis, 195 Systemic teamwork, 159
network’s complexity, 266 Systems Theoretic Accident Model and Process
nodes relationship, 7, 235 (STAMP), 3, 81–282, 207
non-CAV and CAV, 194–196 accident analysis, 208–209
non-level metrics, 234 control theory and, 209–212
numerical analysis, 268 generic control structure model, 210–212
organization, 4 integrating EAST and, 212–213
road transportation systems, 189, 194–196 limitations, 211–212
role of DD, 175–176 Systems thinking approach, 3
role of DM, 175–176 System-Theoretic Process Analysis (STPA), 230
system agents, 194
Social Organisational Cooperation Analysis – Tactical tasks, 173
Contextual Activity Template Talisman Command Team Trainer, 87
(SOCA‑CAT), 264 Task networks, 4, 25–27, 159–161, 172–175, 190,
Sociometric status, 72, 88, 173, 176, 179–180, 214, 216, 230, 236–237
190, 198, 214, 235, 243, see also AGNA analysis, 192
‘Broken links’ network combined analysis process, 14
Sociotechnical systems (STS), 188, 229 DD role, 175
assessing risk, 229 driving, 173–175
aviation accidents, 231–232, 243 elite women’s cycling team, 159–161
‘breaking’ links, 230 goals, 4
distributed cognition (DCOG), 188 issuing of RtL document, 236, 243
macro-level, 188 linkages between sub-tasks, 173
modifications to EAST, 230, 243 metrics analyses, 8
Speed, 23 military command, 40–41
Split architecture, 32 nodes relationship, 7
Sport, 4 non-CAV and CAV, 193, 196
STAMP, see Systems Theoretic Accident Model perspective of DM, 175
and Processes railway transport systems, 216
Standard Operating Procedures (SOPs), 255 revision of safe flying altitude, 236, 243
STPA, see System-Theoretic Process Analysis road transportation system, 190
(STPA) tasks spanning, 192
Strategic tasks, 173 terminologies, 173
Subject matter expert (SME), 194, 233–235 Team debriefs, 160
Submarine Control Room, 83–106 Team race planning meetings, 160
analysing distributed cognition, 83–85 Technique for Human Error Rate Prediction
analysis, 86–89 (THERP), 229
combining network models, 95–100 Technique for the Retrospective and Predictive
information and social networks, 99–100 Analysis of Cognitive Errors
task and social networks, 96–99 (TRACEr), 229
data collection, 86–89 Technology Matrix, 31, 53
information network analysis, 93–95 The Society for Automotive Engineers (SAE), 170
social network analysis, 92–93 THERP, see Technique for Human Error Rate
system properties of east, 85–86 Prediction (THERP)
298 Index

Time trial, 25, 159 V2X, see Vehicle-to-Anything (V2X)

TMC, see Traffic Management Centres (TMC) communication
Tottenham, 68 Vehicle-to-Anything (V2X)
TRACEr, see Technique for the Retrospective communication, 188
and Predictive Analysis of Cognitive Vehicle-to-Infrastructure (V2I)
Errors communication, 187
Traffic Management Centres (TMC), 188, 191 Vehicle- to-Vehicle (V2V) communication, 187
Traffic management centres (TMC), 197 Verbal transactions, 165
Transactions, 100
Watch Manager level, 32
UD, see Urban Design Weather conditions, 23
Uninhabited Air Vehicles (UAVs), 272 WESTT software packages, 70
University of the Sunshine Coast’s Human Ethics Work Domain Analysis (WDA), 260
Committee, 160 functional purposes, 260–261
Unmanned Air Vehicles (UAVs), 255 object-related processes, 262
Unscheduled traffic, 23 physical objects, 261–262
Urban Design (UD), 5, 13 purpose-related functions, 262
Urban Traffic Management Control (UTMC), 194 values and priority measures, 261
UTMC, see Urban Traffic Management Control Workload, 25
(UTMC)
Yoga BM-26D boundary microphone, 87
V2I, see Vehicle-to-Infrastructure (V2I)
V2V, see Vehicle- to-Vehicle (V2V) Zonal hazard (safety) analyses, 233

Systems Concepts in Action
100% (4)
Systems Concepts in Action
336 pages
Paul M. Salmon, Neville A. Stanton, Guy H. Walker, Adam Hulme, Natassia Goode, Jason Thompson, Gemma J.M. Read - Handbook of Systems Thinking Methods-CRC Press (2022)
100% (1)
Paul M. Salmon, Neville A. Stanton, Guy H. Walker, Adam Hulme, Natassia Goode, Jason Thompson, Gemma J.M. Read - Handbook of Systems Thinking Methods-CRC Press (2022)
331 pages
Fiber Glass Protection
100% (1)
Fiber Glass Protection
679 pages
Practice Questions - Sign Convention - Spherical Mirrors DONEE
0% (1)
Practice Questions - Sign Convention - Spherical Mirrors DONEE
2 pages
Emma Mason - The Cambridge Introduction To William Wordsworth-Cambridge University Press (2010)
100% (6)
Emma Mason - The Cambridge Introduction To William Wordsworth-Cambridge University Press (2010)
151 pages
Systems+Thinking As of March 2024
No ratings yet
Systems+Thinking As of March 2024
154 pages
System Thinking
50% (2)
System Thinking
18 pages
Sanet - St-Introduction To Safety Science People Organisations and Systems
No ratings yet
Sanet - St-Introduction To Safety Science People Organisations and Systems
305 pages
Ysj Dissertation
100% (2)
Ysj Dissertation
4 pages
Series D1MW Characteristics: Technical Features
No ratings yet
Series D1MW Characteristics: Technical Features
6 pages
Libro para Practicarpdf
No ratings yet
Libro para Practicarpdf
138 pages
Minutes of Meeting Attendance: Present
No ratings yet
Minutes of Meeting Attendance: Present
3 pages
Systems Thinking
No ratings yet
Systems Thinking
5 pages
CH 4 Force System Resultant
No ratings yet
CH 4 Force System Resultant
50 pages
Competency Mapping: Asst Professor, Amity University Noida Asst Professor, Amity University Noida
No ratings yet
Competency Mapping: Asst Professor, Amity University Noida Asst Professor, Amity University Noida
3 pages
Photovoltaic Systems - Artificial Intelligence-Based Fault - K - Mohana Sundaram, Sanjeevikumar Padmanaban, Jens Bo - CRC Press (Unlimited), (S - L - ), - 9781000545852
No ratings yet
Photovoltaic Systems - Artificial Intelligence-Based Fault - K - Mohana Sundaram, Sanjeevikumar Padmanaban, Jens Bo - CRC Press (Unlimited), (S - L - ), - 9781000545852
151 pages
Libro. System Dynamics Review - 2001 - Bathtub Dynamics Initial Results of Asystems Thinking Inventory
No ratings yet
Libro. System Dynamics Review - 2001 - Bathtub Dynamics Initial Results of Asystems Thinking Inventory
38 pages
Shared Module 4 5-Inventory, QM, SCM-Logistics
No ratings yet
Shared Module 4 5-Inventory, QM, SCM-Logistics
107 pages
Nic J. T. A. Kramer, Jacob de Smit (Auth.) - Systems Thinking - Concepts and Notions-Springer US (1977)
No ratings yet
Nic J. T. A. Kramer, Jacob de Smit (Auth.) - Systems Thinking - Concepts and Notions-Springer US (1977)
146 pages
LS2 - CSC 107 SlidesNotes 2
No ratings yet
LS2 - CSC 107 SlidesNotes 2
39 pages
Regional Plan - 2021 For National Capital Region: Addressing The Planned Growth of Delhi by Adopting Regional Approach.
No ratings yet
Regional Plan - 2021 For National Capital Region: Addressing The Planned Growth of Delhi by Adopting Regional Approach.
14 pages
Y.garud Multiaxial Fatigue
No ratings yet
Y.garud Multiaxial Fatigue
27 pages
Apple Supplier List 2013
No ratings yet
Apple Supplier List 2013
33 pages
Characteristics of Letters
No ratings yet
Characteristics of Letters
22 pages
Unit 1 Notes
No ratings yet
Unit 1 Notes
23 pages
Aurora Geo Report
No ratings yet
Aurora Geo Report
86 pages
Richmond Systems Thinking
No ratings yet
Richmond Systems Thinking
25 pages
Module SAD
No ratings yet
Module SAD
20 pages
Technical Paper (Seismic Fragility Curves) - Dec2018
No ratings yet
Technical Paper (Seismic Fragility Curves) - Dec2018
32 pages
RArnold Systems Thinking Definition
No ratings yet
RArnold Systems Thinking Definition
11 pages
HSBC Digital Starter Kit Masterbrand HBPH
No ratings yet
HSBC Digital Starter Kit Masterbrand HBPH
27 pages
IGCSE Chemistry AO3 G10-2 Sungbeen Hong
No ratings yet
IGCSE Chemistry AO3 G10-2 Sungbeen Hong
14 pages
Similar PDF
No ratings yet
Similar PDF
12 pages
Author Biography:: System Dynamics/Systems Thinking: Let's Just Get On With It
No ratings yet
Author Biography:: System Dynamics/Systems Thinking: Let's Just Get On With It
25 pages
Section 7 Gravitational Fields
No ratings yet
Section 7 Gravitational Fields
39 pages
Systemic Thinking
No ratings yet
Systemic Thinking
11 pages
Likap Notebook Manual
No ratings yet
Likap Notebook Manual
8 pages
Arnold 2015 Systems Thinking Definition
No ratings yet
Arnold 2015 Systems Thinking Definition
10 pages
The Art and Practice of Systems Thinking (Harvard 2019)
No ratings yet
The Art and Practice of Systems Thinking (Harvard 2019)
4 pages
Advanced Degree
No ratings yet
Advanced Degree
2 pages
CTU Memo No. PF13 349A S. 2023
No ratings yet
CTU Memo No. PF13 349A S. 2023
6 pages
Sdasdasdas
No ratings yet
Sdasdasdas
5 pages
WiNG 5.0 Cheat Sheet - RF Domains
No ratings yet
WiNG 5.0 Cheat Sheet - RF Domains
6 pages
2425 - Pgdlma - Elscon - Mock - Assessment - Tagged
No ratings yet
2425 - Pgdlma - Elscon - Mock - Assessment - Tagged
4 pages
Modeling & System Handout
No ratings yet
Modeling & System Handout
4 pages
System Thinking
No ratings yet
System Thinking
4 pages
Franck Hertz
No ratings yet
Franck Hertz
6 pages
Tetra 350-450MHz Mini Repeater Brochure, V1.04
No ratings yet
Tetra 350-450MHz Mini Repeater Brochure, V1.04
2 pages
Itinerary of Travel
No ratings yet
Itinerary of Travel
4 pages
Ôn Thi Gi A K 2 Anh 10
No ratings yet
Ôn Thi Gi A K 2 Anh 10
4 pages
HDFC 5000 Book4 07to31mar25
No ratings yet
HDFC 5000 Book4 07to31mar25
3 pages
Estimation of A Population Mean
No ratings yet
Estimation of A Population Mean
1 page
The Straits Times Life! - Interviews With Mediacorp Artistes, Pierre PNG, Cynthia Koh and Rebecca Lim
No ratings yet
The Straits Times Life! - Interviews With Mediacorp Artistes, Pierre PNG, Cynthia Koh and Rebecca Lim
1 page
Design and Technology in Today's World: A First Look
From Everand
Design and Technology in Today's World: A First Look
Baz Professor
No ratings yet
Intrusion Detection Honeypots
From Everand
Intrusion Detection Honeypots
Chris Sanders
3/5 (2)
Software Patterns Made Easy
From Everand
Software Patterns Made Easy
Justice Nanhou
No ratings yet

Systems Thinking in Practice - Applications of The Event Analysis of Systemic Teamwork Method (PDFDrive)

Uploaded by

Systems Thinking in Practice - Applications of The Event Analysis of Systemic Teamwork Method (PDFDrive)

Uploaded by

Systems Thinking

Systems Thinking in Practice

For more information about this series, please visit: https://www.crcpress.com/­

© 2019 by Taylor & Francis Group, LLC

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-138-09787-2 (Hardback)

Library of Congress Cataloging‑ in‑ Publication Data

Names: Stanton, Neville A. (Neville Anthony), 1960- author. | Salmon, Paul

Visit the Taylor & Francis Web site at

and the CRC Press Web site at

Section I Overview of EAST

Chapter 1 The Event Analysis of Systemic Teamwork (EAST) Method............... 3

Section II Applications of EAST

Chapter 2 EAST in Air Traffic Control............................................................... 19

Chapter 3 EAST in Military Command and Control.......................................... 39

Coordination Demand Analysis (CDA)..........................................44

Chapter 4 EAST in Energy Distribution Operations........................................... 65

Chapter 5 EAST in a Submarine Control Room.................................................. 83

Information and Social Networks..................................................99

Chapter 6 EAST in Railway Maintenance......................................................... 109

Chapter 7 EAST at Road Intersections.............................................................. 135

Chapter 8 EAST in Elite Women’s Cycling Teams........................................... 157

Chapter 9 EAST in Automated Driving Systems.............................................. 169

Chapter 10 EAST in Future Road Transportation Systems................................. 187

Section III Future Developments in EAST

Chapter 11 STAMPING on EAST: Adding a Control Network to EAST

Chapter 12 The EAST ‘ Broken-Links’ Approach: Assessing Risk in

Chapter 13 From CWA to SNA: Modelling Future Flight Decks....................... 251

Physical Objects........................................................................... 259

Chapter 14 Future Directions for the Event Analysis of Systemic Teamwork........275

RECOMMENDED FURTHER READING ON EVENT

THERE ARE ALSO THESE BOOKS:

author/co-author of over 100 peer-reviewed journal articles and 13 books. He and

Chris Baber Huw Gibson

Tony Carden Don Harris

Ashleigh J. Filtness Michael G. Lenné

Richard McMaster Alison Starr

Anjum Naweed Nicholas J. Stevens

Paul M. Salmon Guy H. Walker

• Why: the goals of the system, sub-system[s] and actor[s]

To assist researchers and practitioners explore these constructs, EAST (Stanton

BACKGROUND AND APPLICATIONS

• aviation (Chapters 2, 12 and 13)

Underpinning the approach is the notion that distributed teamwork can be

FIGURE 1.1 Network of networks approach.

APPLICATION IN LAND USE PLANNING

PROCEDURE AND ADVICE

Step 2: Define the Task/System Under Analysis

Step 3: Data Collection

Step 4: Transcribe Data

Step 5: Construct Task Network

Network Nodes Relationships Examples

Step 6: Conduct Social Network Analysis

Step 7: Construct Information Networks

Step 8: Construct Composite Networks

• Task by agents network (combined task and social network);

Step 9: Analyse Networks

1. Network Density (overall network): Network density represents the level

APPROXIMATE TRAINING AND APPLICATION TIMES

RELIABILITY AND VALIDITY

FIGURE 1.3 Task network of the generic site analysis.

FIGURE 1.5 Information network for the site analysis task.

Distributed Situation Awareness

1. The outputs of ethnographic approaches (as they are used currently)

THE AIR TRAFFIC CONTROL WORK SETTING

• Keeping in radio and/or radar contact with aircraft.

Charts and Standard Routes

Flight Data Strip

Flight Data Strip Bay

Aircraft Call Signs

DISTRIBUTED COGNITION METHODOLOGY:

Descriptive vs. Formative Methods

(EAST comprises seven individual methodologies, CWA comprises five ‘phases’)

AIR TRAFFIC CONTROL SCENARIOS

Applying the Method

For more information about this series, please visit: https://www.crcpress.com/