Embed Size (px)
Citation preview
Title Mobility Behavior Change Support System for SustainableCampus Commuting( Dissertation_全文 )
Author(s) Sunio, Varsolo Cornago
Citation Kyoto University (京都大学)
Issue Date 2018-03-26
URL https://doi.org/10.14989/doctor.k21086
Right 許諾条件により本文は2018-04-01に公開
Type Thesis or Dissertation
Textversion ETD
Kyoto University
Mobility Behavior Change Support System
for Sustainable Campus Commuting
Sunio Varsolo Cornago
2018
i
Mobility Behavior Change Support System
for Sustainable Campus Commuting
持続可能な通学のための
交通行動変容支援システム
Sunio Varsolo Cornago
A dissertation submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Engineering
in the
Department of Urban Management
Graduate School of Engineering
Kyoto University
2018
ii
iii
Acknowledgment
First, I am immensely grateful to EACOMM Corporation, particularly to
Mark Ernest Matute and Mikhail Joseph Torres, the Managing Director. No
words can express how indebted I am to them for their help.
I also acknowledge the support of Ateneo Java Wireless Competency Center
in the development and deployment of the mobile application. My sincerest
appreciation goes especially to Dr. Regina Estuar, Bon Lemuel dela Cruz,
Dennis Villamor and Jasmine Javier. I appreciate the collaboration.
Many thanks to the Blaze logo designers, Adrian Lim and Jose Alvaro Adizon.
I thank the faculty of the Ateneo de Manila University who helped me in
implementing this study: Christopher Peabody, Jewel Unson, Raphael
Guerrero, Johanna Indias, Armando Guidote, Catherine Lee-Ramos, Kendra
Gotangco, Abigail Favis, Gino Trinidad, Arjan Aguirre, Genejane Adarlo, Ian
Navarrete, James Araneta, Joel Maquiling, Remmon Barbaza. I also
appreciate the help by many others I failed to name.
I acknowledge the various assistance extended to me by staff of Lauan Study
Center for my five months of stay in Manila over the research period: Rodolfo
Camaclang, Willy Ongsitco, Florencio Ballesteros, Manfred Salandanan,
Russel Ong, Mikhail Gallego, and Fr. Laurence Salud.
I thank the following in Kyoto University for their help in data analysis: Hsu-
Sheng Hsieh, Junghwa Kim, Fajar Belgiawan, and Yeun Touh Li.
My heartfelt gratitude goes to the faculty and staff of the Intelligent Transport
Systems Laboratory: Prof. Nobuhiro Uno, Assoc. Prof. Jan-Dirk Schmöcker,
iv
Asst. Prof. Toshiyuki Nakamura, Asst. Prof. Hiroki Yamazaki, Ms. Yuuko
Shii, and Ms. Mie Nishimura. Thank you for all the intellectual, material,
and social assistance. Thank you also to the past student members: 中村,
山村 and 吉本 (2015), 及, 清水, 仙田, 冨永, 中山, 藤井 and 増本
(2016), 森井, 渡邉, 出水 and Chris (2017). Special thanks to the current
members: Kim-san, 李, Saeed, 張, 孫, 川上, 園部, 西垣, 丹羽, 高, 後
藤, 中川, 西原, 松本, 金, 姚, 尾方, 田中, 内海, and 沈.
I also appreciate the various help given to me by friends in Kyoto: Sou, Jumpei,
Shun, Ko, Keiichiro, Shota, Makoto T., Miyabi, Shimono, Sanetaka, Ryosuke,
Deguchi, Hattori, Kenji, Masanobu, Shumpei, Satoshi, Wataru, Takuya,
Shibuya, Takeshi, etc.
My thanks also to Asst. Prof. Yuichiro Kawabata for the valuable help in the
translation from English to Japanese, and to my panelist, Prof. Satoshi Fujii.
Special thanks to all the staff of Yoshida Student Center, my home in Kyoto:
Yoshihiro Nakazato, Toshimi Nakai, David Sell, Makoto Nishida, Fr.
Katsushi Sasano, David Kolf, Fernando de Lecea, and Fr. Jesus Ramos. I also
dedicate this work to my intercessor, Fr. Joseph Muzquiz.
I also thank my parents, Victor and Remy, and my siblings, Verlen, Victor
and Monalisa, and their respective families.
And last but not the least, my sincerest thanks to my dear supervisor, Assoc.
Prof. Jan-Dirk Schmöcker. At many stages in the course of this study, I
benefited from his mentorship and intellectual inputs. I also appreciate his
careful editing of my dissertation as well as of all the manuscripts I submitted
to conferences and journals.
v
Abstract
Advanced technology is increasingly being leveraged to
promote pro-environmental travel behavior. Referred as “mobility
behavior change support systems” or mBCSS, they are information
systems designed to deliver interventions aimed at non-coercively
modifying mobility behavior towards sustainability. These mBCSSs
can be deployed as a travel demand management measure in the
universities, especially in the developing world, at a relatively lower
cost. Many of these mBCSSs, however, are developed based on
principles of persuasive technology, rather than on any behavior change
theory. Moreover, there is no controlled evaluation study of their
effectiveness; only usability assessment based on small samples. The
purpose of this dissertation is to design a theory-driven mBCSS,
evaluate its efficacy, and understand the process by which it is able to
achieve its effect. In the design, evaluation, and mediation analysis,
the stage model of self-regulated behavior change (SSBC) is used as a
theoretical framework.
In the design of the mBCSS, a menu-based approach based on
the stage model is systematically used to develop a “menu of
interventions”. In our mBCSS, called Blaze, the menu includes ten
components: map, goal setting, summary, readiness to blaze, travel
time, mode use, slideshow, ride plan, mode options, and diagnostic
report. The menu-based approach is an alternative to the more
prominent strategy, namely stage-tailoring, which is a design approach
that strictly matches the interventions to the stages. In the menu-based
approach, individuals select components from the menu that are most
relevant to them using self-assessments, hence allowing self-tailoring
to a certain extent. Results of evaluation by 71 users suggest that the
vi
relevance or usefulness of these ten components is stage-dependent.
In other words, individuals tend to self-tailor and choose only those
components in the menu that are useful for their goal pursuit.
In evaluating its efficacy, Blaze is tested for about four weeks
among 126 students of the Ateneo de Manila University, an educational
institution in Quezon City, Philippines. A control group of 115 non-
users is also included. The variables postulated by the stage model are
used in the assessment of effectiveness of the system. The field
experiment shows that the system is able to induce reduction in the car
use, progress in stages, increase in intentions, and positive change in a
number of determinants, relative to the control group. A car use
reduction of 20.38% is estimated due to the relative effect of the
intervention. The experimental group also progressed .43 stages (μ
=+.43). On the other hand, the control group shows regression of .15
stages (μ =-.15). It seems therefore that the intervention not only
induces stage progression, but also inhibits stage regression.
In an attempt to validate the SSBC, we construct a model using
path analysis. Our main findings suggest that, in agreement with SSBC,
travel behavior change is achieved by a transition through a temporal
sequence of four stages: predecision, pre-action, action and post-action.
Transition is achieved by formation of three intention types: goal,
behavioral and implementation intentions. In an extension to the stage
model, evidence is found to distinguish between “initiation of action”
and “maintenance of behavior” in the final stage of the behavior change
process. We observe that the former (initiation) is characterized by
instability (either relapse or progress), while the latter (maintenance) by
stability.
vii
Finally, SSBC is a useful theoretical framework in
understanding the mechanism of effect from intervention to behavior
change of the individuals. In mediation analysis, a causal pathway
linking the effect of the technology intervention to its behavioral
outcome through the mediation of a number of variables is identified.
The validated stage model of self-regulated behavioral change (SSBC)
is used as a theoretical framework to understand how the outcome may
be influenced by determinants (called conceptual theory), and how the
determinants may be activated by different intervention types (termed
action theory). Our findings show that the action and conceptual
components differ by stages: individuals belonging to later stages
change their behavior via the mediation of a change in implementation
intention, while those in the early stages undergo behavior change
through the mediation of a change in self-efficacy. Implications of the
results of this study are discussed.
Keywords: campus sustainability; mobility behavior change support
system; stage model; developing world cases; intervention development
viii
概要
本論文は、持続可能な通学を促進するための交通行動変容支援システ
ム（mobility behavior change support system; mBCSS）を提案するもので
ある。交通行動変容支援システムとは、環境に優しい交通行動を促す
ように設計された情報システムである。既存の支援システムの多くは、
行動変容理論ではなく、説得技術の原則に基づいて開発されている。
行動変容理論に基づくいくつかのシステムについても、その理論が介
入の系統的な開発にどのように用いられたのかが明確にはされてい
ないし、理論上の構成概念と行動変容介入との間の明示的な対応付け
は行われていない。さらに、そうしたシステムの有効性を適切な統制
の下で評価した研究は存在せず、小規模なサンプルに基づくユーザビ
リティ評価が行われているのみである。しかし、そのような評価がな
ければ、技術に基づく介入が人間の行動に変化を及ぼすメカニズムを
明らかにすることはできない。本論文の目的は、行動変容理論に基づ
く支援システムを設計し、その有効性を評価し、その行動変容のプロ
セスを理解することである。本システムの設計、評価、媒介分析
(mediation analysis) においては、理論的枠組みとして自己調整行動変
容ステージモデル（Stage model of self-regulated behavior change; SSBC）
というステージ（段階）モデルを用いており、これが本研究の特長で
もある。
交通行動変容支援システムの設計においては、ステージモデルに基づ
くメニューアプローチが、「介入のメニュー」を開発する上で体系的
に使用されている。メニューアプローチは、ステージに合わせた介入
方略に代わるものである。システムの有効性を評価するために、学生
被験者を対象とした約 4週間に渡るシステム検証実験を行った。効果
の適切な検証のため、統制群（control group）も用意した。このフィー
ルド心理実験により、この支援システムが、統制群と比較して、自動
ix
車の使用の減少、意図の増加、および行動変容ステージの進行を引き
起こしていることが示された。さらに介入の効果と行動変容結果とを
結びつける、複数の変数に媒介された因果経路を特定した。また、因
果経路が行動変容プロセス上のステージによって異なることも示さ
れた。後期ステージに到達している個人は実行意図の変容を媒介して
行動を変えるのに対し、初期ステージ段階に留まっている個人は自己
効力感の変化を媒介として行動変容を起こすことが明らかになった。
そして、SSBCの拡張において、行動変容プロセスの最終段階で、「行
動の開始」(initiation) と「行動の維持」(maintenance)を区別できること
が示される。
x
Preface
Parts of this dissertation have been published in journals, presented in
conferences or else submitted for review:
PUBLICATIONS
I. Sunio, V., Schmöcker, J. D., Estuar, M.R.E, Dela Cruz, B.L.,
& Torres, M.J. (2017). Development And Usability
Evaluation Of Blaze Information System For Promoting
Sustainable Travel Behaviour In Metro Manila. Journal of
Eastern Asia Society for Transportation Studies, 12.
(Accepted) (Chapter 5)
II. Sunio, V., Schmöcker, J. D., Estuar, M.R.E. (2017).
Implementation of a mobility behavior change support
system in Manila Philippines. In Behavioural Change
Support Intelligent Transport Applications. Workshop at
IEEE 20th International Conference on Intelligent
Transportation Systems. IEEE Intelligent Transportation
Systems Society. (Chapter 7)
III. Sunio, V., & Schmöcker, J. D. (2017). Can we promote
sustainable travel behavior through mobile apps? Evaluation
and review of evidence. International Journal of Sustainable
Transportation, 11(8), 553-566. (Chapter 3)
CONFERENCE PRESENTATIONS
I. Sunio, V., Schmöcker, J. D., Estuar, M.R.E, Torres, M.J. &
Robosa, K. (2017). Development Of Blaze: A Mobility
Behavior Change Support System for Promoting
Environmentally Friendly Travel Behavior in Manila. 14th
xi
World Conference on Transport Research. Shanghai,
China. (Chapter 5)
II. Sunio, V., Schmöcker, J.-D., Estuar, R.,Gotangco, C.K. &
Favis, A.M. (2017). Development and Evaluation of
BLAZE Information System. International Conference on
Environmental Psychology. A Coruna, Spain. (Chapters 5
and 7)
III. Sunio, V., Schmöcker, J. D., Kim, J., Estuar, M.R.E. (2017).
Understanding the stages and pathways of travel behavior
change induced by technology-based intervention among
university students. 56th Biannual Conference of the
Committee of Infrastructure Planning and Management,
2017. Committee of Infrastructure Planning and
Management, Japan Society of Civil Engineers. Iwate,
Japan. (Chapters 6, 8 and 9)
IN REVISION/UNDER REVIEW/TO BE SUBMITTED
I. Sunio, V., Schmöcker, J. D., Kim, J. (2017). Understanding
the stages and pathways of travel behavior change induced
by technology-based intervention among university students.
Transportation research part F: traffic psychology and
behaviour (Under Review). (Chapters 6, 8 and 9)
II. Sunio, V., Schmöcker, J. D. (2017). How Can We Use
Stage Models to Inform the Systematic Development of
Computer-Tailored Intervention for Large Scale Travel
Behavior Change? (To be Submitted). (Chapter 4)
xii
Table of Contents
CHAPTER 1 ................................................................................................................................................. 1
1.1. Background ........................................................................................................................................ 1
1.1.1. Sustainable Mobility in the Universities ..................................................................................... 1
1.1.2. Travel Demand Management Measures...................................................................................... 2
1.1.3. Context of universities in developing countries .......................................................................... 3
1.1.4. The Mobile Platform ................................................................................................................... 3
1.2. Research Questions and Objectives ................................................................................................... 4
1.3. Structure of the Dissertation .............................................................................................................. 5
CHAPTER 2 ................................................................................................................................................. 7
2.1. Introduction ........................................................................................................................................ 7
2.2. Travel Demand Management Measures: General Framework .......................................................... 8
2.3. Travel Demand Management Measures in the University Setting .................................................... 9
2.4. Discussion ........................................................................................................................................ 12
2.4.1. Dominantly hard measures ........................................................................................................ 12
2.4.2. Voluntary travel behavior change programs: a case of soft measure ........................................ 15
2.4.3. Issue of Transferability ............................................................................................................. 16
2.5. Mobility Behavior Change Support Systems ................................................................................... 16
2.6. Summary .......................................................................................................................................... 18
CHAPTER 3 ............................................................................................................................................... 19
3.1. Introduction ...................................................................................................................................... 19
3.2. Methods............................................................................................................................................ 22
3.2.1. Eligibility Criteria ..................................................................................................................... 22
3.2.2. Search Method .......................................................................................................................... 23
3.2.3. Evaluation Method .................................................................................................................... 23
3.3. Results .............................................................................................................................................. 30
3.3.1. Overview of BCSSs .................................................................................................................. 30
3.3.2. Persuasiveness Evaluation Using PSD...................................................................................... 32
3.3.3. Assessment of Efficacy Evaluation Studies .............................................................................. 40
3.4. Conclusion and Recommendations .................................................................................................. 42
3.4.1. Explicit Reference to a Behavior Change Theory ..................................................................... 43
xiii
3.4.2. Greater Integration with Shared Mobility Services Platform .................................................... 45
3.4.3. Social Interaction ...................................................................................................................... 48
3.5. Summary and Further Research ....................................................................................................... 49
CHAPTER 4 ............................................................................................................................................... 51
4.1. Introduction ...................................................................................................................................... 51
4.1.1. Effectiveness of tailoring in travel behavior intervention ................................................... 51
4.1.2. Computer-tailored interventions ......................................................................................... 52
4.1.3. Stage theory-driven intervention development ................................................................... 53
4.1.4. Implementation of a stage theory-driven, computer-tailored travel behavior change
interventions ........................................................................................................................................ 54
4.2. Stage Models Applied in Travel Behavior Domain ......................................................................... 55
4.2.1. Transtheoretical model (TTM) .................................................................................................. 55
4.2.2. Stage model of self-regulated behavior change (SSBC) ........................................................... 55
4.3. Literature Review of the Operationalization of Stage-Based Interventions .................................... 56
4.4. Discussion: Stage-Tailored and Menu-Based Interventions ............................................................ 61
4.4.1. Stage-tailored interventions ...................................................................................................... 61
4.4.2. Menu-based interventions ......................................................................................................... 66
4.5. Theoretical Basis: Conceptualizing Stages as Representing Partitions of an Underlying Continuum
of Action Readiness ................................................................................................................................ 67
4.6. Case Study: Blaze Mobility Behavior Change Support System ...................................................... 68
4.7. Summary .......................................................................................................................................... 73
CHAPTER 5 ............................................................................................................................................... 74
5.1. Introduction ...................................................................................................................................... 74
5.2. Stage Model of Self-Regulated Behavior Change (SSBC) Theory ................................................. 77
5.3. Stage-Tailored and Menu-based Interventions ................................................................................ 79
5.4. Stage-Specific Intervention Modules ............................................................................................... 80
5.4.1. Assessment of current stage membership ................................................................................. 81
5.4.2. Intervention module for participants in the pre-decisional stage .............................................. 81
5.4.3. Intervention module for participants in the pre-actional stage .................................................. 83
5.4.4. Intervention module for participants in the actional stage ........................................................ 83
5.4.5. Intervention module for participants in the post-actional stage and for captive public transport
users .................................................................................................................................................... 84
5.5. Standard Interventions ..................................................................................................................... 84
5.5.1. Social traces .............................................................................................................................. 84
5.5.2. Implementation plan ................................................................................................................. 86
xiv
5.5.3. Reflection-on-action.................................................................................................................. 87
5.5.4. Goal-setting ............................................................................................................................... 88
5.6. Usability Evaluation ......................................................................................................................... 89
5.6.1. Trip logging .............................................................................................................................. 91
5.6.2. Standard interventions............................................................................................................... 92
5.6.3. Observable and Latent Change in Behavior .............................................................................. 92
5.7. Summary, Limitations and Further Work ........................................................................................ 93
CHAPTER 6 ............................................................................................................................................... 95
6.1. Study Context................................................................................................................................... 95
6.1.1. Metro Manila ............................................................................................................................ 95
6.1.2. Ateneo de Manila University .................................................................................................... 95
6.2. Field Experiment .............................................................................................................................. 96
6.2.1. Study design .............................................................................................................................. 96
6.2.2. Measurement of Theory-Implied Constructs ............................................................................ 97
6.2.3. Reliability of measurement ..................................................................................................... 100
CHAPTER 7 ............................................................................................................................................. 102
7.1. Recap of Methodology ................................................................................................................... 102
7.2. Equivalence of Control and Experimental Groups at Baseline ...................................................... 103
7.3. Car Usage, Goal, Behavioral and Implementation Intentions, and Socio-cognitive variables per
stage at Baseline .................................................................................................................................... 104
7.4. Effect of Intervention on Car Use .................................................................................................. 110
7.5. Effect of Intervention on Stage Membership, Intentions and Socio-cognitive Determinants ........ 112
7.6. Usefulness Evaluation .................................................................................................................... 118
7.7. Summary, Limitations, Implications and Further Work ................................................................ 120
CHAPTER 8 ............................................................................................................................................. 123
8.1. Introduction .................................................................................................................................... 123
8.2. Association Between Stage and Car Use ....................................................................................... 124
8.3. Stability of Stages .......................................................................................................................... 125
8.4. Association Between Stages and Intention Strengths .................................................................... 126
8.5. Progression in Stage Membership is associated with the formation of three types of intentions .. 127
8.6. Determinants of the three intention types: model structure and parameter estimates .................... 131
8.7. Discussion: Validation and Extension of the Stage-based Behavior Change Model ..................... 134
CHAPTER 9 ............................................................................................................................................. 137
9.1. Introduction .................................................................................................................................... 137
9.2. Pathway from Interventions to Behavior Change via Change in Mediators .................................. 139
xv
9.3. Discussion ...................................................................................................................................... 144
9.3.1. Conceptual component ............................................................................................................ 144
9.3.2. Action component ................................................................................................................... 145
9.4. Limitations, Implications and Future Direction ............................................................................. 145
CHAPTER 10 ........................................................................................................................................... 148
10.1. Summary ...................................................................................................................................... 148
10.2. Main Findings and Contribution .................................................................................................. 151
10.3. Implications.................................................................................................................................. 153
10.4. Limitations ................................................................................................................................... 154
10.5. Recommendation: Towards a Computer-Tailored, Theory-Driven Development of Mobility
Behavior Change Support System ........................................................................................................ 156
10.5.1. Determining the current stage membership and cognition .................................................... 156
10.5.2. Generating automated travel diary and monitoring the level of car use ............................... 157
10.5.3. Reasoning about required changes ........................................................................................ 157
10.5.4. Suggesting relevant interventions in the menu for use ......................................................... 158
References ................................................................................................................................................. 159
xvi
List of Tables
Table 2-1 Examples of TDM measures taken from Garling and Schuitema (2007) ..................................... 9
Table 2-2 TDM Measures Implemented in the Universities ....................................................................... 10
Table 3-1 Coding Protocol used for PSD Evaluation of 9 BCSSs (developed from Kukkonen and
Harjumaa, 2009; Kelders et al, 2012; Lehto and Kukkonen, 2011; Langrial et al, 2012) .......................... 25
Table 3-2 Framework for Characterizing the Effectiveness Evaluation Study (adapted from Graham-
Rowe et al, 2011) ........................................................................................................................................ 30
Table 3-3 Overview of BCSSs reviewed/evaluated in this chapter ............................................................ 31
Table 3-4 Presence of Primary Task, Dialogue and Social Support Features in 9 BCSSs ......................... 38
Table 3-5 Details on Evaluation of Effectiveness of BCSSs ...................................................................... 41
Table 4-1 Operationalization of stage models for intervention development ............................................. 56
Table 4-2 Operationalization of TTM for intervention development in Diniz et al (2015) ........................ 63
Table 4-3 Operationalization of SSBC for intervention development in Bamberg (2013b) and Bamberg et
al (2015) ...................................................................................................................................................... 64
Table 4-4 Ten Standard Interventions within Blaze ................................................................................... 70
Table 5-1 User feedback on Trip Logging .................................................................................................. 91
Table 5-2 Stage Progression of N=22 Users ............................................................................................... 93
Table 6-1. Reliability of Survey Instrument (N=47) ................................................................................. 100
Table 7-1 Distribution across stages of experimental and control groups ................................................ 104
Table 7-2 ANCOVA Results .................................................................................................................... 110
Table 7-3 Stage Memberships .................................................................................................................. 112
Table 7-4 Stage Progression Across Stages .............................................................................................. 113
Table 7-5 Stage Transitions ...................................................................................................................... 113
Table 7-6 Intention Strengths Across Time .............................................................................................. 114
Table 7-7 Socio-cognitive determinants across time ................................................................................ 116
Table 7-8 Average Rating per rating per stage. ........................................................................................ 119
Table 8-1 Stage distribution and transitions. Control and experimental groups combined. T0=baseline;
T1=4 weeks later; PreD=predecision; PreA=preaction; A=action; PostA1=postaction(early);
PostA2=postaction(late) ............................................................................................................................ 126
Table 8-2 Slopes (R2) of the points in Fig 7-1(b), and Figs. 7-3(a)(b)(c) ................................................. 128
Table 8-3 Results of the Ordinal Logistic Regression Using Cumulative and Adjacent Categories
Significance codes: Bold 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘a’ 0.1 ............................................................ 130
Table 8-4 Summary of characteristics of the five stages .......................................................................... 131
Table 8-5 Model structure invariance: Good model fit indices of multiple groups indicate acceptability of
the model and invariance of model structure across groups ..................................................................... 133
Table 9-1 Parameter Estimates of Model 1 ............................................................................................... 140
xvii
List of Figures
Figure 4-1. Content or components of the predecisional report: (a) Response of the individual to prior
questions (topmost); (b) Recommendation: goal suggestion based on previous response (c) Blaze support:
relevant pages in Blaze that are considered meaningful for the person in this stage; (d) Motivational
statements such as quick trivia and what others say. .................................................................................. 69
Figure 5-1 VTBC Programs ........................................................................................................................ 75
Figure 5-2 Stage Model of Self-Regulated Behavior Change (Bamberg, 2013a) ...................................... 77
Figure 5-3 Pre-decisional stage module ...................................................................................................... 82
Figure 5-4 Pre-actional stage module ......................................................................................................... 82
Figure 5-5 Actional Stage Module .............................................................................................................. 83
Figure 5-6 Social Traces as Blaze Map ...................................................................................................... 85
Figure 5-7 Ridesharing and Trip Planning .................................................................................................. 86
Figure 5-8 Summary and Mode Use Graph ................................................................................................ 87
Figure 5-9 Goal-Setting .............................................................................................................................. 88
Figure 5-10 Overview of participation in the usability evaluation ............................................................. 90
Figure 7-1 Mean and 95% confidence intervals for car use per stage at the start of experiment (T0:
baseline) .................................................................................................................................................... 105
Figure 7-2 Goal Intention per stage at T0 .................................................................................................. 106
Figure 7-3 Behavioral Intention per stage at T0 ........................................................................................ 106
Figure 7-4 Implementation Intention per stage at T0 ................................................................................ 107
Figure 7-5 Socio-cognitive determinants per stage at T0 ......................................................................... 108
Figure 7-6 Comparison of responses between two users .......................................................................... 110
Figure 7-7 Car use reduction across different stages ................................................................................ 111
Figure 7-8 Change in goal intention ......................................................................................................... 115
Figure 7-9 Change in behavioral intention ............................................................................................... 115
Figure 7-10 Change in implementation intention ..................................................................................... 116
Figure 7-11 Change in selected socio-cognitive determinants ................................................................. 117
Figure 8-1. (a) Weekly car use and (b) car use change across stage changes ........................................... 124
Figure 8-2 Average intention strengths across stages (with 95% confidence intervals; combined T0 and
T1) ............................................................................................................................................................. 127
Figure 8-3 Changes in: (a) goal intention; (b) behavioral intention; (c) implementation intention relative to
stage progression ....................................................................................................................................... 128
Figure 8-4 Original model (based on Bamberg 2013b, ............................................................................. 132
Figure 8-5 Base model. We combine data from baseline and 4 weeks later (N=482). ............................. 133
Figure 9-1 Model 2, the change model. Path coefficients are reported early stages/late stages. Significant
path coefficients at 95% level are in bold, and at 90% level are underlined.. .......................................... 142
1
CHAPTER 1
Introduction
1.1. Background
1.1.1. Sustainable Mobility in the Universities
Awareness of the impact that universities have on cities has led university managers to find ways
to reduce the social and environmental impact of travel and contribute to community well-being
by promoting sustainable mobility in campuses. Such sustainable mobility campaigns in
universities are not at all trivial (Limanond et al, 2011; Zhou, 2012).
In the first place, universities are among the largest trip attractors/generators in a certain
community (Lovejoy and Handy, 2011; Rotaris & Danielis, 2015). Promoting alternative mobility
strategies can thus help minimize the impacts of the universities on surrounding areas. Secondly,
universities are “privileged places to communicate sustainability” (Balsas, 2003) especially among
the students, who will most likely “progress to occupying influential roles in government,
companies or other organizations” (Tolley, 1996). It is in the universities where “future leaders,
decision-makers and intellectuals of the social, political, economic and academic sectors are
created, formed and shaped” (Lozano, 2006). If universities succeed in forming these future
professionals into advocates of sustainability, they can help trigger positive changes to society.
Finally, universities can also serve as “laboratory for testing and implementing various alternative
transportation strategies” (Balsas, 2003) as well as “transport policy changes” (Bond and Steiner,
2006). Universities successful in their sustainability efforts can become “exemplars for other
employers and even for the society at large” (Zhou, 2012).
To highlight the importance of sustainable mobility in universities, the European Commission in
2016 co-funded a 5-year European project, called U-MOB LIFE (http://u-mob.eu/), aimed at “the
creation of a university network to facilitate the exchange and transfer of knowledge about
2
sustainable mobility best practices among European universities.” From 2016 until 2021, the
project will work “towards the reduction of CO2 emissions thanks to an improved mobility of the
university community.”
1.1.2. Travel Demand Management Measures
Like other types of public and private institutions, universities are among the largest poles of
attraction for commuters: thousands of daily trips are being generated towards and from the
universities. They are major trip attractors or generators that represent a large share of urban traffic,
especially when they are within the city (Lovejoy and Handy, 2011; Rotaris & Danielis, 2015).
Accordingly, it is claimed that commuting constitutes the single largest impact of the university
on the environment (Tolley, 1996; Bonham and Koth, 2010; Rotaris and Danielis, 2015).
Given this, it goes without saying how important the role is of transportation demand management
(TDM) measures as part of a university’s environmental management or sustainability plans.
Implementing such measures is critical especially in the context of large universities. High student
travel demand puts a heavy strain on the infrastructure supply of the universities themselves,
affecting the well-being of the students and employees. It also can have an impact on the levels
of congestion in the area, affecting the well-being of residents and businesses in the university
neighborhood (Danaf, Abou-Zeid, Kaysi, 2014). Hence, promotion of transit and active travel,
aside from being aligned well with the institutional sustainability goals of the university, can
impact the larger community as well (Whalen, Paez, Carrasco, 2013).
Schools or universities constitute one of the most important situations for implementing travel
demand management measures. Unlike other travel patterns which are dynamic and thus are hard
to manage, travel to school “represents a regular, estimable flow of persons to a defined location
generally within a narrow window of time. …the pattern and predictability of such trips should
lend itself well to the evaluation and design of policies, services or infrastructure focused
interventions that can influence the choices made by these individuals and reduce the associated
negative impacts” (Kelly and Fu, 2014).
3
In managing travel demand by car, policy interventions that aim at reducing car use can be
characterized broadly as structural/hard or psychological/soft (Bamberg et al, 2011; Graham-Rowe
et al, 2011). Structural interventions involve modification of the structures surrounding travel
behavior through physical and/or legislative measures (e.g. road pricing and bus priority lanes).
Psychological interventions are designed to modify perceptions, beliefs and attitudes (Graham-
Rowe et al, 2011). Both types of interventions can be effective. In fact, in practice, both must be
used in concert for greater effectiveness: “Individual TDM strategies have a modest impact on the
transportation system, but when multiple strategies are applied in a coordinated manner, the impact
on mode choice can be substantial. Further, when multiple strategies are applied, the negative
impacts on individual users are mitigated” (Bond and Steiner, 2006).
1.1.3. Context of universities in developing countries
Javid et al (2015) argue that many cities in the developing countries are “experiencing difficulty
in determining appropriate sets of policy instruments for reducing transport sector externalities.”
Developing countries are confronted with unique challenges, such as deficits in financial and
technical resources, that are often not experienced by developed countries. In the context of TDM
measures in the university setting, in particular at the American University of Beiruit (AUB) in
Lebanon, Aoun et al (2013) claim that “documented Transportation Demand Management (TDM)
practices at campuses are mostly from the developed world, and contrast markedly with Beirut,
which lacks an organized public transport sector and effective law enforcement. Further, the AUB
campus population comes from wealthier households and has higher car ownership compared to
the rest of Lebanon. Thus, conventional strategies such as subsidizing transit passes and restricting
or pricing parking are not perceived to be appropriate solutions for AUB and many other
developing world cases.”
1.1.4. The Mobile Platform
Classically, sustainable travel behavior has been promoted through various programs without
extensive support systems. Examples include so called “Travel Feedback Programs” (TFP) in
Japan (Fujii and Taniguchi, 2005). In a typical TFP, participants of the program are given feedback
4
based on their reported travel behavior with the aim of modifying their behavior without coercion.
TFPs come in several types, differentiated by the location, techniques, procedures and
communication media used in their implementation (Fujii and Taniguchi, 2006). With respect to
communication media, TFPs have relied so far mostly on traditional technologies, namely face-
to-face communication, regular mail, telephone and email (Fujii and Taniguchi, 2006). This
severely limits the potential of classical TFPs for scaling up (Jariyasunant et al, 2015). Nonetheless,
in recent years, a new platform for persuasion has emerged – the mobile platform – which can
broaden the applicability of travel demand management programs while maintaining their
effectiveness (Meloni et al, 2015). For instance, Quantified Traveler (Jariyasunant et al, 2015) is
a mobile-based computational TFP. These computational systems are simply traditional soft
measures implemented on a technology platform, with operational and functional aspects capable
of automating the whole process for large scale implementation. With their widespread adoption
and persuasive use in society, these technologies can be leveraged to deliver large-scale and cost-
effective behavior-change interventions – an important consideration for developing world cases
(Lathia et al, 2013).
1.2. Research Questions and Objectives
The main objective of this dissertation is to design, develop and evaluate a mobility behavior
change support system (mBCSS) as a travel demand management measure in the university setting
for developing world cases. Nonetheless, though the primary focus is the university setting, we
envision an even broader domain of application, such as the general population. In particular, we
ask the following questions:
How can we leverage the mobile platform in delivering travel behavior change
interventions? How can behavior change theories, in particular stage models, be used to
inform the development of mobile platform interventions in the travel behavior change
domain?
In the literature, studies evaluating the effectiveness of mobility behavior change systems
mostly address proof-of-concept, technicality, and usability concerns, but they do not
5
evaluate intervention-induced behavior change. A few studies have behavioral change
evaluations, but these are run on small sample sizes and short-term field study durations
(see, for example, Jariyasunant et al, 2015 and Meloni et al, 2015). However, a robust
evaluation study is resource-intensive. How can we design a social experiment that can
reliably evaluate the effectiveness of a mobility behavior change support system at a
reasonably acceptable cost?
Given that a mobility behavior change support system is effective in causing shift in
behavior and attitudes, through what processes does the intervention achieve its effects?
How can we use the comprehensive stage models to identify a causal pathway between
technology interventions and behaviour change?
1.3. Structure of the Dissertation
After this Introduction, we review in Chapter 2 some travel demand management (TDM) measures
implemented in universities or schools around the world. In Chapter 3, a review of existing
mobility behavior change systems for promoting sustainable travel behavior is carried out.
Persuasive features embedded in these systems are evaluated using the persuasive systems design
(PSD) model. In Chapter 4, we discuss how behavior change theories can be used to inform the
development of mobility behavior change systems. Specifically, we discuss how stage models
(namely, the transtheoretical model and the stage model of self-regulated behavior change) can be
used to systematically inform the development of travel behavior change interventions at large
scale. From these three reviews (Chapters 2 – 4), we describe in Chapter 5 how the Stage Model
of Self-Regulated Behavioral Change (SSBC) is used in the development of Blaze, a mobility
behavior change support system, consisting of Smartphone and web application. SSBC describes
behavior change process as a transition through four stages: predecision, preaction, action and
post-action. Using SSBC, we systematically develop theory-based interventions. In Chapter 6,
we provide the study context. We deploy Blaze among the students of a tertiary educational
institution located in Metro Manila, Philippines, a developing country. We describe the details of
the field experiment conducted in the university, the survey instrument used and its reliability. In
Chapter 7, we evaluate the effectiveness of Blaze. Our analysis shows that Blaze is able to induce
6
a positive change across a wide range of indicators, as specified by the theory: car use, stage
membership, intentions and socio-cognitive determinants. Furthermore, we conduct a usefulness
evaluation to assess the relevance of the features within Blaze. In Chapter 8, we validate and
extend the stage model of self-regulated behavior change. Our main findings suggest that, in
agreement with SSBC, travel behavior change is achieved by a transition through a temporal
sequence of four stages: predecision, pre-action, action and post-action. In an extension from
SSBC, we further distinguish post-action depending on whether the behavior is on initiation or
under maintenance. In Chapter 9, we describe how our mobility behavior change support system
is able to achieve its effects in changing the travel behavior of university students. We identify a
causal pathway linking the effect of the technology intervention to its behavioral outcome through
the mediation of a number of variables. In Chapter 10, we end by summarizing, presenting the
main contributions of this research, and giving recommendations for further work. We write this
dissertation such that it can be independently read chapter-by-chapter.
7
CHAPTER 2
Travel Demand Management Measures for Sustainable Mobility in
the Universities: A Review
Chapter Summary. We give a brief overview of some travel demand management (TDM)
measures implemented in universities or schools around the world. We begin first with TDM
measures in general, then specifically discuss TDM measures in the universities. We do so to
better position our current work within the existing literature. Two main issues in the literature
emerge from this review: the dominance of hard TDM measures and the transferability of
documented measures to cases in the developing world. We suggest that voluntary travel behavior
change programs, an example of soft measure, implemented as behavior change support systems
may be appropriate for university setting in the developing countries.
Keywords: Campus mobility sustainability; travel demand management measure; developing
world cases; behavior change support systems
2.1. Introduction
Universities can affect the surrounding areas in many ways, but commuting activities, which
represent a large share of urban traffic, constitute the “single largest impact” that universities have
on the environment (Rotaris and Danielis, 2015; Tolley, 1996). Transport use is among the top
three contributors to a university’s ecological footprint (Bonham and Koth, 2010). Given this, it is
to the best interests of universities to implement travel demand management (TDM) measures as
part of their environmental management or sustainability plans (Bhattacharjee et al 1997).
In the literature, TDM programs implemented in schools, not to mention residential areas and
workplaces, have been documented (e.g. Fujii and Taniguchi, 2006; Graham-Rowe et al, 2011).
Schools readily lend themselves to evaluation and design of such programs since travel to school
“represents a regular, estimable flow of persons to a defined location generally within a narrow
window of time” (Kelly and Fu, 2014). Nonetheless, these documented TDM solutions in the
8
university setting, which mostly come from the developed world, are considered not appropriate
for many cases in the developing world (Aoun et al 2013; Javid et al 2015).
TDM measures implemented in schools can be physical change, legal policy, economic policy and
information and education (Garling and Schuitema, 2007). Phyical, legal and economic measures
are, in general, considered hard interventions since they introduce modifications to structures
surrounding travel behavior, sometimes coercively. Information or education are also called soft
interventions, since they are non-coercive or voluntary and they do not introduce structural
changes through physical or legislative measures. In this chapter, we argue that soft measures,
since they do not require huge investments, may be appropriate for universities in the developing
world. Ideally, however, hard and soft TDM measures must be used together in concert for greater
effectiveness and more substantial impact (Bond and Steiner, 2006; Garling and Schuitema, 2007).
2.2. Travel Demand Management Measures: General Framework
Garling and Schuitema (2007) summarized four measures for managing travel demand by car:
physical change, legal policy, economic policy and information and education (Table 2-1).
First, physical changes are infrastructure modifications which aim at making alternative modes
relatively more attractive than car. Examples are physical improvements of infrastructures for
public transport, walking and cycling; provision of new transport services or exclusive bus lanes;
relocation of parking places; and increasing energy efficiency of cars.
Second, legal measures refer to restrictions imposed by law to limit car use. These include
prohibiting access to inner cities by car or restricting its use on certain days of the week; driving
restrictions such as speed limit; and parking controls.
Third, economic measures refer to all economic disincentives or sanctions related to car use with
the aim of making it relatively costlier compared to alternatives. Examples include taxation of
fuel and cars, urban transport pricing, congestion charging schemes, and public transport subsidies.
9
These measures are considered important public policy but a precondition for their successful
implementation is public and political acceptability.
Fourth, information and education measures are soft transport policy measures aimed at voluntary
switch to sustainable travel modes through a combination of marketing, information dissemination
and persuasion. Examples are travel feedback programs, personalized travel plan or individualized
marketing, social modeling and information campaign about the negative impact of car use.
Table 2-1 Examples of TDM measures taken from Garling and Schuitema (2007)
TDM Measure Examples
Physical change measures − improving public transport
– improving infrastructure for walking and cycling
– park & ride schemes
– land use planning to encourage shorter travel times
– technical changes to make cars more energy-efficient
Legal policies − prohibiting car traffic in city centers
– parking control
– decreasing speed limits
Economic policies − taxation of cars and fuel
– road or congestion pricing
– kilometer charging
– decreasing costs for public transport
Information and education measures − individualized marketing
– public information campaigns
– giving feedback about consequences of behavior
– social modeling
2.3. Travel Demand Management Measures in the University Setting
In this section, we conduct a survey of the TDM measures implemented in the university setting.
We include in Table 2-2 the reference, the university, the TDM measure implemented/proposed,
and the categorization based on the framework introduced in the previous section. We draw up
the list based on the authors’ own reviews and those by Danaf, Abou-Zeid, Kaysi (2014) and
Rotaris and Danielis (2014; 2015). This list is not meant to be exhaustive.
10
Based on our survey in Table 2-2, 12 of the 15 universities (80%) enforced economic measures to
reduce car use. These measures are mostly in the form of partial or full subsidies for public
transport services (e.g. unlimited access) or increase of car parking costs. Eight (53%) made
physical/technical improvements to encourage use of alternatives, including improvements of
existing transit services or provision of new ones (e.g. express and shuttle bus), construction of
infrastructure (e.g. cycling networks or reserved lanes), and relocation of parking lots. Five (33%)
implemented legal policies, mostly parking restrictions. Only four (27%) carried out education
and information campaign (e.g. TravelSmart or Cycle Safe).
Table 2-2 TDM Measures Implemented in the Universities Reference University TDM Measure
Description
Category
1 Brown et al. (2003) University of
California, Los
Angeles
Implementation of
BruinGO, a pilot fare-free
transit service that allows
unlimited access to public
transport
Economic
2 Dorsey (2005) University of Utah
and Weber State
University (WSU)
Full or partial subsidies for
transit passes (unlimited
access)
Economic
3 Bond and Steiner (2006) University of
Florida
Comprehensive TDM
system that includes
parking restriction, parking
pricing,
transit service
enhancements, and
unlimited-access transit
Legal
Physical
Economic
4 Shannon et al. (2006) University of
Western Australia
Subsidy for public
transport services;
Increase in the cost of
parking;
Improvement of bus
services;
Economic
Physical
11
Construction of cycle
networks
5 Rose (2008) Monash University Participation in the
TravelSmart initiative, a
voluntary travel behavior
change program
Information and
education
6 Ripplinger et al. (2009) North Dakota State
University
Increase in fuel prices;
Provision of express bus
Physical
Economic
7 Miralles-Guasch and
Domene (2010)
Autonomous
University of
Barcelona
Improved public transport
service (e.g. shorter travel
time);
Subsidized fares;
Appropriate parking
policies;
Infrastructure for non-
motorized and public
transport;
Education about the social
and environmental costs of
car use
Economic
Physical
Legal
Information and
education
8 Barata et al (2011) University of
Coimbra
Decreasing the overall
subsidization of parking;
Increasing control over
non-regular parking and
eliminating free on-street
parking
Legal
Economic
9 Zhuo (2012) University of
California, Los
Angeles
Discounted transit pass;
Utilization of information
contagion effects among
students;
Discounted daily permit
for students who usually
commute by an alternative
mode
Information and
education
Economic
10 Gonzalo-Orden et al.
(2012)
University of
Burgos, Spain
Car restriction in some
streets;
Charging parking fees;
Legal
Economic
Physical
12
Reserved bus lanes for
buses
11 Delmelle and Delmelle
(2012)
University of Idaho
(USA)
Increasing the parking cost
Economic
12 Whalen et al (2013) McMaster
University
Promote active modes,
such as cycling (Cycle Safe
Program)
Daily (and not seasonal)
parking passes
Education and
information
Legal
13 Mohammed and Shakir
(2013)
National University
of Malaysia
Improvement in bus
services;
Charging fees on parking
Economic
Physical
14 Aoun et al (2013);
Danaf et al (2014)
American
University of Beirut
Increasing parking fees;
Provision of shuttle
services or taxi sharing
Economic
Physical
15 Rotaris and Danielis (2014;
2015)
University of
Trieste
Subsidizing use of bus;
Increasing parking tariffs;
Reducing the size of
parking lots;
Relocating parking lots
Economic
Physical
2.4. Discussion
2.4.1. Dominantly hard measures
A brief review in Table 2-2 of the TDM measures implemented by universities reveals that
economic measures are widely used as a means to reduce car use or to encourage use of alternatives.
We now examine two specific examples of these economic measures: unlimited access and parking
pricing
The first example is Unlimited Access (UA). In this scheme, an arrangement between the
universities and public transit agencies is reached to provide fare-free transit services to members
of the university community. Typically, the university pays the transit agency an annual lump
13
sum, and all members of the university ride for free (Brown et al, 2001; 2003; Dorsey 2005). UA
has been implemented in 35 universities in the United States, and evaluation of its impact showed
substantial increase in ridership from 71% to 200% during the first year of the program. Aside
from this, UA also reduced demand for parking, improved overall transit service, and increased
transportation equity (Brown et al, 2001). At the University of Utah (U of U), UA is referred to
as Eco Pass. In this program, the university partners with the Utah Transit Authority (UTA), which
allows members of the U of U community, who are holders of the Eco Pass, to ride fare-free any
form of public transportation, such as buses and light rail, offered by UTA (Dorsey, 2005). In
other cases, some universities (e.g. University of Florida) implement the unlimited access transit,
together with parking pricing in order to raise funds, discourage parking, and to encourage the use
of alternative modes (Bond and Steiner, 2006). When unlimited access proves too costly for the
universities (e.g. Autonomous University of Barcelona), subsidization of public transport is
considered an option (Miralles-Guasch and Domene, 2010).
Unlimited access is a pull measure, aimed at increasing the attractiveness of alternatives. A good
strategy is to use pull measures in conjunction with push measures. An example of a push measure,
which is intended to decrease the attractiveness of car, is parking pricing. Many universities (e.g.
Autonomous University of Barcelona) subsidize the costs of parking of students by providing it
free of charge, or at lower prices that do not accurately reflect the actual costs (Barata et al, 2011).
Hence, it is necessary to reduce or even eliminate the over-all subsidization. Some proposals
include: increase parking fees for single occupant vehicles, reduce parking fees for carpool drivers,
and eliminate on-street free parking. Whalen et al (2013) reported that at McMaster University,
replacing seasonal parking passes with (discounted) daily passes can also make the car owners
more sensitive to parking costs, which thus can discourage solo driving.
These two measures (unlimited access and parking pricing) are also found by Aoun et al (2013) as
the most popular TDM strategies at campuses. We now look at examples of physical measures,
adopted by more than half of the universities considered in this review.
The first example is improvement of public transport service. The main barrier to using public
transport on a large scale is the long commute time (Miralles-Guasch and Domene, 2010). This
14
can be because of long in-vehicle travel time, long wait time, and long travel distance. In the
context of the University of Florida, improved service refers to increased frequency and direct
routes from home to school (Bond and Steiner, 2006). At the American University of Beirut, the
provision of express shuttles or taxis has been proposed (Aoun et al, 2013; Danaf et al, 2014),
similar to express bus services shuttling between campus and off-campus areas with high student
populations suggested at the North Dakota State University (Ripplinger et al., 2009).
A second example is construction of infrastructures for non-motorized modes in order to promote
active travel (walking and cycling). Shannon et al (2006) reported about the Active Commuting
Project undertaken at the University of Western Australia, aimed at encouraging active commuting.
Appropriate strategies proposed for increasing active commuting include the improvement of “the
pedestrian and bicycle network leading to campus”. However, since walking or cycling are only
feasible for short distances, student housing on or near campus is also suggested. Miralles-Guasch
and Domene (2010) also observed that at the Autonomous University of Barcelona around “10.0%
of the university community lives within a reasonable walking or cycling distance, however,
structural aspects such as the suburban character of the campus and the lack of paths for pedestrians
and bicycles, represent obstacles to changing to these modes, particularly for pedestrians.” Hence,
by constructing paths for pedestrians and cyclists, the modal share of active modes can be increased.
Finally, relocation or size reduction of parking lots is also cited as an example of physical change
measure in Rotaris and Danielis (2014; 2015). In their study simulating the impact of various
transportation demand management measures on commuting to the University of Trieste, Rotaris
and Danielis (2015) found that parking regulations, such as the number of parking spaces and the
location of the parking lots, has a big effect on mode choice in favor of bus use. Reducing the size
of the parking lots increases bus ridership by 56% while relocating them away from the university
buildings contributes to a bus ridership increase by 69%.
Both economic and physical change measures – including legal measures – are classified as hard
(or structural) interventions. Structural interventions involve modification of the structures
surrounding travel behavior. Hard interventions are, as we have seen, the dominant measures used
in travel demand management in the university setting. Psychological or soft interventions such
15
as TravelSmart (Rose, 2008) and Cycle Safe (Whalen et al, 2013), designed to modify perceptions,
beliefs and attitudes, are poorly represented and rarely used among the university travel demand
management measures. We consider next a voluntary travel behavior change program, called
TravelSmart, implemented in Monash University (Rose, 2008). TravelSmart is an example of a
soft measure.
2.4.2. Voluntary travel behavior change programs: a case of soft measure
Voluntary travel behavior change or VTBC programs are “designed to enable individuals to
become more aware of their travel options and, where possible, exercise choices that reduce the
use of private motor vehicles” (Rose and Ampt, 2003). An example is TravelSmart. In this
program, no additional transport or infrastructure, nor improvements in public transport services
is provided. Instead, it aims to facilitate behavior change within the existing urban transport and
land-use infrastructures. It targets workplaces, communities, and educational institutions (Rose,
2008).
In the academic year of 2004-2005, the Clayton Campus of Monash University implemented the
TravelSmart initiative. Target participants of this program are first-year students, who are selected
because they are thought to be more open to alternative suggestions. These students were given
travel information delivered at a single point in time – during the enrollment process. The
information package includes: generic cover letter; local area map showing cycling, walking and
bus routes; public transport map of Melbourne; student public transport concession card
application form; and carpool postcard. Students also had a face-to-face active dialogue and
interaction with travel officers who provided them with tailored information such as appropriate
bus and/or train timetables, daily public transport tickets, and other information. Nonetheless, the
students were not asked to make a behavioral plan as basis of changing their behavior. They only
verbally indicated the most likely travel mode they will use in going to campus.
Two surveys were conducted among the first-year students, the first in October 2003, and the
second in May 2004 in order to evaluate the effectiveness of the VTBC program. These surveys
represent the period before and after the TravelSmart initiative was run. Before and after
16
evaluation results showed a “reduction in car driver trips of about 9 percent and an increase in
public transport trips of about 6 percent” (Rose, 2008).
2.4.3. Issue of Transferability
In Table 2-2, we observe that the measures implemented or proposed so far are dominated by “hard”
interventions, mostly through economic measures (e.g. parking and pricing, and subsidizing).
Only in one university (the Monash University) is a voluntary travel behavior change program
implemented, called TravelSmart (Rose, 2008). Moreover, these measures are implemented in
universities in the developed countries, with the exception of the American University of Beirut
(AUB) and National University of Malaysia. As Aoun et al (2013) have pointed out, these
strategies that are found to be successful in the developed world, are not perceived to be
appropriate solutions for – and therefore not readily transferable to – many developing world cases,
including Beirut in Lebanon. Some features pointed out by Aoun et al (2013) that make Lebanon
different from developed world in terms of transport context include: higher densities and traffic
congestion, weak governance and informal public transport sectors. Furthermore, they argued that,
“South American, Asian, and African cities might better compare with Beirut in terms of their
densities, congestion, weak governance, and informal sectors; yet, literature on campus TDM
applications in these regions is slim to nonexistent.”
2.5. Mobility Behavior Change Support Systems
In this section, we introduce Quantified Traveler (Jariyasunant et al, 2015), a mobility behavior
change support system (mBCSS; definition below) piloted at the University of California in
Berkeley (UCB). Though not introduced with the intention of implementing it as a travel demand
management measure in the university, QT has the potential to persuade students in UCB to take
alternative modes to car. We argue in this section that mBCSS, because of their many features
which we shall elaborate shortly, may be the appropriate voluntary behavior change program for
cases in the developing world.
17
A behavior change support system (BCSS) is an “information system designed to form, alter, or
reinforce attitudes, behaviors or an act of complying without using deception, coercion or
inducements” (Oinas-Kukkonen, 2010). The prefix “mobility” attached to BCSS thus just denotes
the target of study – namely, mobility. mBCSS refers to an information support system to promote
environmentally friendly everyday mobility behavior (c.f. Bamberg et al, 2015).
Quantified Traveler, or QT for short, is a behavior change support system designed to change
travel behavior. As in Monash University, QT is a soft measure encouraging a voluntary
modification in travel behavior, but unlike TravelSmart, it is implemented as a computational
travel feedback system that aims to change mode or trip choice, without need for traditional travel
counselors. The system automatically collects trip data, converts them into a travel diary, and gives
quantitative feedback (time and money spent, calories burned and CO2 emitted) to the traveler.
QT was tested for use by 135 subjects from University of California, Berkeley from March 18 to
April 7, 2012. Before-after comparison of key measured variables shows that QT induces a
significant change in travel distance (i.e. driving distance), awareness, attitudes and intentions over
three-week duration as a result of the intervention (Jariyasunant et al, 2015).
As a soft measure, mBCSSs do not require huge financial investments for infrastructures or
operations, which certainly are a major consideration for universities in the developing world.
With limited investments on infrastructure and public transportation improvements in developing
countries, “soft measures” such as voluntary travel behavior change programs, which do not
require huge funding, may be the only viable option at the moment to encourage a shift towards
sustainable transportation. Moreover, by using a technology platform, mBCSSs have the potential
to deliver interventions at a large-scale in a cost-effective manner. In classical voluntary travel
behavior change programs such as TravelSmart, the need for travel counselors incurs huge cost,
and hence imposes limitation on scalability. The mobile platform, however, utilized by BCSSs,
can broaden the applicability of VTBC programs while maintaining their effectiveness (Meloni &
di Teulada, 2015). Furthermore, the rise in smartphone penetration rate and internet usage in many
developing countries (c.f. Poushter, 2016) poses the possibility of deploying mBCSSs among
larger segments of the population, especially among young people. Majority of the members of
the university community are the youth, who are known to have high Smartphone ownership and
18
usage. Finally, “public transportation infrastructure may be more facilitative of commuting trips
to the university than to other destinations in the city, making the switch to alternative transport
for commuting trips an easier or more convenient choice than for non-commuting trips” (Kormos
et al, 2014). This is an important consideration since the effectiveness of mBCSSs, or soft
transport policies in general, will be strengthened if combined with hard transport policy measures
or if there exist good infrastructure supporting sustainable travel behavior (c.f. Garling and
Schuitema, 2007).
2.6. Summary
A review of travel demand management measures implemented in the university setting is
presented. The general framework by Garling and Schuitema (2007) is used to classify these TDM
measures into physical, economic, legal and information measures. Two main issues in the
literature emerge from this review: the dominance of hard TDM measures (i.e. economic and
physical changes) and the transferability of documented measures to cases in the developing world.
We then recommend that voluntary travel behavior change programs, an example of soft measure,
implemented as behavior change support systems may be appropriate for university setting in the
developing countries.
19
CHAPTER 3
Can we promote Sustainable Travel Behavior Through Mobile
Apps? Evaluation and Review of Evidence
Chapter Summary. Mobile phone applications to monitor and influence one’s behavior are
numerous. Most developed appear to be health applications but in the past decade, “persuasive
technology” has also been leveraged and applied to promote sustainable travel behavior. We
discuss the health applications and review and evaluate existing behavior change support systems
(BCSS) designed to promote sustainable travel behavior. We extract the persuasive features
embedded in these systems and evaluate their persuasive potential by using the persuasive systems
design (PSD) model that has been used to evaluate BCSSs in the health domain. Our evaluation
reveals that some features crucial for successful travel behavior change, such as tunneling,
rehearsal and social facilitation, are missing. Furthermore, we assess studies conducted to evaluate
the effectiveness of these BCSSs in changing behavior and find indications that effect sizes are
mostly small though methodologically robust studies are largely missing and hence no definitive
conclusion yet can be derived. Based on these findings as well as literature related to public health
where BCSSs appear to be further developed, we then derive three important suggestions on
research needs and applications for further development of BCSSs in the transport policy realm.
Keywords: Smartphone; behavior change support system; sustainability; travel behavior;
persuasive technology.
3.1. Introduction
There has been a burgeoning interest in using technology to deliver interventions to change
behavior ever since Fogg (2002) introduced his pioneering work on “persuasive technology”. This
is defined as technology designed to change attitudes and behaviors of users through persuasion
(Fogg, 2002). Recently, Oinas-Kukkonen (2010; 2013) coined the concept of behavior change
support system (BCSS), which builds on this tradition of persuasive technology. Oinas-Kukkonen
20
(2010) defines BCSS as an “information system designed to form, alter, or reinforce attitudes,
behaviors or an act of complying without using deception, coercion or inducements”. In the design
of the information system, a web- or mobile- based platform is often used. Here, smartphones as a
medium of intervention are particularly noteworthy and promising. With their widespread
adoption and pervasive use in society, smartphones can be leveraged to deliver large-scale and
cost-effective behavior-change interventions (Lathia et al, 2013). Indeed, as Fogg and Eckles
(2007) asserted, the future platform for persuasion and behavior change is mobile.
BCSSs have been implemented in domains such as healthcare/wellbeing (e.g. Lane et al, 2011),
energy use conservation (e.g. Weiss et al, 2011), education (e.g. Mintz and Aagaard, 2012; Mintz
et al, 2012), and travel (e.g. Froehlich et al, 2009), to name only a few. In the healthcare/well-
being domain, extensive reviews have already been conducted. These studies include evaluations
on the persuasiveness of the design features of the health/well-being BCSSs (Langrial et al., 2012;
Lehto & Oinas-Kukkonen, 2010; Lehto & Oinas-Kukkonen, 2011); content analysis on the extent
of inclusion of health behavior theory constructs (e.g. Cowan et al, 2012); characterizations of
behavior change techniques implemented (e.g. Conroy et al, 2014; Yang et al, 2015); and
assessment of their effectiveness (e.g. Wang et al, 2014). Although a number of reviews like these
have been done in the healthcare/wellness domain, we observe that, in the domain of travel
behavior, no review of this kind has been carried out yet.
Moreover, in recent years, Fogg’s (2002) framework has also been applied to the topic of
environmental sustainability (e.g. Di Salvo et al, 2010). A broad range of environmental
sustainability issues are being addressed or tackled, such as energy consumption, water and fuel
use, indoor air quality and transportation (Brynjarsdottir et al, 2012). The primary aim is to
promote environmental sustainability through persuasive technology.
In this chapter, our interest is on persuasive technology, in particular BCSSs, designed to promote
“sustainable travel” or “sustainable mobility”. Technology – or in particular, information and
communication technology (ICT) – can promote sustainable urban mobility in various ways by
changing travel demand, travel patterns and urban forms (see, for example, Cohen-Blankshtain &
Rotem-Mindali, 2016), but here we only consider ICT potential effect on travel behavior.
21
Moreover, we use “travel” and “mobility” interchangeably. Definitions of sustainability vary in
literature, but it is generally considered as encompassing environmental, economic and social
dimensions. In this chapter, we regard sustainability in its narrow environmental sense of simply
minimizing the amount of car travel or carbon emissions. Unsustainability of travel is thus usually
equated with car use.
To the best of our knowledge, work on sustainable travel BCSSs began with Ubigreen
Transportation Display (Froehlich et al, 2009). Its purpose is to increase awareness about the
user’s sustainable travel behavior. Each time the user takes greener alternatives, Ubigreen senses
this and feeds it back as ambient changes to the background graphics of the user’s phone. Later
on, other sustainable travel BCSSs were developed: Peacox (Schrammel et al, 2012), Quantified
Traveler (Jariyasunant et al, 2015) and MatkaHupi (Jylhä et al, 2013).
Traditionally sustainable travel behavior has been promoted through various voluntary behavior
change programs (VTBC) without extensive support systems. Examples include so called “Travel
Feedback Programs” (TFP) in Japan (Fujii and Taniguchi, 2005). In a typical TFP, participants of
the program are given feedback based on their reported travel behavior with the aim of modifying
their behavior without coercion. TFPs come in several types, differentiated by the location,
techniques, procedures and communication media used in their implementation (Fujii and
Taniguchi, 2006). With respect to communication media, TFPs have relied so far only on
traditional technologies, namely face-to-face communication, regular mail, telephone and email
(Fujii and Taniguchi, 2006). This severely limits the potential of classical TFPs for scaling up
(Jariyasunant et al, 2015). Nonetheless, as argued earlier, a new platform for persuasion has
emerged recently – the mobile platform – which TFPs, or VTBC programs in general, appear to
not yet have taken full advantage of. The mobile platform can broaden the applicability of VTBC
programs while maintaining their effectiveness (Meloni et al, 2015). For instance, Quantified
Traveler (Jariyasunant et al, 2015), mentioned earlier, is a mobile-based computational TFP.
Besides their potential in promoting sustainable behavior change, many of these systems also
function as automated trip diaries, capable of generating travel surveys at large scale. Smartphones
are equipped with sensors that can be used to collect trip data from individuals, often with minimal
22
respondent burden. These data can then be processed to derive detailed information about mobility
patterns (Roxas et al, 2016; Sadeghvaziri et al, 2016; Patterson and Fitzsimmons, 2016) and urban
activities (Demissie et al, 2015).
In this chapter, we aim to extract and evaluate the persuasive elements embedded in travel behavior
change support systems by applying the Persuasive Systems Design (PSD) model (Oinas-
Kukkonen and Harjumaa, 2009; explained below). We are motivated by the need to identify
important persuasive design elements in these BCSSs. Davies (2012) observed that in non-
technology-based VTBC programs, “little regard has been paid to the actual design of VTBC
campaigns and their individual elements… However, there is also a need to understand and
theorise the design of campaigns, both singularly and collectively, as part of overall policy to
change travel behavior on a wide scale”. Hence, it is necessary to evaluate the design of the VTBC
program itself, in addition to its impact in changing behavior. Finally, we conclude with some
propositions and suggestions to advance the field.
3.2. Methods
3.2.1. Eligibility Criteria
There are clearly a large number of studies aiming to change the travel behavior through various
information feedback (Ampt, 2003; Fujii and Taniguchi, 2005; Richter, Friman, & Gärling, 2011).
We limited our review by the following criteria: A study was included if (1) it aimed to change
travel behavior by influencing the traveler’s trip, mode, departure time, or route choices; (2) the
intervention used a mobile/smartphone application and/or website, and (3) it is published in
journals or conferences. Studies that use smartphones as mobile sensors to automatically detect
trip information or construct travel diaries, but do not (explicitly) aim to change behavior, are
excluded. Moreover, those which use smartphones in designing behavioral modification
interventions but have no corresponding publication are likewise excluded. For example, “in-
Time”, “Wiewohin”, “Kongressnavigator” and “EcoWalk”, which are cited in Busch et al (2012),
are not included in this review because we could not identify sufficient material describing these.
23
3.2.2. Search Method
We conducted our literature search in several databases using keywords such as ‘‘smartphone”,
“travel”, “behavior”, “change”, “persuasive” and “intervention”. We searched articles published
between January 2005 and June 2015. We screened the articles based on their titles and abstracts,
and those that seemed to meet the inclusion criteria were selected for further examination. After
we verified that they indeed met the necessary criteria, they were then included in the final list. A
snowball review was then conducted, in which selected articles that emerged from the literature
search were screened and checked for potential eligible studies. In case when some papers obtained
through this search method are only a part of a bigger study, supplementary papers are also
obtained.
3.2.3. Evaluation Method
3.2.3.1. Persuasiveness
Several frameworks have been introduced to evaluate persuasiveness of design. The BCSSs
included in the study were evaluated based on the framework introduced by Oinas-Kukkonen and
Harjumaa (2009), known as the Persuasive Systems Design (PSD) model. The PSD represents an
extensive conceptualization for technology-based persuasion and the most sophisticated
evaluation method available (Lehto & Oinas-Kukkonen, 2011). In general, PSD can be used in a
variety of settings: from evaluating the software design specifications to assessing the literature in
some problem domain (Oinas-Kukkonen, 2013).
The PSD model is a conceptualization for designing, developing and evaluating BCSSs. In
evaluating BCSSs using PSD, we first analyze the persuasion context and then the persuasive
system features. Persuasion context analysis includes identifying the intent (Who is the persuader?
What type of change does the persuader target?), the event (use, user and technology contexts) and
the strategy (message and route). Next, we analyze the persuasive system features. In the PSD
model, there are four categories for the features: primary task support, dialogue support, system
credibility support and social support. Primary task support helps the user achieve or carry out the
primary task or target behavior by means of seven principles, namely reduction, tunneling,
tailoring, personalization, self-monitoring, simulation, and rehearsal. Dialogue support enables
24
user-system interaction to keep the user active and motivated in using the system. Praise, rewards,
reminders, suggestion, similarity, liking, and social role are seven principles of dialogue support.
System credibility support aims to make the system more believable and thereby more persuasive.
It includes trustworthiness, expertise, surface credibility, real world feel, authority, third party
endorsements and verifiability. Social support leverages social influence to motivate users and
employs social learning, social comparison, normative influence, social facilitation, cooperation,
competition and recognition (Oinas-Kukkonen and Harjumaa, 2009; Oinas-Kukkonen, 2010;
Oinas-Kukkonen, 2013).
Two methodologies can be used to carry out BCSS evaluation according to the PSD model. The
first, and the most rigorous, is to have two or more research scientists independently carry out
feature-by-feature evaluation of the applications using the persuasive features specified by the PSD
Model. After the independent review is made, the research scientists meet to discuss findings and
to resolve disparities in their evaluations. In Langrial et al (2012), four scientists carry out
independent evaluations while in Lehto and Kukkonen (2010) and Kelders et al (2012), only two
perform independent review. The second, which is less rigorous, consists of one research scientist
preparing a comprehensive evaluation, which is then verified and commented by one or more
research scientists. In Lehto and Kukkonen (2011), this is the methodology employed: one
research scientist did the coding, which was then checked by the second scientist.
Moreover, feature-by-feature evaluation can be done by simply coding descriptions based on
published literature without using the applications (e.g. Lehto and Kukkonen, 2011; Kelders et al,
2012) or by actually using the applications and performing representative tasks (e.g. Lehto and
Kukkonen, 2010), or both (e.g. Langrial et al, 2012).
In our study, four research scientists performed the PSD evaluation. We first developed a common
coding framework from the seminal paper on PSD by Kukkonen and Harjumaa (2009), and papers
that use the PSD model to evaluate applications (i.e. Lehto and Kukkonen, 2010; Lehto and
Kukkonen, 2011; Kelders et al, 2012; Langrial et al, 2012). This coding framework (see Table 3-
1) was used by the first author to prepare a comprehensive feature-by-feature evaluation of the 9
BCSSs as described in the published reports. The resulting entries were then put in tabular form,
25
and independently reviewed and commented by the second author and two other research scientists.
Any disparity was then resolved by all the four evaluators through a rigorous discussion. It was
agreed that persuasive features would only be reported if all four evaluators reached a consensus.
Our evaluation is limited only to published reports, since many of the applications are not
downloadable from the online stores, and if ever they are, they cannot be used in Japan. Lastly,
in our evaluation, as in Kelders et al (2012), we omitted system credibility support because the
published studies do not sufficiently report these principles, making any evaluation difficult to
carry out in a manner that is as objective as possible.
Finally, we note that evaluating persuasiveness of design using PSD is based on “interpretive
categorization” (Lehto and Kukkonen, 2011). In many articles cited in this review, their authors
did not state explicitly the persuasive features; in few others, their authors clearly stated these
features even though they did not necessarily use the same terminologies in the PSD. Hence, the
authors of the present review, in extracting and categorizing persuasive features, had to use their
subjective judgment – and herein lies potential bias.
Table 3-1 Coding Protocol used for PSD Evaluation of 9 BCSSs (developed from Kukkonen and
Harjumaa, 2009; Kelders et al, 2012; Lehto and Kukkonen, 2011; Langrial et al, 2012)
Persuasion
Strategy
Description
Coded as element included
when the BCSS:
Example Implementation
Reduction A system that reduces
complex behavior into simple
tasks helps users perform the
target behavior, and it may
increase the benefit/cost ratio
of a behavior.
Specifically divides the target
behavior into small, simple
steps
A support system for weight
management includes a diary for
recording daily calorie intake,
thereby dividing the target
behavior (reducing calorie
intake) into small, simple steps
of which one is recording calorie
intake
Tunneling Using the system to guide
users through a process or
experience provides
opportunities to persuade
along the way. System
Delivers content in a step-by-
step format with a predefined
order
A support system for the
prevention of depression that
delivers the content in sequential
lessons that can only be
accessed when the previous
lesson is completed
26
Tailoring Information provided by the
system will be more
persuasive if it is tailored to
the potential needs, interests,
personality, usage context, or
other factors relevant to a user
group.
Provides content that is
adapted to factors relevant to a
user group, or when feedback
is provided based on
information filled out by a
participant
A support system for personal
training provides different
information content for different
user groups, e.g. beginners and
professional
Personalization A system that offers
personalized content or
services has a greater
capability for persuasion
Provides content that is
adapted to one user (ie, the
name of the user is mentioned
and/or the user can adapt a part
of the intervention)
A support system personalizes
service and content based on
user-inputs and other known
variables e.g. name, gender, age,
location, language
Self-
monitoring
A system that keeps track of
one’s own performance or
status supports the user in
achieving goals.
Provides the ability to track
and view the user’s behavior,
performance or status
A support system for smoking
cessation allows participants to
review their smoking by sending
immediate feedback forms and
copies of the personalized
assessments to their email
accounts.
Simulation Systems that provide
simulations can persuade by
enabling users to observe
immediately the link between
cause and effect.
Provides the ability to observe
the cause-and-effect
relationship of relevant
behavior
A support system for smoking
cessation includes an interactive
smoker’s risk tool that simulates
changes in the risk of death due
to smoking based on the
smoker’s history and time of
quitting.
Rehearsal A system providing means
with which to rehearse a
behavior can enable people to
change their attitudes or
behavior in the real world.
Provides the ability and
stimulation to rehearse a
behavior or to rehearse the
content of the intervention
A support system that includes a
flying simulator that helps flight
pilots practice for severe
weather conditions
Praise By offering praise, a system
can make users more open to
persuasion.
Offers praise to the participant
on any occasion
A support system that aims to
promote healthy nutritional
habits compliments participants
when they have eaten 2 pieces of
fruit for 5 days
27
Rewards Systems that reward target
behaviors may have great
persuasive powers.
Offers some kind of reward
when the participant performs
a target behavior relating to the
use or goal of the intervention
A support system for the
treatment of social phobia gives
points to participants when they
engage in exposure exercises
Reminders If a system reminds users of
their target behavior, the users
will more likely achieve their
goals.
Provides reminders about the
use of the intervention or the
performance of target behavior
A support system to support
self-management among
patients with rheumatic arthritis
sends an automatic email
message to remind the
participant that the new lesson
may begin
Suggestion Systems offering fitting
suggestions will have greater
persuasive powers.
Provides a suggestion to help
the participants reach the
target behavior
A support system for weight
management provides low-
calorie recipes
Similarity People are more readily
persuaded through systems
that remind them of
themselves in some
meaningful way.
Is designed to look familiar
and designed especially for the
participant
A support system for the
treatment of panic disorder in
teenage girls explains the
exercises through a teenage girl
with panic problems
Liking A system that is visually
attractive for its users is likely
to be more persuasive.
Is visually designed to be
attractive to the participants
A support system that aims at
encouraging children to take
care of their pets properly has
pictures of cute animals
Social Role If a system adopts a social
role, users will more likely use
it for persuasive purposes
Acts as if it has a social role
(eg, a coach, instructor, or
buddy)
A support system to support
self-management among
patients with migraine
incorporated an avatar to guide
the participant through the
intervention
Social learning A person will be more
motivated to perform a target
behavior if (s)he can use a
system to observe others
performing the behavior.
Provides the opportunity and
stimulates participants to see
others using the intervention or
performing the target behavior
A support system for weight
management provides the
option, and stresses the
importance, of posting physical
activity self-monitoring data on
the discussion board and
28
commenting on the performance
of others
Social
comparison
System users will have a
greater motivation to perform
the target behavior if they can
compare their performance
with the performance of
others.
Provides the opportunity for
participants to compare their
behavior to the target behavior
of other participants and
stimulates them to do this
A support system for drug abuse
prevention for teenagers
automatically compares the
response of the participant to
other users of the intervention
Normative
influence
A system can leverage
normative influence or peer
pressure to increase the
likelihood that a person will
adopt a target behavior
Provides normative
information on the target
behavior or the usage of the
intervention
A support system to promote
self-management among
patients with COPD provides
feedback on the level of physical
activity of the participant by
comparing it to the physical
activity of well-managed COPD
patients
Social
facilitation
System users are more likely
to perform target behavior if
they discern via the system
that others are performing the
behavior along with them.
Provides the opportunity to see
whether there are other
participants using the
intervention
A support system that provides
opportunity to contact others
using the same intervention (e.g.
discussion group, peer-to-peer
forums, chat rooms)
Cooperation A system can motivate users to
adopt a target attitude or
behavior by leveraging human
beings’ natural drive to co-
operate.
Stimulates participants to
cooperate to achieve a target
behavior
A support system for the
promotion of physical activity
stimulates participants to form
groups and to achieve the group
goal of a certain number of steps
each week
Competition A system can motivate users to
adopt a target attitude or
behavior by leveraging human
beings’ natural drive to
compete.
Stimulates participants to
compete with each other to
achieve a target behavior
A support system for diabetes
management among children
includes a leaderboard in which
the children who enter blood
glucose levels at the right times
receive the highest place
Recognition By offering public recognition
for an individual or group, a
system can increase the
Prominently shows (former)
participants who adopted the
target behavior
A support system for treatment
of anxiety includes a testimonial
29
likelihood that a person/group
will adopt a target behavior.
page where successful users of
the intervention tell their story
3.2.3.2. Effectiveness
We then characterize the evaluation studies conducted to assess the effectiveness of the identified
BCSSs. In our characterization, we adapt parts of the framework used by Graham-Rowe et al
(2011) to assess the efficacy of 77 car-use reduction interventions. In this framework, evaluation
studies are characterized in terms of intervention strategy (structural, psychological),
methodological quality of study design (high, low), and measure type (distance, mode change,
trips/frequency, time/duration). Structural interventions involve modification of the structures
surrounding travel behavior through physical and/or legislative measures (e.g. road pricing and
bus priority lanes). Psychological interventions are designed to modify perceptions, beliefs and
attitudes (Graham-Rowe et al, 2011). Since we are interested only in BCSSs, we only consider
psychological intervention strategies.
Three study designs are considered of high methodological quality: experimental, quasi-
experimental, and cohort-analytic (with control). In experimental designs (e.g., randomised
controlled trials), individuals are allocated randomly to either intervention or control groups. In
quasi-experimental designs, matched but not randomized control groups are used. In the cohort-
analytic (with control) method groups exposed and unexposed to an intervention are compared. In
this method, the investigator does not control the intervention exposure but only ‘observes’
(observational study design). These three designs are considered of superior quality because they
provide strongest control of confounding variables.
Low quality study designs include case controlled/cross sectional and cohort uncontrolled. In
case-control or cross-sectional between-group evaluation, two groups, without pre-intervention
measures, are compared post-intervention. One group is exposed to the intervention, and the other
is not. In cohort-uncontrolled, before- and after-intervention measures are recorded, but there is
no control group for comparison. In the original framework by Graham-Rowe et al (2011), medium
and medium/low study designs are also cited, but for our purposes, we deem it unnecessary to
include them here.
30
The outcome measures are categorized as: (i) distance travelled, (ii) number of car trips or
frequency or car use, (iii) time spent in a car and (iv) measures of modal shift away from car use
or away from single car occupancy. Section 3.3 presents the results of our assessment of the
efficacy evaluation studies of the BCSSs using this categorisation.
Table 3-2 Framework for Characterizing the Effectiveness Evaluation Study (adapted from
Graham-Rowe et al, 2011) Intervention Strategy Methodological Quality Outcome Measure
Psychological
Structural
High:
Experimental, quasi-experimental,
and cohort-analytic (with control)
Low:
Case controlled/cross sectional and
cohort uncontrolled
Distance
Mode change
Trips/frequency
Time/duration
3.3. Results
3.3.1. Overview of BCSSs
Our search method yielded nine unique BCSSs that met our inclusion criteria and that are listed
and described in Table 3-3. We note that these BCSSs differ quite significantly in the size of the
consortium involved and hence probably also in the amount of resources available for their
development. Five out of these nine BCSSs (Peacox, SuperHub, Tripzoom, Matkahupi, iTour) are
part of larger projects, sponsored by for example the European Commission. In addition PEIR is a
collaboration between academics and industry. The remaining three BCSSs (IPET, Ubigreen and
QT) appear to be smaller scale academic projects. Moreover, all nine BCSSs are from the
developed countries in Europe (Italy, Finland and Austria) and the United States only, and none
from developing countries.
31
Table 3-3 Overview of BCSSs reviewed/evaluated in this chapter Name of BCSS,
implementation country
and main reference
Description of main functionality and objective
IPET, Italy,
(Meloni et al, 2014)
IPET is an Individual Persuasive Eco-Travel Technology. IPET is not, strictly
speaking, a BCSS but a technological platform for large-scale implementation of
voluntary travel behavior change program. The BCSS (or the mobile application) is
called Activity Locator. The system collects relevant trip information, converts the
data into an activity-travel diary and provides, through mail or a website, a
personalized travel plan in place of car.
MatkaHupi, Finland,
(Jylhä et al, 2013)
MatkaHupi automatically tracks the carbon emissions of the transportation modes
and uses this information to recommend the traveler a set of challenges, such as
“Reduce this week’s CO2 by 10%” or “Walk 3km”.
QT, United States,
(Jariyasunant et al, 2015)
QT (Quantified Traveler) is a computational travel feedback system that aims to
change mode or trip choice, without need for traditional travel counselors. The
system automatically collects trip data, converts them into a travel diary and gives
quantitative feedback (time and money spent, calories burned and CO2 emitted) to the
traveler.
Peacox, Austria,
(Schrammel et al, 2013)
PEACOX, or Persuasive Advisor for CO2-reducing cross-model trip planning, is a
multi-modal navigation smartphone application that aims to help users travel with
lower carbon impact. Suggestions can be “We estimated that you could walk 29% of
your car – and 41% of your PT trips to save more CO2” or “Improve your Rank by
reducing your CO2 Consumption. We recommend walking instead of using your
car/PT for short trips.”
Tripzoom, Netherlands,
(Broll et al, 2012)
Tripzoom is a mobile application that constructs individual mobility profiles and
patterns from mobile sensing. Based on this data, it then encourages the users to
change their behavior by offering appropriate incentives and providing feedback from
the community.
SuperHub, Italy,
(Carreras et al, 2012)
SUPERHUB (SUstainable and PERsuasive Human Users moBility in future cities) is
a mobile application and open source platform that aims to raise in citizens a personal
awareness of the carbon impact of their daily mobility, thus fostering a more
environmentally-friendly behavior. As an open-source platform, it collects, mines
and aggregates data from a variety of mobility sources/providers, then builds eco-
friendly and suitable multi-modal journey plans for citizens.
i-Tour, Italy,
(Magliocchetti et al, 2011)
i-Tour, or intelligent Transport system for Optimised Urban Trips, is a personal
mobility assistant that aims to promote sustainable travel choices. To encourage use
32
of public transport, the system supports routing across a multimodal transport
network.
PEIR, United States,
(Mun et al, 2009)
PEIR, the Personal Environmental Impact Report, is an application that automatically
calculates and provides estimates of one’s environmental impact (carbon and
sensitive site impact) and exposure (smog and fast food exposure).
Ubigreen, United States,
(Froehlich et al, 2009)
UbiGreen Transportation Display uses iconic feedback (tree or polar bear) and
ambient changes in the graphics of the user’s mobile phones to provide awareness
about green mobility behavior. Rewards can be earned by taking sustainable
transportation.
3.3.2. Persuasiveness Evaluation Using PSD
3.3.2.1. Persuasion Context
The persuaders behind the BCSSs (intent) were either consortia of public and private partners, or
stand-alone universities. All the BCSSs target some behavioral and attitudinal change (change
type). The event contains the use, user and technology contexts. Some BCSSs explicitly identify
the use scenarios (use context) though for all nine target users are assumed to be the population at
large (user context). Further all nine use website and/or smartphones (technology context). The
strategy in the PSD model includes two elements, the message and the route.
In so far as message is concerned, we observe that all the BCSSs use carbon emissions as one of
the attributes for feedback/feedforward information to induce sustainable mobility behavior. We
can observe the increasing popularity of utilizing carbon emission information also from other
transport literature. While there is no conclusive evidence yet regarding the effect of carbon
information on behavior change of a population at large (Avineri and Waygood, 2013), three
experiments conducted by Gaker et al (2011) among UC Berkeley undergraduates suggest that
presenting carbon impact of transport alternatives can influence transport decisions such as mode
choice, and therefore, can potentially be used to promote sustainable transport behavior. In
addition, other attributes are also used in BCSSs: time, cost, calories, environmental impact and
exposure. Time and cost are associated with the utility of transport modes. Calories burned are
33
increasingly being included to promote active transportation as a means to combat the global
problem of obesity. The rest of the attributes are BCSS-specific.
The route of aiming to influence users is both direct and indirect depending on the BCSS: Some
explicitly and directly ask the users to reduce carbon emissions (direct route) whereas others
indirectly do this by simply providing carbon feedback, without explicitly asking the users to
reduce their footprint (indirect route).
3.3.2.2 Persuasive Features
The distinctive persuasive features assimilated in the nine selected mobility BCSSs are described
next. We begin with BCSSs that contain most of the persuasive features: PEACOX, SuperHub,
IPET and Tripzoom.
PEACOX has 7 primary task support, 6 dialogue support and 6 social support features. It is a
multi-modal navigation smartphone application that guides users through a step-by-step process
of route searching (tunneling) and aims to help them travel with lower carbon impact. It is a route
choice planner (rehearsal), which is able to provide prognosis of emitted carbon for each trip
option (simulation). Routes which help users save more emissions are highlighted (suggestion,
reduction). PEACOX also gives other tailored and personalized recommendations (tailoring and
personalization) such as “We estimated that you could walk 29% of your car – and 41% of your
PT trips to save more CO2” or “Improve your Rank by reducing your CO2 Consumption. We
recommend walking instead of using your car/PT for short trips.” PEACOX calls this “shift
potential”. Feedbacks on user’s performance through detailed CO2 and PM10 statistics or a tree
showing over-all CO2 status are also provided (self-monitoring). Users who saved carbon
emissions or completed challenges are praised or given positive feedback (praise) and badges
(rewards). They are also regularly reminded to save carbon emissions (reminder). Users also
have their own personal accounts (similarity), with attractive features (liking). PEACOX allows
comparison of one’s performance with others (social comparison), especially with one’s in-group
(normative influence), and viewing the details of others’ trips (social learning). It also calls for
34
the best to win (competition) or for everyone to cooperate (cooperation). Finally, it displays
statistics leader board (recognition).
SuperHub also includes many support features: 7 primary task support, 6 dialogue support and 5
social support. It seriously profiles the users in order to “tailor, customize and rank mobility offers”
(reduction) so that the solutions are close to the users’ needs (tailoring). It also allows users to
self-monitor their progress through personalized statistics and charts (personalization) or track
their behavior change status or past trips (self-monitoring). Moreover, it supports a multi-modal
multi-criteria journey planner (rehearsal) and allows calculation of CO2-footprints using various
transport modes (simulation). Once the user is close to achieving his goals (or has achieved them),
s/he is congratulated (praise) or given incentives including discounts or concessions on travel or
within local establishments (reward). He is also prompted to set and review his individual,
mobility-related behavioral goals (reminder and suggestion). SuperHub continuously updates
itself over time of user’s preferences in terms of mobility options (similarity). One unique feature
of SuperHub is its persuasive games, which offer the users opportunities to learn more about
sustainability through games (tunneling). Its interface is also visually appealing (liking). Users
can also compare their scores with their friends or others in the community (social comparison,
normative influence, competition, recognition) and share trip plans (social learning).
IPET is embedded with 7 primary task support, 6 dialogue support and 4 social support features.
Under primary task support, IPET guides the users through a process (tunneling), consisting of
three steps: data collection of car users’ actual behavior, construction of activity travel diaries and
provision of a personalized travel plan (PTP). The PTP is a detailed plan which can be used to
rehearse the target sustainable behavior (rehearsal). It also “simplifies the complex process of
considering different alternatives of transport (reduction)” by identifying a prospective sustainable
travel behavior that is “highly customized and based on the individual’s particular needs and
characteristics (tailoring)”. Moreover, the PTP makes immediate comparison of the unsustainable
car and the proposed sustainable alternative modes using four measures: travel time, cost, distance
traveled, and calories burned or carbon emitted (simulation). Users can also “monitor their own
behavior (self-monitoring) which allows them to view their movements and feedback quantities”
in a personal website (personalization). Under dialogue support, IPET first identifies an
35
alternative transport solution (suggestion) and then monitors and compares the actual behavior
with the suggested behavior. If users follow the sustainable advise, they will be congratulated
(praise) and given badges (rewards). If they continue to use car for their trips, they will receive
regret messages, and hence will be continuously reminded to try traveling sustainably (reminder).
In designing the messages and PTP, “persuasive graphics are used” in order to attract the
participants (liking). All the information is displayed on a dedicated personal webpage (similarity).
Finally, under social support, IPET automatically calculates the scores, so that participants can see
the results of their performance and those of others (social learning), compare their scores (social
comparison) and their rank with others (competition). The names of top-scorers can be seen by
other users (public recognition).
Tripzoom embeds 4 primary task support, 6 dialogue support and 5 social support features.
Tripzoom proposes a set of challenges to users (e.g. Take the bike to work, Go for a walk during
your lunch break), which dare them to improve their regular travel behavior (reduction, suggestion,
tailoring). Their performance is then translated in terms of saved money, CO2, health and
collected points, which can serve as feedback (self-monitoring). TripZoom has a “Me” tab, where
detailed information about the user’s mobility profile, including visited places, trails or statistics
is provided (personalization and similarity). Users who master challenges are congratulated
(praise) or given incentives (rewards). There are also illustrations – which can be either positive
or negative – which can remind them of their goals (reminder). Tripzoom also exploits several
social support features. Through the “Community Tab”, one can compare his performance with
that of others (social comparison). Aggregated data and comparison with averages can also give
clues about community norms (normative influence). In the “Friends Tab”, one can share their
travel behavior with friends or obtain concrete and more detailed mobility behavior of others
(social learning) or view ranking among friends (a source of competition and public recognition).
The next three BCSSs we describe have strong support in other features, but have very weak social
support: iTour, MatkaHupi and Ubigreen.
iTour has 6 primary task support and 3 dialogue support features but only 1 social support feature.
iTour is a personal multi-modal travel assistant. It guides users through a route selection search
36
process (tunneling) using various visualization modes, which can be selected to identify the best
route (reduction). The visualization mode is a visually attractive circular graph-like structure
(liking) showing routes to final destination (rehearsal) that is dynamically adjusted depending on
user preferences and other contextual information (tailoring, personalization and similarity), such
as time, distance, cost or emission generated to reach the final destination (simulation). A
recommended route option (providing fastest, most sustainable, shortest or cheapest solution) is
highlighted, while less favorable options are not (suggestion). With iTour, one can also see if others
are traveling by the same leg of a particular route (social facilitation).
MatkaHupi contains 5 primary task support, 5 dialogue support and no social support features. By
means of automatic sensing of behavior and trips taken by the user, MatkaHupi targets behavior
change through challenges that are tailored according to individual behavior (tailoring). The
challenges are simple and direct (reduction), such as “walk 3km, cycle 3k and tram 3km”. Through
an appealing visual feedback (liking), the users can track their progress towards their
goals/challenges (self-monitoring). Moreover, users can also view their trip history
(personalization and similarity). After each detected trip, MatkaHupi checks if the same trip can
be made faster and with lesser emission (simulation), and if so, proposes an alternative route plan
for the future (suggestion). It also works as a journey planner for public transportation (rehearsal).
After the user completes the challenges, he is congratulated (praise) and awarded a badge and
some points (reward).
UbiGreen Transportation Display has 2 primary task support and 5 dialogue support features but
assimilates no social support feature. It is a mobile phone application that provides iconic feedback
about one’s green transportation behavior (self-monitoring). The icon can either be a polar bear
or tree, and the user can select which of the two he can use for his ambient display (personalization).
Ubigreen tracks transport behavior of the user either through automatic sensing or self-report
(similarity), and each time the user rides a green transportation such as bus or train, this green
mode is emphasized with its corresponding benefits (saving money, getting exercise, etc.), thus
serving as a suggestion for his next transport mode (suggestion). Moreover, the user also earns
some small graphical rewards which culminates in a complete tree or bear (praise and reward).
37
Since the ambient wallpaper represents a critical area of persuasion, care is taken to ensure it is
visually attractive (liking).
The last two BCSSs, in addition to the just-described Ubigreen, have weak primary task support:
PEIR and Quantified Traveler.
In PEIR, only 2 primary task support (but 4 dialogue support and 5 social support) features are
present. Personalized estimates of environmental impact (carbon and sensitive site) and exposure
(smog and fast food) are given as feedback (self-monitoring). The feedback can be viewed from
a visually appealing interactive map (liking) and is broken down weekly or daily, as well as via a
time browser, through a personal website (personalization and similarity). If one’s impact and
exposure are low relative to friends, green icons of trees appear; conversely, if they are high, smoky
and smoggy icons appear (praise). These icons can also serve as reminders to travel sustainably
(reminder). Users can also see the performance of others (social learning), and their ranking
relative to their friends, encouraging competition (social comparison, competition and
recognition). Aggregate statistics from friends can also provide information about norms
(normative influence).
Quantified Traveler (QT) assimilates the least number of features: 2 primary task support, 1
dialogue support and 3 social support features. QT gathers data from users and automatically
transforms such raw data into trip diaries and footprints (time, money, calories and CO2 emitted),
which is then presented as personalized feedback to the user (self-monitoring) in his own personal
website (personalization and similarity). QT also exploits social influences: QT users can view
the average performances of other users (social learning), enabling peer comparisons (social
comparison). The average statistics of these peer groups (SF Bay area, US average, Berkeley
students) can provide clues to norms (normative influence).
3.3.2.3 Discussion
Table 3-4 summarizes the presence of support features in the nine evaluated BCSSs.
38
Table 3-4 Presence of Primary Task, Dialogue and Social Support Features in 9 BCSSs Category Persuasive
Feature
PEA-
COX
Super-
Hub
IPET Trip-
Zoom
iTour Matka-
Hupi
Ubi-
green
PEIR QT
Primary
Task
Support
Personalization
(n=9)
● ● ● ● ● ● ● ● ●
Self-monitoring
(n=8)
● ● ● ● ● ● ● ●
Reduction
(n=6)
● ● ● ● ● ●
Tailoring
(n=5)
● ● ● ● ●
Simulation
(n=5)
● ● ● ● ●
Rehearsal
(n=5)
● ● ● ● ●
Tunneling
(n=4)
● ● ● ●
Dialogue
Support
Similarity
(n=9)
● ● ● ● ● ● ● ● ●
Liking
(n=8)
● ● ● ● ● ● ● ●
Praise
(n=7)
● ● ● ● ● ● ●
Suggestion
(n=7)
● ● ● ● ● ● ●
Rewards
(n=6)
● ● ● ● ● ●
Reminders
(n=5)
● ● ● ● ●
Social Role
(n=0)
Social
Support
Social learning
(n=6)
● ● ● ● ● ●
Social
comparison (n=6)
● ● ● ● ● ●
Norm. Influence
(n=5)
● ● ● ● ●
Competition
(n=5)
● ● ● ● ●
Recognition
(n=5)
● ● ● ● ●
39
Social facilitation
(n=1)
●
Cooperation
(n=1)
●
In Table 3-4, under the primary task support category, we observe that personalization (found in
all applications), and self-monitoring (in eight applications) appear to be the most commonly
utilized techniques. Personalization is easily implemented because of the capability of all the
BCSSs to automatically sense travel footprints and present them in the personal accounts of the
users. Self-monitoring enables tracking of behavior or performance with respect to certain
measures (e.g. carbon emission, cost, calories burned, time spent, distance traveled per mode, etc.),
aggregated in a variety of ways, such as by day, week or month. Only iTour does not support self-
monitoring, as this is unnecessary, because this BCSS only promotes sustainable multi-modal
routing options without checking or monitoring the user’s actual behavior.
In the same table, we notice that tunneling is the least utilized feature (n=4). Tunneling means
guiding the user through a process or stages of behavior change. We revisit this point again in our
recommendation and argue that to improve the effectiveness of BCSS, there is a need to
incorporate formal behavior change models, especially those which explicitly take into account
the temporal sequence of stages in a behavioral change process. Tunneling can be used together
with tailoring (in the above table, tailoring is seen to be less utilized as well). BCSSs can guide
users through appropriate stages in a process (tunneling) only if they are sensitive to the users’
profile and characteristics (tailoring).
Moreover, simulation and rehearsal are also less utilized features (n=5 for both). BCSSs which
support rehearsal provide journey or trip plans to the users, while those that support simulation
provide the means of observing the benefit or effect of following such a plan in the form of, say,
prognosis or equivalent points. Later, we argue as a recommendation the need for more BCSSs to
include these features of simulation and rehearsal in order to facilitate an effective behavior change.
40
Under the human-computer dialogue category, similarity and liking are supported in all or most
applications. Persuasive technology places strong importance on the appeal of the design (liking).
Since the applications also have sensing capability, it is easy for them to mimic their users in some
way (similarity). Rewards are implemented in the form of badges, points/icon, discounts or
concessions given as incentives every time the desired behavior is performed. Suggestions are
recommendations to try alternative modes, routes, travel plans or pursue mobility challenges or
goals. Reminders enable the user to continue pursuing their goal or to review them. Praises are
congratulatory remarks as a form of approval for desired behavior. None of the applications
supports a social role (i.e. it does not support communication between users and real agents
offering advice). Lehto and Kukkonen (2011) also found that none of web-based alcohol and
smoking interventions support social role per se.
Finally, the design principles that belong under the social support category are social facilitation,
social comparison, normative influence, social learning, cooperation, competition, and recognition.
Table 3-4 shows that some BCSSs (iTour, MatkaHupi, Ubigreen) do not have any social support
features at all (or only have one). Moreover, social facilitation and cooperation (n=1 for both) are
not present in almost all applications. In our conclusions and recommendations, we also elaborate
further how BCSSs can further enable social facilitation and cooperation, as they have a significant
influence for behavior change.
3.3.3. Assessment of Efficacy Evaluation Studies
In the previous section, we carried out an evaluation of the persuasive features of the BCSSs to
determine to what extent they assimilate the Persuasive Design Model. This gives us an idea of
the “persuasive potential” of the systems. While it is true that for “achieving better outcomes from
BCSSs, they should be designed by using persuasive systems design frameworks and models”
(Oinas-Kukkonen, 2010), separate efficacy evaluation of these persuasive technologies needs to
be carried out.
One of the issues with studies on persuasive sustainability systems (e.g. energy conservation,
responsible resource consumption, etc.) is that they lack user evaluation (Brynjarsdottir et al, 2012).
41
Many of them report system evaluation that addresses proof-of-concept, technicality and usability
concerns, but they do not evaluate intervention-induced behavior change. A few studies have
behavioral change evaluations, but these are run on small sample sizes, short-term field study
durations, with limited evidence of lasting behavioral impact.
Of the nine mobility BCSSs reviewed in this study, seven include system usability evaluations.
However, only three (QT, Peacox, SuperHub) attempt to evaluate the effectiveness of behavioral
change interventions. We are aware that the development of these mobility BCSSs is still on-
going, with design issues as their foremost concern at the moment. We present in the table below
some details of the behavior change evaluation carried out by these three studies.
Table 3-5 Details on Evaluation of Effectiveness of BCSSs
** Only the result of the second trial is reported since the document for the final trial is not yet published.
BCSS
(LOCATION)
QUALITY OF
STUDY DESIGN
INTERVENTION
MEASURE TYPE OUTCOMES
QT (California) Low (cohort-
uncontrolled)
N = 135 subjects;
Duration of 3 weeks
Distance
Pro-environmental
attitudes
Significant: driving distance (-),
awareness (+), attitudes (+),
intention (+)
Not significant: distance of
walk/bike/transit
Peacox
(ITS Vienna)
Low (cohort-
uncontrolled)
N = 24 subjects;
Duration of 8 weeks
Modal change
Pro-environmental
attitudes
Significant: attitude towards
transport modes (+)
Not significant: environmental
concern, modal change
SuperHub**
(Barcelona,
Helsinki, Milan)
Low (cohort-
uncontrolled)
N = 471 subjects;
Duration of 8 weeks
Trips
Carbon emissions
Pro-environmental
attitudes and motivation
Not significant: environmental
attitudes, behavior
42
The three studies are similar in many aspects. All assess not only changes in actual behavior but
also in pro-environmental shifts in attitudes and motivations. Only QT reports a significant change
in travel behavior (driving distance) as a result of the intervention, while Peacox and SuperHub
report no significant change. Since QT’s field study duration is short though (3 weeks), this raises
doubts regarding the long-lasting impact of the intervention. Finally, as all employ cohort
uncontrolled evaluations (e.g. before-after comparison, without adequate control groups), these
must be considered of low methodological quality (Graham-Rowe et al, 2011; Fujii et al, 2009)
and threatens the validity of any result.
It seems that there is, at this point, no possibility for a more robust evaluation of the efficacy of
persuasive sustainability systems, as many are still in their early or ongoing stages of design
development. In contrast, in the health domain, a number of BCSS studies include effectiveness
evaluations, based on sufficient sample size with control groups (e.g. Wang et al, 2014).
3.4. Conclusion and Recommendations
We considered in this review nine mobility BCSSs, and examined the persuasiveness of their
design features using the Persuasive System Design Model. We also assessed the quality of the
study conducted to evaluate the effectiveness of the BCSSs in changing behavior.
In assessing the quality of the studies, we find that of the three studies that include behavior change
evaluations (QT, Peacox and SuperHub), two report no significant shift towards sustainable travel
behavior as a result of the BCSS interventions. No definitive conclusion can be derived from these
studies regarding the effectiveness of the BCSSs though since a robust evaluation was not
conducted in the first place. Their study design employs only (1) a small sample size, partly
because of large drop outs and (2) uncontrolled cohorts (before-after comparison, without any
control group). Although the foremost concern of BCSSs right now is design development, we
recommend proper evaluations of BCSSs using methodologically robust study designs such as
Randomised Controlled Trial, Cluster Randomised Controlled Trial or Controlled Before and
After studies as suggested by Arnott et al (2014) and Graham-Rowe et al (2011).
43
In the absence of proper evaluation studies of the efficacy of mobility BCSSs, it is instructive to
look into studies on health BCSSs, where evaluation studies appear to be more developed. We
find that, if the meta-analysis of health BCSSs is any indication, we can reasonably assume that,
in general, the intervention effect size of BCSSs is significant but small (e.g. Bamberg et al, 2015).
Hence, the challenge for future studies is to develop BCSSs in the domain of travel with greater
intervention effect. In the following, based on our review, we suggest three propositions that can
be pursued in future work, which address the small effect size of BCSSs. These
recommendations will be further supported by empirical evidences from studies in the health
domain.
First, we observe that tunneling and tailoring features are less utilized. Thus, as will be explained
below, we suggest to incorporate a stage-based behavior change model in the design of the BCSSs.
Stage-based models explicitly specify the stages in the process of behavior change that users will
go through leading to the desired behavior. These models can be used to first determine the users’
current stage membership (tailoring) and then to guide them to progress towards the next stages in
the process (tunneling). Second, we also notice that the simulation and rehearsal features are
underutilized. Behavior change in mobility requires formation of an implementation intention,
which is greatly facilitated by a provision of an appropriate travel plan (rehearsal) with its
corresponding benefits (simulation). Third, our persuasive design evaluation reveals that social
facilitation and cooperation are least supported. We therefore suggest, following again findings
from health BCSSs as well as other literature, to incorporate these social support features in the
design of the mobility BCSSs. We discuss each of these recommendations further in the next
sections.
3.4.1. Explicit Reference to a Behavior Change Theory
As pointed out by some authors, many BCSSs were developed using techniques solely drawn from
persuasive technology literature (Klein et al, 2014; Arnott et al, 2014; Bamberg et al, 2015). Too
little effort is given on grounding the BCSSs in an explicit behavior change theory. Among the
mobility BCSSs included in this chapter, we note that only one BCSS, the Quantified Traveler, is
grounded explicitly on a behavior change theory.
44
In the literature, this seems to be commonplace though not only for transport applications. Klein
et al (2014) notes that “intelligent support systems have become increasingly popular for the use
of behavior interventions in recent years, [but] those systems are rarely based on models of
behavior change”. A notable exception to this is the eMate system, an intelligent behavior change
support system for therapy adherence for patients with type 2 diabetes, HIV and cardiovascular
diseases (Klein et al, 2011; 2014). This health BCSS is based on an integrated model of human
behavior change, called Computerized Behavior Intervention or COMBI (discussed below).
Bamberg et al (2015) is also in the process of developing a BCSS of this type in the mobility
domain, called “PrimaKlima Bielefeld”.
Arnott et al (2014) suggests that developing a successful behavior change support system (BCSS)
depends not only on the creative and appropriate implementation of the behavior change
techniques, but also on explicitly grounding it on established theoretical constructs from behavioral
theories. In the health sector, a few meta-analyses seem to support this conclusion (Bamberg et al,
2015). One significant meta-analysis is by Webb et al (2010), which finds that “the extent of a
BCSS’s theoretical foundation [is] positively correlated with its effectiveness.” An example of a
theory-based BCSS demonstrated to be effective through randomized controlled trial is Happy
Ending, a digital program on nicotine withdrawal (Brendryen et al, 2008).
By using a formal behavior theory in the design of a support system, we can “understand the
underlying mechanisms of behavior change and how these mechanisms can be influenced to
establish the desired behavior” (Klein et al, 2014). This is especially important because empirical
evidence increasingly shows that “modality styles” of individuals are very much influenced by
deeply entrenched habits and ingrained lifestyles, and are therefore difficult to change (e.g. Vij et
al 2013). In fact, this may be the reason why travel behavior change programmes to date have
statistically significant but only small intervention effects (Fujii et al, 2009; Möser and Bamberg,
2008). Hence, by carefully understanding the possible mechanisms of behavior change, we can
exploit them to increase the effectiveness of these programmes in changing behavior.
45
In addition, as pointed out by Brynjarsdottir et al (2012) in their critical review of 36 persuasive
sustainability systems, many of these systems design their persuasion around a narrow
understanding of sustainability as mere “optimization of simple [behavioral] metrics”, and
consequently, their common persuasion tactic is simply to “tweak behaviors with the goal of
adjusting actions to be more in line with benchmarks of sustainability”. However, this may
sidestep “difficult lifestyle choices that may in fact be necessary to work toward a more sustainable
society”. By adopting a holistic model of behavior change, we can consider multiple metrics of
sustainability that allow us to see the bigger picture.
QT, the only BCSS explicitly informed by a behavior change theory, is based on the theory of
planned behavior (TPB). One criticism against this theory, however, is that it fails to take into
account the time dimension of behavior change. As other models such as e.g. the “Transtheoretical
Model” suggest, behavior change is a transition through a temporally-ordered sequence of different
stages. Moreover, as Riley et al (2011) point out, static behavior change models, such as TPB,
seem “inadequate to inform mobile intervention development as these interventions become more
interactive and dynamic.” Recently, Bamberg (2013) introduced a cumulative theory that
incorporated the dynamic stage concept, called the Stage Model of Self-Regulated Behavior
Change Theory (SSBC). Because it is a dynamic theory, it may prove suitable in the BCSS
platform. To date, however, except for the aforementioned “PrimaKlima” that appears to be under
development, no mobility BCSS based on this theory has been developed so far to the best of our
knowledge. Other models, such as the Computerized Behavior Intervention or COMBI (Klein et
al, 2011; 2014), may also be suitable to be implemented in the BCSS platform. COMBI is a
computational model of behavior change that integrates constructs from the most influential
theories such as the Transtheoretical Model, Social Cognitive Theory, Theory of Planned Behavior,
Attitude Formation, Self-Regulation Theory, Relapse Prevention Model and Health Belief Model.
3.4.2. Greater Integration with Shared Mobility Services Platform
It is well-known that goal initiation is not sufficient for a successful goal achievement. Formation
of implementation intention is also important (Gollwitzer 1999). This is also true in a successful
mobility behavior change (e.g. Fujii and Taniguchi, 2005). Implementation intention, which
46
entails a plan for when, where and how to implement the target mobility behavior, mediates the
effect of behavioral intention on behavior (Gärling and Fujii, 2002). Mobility BCSSs can facilitate
the formation of implementation intention by exploiting the rehearsal feature, i.e. by providing
people a trip/travel plan which they can use to rehearse the target behavior. However, as the
previous section shows, very few mobility BCSSs have this feature: out of the 9 BCSSs we
examined here, only five support the rehearsal feature in the form of personalized travel plans
(IPET), journey planner (MatkaHupi) or dynamic trip plans (PEACOX, SuperHub and iTour).
The rehearsal feature can be combined with the simulation feature by showing potential outcomes
of the trip plan to the users.
In contrast, in health-based BCSSs, as pointed out by the review of Lehto and Kukkonen (2010),
simulation and rehearsal are rather common features. In these health BCSSs, a typical example of
simulation “was calculating how much calories a specific physical activity burns, or the type and
duration of exercise needed to burn the calories from, e.g., a chocolate bar. The rehearsal feature
was supported by providing workout plans and exercise ideas to the user. As a highlight, [two
BCSSs] provided extensive video-based, customizable workout builders.” In another review of
mobile health applications for physical activity Conroy et al (2014) observed that most of them
have features that “provide instruction on how to perform the behavior” and “model/demonstrate
the behavior”, which support the formation of implementation intention. Moreover, there have
been studies in the health sector showing the effectiveness of implementation intentions in
changing behavior (e.g. Belanger-Gravel et al, 2013). As a consequence, one health BCSS recently
developed for cardiac rehabilitation explicitly included “implementation intention” in the design
approach (Antypas and Wangberg, 2014).
How can mobility BCSSs support further the rehearsal feature? Lately, due to advances in
technology, both hardware and software, multi-modal integration – or the seamless connection of
various modes – is becoming a reality. These technological advances, together with the ubiquity
of internet-enabled smartphones, enable users to plan and organize their trips on a very short notice
or even en-route. For instance, PEACOX, as a journey planner application, is able to provide
cross-modal trip plans to users. Such provision of trip plans is akin to making implementation
plans in the traditional VTBC programs. Tang and Thakuriah (2012) showed that mere provision
47
of real-time bus information, through a system called CTA Bus Tracker, can increase transit
ridership.
An important application are the recent developments in ridesharing, as reviewed by several
authors (e.g. Agatz et al, 2012; Chan and Shaheen, 2012; Siddiqi and Buliung, 2013; Furuhata et
al, 2013). In ridesharing, travelers are grouped together into common trips by car or van (Chan
and Shaheen, 2012). In recent years, as ridesharing becomes more dynamic, i.e. real-time, it is
being considered as one travel demand management (TDM) strategy that can alleviate congestion,
while maintaining an acceptable level of service (Siddiqi and Buliung, 2013). In the United States,
ridesharing is increasingly being discussed as a powerful strategy to reduce congestion, emissions
and fossil fuel dependency.
Dynamic ridesharing is being touted as a promising and attractive alternative to private car usage
because of its potential to provide immediate access to door-to-door transportation. Moreover, it
allows users to share costs on car-usage. Furuhata et al (2013) summarize it best: “Conceptually,
ridesharing is a system that can combine the flexibility and speed of private cars with the reduced
cost of fixed-line systems”. In other words, dynamic real-time ridesharing combines tailored trip
and cost-sharing.
In the coming years, ridesharing is likely to take greater modal share as it tries to include greater
technology interoperability and multi-modal integration (Chan and Shaheen, 2012).
Interoperability here means “allowing open source data sharing among ride-matching companies,
which will enable members to find matches across all databases.” Integration means “seamless
connection of ridesharing with other transportation modes, such as public transit and carsharing.”
Hence, with further technological advances, ridesharing can capture more market share, as it
positions itself as a very attractive alternative to private car use.
In summary, a mobility BCSS is hypothesized to have a greater intervention effect if it includes
provision of a dynamic, cross-modal trip planning tool (rehearsal), together with potential effects
or outcomes (simulation feature). There is no available evidence yet for this, but if empirical data
from the health BCSSs and the increasing adoption of ridesharing in the recent years are any
48
indication – a trend greatly facilitated by the fact that ridesharing has become more dynamic and
cross-modal – we can be confident that a mobility BCSS that supports cross-modal planning in
real time may have greater chances of effecting behavior change.
3.4.3. Social Interaction
Various studies have already shown that social information drawn from one’s social network can
be a potentially powerful tool to trigger sustainable travel behavior change (e.g. Bamberg et al,
2011; Ettema, Arentze, Timmermans, 2011; Abou-Zeid and Ben-Akiva, 2011; Axsen and Kurani,
2012; Kormos et al, 2014; Zhang et al, 2015). This review, in fact, shows that the majority of the
BCSSs have assimilated at least five of the seven social support features. Nonetheless, we also
observe that two of these social support features – social facilitation and cooperation – are largely
missing in most applications.
Social facilitation means facilitating social interaction among the users. In health-based BCSSs,
the most common means of social facilitation are “asynchronous peer discussion forums and
synchronous chat rooms”, for example, discussion forum, peer-to-peer forum, chat room, online
and support community (Lehto and Kukkonen, 2011). In contrast to mobility BCSSs, social
facilitation is a widely used element in health BCSSs (e.g. Lehto and Kukkonen, 2010; Kelders et
al, 2012). Nonetheless, these studies are inconclusive yet on whether social facilitation is effective
in changing behavior (for instance, better adherence to interventions). Even though conclusive
studies are not yet available, we consider social facilitation a promising direction to pursue as this
offers social support, which is identified by Ploderer et al (2014) in their review as one of the key
approaches for future work of BCSSs in general. Social facilitation by means of exchanges in
online and offline communities can provide social support such as “esteem support, intimacy,
companionship and validation” or “material aid as well as informational support like advice and
help in problem-solving” (Ploderer et al, 2014).
In contrast to this, cooperation is largely not supported, not only in mobility BCSSs, but also in
health BCSSs (e.g. Lehto and Kukkonen, 2010; Lehto and Kukkonen, 2011; Kelders et al, 2012)
and therefore provides area for improvement. Ploderer et al (2014), in the same review, present
49
some guidelines on how cooperation can be implemented in BCSSs. They note that many systems
are designed for individual use, though some have partially collective orientation which facilitate
interaction between individuals. They thus suggest developing BCSSs for collective, rather than
the usual single, use. They cite as an example the Tidy Street project (Bird and Rogers, 2010),
which focuses on the average electricity consumption of households in the street, instead of
individual households, in order to foster cooperation and collaboration. In the domain of travel,
users can register in travel behavior change or shared-use mobility programs in groups (e.g. as a
family or co-workers) instead of merely as single participants.
3.5. Summary and Further Research
There are numerous behavior change support systems (BCSS). Most developed appear to be health
applications, but the last ten years also saw the development of BCSSs for travel behavior. Using
the Persuasive Systems Design (PSD), we evaluate the persuasive potential of nine behavior
change support systems designed to promote sustainable travel behavior. We extract the
persuasive features embedded in these support systems and find that tunneling, tailoring, rehearsal,
simulation, social facilitation and cooperation are not widely present. In contrast, in the health
domain, these features, except cooperation, are commonly used. Furthermore, we assess studies
conducted to evaluate the effectiveness of these BCSSs in changing travel behavior and find
indications that effect sizes are mostly small though methodologically robust studies are largely
missing and hence no definitive conclusion can be derived yet. We then propose three suggestions
on research needs and applications for further development, especially addressing the small size
of the intervention effects.
In proposing that smart devices be utilized as medium of intervention for promoting sustainable
behavior change, a caveat must be mentioned. Aside from the expected bias towards a younger
population group, there is some evidence that usage of these devices generates a negative effect
against sustainable behavior. While many studies (e.g. Guo et al, 2015) suggest that usage of such
devices can enrich use of travel time on public transport such as buses, bringing them positive
utility, a study by Julsrud and Denstadli (2017) finds that active users of smart devices (“equipped
travelers”) tend to bear critical (negative) attitudes to public transport. Moreover, the amount of
50
time spent in the usage of ICT may generate more motorized travel, including by car (Hong and
Thakuriah, 2016). This is, however, beyond the scope of our present study and should be addressed
in future research.
51
CHAPTER 4
How Can We Use Stage Models to Inform the Systematic
Development of Computer-Tailored Intervention for Large Scale
Travel Behavior Change?
Chapter Summary. This chapter discusses how stage models (the transtheoretical model and the
stage model of self-regulated behavior change) can be used to systematically inform the
development of travel behavior change interventions at large scale. A review of intervention
studies applying the stage models reveals little guidance from existing literature on the systematic
application of the theory in the design of interventions. Stage-tailoring, a design approach that
matches the interventions to the stages, is the most prominent strategy among the few studies that
do systematically apply the theories, but an alternative approach, called menu-based, is also
introduced, which is grounded on the theoretical conceptualization that stages represent an
underlying continuum of action readiness. As illustration on how a menu-based approach may be
applied, we present Blaze mobility behavior change system.
Keywords: Stage model; intervention development; menu-based intervention; tailoring; behavior
change support system
4.1. Introduction
4.1.1. Effectiveness of tailoring in travel behavior intervention
It is widely claimed that tailored information or personalized messages are more effective than
standard, generic information in encouraging behavioral changes in a wide variety of domains. In
the health domain, for instance, reviews and meta-analyses have consistently reported the
advantages of tailored over non-tailored interventions (Lustria et al, 2009). Tailoring is a “process
for creating individualized communications by gathering and assessing personal data…in order to
52
determine the most appropriate information or strategies to meet that person’s unique needs”
(Rimer and Kreuter, 2006).
In the travel behavior domain, a prominent tailoring strategy in changing travel behavior is the
personal travel planning (PTP), or individualized travel marketing. PTP involves “directly
contacting individuals with the offer of information, assistance, incentives and motivation to
enable them to voluntarily alter their travel choices” (Chatterjee, 2009). In PTP, tailored
information based on the needs of the individual is provided (Meloni et al, 2016). In Japan, PTP
is referred to as travel feedback programs (Fujii and Taniguchi, 2005). Several meta-analyses have
demonstrated the effectiveness of PTPs, with standardized effect size of Cohen’s h = 0.12-0.15
(Möser & Bamberg, 2008; Bamberg and Rees, 2017).
Nonetheless, the strength of PTPs is at the same time its own weakness: personalized or tailored
programs can only be deployed at small scale. Implementing highly customized PTPs on a larger
scale, at population level, is a difficult and challenging task (Meloni et al, 2015).
4.1.2. Computer-tailored interventions
With advances in computing and web technologies, it is now becoming possible to enlarge the
scale of reach of personalized and tailored interventions. Collection of personal information can
be automated through the aid of technology, which can then be processed and used in providing
individualized feedback. Moreover, the web technology allows wider access to the interventions
and presentation of the tailored information in a number of formats and modalities. Luistra et al
(2013) summarize it best: “Web-based, computer-tailored interventions have multiple advantages
over single-mode, static interventions: (1) wider access to expert care and feedback; (2) ability to
toggle between modalities and formats to suit different learning styles and literacy levels; (3)
option to communicate synchronously and asynchronously thus enabling convenient scheduling
of interactions and delivery of reminders and messages; (4) a wide array of interactive components
to enhance user experiences and support skills development, behavior/goal monitoring, and
progress tracking (e.g., e-journals, simulations, games, etc.).”
53
Recently, computational PTPs, namely IPET (Meloni et al, 2015) and Quantified Traveler or QT
(Jariyasunant et al, 2015), have already been implemented with some evidence of success. These
computational systems are simply PTP or TFP implemented on a technology platform, with
operational and functional aspects capable of automating the whole process for large scale
implementation. Although both computational systems are grounded on some behavior change
theories, it is not clear how those theories are explicitly used in the systematic development of the
interventions. There is no explicit mapping between the theoretical constructs and the behavior
change interventions. In other words, although both IPET and QT are theory-inspired, what is
desired are theory-based interventions that use an explicit causal pathway (Michie et al, 2008). An
explicit theory basis is needed for systematic development of interventions.
4.1.3. Stage theory-driven intervention development
In the travel behavior change domain, two theories stand out as the most prominent: the theory of
planned behavior (TPB) (Ajzen, 1991) and the norm-activation theory (Schwartz, 1977). Many
studies have demonstrated their robustness and ability to explain behavior (Bamberg et al, 2011).
Nonetheless, one wonders if they constitute an adequate theoretical framework, especially if one
conceptualizes behavior change as a temporal process through different stages (c.f. Bamberg,
2007).
To address this temporal aspect in the literature, two stage models have been used in understanding
travel behavior change as a transition through stages: the transtheoretical model (TTM; e.g.
Prochaska & Velicer, 1997) and the stage model of self-regulated behavior change (SSBC;
Bamberg, 2013a) (to be discussed in more detail in the next section). An attractive feature of both
TTM and SSBC is that they provide a framework for developing interventions tailored to the
individual’s stage membership. TTM, for instance, in addition to qualitatively characterizing the
different stages, specifies at the same time various stage-specific processes of change that may
trigger transition to advanced stages (Prochaska, DiClemente & Norcross, 1992). These processes
of change are useful in guiding the design of stage-matched interventions.
54
Moreover, both stage models can be used to inform the degree of tailoring that is necessary. SSBC,
for instance, postulates only four stages in the process of change; hence it is sufficient to deploy
only four tailored interventions to target populations. A very high degree of
tailoring/customization is thus not necessary, and for non-computer-tailored interventions this may
even be too costly. This added advantage of stage-based interventions enables its implementation
on a large scale.
Unlike TTM, SSBC also enumerates a number of socio-cognitive constructs underlying stage
transitions. For example, it postulates that transitioning from the first to the second stage is
triggered by formation of a goal intention, which in turn is activated by three constructs: felt moral
obligation to fulfill a personal norm, anticipation of feeling positive emotions, and perceived goal
feasibility. The SSBC framework can therefore be used as a blueprint to guide systematic
intervention development, providing insights on possible links or pathways between the
interventions to the causal determinants of behavior change. This will allow us later on to identify
the elements, components or aspects of the intervention that produced the change (c.f. Ritterband
et al, 2009). Identifying the elements within the intervention that effect the change is essential in
informing and guiding future intervention developments.
4.1.4. Implementation of a stage theory-driven, computer-tailored travel behavior change
interventions
In this chapter, our main objective is to present a conceptualization of a stage theory-driven,
computer-tailored travel behavior change intervention on a technology platform for large scale
implementation. We ask the following research questions:
How can stage models (TTM and SSBC discussed further in Section 4.2) be used to inform
the development of interventions in the travel behavior change domain? Existing literature
is a poor resource for this since many studies, though they claim to ground their
interventions on stage models, fail to use these stage models for systematic intervention
development (Section 4.3). Moreover, the few studies that are systematically based on
stage models recommend stage-tailored approaches, but these require a high degree of
customization that may be too costly from the viewpoint of development.
55
Are there any alternative approaches to stage-tailoring? We argue that there is (Section 4.4),
and we then present a theoretical basis for this approach (Section 4.5), and provide a case
example (Section 4.6). We end with a brief summary (Section 4.7).
4.2. Stage Models Applied in Travel Behavior Domain
4.2.1. Transtheoretical model (TTM)
The TTM consists of five major stages: precontemplation, contemplation, preparation, action, and
maintenance (Prochaska & Velicer, 1997). In the first stage – precontemplation - the individual
performs the (problem) behavior in a habitual way, without much deliberation and intention of
changing it. In the second stage – contemplation – the individual is thinking of changing his/her
behavior within six months, but since he/she overestimates the efforts required, he remains
uncommitted to a goal. During the next stage – contemplation – the person thinks of changing in
the immediate future, typically in the next month, and has taken some small steps to reach his/her
goal. Individuals in the action stage have changed their behavior within the past six months. Those
in the maintenance stage have sustained the behavior change for a long time and are trying to avoid
relapse. A final stage – termination – is sometimes included in the TTM. In this stage, the person
is sure never to relapse to the problem behavior.
4.2.2. Stage model of self-regulated behavior change (SSBC)
The SSBC theory posits that behavioral change is achieved by a transition through a temporal
sequence of four stages: predecision, pre-action, action and post-action (Bamberg, 2013a).
Transition to the next stages is marked by formation of goal, behavioral and implementation
intentions. The formation of goal intention marks the individual’s transition from pre-decisional
to the pre-actional stage. Similarly, the formation of a behavioral (implementation) intention marks
the transition to the actional (post-actional) stage of the behavioral change process. SSBC also
includes stage-specific affective and socio-cognitive constructs. These variables, according to
SSBC, influence the formation of the three intention types (more details in the next chapter).
56
Although SSBC draws from a different theoretical framework, SSBC and TTM have a
considerable overlap (Bamberg, 2007). The predecisional phase of SSBC is similar to the pre-
contemplation and contemplation stages. The preactional phase corresponds to the preparation
stage. The actional phase is the actional stage. And the post-actional phase is the same as the
maintenance stage. In the next section, we provide some studies making use of TTM and SSBC
in designing interventions aimed at changing travel behavior.
4.3. Literature Review of the Operationalization of Stage-Based Interventions
We examine how the two stage models were used to operationalize or inform the development of
interventions. We use the list drawn up by Friman et al (2017) in their integrative review plus two
studies identified by the authors. We provide the reference, description of the intervention, and
remarks on how the stage model is used. In general, we find three patterns of use: (1) assessment
of effectiveness; (2) analytical framework; and (3) explicit and systematic use for intervention
development.
Assessment of effectiveness refers to using the stages of change as a variable by which to measure
the effectiveness of the intervention. An intervention is effective if, post intervention, it is able to
induce the individuals to progress to advanced stages. Analytical framework means using the
model to segment the participants into homogenous groups defined by the stages in the model then
analyze the patterns of behavior of each group. Explicit and systematic use means grounding the
design of interventions based on the model’s theoretical constructs.
Table 4-1 Operationalization of stage models for intervention development Reference Description of intervention Remarks Use of stage model
Cooper (2007) Community-based social marketing
technique, which consists of
providing neighborhood residents
with incentives to try driving less,
raising individual awareness of
alternative travel options, and helping
break the automatic reflex to drive for
all trips.
Stages of change measured,
but TTM did not inform
systematic intervention
development
Assessment of
effectiveness
57
Diniz et al. (2015) Participants are grouped into two
groups. The first group (‘not cycling
to work’) consists of individuals in
the precontemplation, contemplation
and preparation stages. The second
group (‘cycling to work’) consists of
those in the action and maintenance
stages. Different interventions are
given to the two groups. The
intervention consisted of a number of
activities (lectures, video
presentations, games, distribution of
traffic safety checklists, safety
equipment and prizes) targeting
various issues: health and
environmental benefits of cycling and
safe cycling.
TTM was used to develop
intervention strategies
Explicit/ systematic
use
Gatersleben &
Appleton (2007)
A two week-cycling study was done
after a travel to work survey was
conducted
The TTM was not used for
intervention development,
but to examine the views of
commuters in different
stages of change
Analytical
framework
Hemmingsson et al.
(2009)
Participants in the intervention group,
aside from being given standard care,
had an intensive care consisting of: 3
individual counseling sessions with a
physician at baseline, 6 months and
12 months; 2 group counseling
sessions; and a new bicycle. Standard
care consisted of a low-intensity,
pedometer-driven walking
intervention with two 2-h group
counselling sessions at baseline and 6
months.
The paper states that the
principles of behaviour
change were founded on
the Transtheoretical model,
but it is not adequately
described how TTM was
used to develop the
interventions.
Inadequate/ unclear
McKee et al. (2006) Travelling Green, a school-based
active travel project, was
implemented. The classroom teacher
Stages of change measured,
but TTM did not inform
Assessment of
effectiveness
58
and the school children and their
families used a set of written
interactive resources during the
Travelling Green project. Active
travel was integrated into the
curriculum and participants used
interactive travelplanning resources
at home.
systematic intervention
development
Meloni et al. (2013) Personalized travel plan (PTP) TTM not used explicitly for
intervention development.
A stage model (i.e. decision
to change behavior) was
used rather in the analysis:
in particular, a multinomial
logit (MNL) model was
built to identify significant
attributes that may explain
the propensity of the
individuals to change
behavior voluntarily in
stages
Analytical
framework
Molina-Garcia et al.
(2013)
A new public bicycle share program
(PBSP) was introduced. Information
sessions were also given to the
recruited participants.
Stages of change were
measured as an indicator of
effectiveness of the
intervention, but TTM did
not inform systematic
intervention development
Assessment of
effectiveness
Mundorf et al.
(2013)
Two interventions were mentioned:
video and computer-tailored
intervention (CTI).
Video intervention: A four-minute
video was made targeting those in the
precontemplation and contemplation
stages. The video showed transit
riders getting on the bus at an urban
terminal, walkers, and bicyclists on
Stages of change were used
for instrument
development to assess the
impact of campus culture
and policies on commute
patterns;
Assessment of
effectiveness
Explicit/ systematic
use only for 2 stages
(video)
Explicit/ systematic
use on the
development of
59
campus, along with brief interview
clips focusing on the benefits of these
transportation modes, as well as
carpooling. The video emphasized
the pros (benefits) of change, in
particular saving money, getting
exercise, and relaxing, listening to
music, or getting work done on the
commute. The video also emphasized
efficacy in that peers were shown
walking, biking, and riding the bus.
CTI: the CTI proceeded with
alternating assessments and
individualized feedback on
transportation behaviors and tailored
feedback based on TTM constructs.
The CTI feedback concluded with
stage-matched feedback and
transportation tips.
CTI, but in the paper
the description of
these interventions
is limited
Mutrie et al. (2002) A self-help intervention, called
“Walk in to Work Out”, which is
based on the transtheoretical model of
behavior change, is given to
contemplators and preparers.
The pack “Walk in to Work Out”
contains: a booklet with written
interactive materials based on the
transtheoretical model of behaviour
change; and information on: choosing
routes, maintaining personal safety,
shower and safe cycle storage
information, and useful contacts. The
pack also included an activity diary in
the form of a wall chart, a workplace
map, distances from local stations,
The description about the
“Walk in to Work Out”
pack is very limited. Thus,
we cannot deduce how
TTM was used to inform
the development of the
self-help intervention.
Inadequate/ unclear
60
local cycle retailers and outdoor
shops, contacts for relevant
organisations, local maps, and
reflective safety accessories.
Rissel et al. (2010) The Cycling Connecting
Communities (CCC) Project is a
community-based cycling promotion
program that included a range of
community engagement and social
marketing activities, such as
organised bike rides and events,
cycling skills courses, the distribution
of cycling maps of the area and
coverage in the local press.
It is explicitly stated in the
paper that the intervention
program is based on
behavior change theories
like TTM, but it is not clear
how TTM was used exactly
to inform intervention
development.
Inadequate/ unclear
Rose & Marfurt
(2007)
‘Ride to Work Day’ (RTWD), an
annual event, aims to actively
promote riding to and from work,
inform participants about the existing
cycling infrastructure that is available
to them and inform workplaces about
making their workplace more
‘cycling friendly’. Organizers of
RTWD promote this event by emails
& posters/ postcards, and provide
information to participants on how to
ride to work and a map of bicycle
facilities.
TTM did not inform
intervention development.
It was used instead to
assess the effectiveness of
the pilot study in promoting
stage progression
Assessment of
effectiveness
Wen et al (2005) Social marketing (via email, poster
and events) to raise awareness of
active transport, and individualized
marketing to address barriers to
active transport for 64 participants
TTM did not inform
intervention development.
It was used instead to
assess the effectiveness of
the pilot study in promoting
stage progression
Assessment of
effectiveness
Wilson et al. (2016) The Active Lions campaign promoted
Active Transport (AT) to and on a
large university campus for
employees and students. The
It is written explicitly that
the campaign utilized
behavior change theories
like TTM, but it was not
Inadequate/ unclear
61
campaign included local events
promoting AT, a smartphone app, and
social media postings (Facebook,
Twitter) from August 2014 to August
2015.
clear how it was exactly
used.
Bamberg (2013b) Four standardized stage-tailored
dialog modules were used during
phone contact with participants in the
predecisional, preactional, actional,
and postactional stages
SSBC was used explicitly
to inform intervention
development. Four
modules, corresponding to
the four stages, are
developed.
Explicit/ systematic
use
Bamberg et al
(2015)
A behavior change support system
was developed, consisting of four
modules. Participants first chose a
diagnostic statement, and depending
on their answer, they were assigned to
one of the four modules of
intervention.
SSBC was used explicitly
to inform intervention
development. Four
modules, corresponding to
the four stages, are
developed.
Explicit/ systematic
use
Many of the articles mentioned in the review by Friman et al (2017) used the TTM, but mostly to
assess the effectiveness of the intervention, and not to inform any systematic intervention
development. Other articles which claim to have based their intervention development
systematically on TTM do not include an adequate description of the intervention. From the table,
it is evident that that there is a dearth of systematic development of interventions based on the
stage model. Only Diniz et al (2015), Bamberg (2013b) and Bamberg et al (2015) used the stage
model for systematic intervention development, which we will examine more closely in the next
section.
4.4. Discussion: Stage-Tailored and Menu-Based Interventions
4.4.1. Stage-tailored interventions
Stage models suggest that individuals can be assigned to qualitatively different stages depending
on their mindsets, and that behavior change interventions are most effective if they are tailored to
62
match the needs, deficits and cognitions associated with the stages. Bamberg (2013a) claims the
“need to develop stage-tailored intervention modules matching the specific needs of individuals in
the four different stages of behavioral change.” For instance, individuals in the predecisional stage
of the behavior change process postulated by SSBC can be influenced effectively by interventions
aimed at “enhancing problem awareness; increasing acceptance of personal responsibility; making
social norms salient; strengthening perceived ability to change current behavior; and promoting
the formation of a clear and challenging, but not excessive personal change goal” (Bamberg,
2013b). SSBC further claims that asking individuals in this stage to change their behavior with a
direct request may trigger negative reactance. Stage models also propose that delivering stage-
mismatched interventions, or targeting stage-inappropriate determinants of change, may not only
be ineffective but may even be counter-productive.
In Diniz et al (2015), Bamberg (2013b) and Bamberg et al (2015), the intervention approach
consists, in general, of two phases: (a) diagnosis of current stage membership; and (b) delivery of
the stage-matched intervention module.
In Bamberg (2013b), the assessment of the current stage membership is measured using a
questionnaire consisting of six statements. Individuals are asked to choose one among the 6
statements that best describes their level of car use: (Ia) At the moment, I use the motor car for
most of my trips. I am happy with my current level of motor-car use and see no reason why I
should reduce it. (Ib) At the moment, I still use the motor car for most of my trips. I would like to
reduce my current level of motor-car use, but, at the moment, I feel it would be impossible for me
to do so. (II) At the moment, I use the motor car for most of my trips. I am currently thinking about
changing some or all of these trips to non-motor-car modes, but at the moment I am not sure about
how I can replace these motor-car trips, or when I should do so. (III) At the moment, I use the
motor car for most of my trips, but my aim is to reduce my current level of motor-car use. I already
know which trips I shall replace and which alternative transport mode I shall use, but, as yet, I
have not actually put this into practice. (IV) Because I am aware of the many problems associated
with motor-car use, I already try to use non-motor-car modes as much as possible. I shall maintain
or even reduce my already low level of motor-car use over the next few months. (V) Because I do
not own/have access to a motor car, reducing my level of motor-car use is not a current issue.
63
Those who choose (Ia) or (Ib) are assigned to the first predecisional stage; (II) to the second
preactional stage; (III) to the actional; and (IV) to the post-actional stage. Choosing statement (V)
means the person is a captive non-car user. Bamberg et al (2015) use the same stage measure,
with slight variation.
In Diniz et al (2015), the stage membership is evaluated by the following question: “Do you cycle
to and from work?” The possible answers were a) I have been cycling to work for more than 6
months (maintenance); b) I have been cycling for less than 6 months (action); c) I do not cycle but
intend to do so in the next 30 days (preparation); d) I do not cycle, but intend to do so in the next
6 months (contemplation); and e) I do not cycle nor intend to do so in the next 6 months
(precontemplation). A dichotomous outcome was then considered that determined whether the
person cycles to work or not: those who answered (a) or (b) were classified as cycling to work and
those who chose otherwise were categorized as not cycling to work.
After the stage membership diagnosis, individuals are then given stage-matched intervention. In
delivering the stage-tailored interventions, Diniz et al (2015) used various delivery modes (Table
4-2). On the other hand, Bamberg (2013b) made phone calls to the participants, while Bamberg
et al (2015) used a behavior change support system (i.e. web application) (Table 4-3). We present
in the tables below the main elements of the content of the intervention modules used.
Table 4-2 Operationalization of TTM for intervention development in Diniz et al (2015)
Stage Activities
Precontemplation
Contemplation
Preparation
(Not cycling to work group)
Theme exposition through dialogue and group formation activity (for
raising consciousness)
Lectures (e.g. informative booklets and testimonies of individuals who used
to use cycling as an active means of transportation)
Processing activity of advantages and disadvantages of using a bicycle
64
Experimental activities (e.g. the individual doesn’t have to go to work by
bike every day, but he/she can, for instance, commit to doing it whenever
the weather is nice or to cycling on weekends)
Ludic games; Bike circuit (to enable people see the positive effects of
cycling to the environment)
Action
Maintenance
(Cycling to work group)
Theme exposition through dialogue, group formation activity and
cooperative games (for raising consciousness)
Games using people as markers; Bike circuit–friendly bike (to allow the
individuals to experience using the bicycle with consciousness and safety)
Bike club; video testimonials and activity with challenging game; Raffle of
bicycle equipment
Table 4-3 Operationalization of SSBC for intervention development in Bamberg (2013b) and
Bamberg et al (2015)
Stage Bamberg (2013b) Bamberg et al (2015)
Predecisional The dialog module begins by raising the
issue of climate protection, which is a task
that the majority of the people are willing to
personally contribute to by changing an
aspect of their own mobility behavior. The
individual is then invited to make his own
(small) personal contribution.
Then it asks by identifying 1-2 car trips that
can be replaced by alternative modes, and
to think of advantages/disadvantages and
barriers to switching.
Then it recommends carrying out a one-off
test of the alternative.
The web module brings to the individual’s
attention his lack of awareness concerning
behavior or habits that are environmentally
harmful.
Then behavioral alternatives, together with their
positive aspects, are given as suggestions, at the
same time emphasizing that carrying them out is
a lot easier. The individual is then invited to
“try”.
A typology of lifestyles is also used
65
Preactional The dialogue module begins by
complimenting the individual for his
interest to try alternatives.
Then it asks to specify trips made by car but
that can be replaced by alternatives and to
think of advantages or disadvantages.
Then, even more specifically, it asks the
individual to think of one particular trip X.
A follow-up question is then asked if the
person is willing to test making the trip by
an alternative mode. If cycling/ walking is
chosen, a proposal will be sent to the person
(including cases of bad weather). If public
transportation is chosen as a test mode, a
test ticket will be given as well.
An overview of potential behavioral alternatives
(walking, bicycling, public transportation, car-
sharing, etc.) and some details about each option
(for example, for cycling, routes and travel time)
are presented
Costs and carbon emissions (or carbon savings)
associated with each of the alternatives are also
given
Potential obstacles/barriers that could arise when
switching from car use to the chosen mobility
alternative, and specific solutions to overcome
each of these problems, are considered
The individual is then asked to rate the subjective
strength of his intention to actually try out the
chosen mobility alternative instead of using the
car.
Actional The dialog praises the individual for the
desire to reduce motor car use.
Then it asks the person to identify the trips
that can be done by alternative modes, and
to think of benefits/barriers of doing so.
Then the person is asked to prepare a
specific plan to achieve the goal using a
particular alternative mode: day, time,
route, trip, etc.
The individual first receives compliments for his
intention to switch from car. Then he is asked to
choose one of the behavioral alternatives, and to
write down potential obstacles or barriers and to
generate solutions to overcome these problems.
He then gives a rating to the subjective strength
of his current intention. If the rating is high, he
is asked to write down a precise implementation
intention, i.e., to specify the exact date, time and
context of the new behavior
Postactional A compliment is given to the individual
who regularly uses alternatives, then an
offer of a token – a free pass for a day.
Then a suggestion is made to purchase a
season ticket.
The individual is complimented for his good
behavior and receives a symbolic token.
66
In Table 4-3, content analysis of the interventions reveals that the predecisional module places a
strong emphasis on triggering self-doubt within the individual regarding his current lifestyle and
on helping him bind himself to a personal goal of contributing to climate protection by reducing
car use. The preactional module, on the other hand, focuses on helping the individual select a
specific behavioral alternative to car use. Here, benefits/dis-benefits are presented for each
alternative so the person can choose the best mode that suits him. The actional module aims to
help the individual prepare and implement an action plan, consisting of day, time, route, mode, etc.
Finally, the postactional module tries to help the individual sustain his good behavior and avoid
relapse.
Although the intervention modules contain characteristic features associated strongly with specific
stages, we also observe that same elements are found in more than one stage, albeit with differential
emphasis. For example, in Bamberg et al (2015), the presentation of “behavioral alternatives” is
found in predecisional, preactional and actional stages. As another example, the individual in the
predecisional stage is asked to carry out a one-off test of the alternative, or to try a more
environmentally friendly kind of mobility. In the preactional stage, the same request is also made
to test making the trip by an alternative mode. In the actional stage, the individual is asked to
make an implementation plan, down to the very specific details. In other words, the same request
of trying out the desired behavior is asked from the individual, regardless of his stage membership,
albeit in varying degrees of subtlety or directness. The same observations can be said about Table
4-2, where there is a complete overlap of interventions among the first three stages, as well as
identical interventions for the last two stages. There is also overlap among all the five stages of
TTM such as the activity theme exposition through dialogue and group formation (for raising
consciousness).
4.4.2. Menu-based interventions
The fact that similar elements are found in multiple stage-specific interventions suggests that
certain processes of change are not strictly connected to each stage, and therefore design strategies
may spread over multiple stages (c.f. Ludden and Hekkert, 2014). Given this conceptualization,
67
it can be thought that stage models propose an alternative intervention development approach to
strict stage-tailoring. Abraham (2008) coins this alternative approach the “menu-based
intervention”.
In menu-based intervention approach, “rather than preintervention screening followed by separate
stage-matched interventions for different groups, one could provide generic interventions that
allow individual selection of intervention techniques corresponding to individual action readiness”
(Abraham, 2008). In other words, a menu of interventions is provided to all individuals, from
which he can select only those that are most relevant to him using self-assessments. It is possible
that the same intervention in the menu is meaningful to individuals in different stages (c.f. Ludden
and Hekkert, 2014). In Section 4.6, we provide a case example of how menu-based intervention
was implemented in the web platform in the domain of travel behavior change.
4.5. Theoretical Basis: Conceptualizing Stages as Representing Partitions of an Underlying
Continuum of Action Readiness
Stage models posit that the behavior change process is a transition through a sequence of
qualitatively different stages. Each stage is associated with its own set of determinants or
predictors. Although some evidence points to the observation of discontinuity patterns in the
different determinants, which may confirm the existence of discrete stages (e.g. Lippke,
Ziegelmann and Schwarzer, 2005), a competing interpretation is that these stages are simply
divisions of an underlying continuum (Abraham, 2008).
Take the SSBC for example. Though SSBC identifies four distinct stages, each with its own set
of determinants, the demarcation among these stages can be fuzzy, unstable and less clear-cut (c.f.
Abraham, 2008). Hence, behavior change can be theorized not so much as a transition through
distinct stages, but as progress along a “continuum of action readiness” (c.f. Abraham, 2008). This
continuum, generated by multiple determinants interacting, runs through the postulated four stages
of SSBC. An individual’s position along this continuum may shift with changes in the
determinants.
68
Given this conceptualization, “multi-target menu-based interventions” (Abraham, 2008) that
simultaneously target cognitions associated with all stages, albeit allowing differential focus
depending on the position of the individual in the action-readiness continuum, are most appropriate.
4.6. Case Study: Blaze Mobility Behavior Change Support System
As a case study, we present how a menu-based intervention approach can be used in the
development of a mobility behavior change support system. We call this support system Blaze.
A fuller description of Blaze will be presented in the next chapter. In this section, we simply
provide a conceptualization of how a menu-based approach can be implemented.
The development of Blaze is grounded explicitly on a behavior change theory, the Stage Model of
Self-Regulated Behavioral Change (SSBC). Blaze is implemented as a mobility behavior change
support system, consisting of web and Smartphone applications (the Smartphone app is almost a
clone of the web version). In designing SSBC-based interventions, the proximal causal
determinants directly triggering transitions to the next stages, as postulated by the theory, are
activated. A stage-tailored feedback intervention, directing the attention of the individual to
specific tasks for progress through stages to occur, aims to accomplish this. However, this does
not mean that only these proximal variables are targeted, while the other distal variables are
entirely ignored. In Blaze, apart from the stage-matched intervention, we also subject the users to
various standard interventions that simultaneously target all the variables in the model, albeit in
differential degrees of emphasis depending on the current membership stage of the user. We design
the standard interventions so that the same intervention could mean and work differently for
individuals in different stages (c.f. Ludden and Hekkert, 2014). Hence, after a brief self-assessment
during which the individual is provided a diagnostic report, he is given a menu of intervention,
from which the individual may select only those interventions deemed meaningful. In menu-based
intervention, the same (generic) menu is provided to all, but this does not constitute a one-size
approach, since the individual, through the diagnostic report, is directed to components in the menu
that are most relevant to him, allowing one to self-tailor.
69
Figure 4-1. Content or components of the predecisional report: (a) Response of the individual to prior questions
(topmost); (b) Recommendation: goal suggestion based on previous response (c) Blaze support: relevant pages in
Blaze that are considered meaningful for the person in this stage; (d) Motivational statements such as quick trivia
and what others say.
Just like in Bamberg (2013b), the individual first takes a self-assessment to determine his stage
membership. Afterwards, he is given a stage-tailored diagnostic report, depending on his stage
membership (see Fig. 4-1 for sample report for predecisional individuals). This is the only stage-
tailored feedback intervention in Blaze. After this brief self-diagnosis, the individual is directed
to the website, which shows a menu of different interventions.
The website version of Blaze has six main tabs: Home, Ride Plan, Summary, Dashboard,
Performance and Diary (see screenshot figure of Map in Table 4-4). Each tab is associated with a
particular intervention. Home implements social traces; Ride plan enables formation of an
implementation plan; Summary and Diary allow reflection-on-action; and Performance facilitates
goal-setting. Dashboard, a non-intervention, is simply the tab for trip logging where the user can
input or change his trip data.
70
In Blaze, we implement social traces (in Home tab) mainly as a map showing the points of origin
of the students, who all travel to the same destination (a university). For privacy reasons, the points
of origin are not shown as exact locations, but are aggregated to barangay level (smallest political
administrative unit in the Philippines). The barangays come in various shades of three colors (red,
yellow, green). If a higher proportion of students from a barangay commute to school using
alternative modes, the barangay is tagged with (darker shade of) green. Barangays with a higher
percentage of car-driving students are tagged with (darker shade of) red. Barangays tagged with
yellow have equal proportion of car and alternative mode use.
Moreover, in Blaze, we trigger reflection-on-action on personal mode use (in Summary and Diary
tabs). We provide two types of data representations to engage the person in reflection: the first is
a summary page of the person’s mode usage (latest, frequent and low-carbon), and the second is a
graph of one’s modal split. Reflection is mainly activated through comparisons (latest and frequent
versus low-carbon; one’s own mode use versus relevant others).
Furthermore, to promote weekly goal-setting (in Performance tab), we ask the individual to report
his stage membership at the end of each week. These self-reports are then plotted in a graph, which
can implicitly encourage him to make progress to the next stages in the following weeks. Apart
from goal-setting, we also urge the individual to implement his goal by promoting ridesharing and
trip planning online (Ride Plan tab). Blaze will be described in more detail in the next Chapter.
Table 4-4 Ten Standard Interventions within the Blaze “menu”
Element Screenshot Description
Map
Map showing the points of origin
of the students, who all travel to
the same destination (a university),
aggregated to the barangay level
71
Goal Setting
At the end of each week, the user
chooses a goal to pursue in the
next week
Summary
Summary page of the person’s
mode usage (latest, frequent and
low-carbon)
Readiness to
Blaze
Each week, the individual sends a
self-report of his stage
membership. These self-reports
are then plotted in a graph, which
can implicitly encourage him to
make progress to the higher stages
in the following weeks. The graph
thus shows changes in his
readiness to blaze (i.e. in taking
alternative modes).
Travel Time
Graph of the individual’s travel
time (going to school only) with
social comparisons
72
Mode Use
Graph of the individual’s modal
split with social comparison
Slideshow
(motivational
messages)
20 different images, displayed as
slideshow, subtly motivating the
individual to take action
Ride plan
Ridesharing and trip planning
Mode Options
Modes (in percentage) the students
in the same barangay use to travel
to and from the university
73
Diagnostic report
After the individual completes a
brief self-assessment, he receives a
stage-matched diagnostic report
4.7. Summary
We give an overview in this chapter how two stage models, the transtheoretical model and the
stage model of self-regulated behavior change, have been used to guide the development of
interventions in the domain of travel behavior change. A review of 15 intervention studies
informed by these two stage models reveals that, save three studies, there is a lack of systematic
application of the theory in the design of interventions. Content analysis of the interventions found
in these three studies shows that, though the intervention modules contain characteristic features
associated strongly with specific stages, same intervention elements are also found in more than
one stage. This indicates that we can conceive an alternative to strict stage-tailoring approach to
intervention development, called menu-based intervention, which assumes an underlying
continuum of action readiness. As case study, we present Blaze mobility behavior change system,
which applies this menu-based approach to intervention development.
74
CHAPTER 5
Development and Usability Evaluation of Blaze Information System
for Promoting Sustainable Travel Behavior in Metro Manila
Chapter Summary. A number of programs aimed at promoting sustainable travel behavior have
already been developed. These programs are implemented and delivered through various
communication media. However, their development lacks an explicit theoretical foundation. Here
we describe how the Stage Model of Self-Regulated Behavioral Change (SSBC) is used for the
first time in the development of Blaze, consisting of Smartphone and web application. SSBC posits
that behavioral change is achieved by a transition through a temporal sequence of different stages.
Using SSBC, we systematically develop theory-based interventions. We evaluate the usability and
potential effectiveness of Blaze among university students in Metro Manila, Philippines. Our
results suggest that Blaze is relatively user-friendly and can provoke reflection about personal
travel behavior. We also report a 4.6% reduction in car use. Some students have also progressed
to more advanced stages in the behavior change process. Blaze thus has great potential in inducing
travel behavior change.
Keywords: Voluntary Travel Behavior Change, Sustainability, Behavior Change Support System,
Smartphone, Persuasive Technology, Stage Model
5.1. Introduction The many environmental, economic and social problems associated with increasing levels of car
use have already been sufficiently noted and documented in literature (e.g. Fujii et al, 2009). Policy
interventions that aim at reducing car use can be characterised broadly as structural/hard or
psychological/soft (Bamberg et al, 2011; Graham-Rowe et al, 2011). Structural interventions
involve modification of the structures surrounding travel behavior through physical and/or
legislative measures (e.g. road pricing and bus priority lanes). Psychological interventions are
designed to modify perceptions, beliefs and attitudes (Graham-Rowe et al, 2011). Both
interventions can be effective. However, in cases where infrastructure investments are limited and
75
pricing policies are politically sensitive, soft interventions may be the only viable option.
Increasing policy importance of soft measures can be seen in growing interest in the so-called
voluntary travel behavior change (VTBC) programs (Ampt, 2003).
An example of VTBC program includes the so-called “Travel Feedback Programs” (TFP) in Japan
(Fujii & Taniguchi, 2005). In a typical TFP, participants of the program are given feedback based
on their reported travel behavior with the aim of modifying their behavior without coercion. TFPs
come in several types, differentiated by the location, techniques, procedures and communication
media used in their implementation (Fujii and Taniguchi, 2006). With respect to communication
media, TFPs have relied so far only on traditional technologies, namely face-to-face
communication, regular mail, telephone and email (Fujii & Taniguchi, 2006). This severely limits
the potential of classical TFPs for scaling up (Jariyasunant et al, 2015). Other VTBC programs
that rely on traditional media are TravelSmart (Seethaler & Rose, 2006), SmarterChoices (Cairns
et al, 2008), and IndiMark (Brög et al, 2002).
Recently, VTBC programs have begun to be implemented as mobility information systems using
advanced technology platforms. An example is Quantified Traveler (Jariyasunant et al, 2015),
which is simply a TFP implemented on the web and Smartphone platform. Other examples include
Figure 5-1 VTBC Programs
76
Ubigreen Transportation Display (Froehlich et al, 2009), Peacox (Schrammel et al, 2013), IPET
(Meloni et al, 2014) and MatkaHupi (Jylhä et al, 2013). Since they are supported by advanced
communication media, such as the mobile platform, they have the potential to deliver large-scale
and cost-effective behavior-change interventions (Meloni et al, 2015).
Nonetheless, a general lack of theoretical grounding of a number of these VTBC programs has
been noted in literature (Richter et al, 2010; also discussed in Chapter 3). These programs were
developed using techniques solely drawn from persuasive literature while too little effort is given
on grounding them in an explicit behavior change theory. To the best of our knowledge, among
the current VTBC programs, only three – namely, TFP, Quantified Traveler, and PrimaKlima –
are explicitly supported by a behavior change theory. TFP and Quantified Traveler are explicitly
grounded on the Theory of Planned Behavior or TPB (Ajzen, 1991). Nonetheless, theories like
TPB represent “continuum models” of behavior change that typically imply a “one-size-fits-all”
intervention approach. In contrast, stage models conceptualize behavior change as a progress
through a series of distinct stages, and interventions tailored for each stage are considered most
effective in changing behavior (Schwarzer, 2008). Only PrimaKlima (Bamberg et al, 2015) is
based on a stage model.
In this chapter, we describe the efforts put into the development of a voluntary travel change
programme that is (1) grounded on a stage model; and (2) supported by advanced communication
technology. Our VTBC programme, called Blaze, is based on the Stage Model of Self-Regulated
Behavioral Change (SSBC) theory (Bamberg, 2013a). Moreover, it is implemented on a
Smartphone- and Web- platform; hence, it can also be called a “behavior change support system”
(Oinas-Kukkonen, 2010; Oinas-Kukkonen, 2013). Fig. 5-1 shows the positioning of Blaze vis-à-
vis other VTBC programs.
The structure of the remainder of this chapter is as follows: Section 5.2 explains the SSBC in more
detail. Here we describe the cognitive/social variables postulated by SSBC. In Section 5.3, we
discuss the principles used to develop interventions based on SSBC. Here we introduce the
77
concept of “menu-based interventions” (c.f. Abraham, 2008). In Sections 5.4 and 5.5, we specify
the stage-specific and standard interventions used to influence these variables assumed by SSBC.
Section 5.6 presents the results of the usability testing we conducted. We summarize and conclude
in the final section.
5.2. Stage Model of Self-Regulated Behavior Change (SSBC) Theory
Fig. 5-2 shows the four stages of behavioral change (predecisional, pre-actional, actional and post-
actional) and the stage-specific affective and social-cognitive constructs. These variables,
according to SSBC, influence the formation of the three intention types necessary for transition to
the next stages (the three critical transition points): goal intention, behavioral intention and
implementation intention. The theoretical model represents an explicit causal pathway underlying
the process of behavior change.
In the pre-decisional stage, the individual performs the environmentally harmful behavior in a
habitual way, without much deliberation. SSBC specifies the social-cognitive variables that
motivate the individual to bind himself to a goal, i.e. to form a goal intention. When a person
becomes aware of the negative consequences of his current behavior and accepts personal
responsibility for causing harm, he may feel negative emotions. These negative moral feelings may
activate his personal norm and urge him to act in accordance with his personal moral standards.
Figure 5-2 Stage Model of Self-Regulated Behavior Change (Bamberg, 2013a)
78
At the same time, the ascription of personal responsibility may give rise to concerns about
expectations of important social referents (salient social norms). Fear of their social disapproval
may further activate his personal norm. The felt moral obligation to fulfill the personal norm may
elicit positive emotions, which, together with personal norm, urge him to form a goal. Nonetheless,
whether he actually binds himself to a goal depends on perceived goal feasibility. The formation
of goal intention marks the individual’s transition to the second (pre-actional) stage.
In the pre-actional stage, the individual has the general goal of changing his current behavior (high
goal intention); however, he does not yet have a personally most suitable means to achieve this
goal. Because several actions can lead to the same goal, the task that the individual is confronted
with is to select the most personally suitable behavioral strategy among various options by
weighing their pros and cons. SSBC assumes that goal intention and two other factors influence
the formation of a behavioral intention. These factors are attitudes toward behavioral alternatives
and perceived behavioral control over behavioral strategies. The formation of a behavioral
intention marks the transition to the third (actional) stage of the behavioral change process.
In the actional stage, the individual has strong goal and behavioral intentions; however, he has not
yet put his behavioral strategy to practice. The task therefore for the individual is to actually
implement the behavioral strategy chosen for goal achievement (i.e. he must form an
implementation intention). SSBC assumes that behavioral intention and three additional factors –
action planning, coping planning and maintenance self-efficacy – promote the formation of an
implementation intention. Action planning refers to the specific situation parameters, e.g. “when”
and “where”, and the sequence of actions necessary to implement the target behavior, i.e. “how”.
Coping planning refers to the ability to anticipate situations that may hinder the individual to
perform the target behavior and to develop a plan to cope with such situations. Maintenance self-
efficacy refers to the person’s confidence in his ability to maintain the difficult behavior. The
formation of an implementation intention marks the transition to the third and final stage, the post-
actional stage.
79
In the post-actional stage, the individual has successfully implemented the new behavior.
Therefore, the task he is confronted with is to maintain his behavior and avoid relapse into old
behavioral patterns (Bamberg, 2013a). Aside from the well-formed implementation intention,
SSBC views “a person’s confidence in being capable of resuming a difficult behavior after a
relapse (so-called recovery self-efficacy) as an additional factor promoting the maintenance of the
new behavior” (Bamberg, 2013a).
Bamberg (2013a) and Klöckner (2014) have reported that while some individuals progress through
these four stages stages in a linear pattern, change can also occur in spiral or non-linear fashion.
In other words, individuals can cycle and recycle back and forth through the stages or regress to
earlier stages from later ones.
5.3. Stage-Tailored and Menu-based Interventions
As discussed in the previous chapter (c.f. Chapter 4.4), in designing SSBC-based interventions,
the proximal causal determinants directly influencing the formation of a particular intention (goal,
behavioral, implementation) must be activated in order for the individual to transition to the next
stage. A stage-tailored feedback intervention, directing the attention of the individual to specific
tasks for progress through stages to occur, aims to accomplish this. However, this does not mean
that only these proximal variables are targeted, while the other distal variables are entirely ignored.
Though SSBC identifies four distinct stages, each with its own set of determinants, the
demarcation among these stages can be fuzzy, unstable and less clear-cut (c.f. Abraham, 2008).
Hence, behavior change can be theorized not so much as a transition through distinct stages, but
as progress along a “continuum of action readiness” (c.f. Abraham, 2008). This continuum,
generated by multiple determinants interacting, runs through the postulated four stages of SSBC.
Given this conceptualization, “multi-target menu-based interventions” (Abraham, 2008) that
simultaneously target cognitions associated with all stages, albeit allowing differential focus
depending on the position of the individual in the action-readiness continuum, are most appropriate.
In Blaze, apart from the stage-matched intervention, we also subject the users to various standard
80
interventions that simultaneously target all the variables in the model, albeit in differential degrees
of emphasis depending on the current membership stage of the user. We design the standard
interventions so that the same intervention could mean and work differently for individuals in
different stages (c.f. Ludden and Hekkert, 2014). Hence, after a brief self-assessment (Section
5.4.1) during which the individual is provided a diagnostic report (see Figs. 5-3 to 5-5), he is given
a menu of intervention (Section 5.5), from which the individual may select. In menu-based
intervention, the same (generic) menu is provided to all, but this does not constitute a one-size
approach, since the individual, through the diagnostic report, is directed to components in the menu
that are most relevant to him, allowing one to self-tailor.
5.4. Stage-Specific Intervention Modules
We develop Blaze to be implemented among students of the Ateneo de Manila University (Quezon
City, Metro Manila, Philippines). The name “Blaze” comes from “trailblazers”, individuals who
show the way or new behavior to a larger population. We choose to implement it in a university to
take advantage of the fact that all the students in the university have the same destination, and
many of them also have the same origin. Moreover, their sense of belongingness to the same in-
group can facilitate social learning. Finally, “public transportation infrastructure may be more
facilitative of commuting trips to the university than to other destinations in the city, making the
switch to alternative transport for commuting trips an easier or more convenient choice than for
non-commuting trips” (Kormos et al, 2014). For instance, various carpooling, ridesharing and
shuttle services are provided. In 2016, the Ateneo launched the Ateneo P2P+, a shuttle service
which picks up and drops off members of the university at strategic hubs. In addition, the university
also offers carpool services and encourages students to organize their own carpool groups.
In BLAZE, as in many other travel plans, our aim is, first of all, to persuade the traveler to use
alternative transport modes or services that result in higher per vehicle occupancy (the “new habit”)
(Enoch and Zhang, 2008). In many cases, this means reducing use of cars or frequency of drive-
alone trips, and instead taking shared rides, including public transport.
81
SSBC, however, focuses not only on the observable behavioral change, but also on the latent
psychological change, which can be measured via change – better progression – in stage
membership or in the set of social/cognitive variables. Hence, if open behavioral change among
the travelers is hard to achieve, our aim will be to influence them to progress to later stages (e.g.
from predecisional to pre-actional).
We now describe the stage-tailored interventions designed to promote car use reduction and the
theoretical assumptions derived from the SSBC underlying its development.
5.4.1. Assessment of current stage membership
First, the participant creates an account, and provides some background information (e.g.
demographic, mobility and home location) about himself. Next, he answers a survey on his typical
mode use, a stage diagnosis tool and a questionnaire measuring a number of latent affective and
social-cognitive variables. The answer of the participant to the stage diagnosis tool determines the
kind of interventions he will receive next. After the participant completes these steps, he receives
a stage-appropriate feedback module (e.g. Figs 5-3 to 5-5).
5.4.2. Intervention module for participants in the pre-decisional stage
The SSBC assumes that people in the pre-decisional stage have the following characteristics: (a)
They perform the problem behavior on a regular, habitual basis; (b) They are not fully aware of
the negative consequences associated with their behavior; (c) They see no reasons for behavioral
change; (d) There is a risk of triggering reactance if they are confronted with a direct request to
change their behavior; and (e) They are convinced that behavioral change is not feasible or at least
not advantageous for them personally. The main intervention goal in this stage is to persuade the
82
individual to form a goal intention, which marks a transition to the pre-actional stage (Bamberg,
2013b).
In the early pre-decisional stage, interventions aimed at initiating a process of reflection about
consequences of behavior and about one’s ability to perform the desired action are chiefly seen as
most important.
In the feedback module, shown in Fig.(5-3), we first try to push the individual toward reflecting
about his current behavior and its possible impact to the environment and our cities. At the same
time, we also promote behavior change but since there is a risk of eliciting reactance if a direct
request is made, we reduce reactance by asking for only a “small” reduction in the person’s
footprint. We also include messages (“quick trivia” and “what others say”) to further motivate the
person. Finally, we also describe how Blaze can support individuals in the pre-decisional stage. In
particular, they can check the Blaze map, summary, diary and performance tabs of the Blaze
system (described in Section 5.5), which we consider as most useful for individuals in this stage
of behavior change process.
Figure 5-4 Pre-actional stage module Figure 5-3 Pre-decisional stage module
83
5.4.3. Intervention module for participants in the pre-actional stage
The SSBC assumes that people in the pre-actional stage have the following characteristics: (a)
They already have the general goal of changing their
current behavior; however, (b) they do not yet have a
personally suitable behavioral strategy to achieve this
goal (Bamberg, 2013a). The main intervention goal in
this stage is to persuade the individual to form a
behavioral intention and select a personally suitable
replacement for car for some of his trips. We promote
behavioral change by asking individuals in this stage
to “explore alternative options” and “find a suitable
replacement for car” (see Fig. (5-4)). Additional
information (“quick trivia” and “what others say”) are
also included to further motivate the person. Finally,
we also describe how Blaze can support individuals in
the pre-actional stage.
5.4.4. Intervention module for participants in the actional stage
The SSBC assumes that people in the actional stage have the following characteristics: (a) They
have strong goal and behavioral intentions; however, (b) they have not yet put their plan to practice
(Bamberg, 2013). The main intervention goal in this stage is to persuade the participants to try
implementing this plan, that is, to translate their plan into action.
In the feedback module (Fig. (5-5)), we encourage individuals in the actional stage to “try to
succeed in actually using (more) public transport or rideshare” within the next seven days. We also
Figure 5-5 Actional Stage Module
84
describe how they can be supported in this task through Blaze and a trip planner called sakay.ph.
Finally, we also include messages (“what others say”) to further encourage them to change.
5.4.5. Intervention module for participants in the post-actional stage and for captive public
transport users
Post-actional individuals and captive public transport users are generally classified as blazers.
Hence, we offer them commendation and praise for their good behavior. For individuals in the
post-actional stage, the main intervention goal is to help them maintain their behavior. Individuals
in this stage and captive public transport users are also given similar feedback modules.
5.5. Standard Interventions
We now describe the standard (generic) interventions in Blaze, common to all users. The website
version of BLAZE has six main tabs, shown in Fig, 5-6: Home, Ride Plan, Summary, Dashboard,
Performance and Diary (the Smartphone app is almost a clone of the web version). Each tab is
associated with a particular intervention. Home implements social traces (described in 5.1); Ride
plan enables formation of an implementation plan (5.2); Summary and Diary allow reflection-on-
action (5.3); and Performance facilitates goal-setting (5.4). Dashboard, a non-intervention, is
simply the tab for trip logging where the user can input or change his trip data. As earlier stated,
standard interventions are received by all users; nonetheless, how these are interpreted or
emphasized differs in each stage (c.f. Ludden and Hekkert, 2014).
5.5.1. Social traces
Social traces are traces or patterns of other users. Ploderer et al (2014) identifies social traces as a
promising approach to behavior change since they can bring a range of benefits. Firstly, they can
provide simple awareness of the issues concerned with behavior change as well as the impact of
collective behavior change. Secondly, they can indicate social norms which, when internalized in
85
time, become personal norms. Third, they can be used to encourage comparison. Comparison “with
less successful others (downward comparisons) in self-relevant domains” can often make
individuals “feel better about themselves and show increased motivation” (Fitzsimons and Finkel,
2010). Fourthly, social traces can provide information regarding “the existence/availability of
alternatives, their characteristics and the utility one derives from them” (Ettema et al, 2011). For
example, in Germany, as the bicycle mode share increases, this may send a signal that “cycling
works”, i.e. it is a feasible mode (Goetze and Rave, 2010). Since the latter three are the immediate
constructs influencing the formation of goal intention, social traces are then a powerful means to
motivate individuals in the pre-decisional stage of the behavior change process.
In Blaze, we implement social traces mainly as Blaze map (Fig. (5-6)), which is a map showing
the points of origin of the students, who all travel to the same destination (a university). The points
of origin are not shown on exact locations, but are aggregated to the barangay level (smallest
political administrative unit). The barangays come in shades of three colors (red, yellow, green).
If a higher proportion of students from a barangay commute to school using alternative modes,
the barangay is tagged with green. Barangays with a higher percentage of car-driving students are
tagged with red. Barangays tagged with yellow are neutral.
Through the Blaze map, we intend to raise awareness about the issue (and negative consequences)
of car travel, trigger reflection that low-carbon travel is possible (perceived goal feasibility) and
Figure 5-6 Social Traces as Blaze Map
86
desirable (personal norms), and encourage comparisons of the car use among students from
different barangays (comparison).
Lastly, social traces can also provide information about available options and the perception of
their attributes (Axsen and Kurani, 2012). Hence, social traces can also be used to encourage
individuals in the pre-actional stage of the behavior change process. In our Blaze system, one can
check the modes (in percentage) the students use to travel to and from the university (Fig. (5-6)).
This can help the students valuate the attributes of options into specific benefits and disbenefits in
a process called translation (Axsen and Kurani, 2012). For individuals in the actional stage, this
same information – that low-carbon travel is feasible – may encourage them to try out the new
behavior, or to maintain it if they are in the post-actional stage.
5.5.2. Implementation plan
It is well-known that goal setting or sometimes also referred to as “goal initiation” is not sufficient
for a successful goal achievement. Formation of implementation intention is also important
(Gollwitzer, 1999). This is also true in a successful travel behavior change (e.g. Fujii and Taniguchi,
2005). Implementation intention, which entails a plan for when, where and how to implement the
target mobility behavior, mediates the effect of behavioral intention on behavior (Gärling and Fujii,
2002). Mobility behavior change support systems can facilitate the formation of implementation
intention by providing people with a trip/travel plan which they can use to rehearse the target
behavior.
Figure 5-7 Ridesharing and Trip Planning
87
We facilitate formation of implementation intention by promoting ridesharing and trip planning
(Fig. (5-7)). This intervention is thus most useful for individuals in the actional stage of the
behavior change process. However, the provision of trip plans can also indirectly affect those in
the earlier stages (pre-decisional and pre-actional) by influencing their perception of goal
feasibility or behavioral control. This same information can support those in the late stages (post-
actional) by helping them dynamically adjust their plans if necessary and thus avoid relapse.
5.5.3. Reflection-on-action
Reflection-on-action is reflection that takes place after the activity has ended, and not during a
particular activity. In many cases, reflection is activated by information received as feedback on
the person’s transport decisions (e.g. Jariyasunant et al, 2015). This feedback is given to the
individuals in situations where they can engage this information with more time and resources.
Hence, “systems that support reflection-on-action can allow for more extensive interaction and
even experimentation with one’s data. These systems can go beyond the mere representation of
sensor information by e.g. showing connections, correlations and (long-term) patterns in the
collected data” (Ploderer et al, 2014).
In Blaze, we trigger reflection-on-action on personal mode use. We provide two types of data
representations to engage the person in reflection: the first is a summary page of the person’s mode
Figure 5-8 Summary and Mode Use Graph
88
usage (latest, frequent and low-carbon), and the second is a graph of one’s modal split. Reflection
is mainly activated through comparisons (latest and frequent vs. low-carbon; one’s own mode use
vs. relevant others).
Reflection-on-action is most useful as intervention for individuals in the pre-decisional stage of
the behavior change process. In both data representations, we activate their personal norms by
making the low-carbon mode and car dependency salient, in order to encourage modal shift to
alternatives. At the same time, through comparisons of one’s mode use with that of relevant others,
we activate positive perception of goal feasibility. Individuals in more advanced stages can
likewise benefit from reflection-on-action in the form of feedback. The positive feedback to the
user in post-actional stage about his low-carbon mode use may further motivate him to maintain
his behavior. Comparative feedback, on the other hand, may help individuals in the pre-actional
stage to find a replacement mode for his car trips and those in the actional stage to try implementing
their plan.
5.5.4. Goal-setting
Since SSBC is self-regulative in nature,
individuals must actively invest effort
in setting and achieving goals
(Bamberg, 2013a). The SSBC assumes
that people in the pre-decisional stage
have no interest in or are unsure of
changing their habitual behavior. The
goal then for them is to raise doubts
about their current behavior and push
them into being persuaded that change is both necessary and possible for them. Those in the pre-
actional stage are considering changing their behavior but do not yet have a plan as to how to
achieve it. The goal is to help them find a personally suitable replacement for car for some trips.
People in the actional stage have a concrete plan already but they have not put it into practice yet.
The goal is then to help them initiate and implement their plan. Those in the post-actional stage
Figure 5-9 Goal-Setting
89
simply need to maintain their behavior. In Blaze, we implement goal-setting as Fig. (5-9). At the
end of each week, we ask the user to report their stage membership. These self-reports are then
plotted in a graph, which can implicitly encourage him to make progress to the next stages in the
following weeks.
5.6. Usability Evaluation
We conducted a usability evaluation, in which design issues, and not efficacy evaluation, were the
primary concern. Hence we limit ourselves to system evaluation that addresses technicality and
usability concerns, and not assessment of intervention-induced behavior change. A small sample
size is therefore sufficient for this purpose. For example, Virzi (1992) reports that only four to
five users are needed to detect 80% of the usability problems. Hwang and Salvendy (2010) suggest
a bigger size: 8-11 subjects (the so-called 10±2 rule). Faulkner (2003) recommends an increased
sample size, though she does not exactly specify the optimal size. Schmettow (2012) argues that
identifying a magic number of sample size is doomed to failure. In our usability testing, we use a
sample size that we consider appropriate for our purposes. To test Blaze, we conducted a usability
evaluation among the students of the Ateneo de Manila University.
We used the trial version of Blaze, slightly different from the one described in preceding sections,
for usability testing. This trial version was also being updated during the course of the evaluation.
Further improvements have also been implemented even after the usability testing.
90
Five classes, consisting of about 50 students each, were briefed on either 3rd or 4th November
2016 and invited to participate in the study. Students were asked to log their trips completely using
Blaze for three weeks (7-25 November 2016). Extra class credits and a chance to win in a raffle
were given as incentives to students who used Blaze.
Fig.(5-10) presents an overview of participation. Out of 250 students who were briefed, only 116
signed up in the website. 36 of the registrants never subsequently used Blaze after registration. Out
of the remaining 80, half of them (40 students) accomplished the post-survey after 3 weeks, which
allowed us to evaluate the trip logging feature and the standard interventions (described in Section
5.5) of Blaze. Thirty-two students gave us complete data. However, since ten of them are captive
public transport users (those with no access to car), we eventually exclude them from our
assessment of the potential effectiveness of Blaze in influencing a change in observable and latent
behavior. Note that the 40 students who filled up the post-survey included not only all the 32
students who gave complete data, but also some of those who gave incomplete data.
Figure 5-10 Overview of participation in the usability evaluation
91
5.6.1. Trip logging
Since our Blaze system is not yet able to automatically sense and collect trips but relies on user
input, we are interested to know the ease of use and the accuracy of data reported. On the scale of
1 (very easy) to 5 (very complicated), with 3 as neutral, we asked the students “How easy was it
to log your trips using the website?” The average score of 40 responses is 2.85, indicating a
relatively easy use. Students also provided additional feedbacks about the trip logging, ranging
from “easy to use” to “very difficult”:
Table 5-1 User feedback on Trip Logging
Category Feedback
Easy to Use “The trip logging is very convenient and easy to use.”
“It is user-friendly.”
“Logging my trips is very simple! The user interface is also easy to
understand.”
Very Difficult “Not very easy to use.”
“I had a hard time using the app and the site.”
“The app was quite complicated and somehow, I did not understand how it
worked.”
Difficult at first but
easy later
“It takes a while to adjust to the system”
“It was easy enough to use, once you got the hang of the interface.”
“It was a bit difficult navigating the website, especially with the logs. Took
me a while to understand it.”
“The program is not user friendly. It was hard to navigate in the website. It
took me a while to figure out how to get to the logs.”
From the feedback, we surmise that initially students had difficulty figuring out by themselves
how to use the system (the so-called learning phase) but once they got the habit, Blaze was
relatively easy to use. Nonetheless, we plan to introduce improvements to the current trip logging
feature.
92
Regarding accuracy of data, we asked “How many percent of your trip logs are complete logs?”
with five possible responses: less than 10%, about 30%, half of the trips, about 70% and more than
90%. 30 out of the 40 students (75%) reported that more than 90% of their trip logs were complete
trip logs.
5.6.2. Standard interventions
We also asked the students to rate the effectiveness or persuasiveness of some of our standard
interventions: the Blaze map in Home page, the Summary page, and the graph of mode use in
Diary page (see Figs. 5-6 and 5-8). On a scale from 1 (not at all) to 5 (very effective), with 3 as
neutral, we posed this question: “Did the following images make you reflect on your travel
behavior and make you at least consider taking alternative modes of transport?” For the Blaze map,
the average of 40 responses is 3.18/5. For summary page and graph of mode use, the averages are
3.25/5 and 3.30/5, respectively, indicating that the interventions have to some degree provoked
reflection about personal travel behavior.
5.6.3. Observable and Latent Change in Behavior
We also examined changes in the observable behavior and latent psychological change of a sub-
sample of students. The observable behavior here refers to the car usage, while the latent change
refers to change in stage membership (change in strengths of intention and socio-cognitive
determinants are also considered latent variables). Detailed analysis on the progressive changes in
the psychological factors requires though a larger sample size and is beyond the scope of this
chapter in which we focus on the Blaze tool itself. Only data from students who gave us complete
data and were not captive public transport users (N=22) were included in this analysis. Before the
students used the system, information about their typical car use was asked. They reported a pre-
intervention car use of 89.6% (percentage of mode use by car). After three weeks of using Blaze,
the same information was asked and they reported a post-intervention 85.0% car use. A decrease
93
of 4.6% in weekly car use was observed. We used paired T-test to evaluate the mean car use
percentage before and after the intervention, and the p-value obtained is p=0.14, indicating that
the 4.6% reported decrease in car use over the three-week period is not significant at all. We note
that the small sample size helps explain the non-significant result.
Since we include in Blaze stage measures to determine the stage membership of the students, we
then cross-tabulated their pre- and post-intervention stage diagnoses in order to observe the
occurrence of any stage progression, regression or stability. These students were subjected to the
trial version of Blaze. Only three students made some progress to later stages, while the rest
remained in the same stage prior to intervention. None of the students, however, progressed to the
post-actional stage.
Table 5-2 Stage Progression of N=22 Users
5.7. Summary, Limitations and Further Work
We report the development of Blaze, an application developed to promote voluntary travel
behavior change (VTBC) among students in Metro Manila. Most VTBC programs are not
explicitly grounded on a behavior change theory, and the few others which are, are based on
“continuum” static models such as the Theory of Planned Behavior. In contrast, Blaze is explicitly
theory-driven and, in particular, stage-based. It draws from the Stage model of self-regulated
behavior change (SSBC) theory in designing its standard and stage-specific interventions. Michie
et al (2008) argue that even with a theoretical framework, there is little guidance on how to develop
theory-based interventions. Thus, another contribution of our work is that it represents one of the
Post-intervention Pre-intervention
Pre-decisional Pre-actional Actional Post-actional
Pre-decisional 10
Pre-actional 2 5
Actional 1 4
Post-actional
94
first attempts in the literature to operationalize interventions explicitly based on SSBC. In
particular, we suggest developing multi-target menu-based interventions (Abraham, 2008). We
also report the usability evaluation conducted to assess the usability of Blaze. Since, at the moment,
we limit ourselves to system evaluation that addresses technicality and usability concerns, and not
assessment of intervention-induced behavioral change through experiment, we only employ a
small sample size of university students. Logging of trips is considered relatively easy to use, but
only after the users get accustomed to using the system. The standard interventions are also able
to provoke some reflection among the users regarding their current travel behavior. Although
usability of Blaze is our foremost concern in conducting the evaluation, we also perform initial
assessment of Blaze in changing behavior. We report a 4.6% decrease in the post-intervention car
use of the students from their pre-intervention car use. Some students have also progressed to
more advanced stages along the behavior change process after three weeks. Nonetheless, for a
great majority of users, their pre- and post-intervention stage diagnoses show no progress in stage
membership. However, no statistical inference can be derived from these results as sample size is
small. Further improvements have been implemented on Blaze based on the results of the usability
evaluation. An experiment with the purpose of evaluating the effectiveness of Blaze using robust
methodology will be conducted and reported in future studies. Deployment of Blaze to larger,
more diverse groups of students in Manila as well as extension to other cities in the ASEAN region,
is also envisioned. Adoption and use of mobility information systems using advanced technology
platforms enables the delivery of behavior change interventions on a large scale and in a cost-
effective manner. In the next chapters, we describe and present the field experiment conducted to
evaluate the effectiveness of Blaze.
95
CHAPTER 6
Materials, Methods and Measures
Chapter Summary. In this chapter, we provide the study context, starting first with Metro Manila
and the Ateneo de Manila University. We then describe the details of the field experiment
conducted in the university, the survey instrument used and its reliability.
6.1. Study Context
6.1.1. Metro Manila
Metro Manila (MM) is the capital of the Republic of the Philippines. It occupies a land area of
620km2 and supports a population of 12 million, which can swell to over 15 million during daytime
because of movements by people from nearby provinces flocking to MM for work and business.
In 2012, the traffic demand is 12.8 million trips. The main public modes of transportation traveling
on fixed routes include: three light rail lines (total length of 50 km), jeepney, and buses. Other
modes operating on flexible routes are: tricycles, taxis, and community taxis (called Asian Utility
vehicles). According to a 2012 study in MM car travel (with average occupancy of 1.7) only
accounts for 30% of person-km trips, and yet it constitutes 72% of road traffic. Road congestion
in MM is very severe: the volume/capacity ratio of roads is at 1.25, and road speed is below 10kph
for 55%-76% of the road km (NEDA, 2014). Car ownership is at about half a million in 2014 and
2015 (National Statistics Office, 2015). According to Nielsen Global Survey of Automative
Demand (2013), 47% of the households have no car.
6.1.2. Ateneo de Manila University
One of the Philippines’ most elite universities where predominantly students from rich families
come to study is Ateneo de Manila University (AdMU) which is situated on a 70-hectare area
located along Katipunan Avenue in Metro Manila. It has four campuses: Loyola Heights, Ortigas,
Rockwell and Salcedo campuses. We implement this study in Loyola Heights Campus (LHC) only.
96
With a total population of 18, 752 students (primary, secondary and tertiary levels combined) and
employees, LHC is a major trip attractor and generator that contributes to heavy congestion along
Katipunan Avenue.
According to a 2012 survey in LHC, about 93% of the population belong to the middle- and upper-
class (ADMU, 2012). The car ownership is high. Less than 1% of the population have no car in
the family, while 88% have two or more cars in their household. For the mode going to LHC, 43%
of the population arrive in LHC by private car, 24% by jeepneys and tricycles, 10% by school bus,
shuttle service or carpool, 4% by rail transit, 2% by bike and the rest by walk (note that those who
walk live within or near the campus). 89% of the population use the same mode of transport for
both going to LHC as well as going home.
6.2. Field Experiment
6.2.1. Study design
In this study, though representativeness of sample is ideal, we resort to convenience sampling
primarily due to resource constraint considerations. We deploy Blaze among students of the
Ateneo de Manila University (AdMU) in Metro Manila Philippines. Through the cooperation of
16 teachers, we briefed 20 classes composed of 20-75 students each. In total, we were able to
invite 1,063 students to participate in the study. Out of this, only 788 students agreed to participate
and accomplish the pre-survey. At baseline, we had a total of 414 students in both control and
experimental groups. 374 students had to be excluded because they did not meet the inclusion
criteria (participants must have access to car) or they gave invalid survey responses. Of the 414
students, 163 (39.37%) are male, 249 (60.15%) are female, and 2 (0.48%) did not provide this
information. Respondents are 18-26 years old, with a mean of 20.1 years (s=1.24). We then
assigned the classes into control or experimental groups (199 students in the control group, and
215 students in the experimental group). We asked those in the experimental group to register and
use our website or Smartphone application for about 4 weeks (1st-24th day) of February 2017.
Incentives for participation were offered in the form of extra credit in class and a chance to be one
of three raffle winners of about 4,500 PHP (1 USD = 50 PHP). The incentive is offered for using
97
the Blaze system, not for changing behavior. Each time the student uses Blaze by performing
certain tasks, he earns system points and is included in the raffle. Four weeks after the first contact,
we again approached the same classes and administered the post-survey. 700 students answered
the post-survey. We then matched the post-survey and the pre-survey for the same users. We
were able to match 138 students in the control group, and 163 students in the experimental group.
Some students had to be excluded from the experimental group, because they never registered in
our website, or used it after registration. In the end, for subsequent analysis, we only considered
a further reduced sample size (control=115; experimental=126) because of cases of missing values
in some responses. The experiment was approved by the University Research Ethics Committee
of the Ateneo de Manila University.
6.2.2. Measurement of Theory-Implied Constructs
Not only is the Stage Model of Self-Regulated Behavior Change used as a blueprint to guide
intervention development. The same model is also used as a framework to understand how the
intervention can induce a change in behavior through the mediation of a number of variables. In
this respect, we measure through a survey all the constructs specified in the model.
Our pre- and post- survey consists of questions asking students the following: demographics, stage
membership, thoughts on car use, and typical main mode used each day of the week both for going
to the university and returning home. The stage membership questionnaire, on the other hand, is
based on Bamberg (2013). We ask the students to choose one among the 7 statements that best
describes their level of car use: (Ia) I often use car to school, either as a student driving alone or as
the only passenger being driven. I feel content with this behavior and I do not see any reason to
change it. (Ib) I use car frequently to school. I am unsure if I need to change this behavior. (II) I
often use car to school, but I am also thinking about taking alternatives like public transport, or
sharing rides with others, but I am not sure whether and how I can achieve this goal. (III) I often
use car to school but it is my aim to change this behavior. I already know which trips I will make
with alternative modes, but, as yet, I have not actually put my plan into practice. (IVa) I often use
car to school but recently, I was able to go to school by other modes. I succeeded in reducing my
car use! (IVb) I mostly share rides or use alternative modes. (V) As I do not own/have access to a
98
car, car use is not an issue for me. We assign those who choose (Ia) or (Ib) to the first predecisional
stage; (II) to the second preactional stage; (III) to the actional; and either (IVa) or (IVb) to the post-
actional stage. Choosing statement (V) means the person is a captive non-car user, and is therefore
excluded from subsequent analyses.
We further distinguish (IVa) early post-actional and (IVb) late post-actional, as the survey questions
show. Those in the former have performed the desired behavior, though not yet habitually. The
latter includes those whose behavior has settled into the new, desired way. Our distinction between
the two post-actions are grounded on related works in the health domain. Schwarzer (2008), in his
Health Action Process Approach, also made some distinction between initiative and maintenance
of health behavior. The same distinction can be found also in Rothman et al (2004), Lippke,
Ziegelmann & Schwarzer (2008), Voils et al (2014), and Kwasnicka et al (2016). In our present
work, we only consider home-school commuting behavior of students, which is only one part of
overall student travel behavior. By doing so, we can easily distinguish between initiation and
maintenance of behavior. However, although the two post-actions are distinguished, there is no
intention associated with transition from the early post-action to late post-action (unlike, for
instance, the transition from predecision to preaction, which is associated with the formation of
goal intention). Thus, the distinction is based only on the position of the individual along a
continuum within the same stage. However, if the two post-actions are to be distinguished as
qualitatively separate stages, then it must be established that an intention, such as, say, “intention
to maintain behavior” (c.f. Luszczynska et al, 2007), is associated with transition from initiative
(early post-action) to maintenance (late post-action) stages. This warrants further research work.
The questionnaire on thoughts on car use consists of 15 questions, measuring on a 11-point scale
the socio-cognitive constructs in the SSBC model. To ensure construct validity, the questions are
derived from Schwarzer (2008), Bamberg (2013b), Klöckner (2014) and Klöckner (2017): (1)
Problem awareness: “There are problems associated with car use: traffic congestion, traffic
accidents, air pollution and global warming; (2) Personal responsibility: Each of us can contribute
in solving problems associated with car use by using car less or ridesharing with others if possible;
(3) Personal norm: Before I make the trip by car, I first consider if I can make the same trip by
alternatives. Cars should be seen as mode of last resort; (4) Social norm: I know some Ateneans
99
who, though have access to cars, go to school by alternatives. They somehow motivate me to travel
like them; (5) Positive emotions: I will feel good about myself if I am able to use alternatives for
my trips to school; (6) Negative Emotions: I will feel bad if I do not use alternative modes. (7)
Perceived behavioral control or goal feasibility: It is practical / possible for me to go to the
university (more often?) by alternative modes; (8) Goal intention: I intend to contribute in taking
cars off the roads, either by taking alternative means or pooling with others car trips going to same
destination; (9) Attitudes towards alternatives: Among the different mode options to go to the
university, there is one option, except driving alone or being the only passenger driven, that is
favorable for me; (10) Behavioral intention: I have decided which mode to use as substitute for
my car for some of my trips to the university. I intend to make a plan on how to go to the university
using this mode; (11) Action plan: I have a commute plan on how to go to school by alternatives,
and I have already run through my head on how to best carry out this plan; (12) Coping plan: I
have anticipated all the possible problems that can occur and hinder me as I put my commute plan
into practice; (13) Maintenance efficacy: I have already mentally developed ways to overcome
problems and obstacles to my commute plan or to be flexible depending on the situation; (14)
Implementation intention: Within the next seven days, I intend to actually use alternatives in going
to the university/home; (15) Recovery efficacy: I will continue to use alternatives to go to school,
even though this may be inconvenient.
As can be seen in our questionnaire, we used single-item measures per construct. This is clearly a
limitation, since we cannot test the internal consistency using indices such as, for example,
Cronbach’s alpha. However, we decided against multiple-item measures per construct in our
survey because they are costly and time-consuming in practice (Christophersen and Konradt, 2011)
and we believe it would have led to a drop in attention and an even higher drop-out rate. Bergkvist
and Rossiter (2007) and Lucas and Donnellan (2012) similarly argue that single-item measures
can be reliable and are preferred especially when respondent burden is primary concern. We also
note that Klockner (2014) in his study, on which some of our survey questions are based, used
one-item measures.
An alternative to reliability testing by examining internal consistency is by using longitudinal data
(Lucas and Donnellan, 2012). To test the reliability of our single-item survey, we assess the
100
stability of the responses of the students (in the control condition) at two different time points: at
baseline and after four weeks. We use a specific cohort of the control group for our analysis: those
students identifying themselves as not having changed their stage membership over four weeks.
This is to ensure that their test-retest responses on all the questions are not likely to change over
time. Only data from postaction-stable (N=47) are used since this constitutes the largest sample
size. Following Thøgersen (2006; 2009), we calculate the t-test and correlation at the two time
points to find any systematic changes over time in this group. If there is not any systematic change,
we can assume that the survey instrument is reliable since the responses to the survey instrument
are stable over time in stable subjects (Vaz et al, 2013).
6.2.3. Reliability of measurement
In order to get a general impression of the reliability of our instrument, we present in Table 6-1
the item-level means at both time points, stabilities (i.e. test-retest correlations) and change (i.e.
paired t-tests) for the postaction-stable cohort group (c.f. Thøgersen, 2006). In Table 6-1, the t-
test shows that, at the Bonferroni corrected significance level of p < .003 (0.05 divided by 15),
only 1 of the 15 paired t-tests – namely, the item negative emotions – is significant. We thus
decide to drop the negative emotions from our subsequent analysis. In Klöckner (2014), negative
emotions are also not considered. In the table, the test-retest correlations reveal a range of 0.298-
0.674. In Thøgersen (2009), the correlation ranges from 0.25-0.69 for test-retest for control group
(over 1-month period) for the variables considered. Hence, insofar as the correlations of our
variables are concerned, we consider them acceptable.
Table 6-1. Reliability of Survey Instrument (N=47) POSTACTION
(STABLE) STAGE
MEAN AT
BASELINE
MEAN AFTER
4-WEEKS
PAIRED
T-TEST
PEARSON
CORRELATION
Positive Emotions 2.85 2.57 0.253 0.673*
Negative Emotions 2.51 0.83 <0.0001** 0.415*
Personal Norm 1.25 2.14 0.065 0.408*
Social Norm 1.49 2.18 0.113 0.443*
Personal Responsibility 4.32 4.17 0.302 0.444*
Problem Awareness 4.53 4.06 0.312 0.312*
Perceived Behavioral Control 4.06 3.28 0.035 0.298*
101
Goal Intention 3.77 3.40 0.215 0.611*
Attitude 3.07 3.32 0.176 0.354*
Behavioral Intention 2.85 3.40 0.089 0.674*
Action Plan 2.91 3.18 0.430 0.669*
Cope Plan 2.60 3.04 0.260 0.536*
Maintenance Efficacy 2.48 3.22 0.038 0.528*
Implementation Intention 3.29 3.13 0.683 0.663*
Recovery Efficacy 3.24 3.62 0.238 0.616*
*significant at 0.05 level
**significant at 0.003 Bonferroni corrected significance level
102
CHAPTER 7
Implementation and evaluation of the mobility behavior change
support system
Chapter Summary. A mobility behavior change support system, Blaze, is deployed among the
students of a university in the Philippines. This is implemented as a travel demand management
measure in a university setting. Blaze is developed based on the stage model of self-regulated
behavior change theory. Our analysis shows that Blaze is able to induce a positive change across
a wide range of indicators, as specified by the theory: car use, stage membership, intentions and
socio-cognitive determinants. Furthermore, we conduct a usefulness evaluation to assess the
relevance of the features within Blaze and associate it with stage membership. We discuss
implications of our results, especially on the development of interventions and transport policies
for cases in developing countries.
Keywords: Mobility behavior change support system; sustainability in universities; travel demand
management
7.1. Recap of Methodology
In this study, we use convenience sampling as sampling method. A total of 1,063 students were
asked to participate in the study. Only 788 students, however, accomplished the pre-survey. At
baseline, we had 414 students in both control and experimental groups. 374 students had to be
excluded because they identified themselves as captive public transport users or captive
pedestrians or gave invalid survey responses. We then assigned the classes into control (199
students) or experimental groups (215 students). We invited those in the experimental group to
register and use our website or Smartphone application for nearly 4 weeks (1st-24th day) of
February 2017. Those in the control group were not told about Blaze, and hence, never used it.
Incentives were offered in the form of extra credit in class and a chance to be one of the three raffle
winners of about 4,500 PHP (1 USD = 50 PHP). The incentive is offered for using our system, and
103
not for changing behavior. Each time the student uses our system by performing certain tasks, he
earns system points and is included in the raffle. Four weeks after the first contact, we again
approached the same classes and administered the post-survey. 700 students answered the post-
survey. We then matched the post-survey and the pre-survey for the same users. We were able to match
138 students in the control group, and 163 students in the experimental group. Some students had
to be excluded from the experimental group, because they never registered in our website, or used
it after registration. In the end, for subsequent analysis, we only considered a further reduced
sample size (control=115; experimental=126) because of cases of missing values in some
responses. The experiment was approved by the University Research Ethics Committee of the
Ateneo de Manila University.
7.2. Equivalence of Control and Experimental Groups at Baseline
Since we used convenience sampling for our sampling method, and since there are drop-outs after
4 weeks, we need to establish first the equivalence of the two groups. We check if the treatment
groups (control and experimental) can be assumed to be equivalent with each other at baseline and
after 4 weeks. The two groups at baseline (control, N=138 and experimental, N=163) are not
statistically different in terms of stage membership and pre-car use. Chi-square test is used to
establish equivalence of stage membership across groups, and Mann-Whitney test for pre-car use.
We fail to reject the null hypothesis that the two samples are drawn from the same population (Chi-
square: χ2=4.975, df=4, p=0.29; Mann Whitney: p=0.88 for predecisional; p=0.19 for pre-actional;
p=0.22 for actional; p=0.90 for post-actional (early) and p=0.71 for post-actional (stable)). We
carried out the same tests for the reduced sample (control, N=115; experimental, N=126) and we
obtain the same result: the two treatment groups at two time points can be assumed to be equivalent
(χ2=0.8391, df=4, p=0.93; Mann Whitney: p=0.61 for predecisional; p=0.45 for pre-actional;
p=0.35 for actional; p=0.60 for post-actional (early) and p=0.38 for post-actional (stable)). There
is no evidence to support selective attrition or self-selection. In what follows, we mostly analyze
only the reduced sample.
104
7.3. Car Usage, Goal, Behavioral and Implementation Intentions, and Socio-cognitive
variables per stage at Baseline
In Table 7-1, we present the distribution across stages of the two groups. Here we observe that
PreD and PostA2 stages are stable. Furthermore, as can be seen in Fig. 7-1, the car-use of the
students at baseline for both groups follow the pattern described by the stage model of self-
regulated behavior change. Those in the stable postaction stage (PostA2) have the lowest car-use
per week, while the car-use for the three early stages are high and almost equivalent (with
overlapping confidence intervals). Those who identify themselves in the early postaction (PostA1)
use car about half as often as those in the early stages.
Table 7-1 Distribution across stages of experimental and control groups
Experimental
T0/T1 PreD PreA A PostA1 PostA2 N
PreD 21 7 3 2 8 41 (32.5%)
PreA 4 11 3 1 3 22 (17.46%)
A 0 3 3 2 2 10 (7.94%)
PostA1 0 0 0 3 6 9 (7.14%)
PostA2 0 0 1 0 43 44 (34.92%)
N
25 21 10 8 62 126
Control
T0/T1 PreD PreA A PostA1 PostA2 N
PreD 32 2 0 0 0 34 (29.57%)
PreA 5 9 2 2 0 18 (15.65%)
A 2 5 0 0 1 8 (6.96%)
PostA1 2 3 0 1 4 10 (8.70%)
PostA2 1 1 0 1 42 45 (39.13%)
N
42 20 2 4 47 115
105
Figure 7-1 Mean and 95% confidence intervals for car use per stage at the start of experiment (T0: baseline)
Looking at the goal, behavioral, and implementation intentions (Figs. 7-2 to 7-4) across stages
compared to the reference group (here we set it to control predecision), we see that the patterns
roughly follow the stage model. As expected, the stage with the lowest goal intention is predecision
(reference group mean = -0.76). This can be observed in both treatment groups. In fact, the theory
claims that those in this stage have not yet formed any goal intention. Similarly, looking at the
relative strengths of behavioral intentions across stages, we find that only those in action and later
stages have formed strong behavioral intentions (reference group mean = -1.18). Finally,
implementation intention is strongest among those in the stable postaction stage (reference group
mean = -1.79).
(2.00)
0.00
2.00
4.00
6.00
8.00
10.00
12.00
14.00
PreD PreA A PostA1 PostA2
car
use
per
wee
k
Control Experimental
106
Figure 7-2 Goal Intention per stage at T0
Figure 7-3 Behavioral Intention per stage at T0
(2.00)
(1.00)
0.00
1.00
2.00
3.00
4.00
5.00
6.00
PreD PreA A PostA1 PostA2dif
fere
nce
to
ref
eren
ce g
rou
p
Control Experimental
(2.00)
(1.00)
0.00
1.00
2.00
3.00
4.00
5.00
6.00
PreD PreA A PostA1 PostA2dif
fere
nce
to
ref
eren
ce g
rou
p
Control Experimental
107
Figure 7-4 Implementation Intention per stage at T0
We also check the pattern at baseline for the other 11 socio-cognitive variables (excluding negative
norms), using the predecision control as the reference group (Fig. 7-5). Here we find the same
trend, as postulated by the stage model, except for personal responsibility and problem awareness
where the strengths of the responses do not vary much across groups.
In general, we find that students in the late stages (postaction) rate higher in all variables. Those
in the early predecision stage, however, score lowest in all the variables. This is consistent with
the theory. Those in the preaction, having had formed their goal intention, should score high in
those variables associated with this type of intention, but should score lower in Attitude (Att),
Action Plan (AP), Cope Plan (CP), Maintenance efficacy (M. Eff) and Recovery efficacy (R. Eff),
since they have not yet formed their behavioral and implementation intentions. We can see this
supported in Fig. 7-5. Those in the action, having had formed both their goal and behavioral
intentions but not yet the implementation intention, should score high in all variables, except AP,
CP, M. Eff and R. Eff, wherein they score lower than the postaction. We observe this is supported
by our data.
(2.00)
(1.00)
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
PreD PreA A PostA1 PostA2dif
fere
nce
to
ref
eren
ce g
rou
p
Control Experimental
108
Figure 7-5 Socio-cognitive determinants per stage at T0
The above figures are based on aggregated data, but for the purpose of illustrating the same idea
at the individual level, we also plot the magnitudes of the responses to all variables of two users in
the predecisional stage (Fig. 7-6). In general, the pattern seen in the data follows that which is
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
C E C E C E C E C E C E
PosEm P. Norm S. Norm P. Resp P. Aware PBC
dif
fere
nce
to
ref
eren
ce g
rou
p
PreD PreA A PostA1 PostA2
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
C E C E C E C E C E
Att AP CP M. Eff R. Eff
dif
fere
nce
to
ref
eren
ce g
rou
p
PreD PreA A PostA1 PostA2
109
postulated by the theory. Specifically, we observe that both users have no goal, behavioral and
implementation intentions yet. Moreover, all the determinants of later stages (preaction, action,
postaction) are negative. However, for the same users, the magnitude of the determinants varies:
whereas User2 anticipates feeling positive emotions when using alternatives, User1 does not much.
While User1 has some perceived control over desirable behavior, User2 does not. Both have not
yet formed goal intention (assuming that the threshold for transitioning to the next preaction stage
is at zero), but the magnitudes are also different for the users. Moreover, they may react differently
to the same intervention: while, for example, User1 may react negatively if we target stage-
inappropriate determinants of action (e.g. AP, CP, M.Eff), User2 may not, and may possibly even
facilitate transitioning to more advanced stages. Here we see that that the users, though belonging
to the same predecision stage, have different needs and deficits. This idea has consequences on
intervention development. Intervention development in mobility and other domains based on the
SSBC has so far recommended stage-tailored interventions (in mobility, see Bamberg (2013a); in
electric car purchase, see Klockner (2014); in food consumption, see Klockner (2017)).
Nonetheless, following Abraham (2008), in contrast to stage-tailored interventions, we suggest
“menu-based behavior change interventions”, providing a range of interventions matching various
deficits through a menu from which individuals can self-tailor. The advantage of using menu-
based interventions is that we can “target cognitions on both sides of purported stage transitions”
(Abraham, 2008).
110
Figure 7-6 Comparison of responses between two users
7.4. Effect of Intervention on Car Use
We now consider the effect of the intervention on the observable car-use behavior. At baseline,
the average weekly car-use of the control group (N=115) is 5.23 (s=4.65), while that of the
experimental group (N=126) is 5.90 (s=4.34). Four weeks later, the mean car-use of the control
group is 5.56 (s=4.49), and of the experimental group it is 5.07 (s=4.35). In general, an increase in
car use is observed in the control group, and a decrease in experimental group. To obtain statistical
significance, we use ANCOVA, where car-use after four weeks is compared, and the car-use at
baseline is entered as a covariate. The ANCOVA results show that there is a significant difference
on car use at T1 after controlling for car use at T0, F (1, 238) = 10.2, p=0.0017.
Table 7-2 ANCOVA Results Adj SS df MS F P(>F)
Covariate (X) 3148.17 1 3148.17 492.79 <.0001
Between (Y) 64.53 1 64.53 10.1 0.0017
Within (Y) 1520.44 238 6.39
Total (Y) 1584.97 239
Total 4733.14 240 X: weekly car use behavior at baseline; Y: weekly car use behavior after 4 weeks
-6
-4
-2
0
2
4
6
mag
nit
ud
e o
f re
spo
nse
User1 User2
111
Figure 7-7 Car use reduction across different stages
Furthermore, we examine the reductions in car use across different stages in the two groups. In
Fig. 7-7, the control action group has increased, on the average, their weekly car usage by 2. We
also observe that, across stages, the experimental group shows greater reduction in car use than
the control group. This reduction is more pronounced in the early predecision stage.
From the ANCOVA table, we can calculate the partial effect size of the intervention. We can
derive it by taking the ratio of the adjusted sum of squares associated with the treatment (between
group at Y) and the total adjusted sum of squares (total at Y). That is, the effect size is
64.53/1584.97 or 0.041. Comparing this result with the effect size of personal travel plan or PTP
unaided by technology (Cohen’s h = 0.12-0.15), as calculated in the meta-analysis of Bamberg and
Rees (2017), the effect size of our technology-based intervention is rather very small. Moreover,
the average weighted effect size of health behavior change support systems, as cited in Bamberg
et al (2015), has a range of d=0.14-0.19. The small effect size of Blaze can be explained by the
fact that in the university where the social experiment was conducted, there is no wide range of
attractive alternative options to car for the students.
We can also compute for the relative effect of the intervention on car use reduction. Following
Hsieh et al (2017), the intervention effect E on a variable can be calculated as:
(2.50)
(2.00)
(1.50)
(1.00)
(0.50)
0.00
0.50
1.00
1.50
2.00
2.50
3.00
PreD PreA A PostA1 PostA2
car
use
red
uct
ion
stage at baseline
Car use reduction
Control Experimental
112
𝐸 = ( 𝑒𝑥𝑝𝑡1 − 𝑒𝑥𝑝𝑡0) − 𝑒𝑥𝑝𝑡0 (𝑐𝑜𝑛𝑡𝑡1− 𝑐𝑜𝑛𝑡𝑡0)
𝑐𝑜𝑛𝑡𝑡0 (Equation 1)
where exptt0 and exptt1 are the means of a variable for an experimental group at two time points t0
and t1; contt0 and contt1 correspond to the control group. The term (contt1 - contt0) is the change
amount due to extraneous factors, which when divided by contt0 yields the change rate. Thus, the
influence of extraneous factors on the experimental group can be eliminated. A relative effect of
intervention (RE) can be calculated as:
𝑅𝐸 = E
𝑒𝑥𝑝𝑡0100% (Equation 2)
The relative effect of the intervention on car use reduction is thus -20.38%.
7.5. Effect of Intervention on Stage Membership, Intentions and Socio-cognitive
Determinants
The stage model not only focuses on the observable behavioral change, but also on the latent
psychological change, which can be measured via change – better progression – in stage
membership or in intentions. First, we examine the stage memberships of the two treatment groups
at two time points (Table 7-3). Here we use a bigger sample size (control=138; experimental=163)
because we have data available:
Table 7-3 Stage Memberships
Stage Control (N=138) Experimental (N=163)
Baseline 4 weeks Baseline 4 weeks
PreD 38 46 52 36
PreA 20 21 31 27
A 8 2 14 12
PostA1 12 7 8 10
PostA2 60 62 58 78
113
We carry out Mann-Whitney and median tests to determine if there is any statistical difference
between these two groups. Our tests show that they do not differ significantly in terms of stage
memberships at baseline, and four weeks later. Similarly, comparing the stage membership of the
control group at the two time points reveals no significant change. However, for the experimental
group, there is a significant stage progression (p=0.02).
We cross-tabulate the pre- and post- intervention stages of the students to gain some idea on the
stage progression and regression. We observe that the predecision and postaction-stable stages do
not change much, and action stage is very unstable, which is in agreement with Bamberg (2013b).
We then examine in which of the five stages the intervention is able to induce a significant effect.
Given the same pre-intervention stage, we compare the stage changes in the two treatment groups
after the period of intervention.
Table 7-4 Stage Progression Across Stages
In Table 7-4, some of those who were originally in the predecision stage (in both treatment groups)
changed their stage membership 4 weeks later. The Mann-Whitney test reveals that the change is
significant when comparing both treatment groups (p<0.01). We also use the 1-tailed Fligner-
Policello test (FP) as it indicates the direction of change (whether progression or regression). The
result shows that, in predecision and action stages, the median of experimental group is greater
than that of the control group, and that the difference is significant (p<0.01 and p=0.01,
respectively).
Table 7-5 Stage Transitions N -4 -3 -2 -1 0 1 2 3 4 μ
Exp. 126 0 0 .8 5.6 64.3 14.3 4.8 4.0 6.3 +.43
Control 115 .9 2.6 4.3 9.6 73.0 7.0 2.6 0 0 -.15
Control (N=138) Experimental (N=163) MW FP
T0 / T1 PreD PreA A PostA1 PostA2 T0 / T1 PreD PreA A PostA1 PostA2
PreD (38) 92.1 7.9 0.0 0.0 0.0 PreD (52) 55.8 19.2 5.8 3.9 15.4 <0.01** <0.01**
PreA (20) 30.0 45.0 10.0 15.0 0.0 PreA (31) 19.4 41.9 16.1 9.7 12.9 0.25 0.11
A (8) 25.0 62.50 0.00 0.00 12.50 A (14) 7.1 28.6 21.4 14.3 28.6 0.06 0.01*
PostA1 (12) 16.7 25.0 0.0 16.7 41.7 PostA1 (8) 0.0 0.0 0.0 37.5 62.5 0.18 0.06
PostA2 (60) 1.7 1.7 0.00 3.3 93.3 PostA2 (58) 0.00 0.0 1.7 0.0 98.3 0.64 0.11
114
In Table 7-5, we show the percentage of students who transition x stages forward or backward
where x=-4, -3, …, +3, +4. We see that 64.3% and 73.0% of the students in the experimental and
control groups, respectively, do not transition at all. Majority therefore remain in their original
stages. The experimental group, however, shows some stage progression with 14.3% of the
students transitioning one stage forward. The 6.3% who transition 4 stages up are those who were
in the predecision at baseline but who are now in the late post-action four weeks later. Taken as a
whole, this group progressed .43 stages (μ =+.43). On the other hand, the control group shows
regression: 9.6% of them transition one stage backward. Overall, this group regressed .15 stages
(μ =-.15). It seems therefore that the intervention not only induces stage progression, but also
inhibits stage regression.
Table 7-6 Intention Strengths Across Time Control (N=115) Experimental (N=126) ANCOVA
Baseline 4 weeks later Baseline 4 weeks later
Goal Intention 1.92 (2.62) 1.54 (2.84) 1.94 (2.63) 2.23 (2.37) F(1,238)=5.45,
p=0.02*
Behavioral
Intention
1.06 (2.64) 0.87 (3.07) 0.81 (3.11) 1.41 (3.31) F(1,238)=4.48,
p=0.04*
Implementation
Intention
0.96 (3.15) 0.24 (3.36) 0.40 (3.33) 1.15 (3.45) F(1,238)=15.67,
p<0.01*
We next examine the effect of the intervention on the three intention types (Table 7-6). To increase
statistical power, we use ANCOVA, where the intentions at baseline are entered as a covariate for
comparing the intentions 4 weeks later. The test shows that the there is a significant difference on
the intentions 4 weeks after controlling for intentions at baseline. The ANCOVA results are: (a)
F(1,238)=5.45, p=0.02 for goal intention; (b) F(1,238)=4.48, p=0.04 for behavioral intention; and
(c) F(1,238)=15.67, p<0.01 for implementation intention. We now examine in which stages in the
change process is the intervention effective in changing the three intentions.
115
Figure 7-8 Change in goal intention
Figure 7-9 Change in behavioral intention
(4.00)
(3.00)
(2.00)
(1.00)
0.00
1.00
2.00
PreD PreA A PostA1 PostA2ch
ange
stage at T0
Control Experimental
(2.00)
(1.50)
(1.00)
(0.50)
0.00
0.50
1.00
1.50
PreD PreA A PostA1 PostA2
chan
ge
stage at T0
Control Experimental
116
Figure 7-10 Change in implementation intention
The intervention succeeds in strengthening the goal intention only for those in the predecisional
stage (Fig. 7-8). The effect is negligible for other stages. This may be realistic since those in the
later stages have already formed goal intention. What is surprising is that those in the actional stage
at T0 and who are in the experimental group shows weakening of the goal intention after 4 weeks.
We do not know, for sure, how this happens. Likewise, the intervention succeeds in strengthening
the behavioral intention across all stages of the experimental group; in the control group, we
observe a weakening of the behavioral intention across all stages after 4 weeks, except in the
postactional-stable group (Fig. 7-9). Finally, the intervention also succeeds in strengthening the
implementation intention across all stages of the experimental group, except the action stage group,
for which there is no change. We observe a weakening of implementation intention in the control
group after 4 weeks (Fig. 7-10). These changes in intentions in the control group imply that there
are uncontrolled extraneous influences during social experiment.
Table 7-7 Socio-cognitive determinants across time Control (N=115) Experimental (N=126) ANCOVA
p-value Baseline 4 weeks later Baseline 4 weeks later
Positive emotions 2.53 (2.10) 2.04 (2.40) 2.64 (2.31) 2.79 (1.99) <0.01***
Personal norm -0.23 (3.20) 0.68 (2.88) 0.04 (2.97) 0.90 (2.72) 0.75
Social norm 0.88 (2.56) 1.36 (2.58) 1.02 (2.52) 1.44 (2.70) 0.96
Personal Responsibility 3.91 (1.25) 4.03 (1.28) 3.67 (1.72) 4.14 (1.38) 0.13
Problem awareness 4.35 (1.19) 4.35 (1.02) 4.35 (1.30) 4.52 (0.88) 0.16
(3.00)
(2.50)
(2.00)
(1.50)
(1.00)
(0.50)
0.00
0.50
1.00
1.50
2.00
PreD PreA A PostA1 PostA2
chan
ge
stage at T0
Control Experimental
117
Perceived control 2.30 (3.08) 1.49 (3.17) 1.99 (3.12) 1.58 (3.20) 0.45
Attitude 1.71 (2.59) 1.98 (2.89) 1.11 (3.13) 2.64 (2.19) <0.01***
Action planning 1.13 (2.91) 0.72 (3.28) 0.88 (2.97) 0.90 (3.69) 0.34
Coping planning 1.25 (2.64) 0.55 (3.13) 0.89 (2.87) 1.19 (3.20) 0.02**
Maintenance efficacy 1.10 (2.63) 0.77 (3.02) 0.55 (2.95) 1.10 (3.26) 0.08*
Recovery efficacy 1.21 (2.87) 0.84 (3.22) 0.89 (3.12) 1.47 (3.32) <0.01***
***significant at 0.01 level ** significant at 0.05 level *significant at 0.1 level
To analyze the effect of the intervention on socio-cognitive determinants, we calculated the mean
of the difference in the responses at two time points for each treatment group, and compared them
using ANCOVA (Table 7-7). Our findings reveal that only positive emotions, attitude, coping
planning and recovery efficacy are significant at 0.05 level or better. However, despite the fact
that not all determinants are influenced significantly by the intervention, the intervention is able to
significantly affect the three intention types. More research is needed to ascertain if these results
mean that (a) the induced changes in intentions are weak, or that (b) the determinants have non-
proportional effect on the intentions, and hence by only activating some determinants with greater
effect, we can in turn already influence the intentions.
Figure 7-11 Change in selected socio-cognitive determinants
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
4.00
C E C E C E C E
PosEm Attitude CopePlan R. Efficacy
chan
ge
PreD PreA A PostA1 PostA2
118
In Fig. 7-11, first we observe that for the control group (denoted by ‘C’), the values are mostly
negative. This means that without the intervention, the magnitude of the responses decreases after
four weeks. Looking at the experimental group (‘E’), the intervention succeeds in strengthening
the positive emotions of the predecision group, the attitudes towards alternatives of those in the
predecision and preaction groups, the coping plan skills of those in the postaction (early and stable)
groups, and the recovery efficacy skills of those belonging to any group, except action group. We
also observe a general weakening of positive emotions, attitudes, coping plan and recovery
efficacy in the action group. What this demonstrates is that our interventions are able to
simultaneously induce activation (or de-activation) of the socio-cognitive variables, which
depends on the current membership stage. Though unintended, the effect can be negative. Caution
must be taken on the reverse effect of the interventions on the variables. To this end, more research
is needed to ascertain how the interventions negatively affect these variables.
7.6. Usefulness Evaluation
We conducted a usefulness evaluation among university students. Two weeks after the first
contact, we asked them to accomplish an online survey, assessing how useful the features within
Blaze are for them. We want to determine if indeed the usefulness or meaningfulness of the
intervention is stage-dependent. In other words, certain interventions will be useful for the person
depending on how relevant the interventions are to the goal the person is pursuing. For example,
if the person is in preactional stage, we assume he is exploring taking alternatives that can replace
his car. He is not yet interested in putting a travel plan into practice. This means an intervention
offering him possible options (i.e. mode options) is considered relevant, while interventions aiming
to help him plan trips (ride plan) is irrelevant. Results may lend support to the menu-based
intervention approach.
Responses were obtained from N=71 students. These students (25% predecisional; 11%
preactional; 17% actional; 47% postactional) used Blaze system for two weeks before they were
asked to evaluate the usefulness of each of the interventions. On a scale of -2 to +2, with 0 as
neutral, the respondents rated each of the 10 features within Blaze (see Chapter 4, Table 4-4). In
119
Table 7-8, we show the average rating for each feature (X1-X10) differentiated by stage
membership (PreD. PreA. A, combined PostA).
Table 7-8 Average Rating per rating per stage.
N: sample size; X1: Map; X2: Goalset; X3: Summary; X4: Readiness; X5: Travel time; X6: Mode Use;
X7: Slideshow; X8: Ride plan; X9: Mode options; X10: Diagnostic report
STAGE (N) X1 X2 X3 X4 X5 X6 X7 X8 X9 X10
PreD (18) 0.11 0.33 0.22 0.11 0.06 0.33 0.50 -0.39 -0.06 0.17
PreA (8) 0.50 0.38 1.13 0.50 0.00 0.88 1.75 -0.13 0.25 1.50
A (12) 0.42 0.92 1.25 0.67 0.83 1.00 0.92 0.67 0.83 0.92
PostA (33) 0.88 0.82 0.97 0.64 0.64 1.09 1.55 0.36 1.06 0.64
We see that most of the interventions are found to be useful by the individuals, regardless of their
stage memberships. This confirms our previous hypothesis that the interventions mean differently
for individuals in different stages. Highlighting the results of the evaluation of mode options and
ride plan reveals an interesting insight. For predecisional users, they find mode options not
relevant. Their average rating is -.06 (less than 0), indicating it is not useful. Rideplan is not useful
for predecisional (-.39) and preactional (-.13) individuals, but it is most useful for actional
individuals (+.67). This somehow gives support to our idea that interventions will be useful for
the person depending on how relevant the interventions are to the goal the person is pursuing.
An alternative interpretation, however, can be proposed regarding the negative values. The fact
that the users give a negative rating on certain features may imply that providing them with
inappropriate interventions may generate negative reactance. In other words, the users may not
simply ignore the irrelevant interventions; they may react negatively on them. For instance,
offering a ride plan to predecisional and preactional individuals, an intervention they consider
irrelevant, may elicit negative reactance, and hence may not only be ineffective but also counter-
productive.
Based on the results of the current study, it is not yet clear whether or not the negative values
indicate that a menu-based intervention approach can be counter-productive, in which case stage-
tailoring is preferable. Abraham (2008) asserts: “Stage-tailored interventions are only likely to be
120
preferable to menu-based interventions when … evidence indicates that administering stage-
mismatched components in menu-based interventions is counter-productive, e.g. evidence
showing that inclusion of a [particular] component weakens the intentions among those with
stronger pre-intervention intentions.” This is one limitation of the present study.
In a menu-based approach, although we give the same generic “menu” to all users, regardless of
their stage membership, this is not a one-size-fits-all approach. “Menu-based interventions are not
a one-size-fits-all approach. Rather the approach assumes that people can be directed to messages
or training materials matching their position on the action readiness continuum. Participants need
not focus on all determinants. Thus menu-based interventions are self-tailored in that participants
select components most relevant to them using self-assessments” (Abraham, 2008).
7.7. Summary, Limitations, Implications and Further Work
The main goal of this work is to evaluate the effectiveness of a travel behavior change support
system, Blaze, which has been deployed among students of a university in a developing country.
Our analysis shows that Blaze is able to induce a positive change across a wide range of indicators,
as specified by the stage model of self-regulated behavior change theory: car use, stage
membership and intentions. However, due to resource constraints, only a convenience sample of
university students is used, and thus we cannot generalize our results to the general population of
the university. Nonetheless, since our methodology includes an adequate control group, this is
considered to be of higher methodological quality than previous evaluation studies of mobility
behavior change support systems, all of which only employ before-after comparison without
adequate control groups.
Furthermore, our work demonstrates useful insights that can be obtained about the travel behavior
of students by using the stage model of self-regulated behavior change (SSBC) theory. Travel
behavior of university students is not well understood yet (Danaf et al, 2014). Using the SSBC as
the framework, we try to gain more insights about the behavior of the university students. From
SSBC, a specific pattern can be postulated about the car use and intentions (goal, behavioral,
implementation), depending on the stage memberships of the students.
121
Finally, our study also shows that the effect of the intervention is largest among those in the early
predecisional stage. Not only the intervention-induced car-use reduction is greatest among students
in this group, their stage progression due to the intervention is also significant. Caution must,
however, be taken since the sample size in this group is larger so it has more statistical power.
Nonetheless, in line with Fujii and Taniguchi (2006), what our results suggest is that more attention
must be given to changing the behavior of “non-riders of public transportation”, in other words,
those in the early stages, since greater behavioral change can be induced from them.
All these have significant implications for transport policy. In the literature, the TDM measures
implemented in the university setting are dominated by hard measures (c.f. Danaf et al, 2014).
Nonetheless, as previously pointed out, these measures may not be appropriate for cases in the
developing countries. An example of a soft measure implemented in the university setting,
TravelSmart, may also not be appropriate as it is costly due to the need for travel counselors. We
suggest here implementing a voluntary travel behavior change program on a technological
platform for greater scalability and cost-effectiveness, an important consideration for developing
world cases as they are confronted by resource constraints. Moreover, we support the idea of
explicitly grounding change programs in a behavior change theory, not only for greater insight into
understanding travel behavior, but also for greater effectiveness. Indeed, it is claimed that a
theoretical grounding is missing in many of the behavior change programs (Richter et al, 2011).
Finally, we note some of the limitations of the present study. We only investigate the home-school
commuting behavior of students, which is only one part of overall student travel behavior.
Moreover, Bamberg (2013) suggests that studies must be conducted to test whether stage transition
is indeed reflected in changes in intention types, which in turn are associated with respective
changes in the stage-specific sets of socio-cognitive variables. Furthermore, this chapter only
tackles behavior outcomes, but adherence of the students to the intervention is not discussed or
investigated (c.f. Cugelman et al, 2011). Finally, though the development of Blaze is grounded
explicitly on SSBC, we are not able to identify which specific interventions within Blaze are able
to induce effect on which variables/constructs specified by the theory. We recommend for future
122
work linking interventions to target socio-cognitive variables. This is of great interest in the
domain of intervention development.
123
CHAPTER 8
Validation and Extension of the Stage Model of Self-Regulated
Behavior Change
Chapter Summary. This chapter aims to characterize the stages of change in terms of car use
level, strength of the three intention types, and stability. Our objective is to describe the
relationships of these variables with each other. We also describe how the three intention types
mark the transition points between stages. We then construct a model specifying the determinants
or predictors of the three intention types. We end by discussing how our results confirm, validate
and extend further the stage model of self-regulated behavior change (SSBC). Our main findings
suggest that, in agreement with SSBC, travel behavior change is achieved by a transition through
a temporal sequence of four stages: predecision, pre-action, action and post-action. In an extension
to SSBC, we further distinguish post-action depending on whether the behavior is on initiation or
under maintenance. We observe that the former (initiation) is characterized by instability (either
relapse or progress), while the latter (maintenance) by stability.
Keywords: Stage of change model; Stage model of self-regulated behavior change; longitudinal
data
8.1. Introduction
In previous chapters, we described in detail the stage model of self-regulated behavior change. In
developing the stage model, Bamberg (2013a) correlated cross-sectional data. The model
postulates that behavior change is a temporal process through four qualitatively different stages:
predecision, preaction, action and postaction. Transition from one stage to the next is marked by
the formation of a particular intention, which in turn is influenced by a number of socio-cognitive
variables (discussed previously). Bamberg (2013a) acknowledged at least two weaknesses of his
study: (a) mere correlation of variables using cross-section data, and; (b) failure to assess some
constructs such as coping/action planning, maintenance- and recovery self-efficacy. Furthermore,
124
related researches in the health domain (e.g. Schwarzer, 2008) suggest a further division of the
postaction stage into initiative and maintenance – a distinction not considered in the current stage
model of self-regulated behavior change. In the present work, longitudinal data measuring the
model variables, including coping/action planning, maintenance- and recovery- self-efficacy,
obtained at two time points were used to confirm and validate the core ideas of the current stage
model, and to further extend it. This work thus represents an independent validation and extension
of the stage model using longitudinal data.
8.2. Association Between Stage and Car Use
We first examine the relationship between stage membership and car use level. We consider the
five stages (PreD=predecisional, PreA=preactional, A=actional, PostA1=early postactional or
initiative, and PostA2=late postactional or maintenance) and the weekly car use (values range from
lowest use at 0 to highest use at 10). We check both cross-sectional data at a single time point and
longitudinal data over two time points.
Figure 8-1. (a) Weekly car use with 95% confidence intervals and (b) car use change across stage changes
In Figure 8-1a we observe that the weekly car usage decreases as we progress towards advanced
stages (mean car use: PreD=9.69, PreA=8.25, A=7.39, PostA1=5.16). Nonetheless, considering
confidence intervals, there is a significant overlap among the first four stages. In the last stage, we
0
2
4
6
8
10
12
PreD PreA A PostA1 PostA2
wee
kly
car
use
at
bas
elin
e
Stages
-12-10
-8-6-4-202468
1012
-4 -2 0 2 4
Car
Use
Ch
ange
Post-stage membership relative to pre-stage membership
PreD PreA A PostA1 PostA2
a b
125
notice a big drop in the car use (PostA2 car use=0.64), suggesting that those in the late postactional
stage have adopted the low car-use behavior thus indicating that, in terms of car usage, PostA1
and PostA2 should be distinguished.
In Figure 8-1b we examine the change in car use with respect to the post-intervention stage
membership. The values on the x-axis denote the post-stage taking the pre-stage as reference. As
an illustration, consider the points on the line labeled PreD. These points represent the car use
change of those belonging to the predecisional stage at T0. After four weeks, some of them remain
in the same stage, while others advance to the next stages. We represent stage progression with
the integers on the x-axis taking as reference the stage at T0. Hence, those who remained in the
same stage after four weeks have their car use change plotted at x=0. The car use change of those
who progress one stage up is plotted at x=1, and of those who regress one stage down is marked
at x=-1. Looking at the PreD plot, we observe that as the students belonging to the predecisional
stage progress to more advanced stages over 4 weeks, they also reduce their car usage. In fact,
those who transition to PostA2 from PreD have decreased their car commuting to university by as
much as 6 trips/week and replaced it with alternative modes. We also observe the same trend in
other longitudinal data: progression through the stages is associated with reduction in car use.
Calculating the slope of the lines connecting these points (Table 8-2), we see that the slopes are
negative, indicating that as we advance through the stages, the car use is reduced as expected.
8.3. Stability of Stages
Table 8-1 presents the general distribution and transitions of stages over time. Following
observations appear important to us: Firstly, consistent with Bamberg (2013b) and Klöckner
(2014), many people “cycle and recycle” across stages. That is, we notice a pattern of progression
to advanced stages as well as regression to earlier stages. For example, out of the 75 individuals
in the predecisional stage at baseline, 8 “jump” to the most advanced stage after 4 weeks. At the
same time, 9 of the 40 preactional individuals descend back to the predecisional stage. Secondly,
we also observe a pattern of stability: some individuals remain at their current stages even after
four weeks. We see that the predecisional and postactional-late stages do not change much, the
pre-actional is moderately stable, and the actional stage is very unstable. Here, the advantage of
126
distinguishing the two postactional stages becomes clear: early post-action is characterized by
instability (especially relapse), while the late post-action by great stability. For instance, in the
early post-action, 5 out of 19 relapsed (26.3%), whereas in the late post-action, only 4 out of 89
did (4.5%).
Table 8-1 Stage distribution and transitions. Control and experimental groups combined.
T0=baseline; T1=4 weeks later; PreD=predecision; PreA=preaction; A=action;
PostA1=postaction(early); PostA2=postaction(late)
T0/T1 PreD PreA A PostA1 PostA2 N
PreD 53 9 3 2 8 75 (31.1%)
PreA 9 20 5 3 3 40 (16.6%)
A 2 8 3 2 3 18 (7.5%)
PostA1 2 3 0 4 10 19 (7.9%)
PostA2 1 1 1 1 85 89 (36.9%)
N
67
(27.8%)
41
(17.0%)
12
(5.0%)
12
(5.0%)
109
(45.2%)
241
8.4. Association Between Stages and Intention Strengths
Next we investigate the relationship between the stages and the strength level of the three intention
types. Looking at the cross-sectional data produced by combining the data from the two time-
points (Figure 8-2) reveals that the (a) predecisional stage is associated with negative levels of
goal, behavioral and implementation intentions; (b) preactional stage is associated with positive
levels of goal, but negative levels of behavioral and implementation intentions; (c) actional with
positive levels of goal and behavioral, but negative level of implementation intentions; and (d)
both postactional stages with positive levels of all three intentions.
Our preceding results are in close agreement with Bamberg (2013b), which claims that
predecisional individuals are characterized by lack of any intention; preactional individuals by
existence of goal intention; actional individuals by behavioral intention, in addition to goal
intention; and postactional individuals by all three intentions.
127
Figure 8-2 Average intention strengths across stages (with 95% confidence intervals; combined T0 and T1)
8.5. Progression in Stage Membership is associated with the formation of three types of
intentions
Next we investigate the relationship between stage progression and intention strengths. Looking
again at Figure 8-2 reveals that all the three intentions (i.e. their means) become stronger as stage
progresses. Nonetheless, if we examine two consecutive stages, we observe that there is both an
overlap and a discontinuity among intentions. For instance, comparing PreD to PreA, we notice
that the behavioral and implementation intentions overlap, but the goal intention does not.
Comparing PreA and A, however, we see that there is discontinuity on behavioral intention but an
overlap in goal and implementation intentions. In A and PostA1, the break is in the implementation
intention. We observe overlap among the three intention types in both postactional stages. These
discontinuity patterns in variables indicate the existence of stages (Armitage & Arden, 2002).
In Figures 8-3(a-c) we see from our longitudinal data that stage progression is associated with
increase in intention strengths. The slopes of these lines in the graphs are positive, indicating
increase in intention strengths with stage progression (Table 8-2).
-4
-3
-2
-1
0
1
2
3
4
5
6
PreD PreA A PostA1 PostA2M
ean
inte
nti
on
str
engt
h
Stages
Goal Int Beh Int Imp Int
128
Figure 8-3 Changes in: (a) goal intention; (b) behavioral intention; (c) implementation intention relative to stage
progression
Table 8-2 Slopes (R2) of the points in Fig 7-1(b), and Figs. 7-3(a)(b)(c) PreD PreA A PostA1 PostA2
Car Use -1.474
(0.825)
-1.200
(0.332)
-2.860
(0.960)
-0.290
(0.470)
-1.980
(0.430)
Goal Intention 0.463
(0.444)
0.643
(0.564)
1.192
(0.745)
-0.005
(0.000)
2.212
(0.868)
Beh Intention 0.959
(0.361)
0.822
(0.625)
1.558
(0.935)
1.272
(0.744)
1.867
(0.499)
Imp Intention 1.029
(0.713)
0.392
(0.192)
2.058
(0.761)
1.648
(0.706)
1.006
(0.259)
-10
-8
-6
-4
-2
0
2
4
6
8
10
-4 -3 -2 -1 0 1 2 3 4
Ch
ange
in g
oal
inte
nti
on
Post-stage membership relative to pre-stage membership
PreD PreA A PostA1 PostA2
-10
-8
-6
-4
-2
0
2
4
6
8
10
-4 -3 -2 -1 0 1 2 3 4
Ch
ange
in b
ehav
iora
l in
ten
tio
n
Post-stage membership relative to pre-stage membership
PreD PreA A PostA1 PostA2
-8
-6
-4
-2
0
2
4
6
8
-4 -3 -2 -1 0 1 2 3 4
chan
ge in
imp
lem
enta
tio
n in
ten
tio
n
Post-stage membership relative to pre-stage membership
PreD PreA A PostA1 PostA2
a b
c
129
The results seem to suggest that progression to later stages is associated with an increase in
strengths in all three intention types, but transition to a particular stage is especially associated
with a specific intention type crossing a certain threshold. The transition from PreD to PreA is
significantly correlated with goal intention only, from PreA to A with behavioral intention only,
and from A to PostA1 with implementation intention only.
To confirm our results, we perform an ordinal logistics regression on the ordered categorical
variable (the time-ordered sequence of stages) using the three intention types as explanatory
variables (Hedeker, Mermelstein, & Weeks, 1999; Bamberg, 2013a). We can posit a continuous
latent “readiness of change” variable, which is divided at certain thresholds (cut-off points) to
make the stages that we observe in our data. In the present case, as we have five stages, we can
conceptualize four thresholds that separate these ordered stages. We can then assess the role of
the predictors – the three intention types – on crossing these thresholds. Since we presume that
predictors have differential effects on the thresholds, we utilize nonproportional logistics
regression (Hedeker & Mermelstein, 1998; Hedeker, Mermelstein & Weeks, 1999; Lippke,
Ziegelmann & Schwarzer, 2005; Bamberg, 2013a). Following Bamberg (2013a), we dichotomize
the three intention variables using median split prior to the analysis. Table 8-3 presents the results
of cumulative and adjacent categories models.
In the cumulative model, the first cumulative logit compares PreD versus the four next stages
combined (PreA-A-PostA1-PostA2). This is to assess the effect of the three intention predictors
on the threshold between predecision to preaction. Since the size of the estimate is difficult to
interpret (Bamberg, 2013a), we only determine which of the intentions are significantly associated
with crossing the thresholds. We see that goal and implementation intentions are significant
predictors at 99% level, and behavioral intention at 90% level. In the second cumulative logit, we
compare the first two stages versus the next three stages (i.e. PreD-PreA versus A-PostA1-PostA2)
to evaluate the association of the three intentions on the preaction-action threshold. All three
intentions are significant predictors at 99% level on the transition from preaction to action.
Similarly, from the third cumulative logit, which compares the three early stages versus the two
latter stages (i.e. PreD-PreA-A versus PostA1-PostA2), we find that all three intention types are
significant. Finally, in the fourth cumulative logit, comparing the first four stages versus the last
130
stage (PreD-PreA-A-PostA1 versus PostA2), we see that only goal and implementation intentions
are significant. The negative sign of the logit estimate indicates a decrease of the threshold
between two stages; that, is an increase in the probability of crossing that threshold (c.f. Bamberg,
2013a). The results confirm our previous observation from the graphs (c.f. Fig. 8-2 and 8-3) that
stage progression is associated with all the intention types.
Next we compare two consecutive stages through adjacent-category ordinal logistic regression.
Having been in a particular stage, we want to determine the significant predictor for progression
(or regression) to the immediate next stage. We first compare predecision versus preaction and
find that only goal intention is closely associated with the transition between the two stages.
Comparing next preaction versus action, we observe that only behavior intention is a significant
predictor. Comparing action and early postaction, we find that only implementation intention is
significant. Finally, if we compare the two post-action stages, we see that none of the intentions
is significant. This confirms our previous result from the graph using cross-sectional data that
transition to a particular stage starting from a stage immediately prior to it is especially associated
with a specific intention type crossing a certain threshold. The transition from PreD to PreA is
significantly correlated with goal intention only, from PreA to A with behavioral intention only,
and from A-PostA1 with implementation intention only.
Table 8-3 Results of the Ordinal Logistic Regression Using Cumulative and Adjacent Categories
Significance codes: Bold 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘a’ 0.1 Cumulative Categories Adjacent Categories
Estimate Standard Error Estimate Standard Error
{PreD} vs. {PreA-A-PostA1-PostA2} PreD-PreA
Intercept 0.4252** 0.1447 Intercept 0.8073*** 0.1667
Goal Intention -2.2061*** 0.4070 Goal Intention -1.8157*** 0.4736
Behavioral Intention -0.5399 a 0.2981 Behavioral Intention 0.2692 0.3871
Implementation Intention -2.4516*** 0.4345 Implementation Intention -0.8550 0.5473
{PreD-PreA} vs. {A-PostA1-PostA2} PreA-A
Intercept 1.8067*** 0.1831 Intercept 1.8928*** 0.3452
Goal Intention -1.6273*** 0.2770 Goal Intention -0.5999 0.4805
Behavioral Intention -1.0504*** 0.2791 Behavioral Intention -1.2591* 0.5252
Implementation Intention -2.5877*** 0.3143 Implementation Intention -0.7829 0.5660
131
{PreD-PreA-A} vs {PostA1-PostA2} A-PostA1
Intercept 2.0972*** 0.1990 Intercept 0.6895 0.5156
Goal Intention -1.3036*** 0.2590 Goal Intention 0.1802 0.5508
Behavioral Intention -0.8275** 0.2823 Behavioral Intention -0.3720 0.6572
Implementation Intention -2.4749*** 0.2875 Implementation Intention -1.0345a 0.6081
{PreD-PreA-A-PostA1} vs {PostA2} PostA1-PostA2
Intercept 2.3010*** 0.2128 Intercept -1.3198** 0.4506
Goal Intention -1.3264*** 0.2574 Goal Intention -0.6479 0.4168
Behavioral Intention -0.4619 0.2930 Behavioral Intention 0.5317 0.5390
Implementation Intention -2.2381*** 0.2833 Implementation Intention -0.7278 0.4839
Model fit (-2LogL) 975.7266 Model fit (-2LogL) 977.6582
Model fit (df) 1912 Model fit (df) 1912
In summary, we can describe the five stages as follows:
Table 8-4 Summary of characteristics of the five stages Stage Car Use Goal
Intention
Behavioral
Intention
Implementation
Intention
Stability Condition for Transition
to next immediate stage
PreD High Low Low Low Stable Goal intention formation
PreA High Moderate-
High
Low Low Moderately
stable
Behavioral intention
formation
A High High Moderate-
High
Low Unstable Implementation intention
formation
PostA1 Moderate High High High Unstable -
PostA2 Low High High High Stable -
8.6. Determinants of the three intention types: model structure and parameter estimates
So far, from the preceding results, we can make the following observations regarding the
mechanism of change in the reduction of car use: decrease in car use is associated with progression
through a sequence of temporal stages, which is also associated with an increase in three intention
types. In this section, we consider the determinants of each of the three intention types.
Our proposed determinants are drawn from the stage model of self-regulated behavior change
theory (Bamberg, 2013a; 2013b). We drop though three determinants from the original model in
132
our subsequent analyses – namely, negative emotions, personal responsibility and problem
awareness. Negative emotion is dropped because of unreliability of the instrument measuring it
as discussed previously. Personal responsibility and problem awareness are dropped because we
find that these are not significant predictors of stage membership.
Figure 8-4 Original model (based on Bamberg 2013b,
adjusted to reliable constructs available in our study)
We test all models using R 3.4.0 software using the maximum likelihood estimation method. First,
we fit our combined data from two time-points (N=482) to the original model (Fig. 8-4) proposed
by Bamberg (2013a); though our previous analysis in general supports the model structure, we
find a poor fit with χ2 (df = 43, n = 482) = 868.337, p <0.001; RMSEA = 0.200; CFI = 0.691; TLI
= 0.590; SRMR = 0.203.1 We therefore make some modifications to the original model. Based on
previous results, we see that the general path of successive formation of three types of goals leading
to the new behavior is supported by our data. Next we specify the determinants of each of the
intentions. We hypothesize that goal intention to reduce car use is activated by personal norms,
positive emotions and perceived goal feasibility, and that attitudes towards a particular alternative
are the main determinant of choosing that alternative over car (behavioral intention). Moreover,
we assume that the ability to make a plan of action to implement the new behavior and to cope
with unexpected situations are a significant predictor of the implementation intention, and the
ability to recover from relapse is the main factor for the maintenance of the behavior.
1 To assess model fit, we report the Chi-square and degree of freedom (χ2 and df), including the p-value, Tucker-
Lewis index (TLI), Comparative fit index (CFI), Root Mean Square Error of Approximation (RMSEA), and
Standardized Root Mean Square Residual (SRMR). A χ2/df ratio of 2 is acceptable, with an insignificant p-value
(p>0.05). However, since χ2 is sample size dependent, we consider other indices. TLI and CFI values greater than
0.95 are acceptable. An RMSEA of <0.07 and SRMR of <0.08 are considered acceptable (Hooper, Coughlan &
Mullen, 2008).
133
After several iterations and confirming modification indices, we obtain the following base model
with parameter estimates (Fig. 8-5). Other constructs from the original model are insignificant and
dropped. We test the base model using combined data from baseline and four weeks later (Group
A in Table 8-5), and we confirm an acceptable model fit. We fit the same path analysis model
against various subgroups, and we see that the model fit is also good. The first two subgroups,
distinguished by exposure to treatment, consist of individuals assigned in the control and
experimental groups. The next two subgroups, distinguished by time-point, consist of pre- and
post- intervention groups. The last two subgroups are distinguished by their stage assignment in
the behavior change process (predecision and preaction are considered early stages; action and
post-action are late stage memberships). The acceptable model fit indices of the different
subgroups establish the acceptability of the base model and invariance of the model structure
regardless of treatment exposure, time-point, and stage membership (Table 8-5).
Figure 8-5 Base model. We combine data from baseline and 4 weeks later (N=482). Fit indices reveal a good fit:
χ2 (df = 15, n = 482) = 55.262, p <0.001; RMSEA = 0.066; CFI = 0.983; TLI = 0.967; SRMR = 0.023. Refer to Table 6-1 for parameter definition. **significant at 0.05 level ***significant at 0.01 level
Table 8-5 Model structure invariance: Good model fit indices of multiple groups indicate
acceptability of the model and invariance of model structure across groups N χ2 df p RMSEA CFI TLI SRMR
A. Complete Sample 482 55.262 18 <0.001 0.066 0.983 0.967 0.023
B. Control group (complete) 230 42.591 18 0.001 0.077 0.976 0.954 0.027
C. Experimental group (complete) 252 38.586 18 0.003 0.067 0.982 0.966 0.023
D. Pre-intervention group 241 18.888 18 0.399 0.014 0.999 0.998 0.017
E. Post-intervention group 241 43.700 18 0.001 0.077 0.979 0.960 0.027
F. Early stages group 115 15.010 18 0.661 <0.001 1.000 1.020 0.022
G. Late stages group 126 21.167 18 0.271 0.037 0.989 0.979 0.035
134
8.7. Discussion: Validation and Extension of the Stage-based Behavior Change Model
Having confirmed structural invariance using cross sectional data2, we now discuss the model
shown in Fig. 8-5.
The core idea of our findings, so far, is that behavior change – i.e. the decrease in car use – is
associated with progression through a temporal sequence of stages: predecision, preaction, action
and postaction. Individuals who transition to advanced stages in the behavior change process show
a corresponding reduction in their car use. Stage progression is, in turn, associated with an increase
in three intention types: goal, behavioral and implementation intentions. Nonetheless, transition
from a particular stage to its immediate higher stage is associated with a specific intention type,
not only increasing, but also crossing a certain threshold.
In the first (pre-decisional) stage, the individual performs the environmentally harmful behavior
in a habitual way, without much deliberation. His behavior is characterized by a high level of car
use, and low levels of goal, behavioral and implementation intentions. In order to encourage
transition to the second (preactional) stage, he must be motivated to develop a strong goal intention.
The felt moral obligation to fulfill his personal norm and the anticipation of feeling good if he does
so, together with a general favorable attitude toward the alternative behavior, may urge him to
form a goal. Nonetheless, whether he actually binds himself to a goal depends on his sense of self-
efficacy or ability to perform the behavior with relative ease (recovery efficacy and perceived
behavioral control). The formation of a strong goal intention marks the transition to the preactional
stage.
In the pre-actional stage, the high car-use individual has the general goal of changing his current
behavior (high goal intention); however, he does not yet have a personally most suitable means to
achieve this goal (hence, low behavioral and implementation intentions). Because several actions
can lead to the same goal, the task that the individual is confronted with is to select the most
2 Although structural invariance of the model is established, invariance of parameter estimates across different
groups cannot be claimed, as multi-group analysis can show.
135
personally suitable behavioral strategy among various options by weighing their pros and cons. It
is assumed that goal intention and two other factors influence the formation of a behavioral
intention. These factors are attitudes toward behavioral alternatives and the action plan. We think
that when a person considers choosing an alternative over a car, he considers not only his general
goal and attitudes toward that alternative, but also his perceived ability to make a trip plan using
that alternative. The formation of a (strong) behavioral intention marks the transition to the third
(actional) stage of the behavioral change process.
In the actional stage, the individual, who still uses car often, has strong goal and behavioral
intentions; however, he has not yet put his behavioral strategy to practice (thus, low
implementation intention). The task therefore for the individual is to actually implement the
behavioral strategy chosen for goal achievement (i.e. he must form an implementation intention).
It is assumed that behavioral intention and three additional factors – action planning, coping
planning and recovery self-efficacy – promote the formation of an implementation intention (c.f.
Hsieh, Kanda, Fujii, 2017). Action planning refers to the specific situation parameters, e.g. “when”
and “where”, and the sequence of actions necessary to implement the target behavior, i.e. “how”.
Coping planning refers to the ability to anticipate situations that may hinder the individual to
perform the target behavior and to develop a plan to cope with such situations. Recovery self-
efficacy, in this stage, refers to the person’s confidence in his ability to sustain the behavior and
resume it after relapse. The formation of an implementation intention marks the transition to the
third and final stage, the post-actional stage.
In the post-actional stage, the individual has successfully implemented the new behavior (low car
use; high goal, behavioral and implementation intentions). Therefore, the task he is confronted
with is to maintain his behavior and avoid relapse into old behavioral patterns (Bamberg, 2013b).
Aside from the well-formed implementation intention, it is assumed that self-efficacy or ease of
performing/maintaining the behavior (recovery efficacy and perceived behavioral control) is an
additional factor in maintaining the desired behavior. Goal intention also still has direct influence
over the behavior (and is not simply mediated by other intentions), which may signify that the goal
or purpose of doing the desired behavior must always be reconsidered. To avoid relapse, people
must always maintain the goal they have set.
136
Our results also suggest that post-actional stage needs to be further distinguished into initiative
(early postaction) and maintenance (late postaction). The former indicates that the action has been
initiated but it has not settled yet as a habit. The latter refers to habitual, sustained behavior.
Although the two post-actions are distinguished, our model as shown in Figure 8-5 does not
associate any intention with transition from initiative to maintenance. Thus, the distinction is
based only on the position of the individual along a continuum within the same stage. However,
if the two post-actions are to be distinguished as qualitatively separate stages, then it must be
established that an intention, such as, say, “intention to maintain behavior” (c.f. Luszczynska et al,
2007), is associated with transition from initiative (early post-action) to maintenance (late post-
action) stages. This is beyond the findings of our current work.
In our model, the recovery efficacy construct has a direct path to goal intention, implementation
intention and behavior. This means it is an important determinant in the predecisional, actional
and postactional stages. This may come as odd since recovery efficacy is usually associated with
recovery from relapse after adopting the new behavior. Nonetheless, since we measure recovery
efficacy as “I will continue to use alternatives to go to school, even though this may be
inconvenient”, we hypothesize that this statement may be interpreted differently by individuals
belonging to various stages. For those in the pre-decisional stage, they may interpret this question
as: “I am a car user now, but if I am to begin using alternatives, will I be able to continue doing
this behavior, even though this may be inconvenient?” In the post-actional stage, this can be
interpreted as: “I am an alternative user now. Will I be able to continue with this?” Moreover, we
observe in our parameter estimates that a large magnitude of recovery efficacy coefficient is
associated with later stages and a smaller value to the early stages, which is in line with our
expectation.
137
CHAPTER 9
Identifying the pathways of travel behavior change induced by
technology-based intervention among university students
Chapter Summary. We describe how a mobility behavior change support system, called Blaze,
is able to achieve its effects in changing the travel behavior of university students. We identify a
causal pathway linking the effect of the technology intervention to its behavioral outcome through
the mediation of a number of variables. The validated stage model of self-regulated behavioral
change (SSBC), described in the previous chapter, is used as a theoretical framework to understand
how the outcome may be influenced by determinants (conceptual theory), and how the
determinants may be activated by different intervention types (action theory). We show that the
action and conceptual components of the Blaze impact pathway on travel behavior change differ
by stages: individuals belonging to later stages change their behavior via the mediation of a change
in implementation intention, while those in the early stages undergo behavior change through the
mediation of a change in self-efficacy. We discuss the implications of our results on the potential
role of technology interventions in mobility management.
Keywords. Mobility behavior change support system; travel demand management; technology-
based intervention; mechanism of change; stage model
9.1. Introduction
With the widespread adoption and pervasive use in society of information and communication
technologies (ICT), technology-based interventions (TBI) to modify behavior, notably in the
health domain, have undergone rapid development in the last decades. Nonetheless, in the domain
of travel behavior modification, less advances have been made (Cohen-Blankshtain and Rotem-
Mindali, 2016). In the literature review in chapter 3, we observed that voluntary travel behavior
change programs have not yet fully taken advantage of the ICT platform, which is unfortunate
since some empirical evidence suggest that TBIs can produce effects that are comparable to, or
138
better than, approaches delivered through conventional methods (e.g. Jariyasunant et al, 2015;
Bamberg et al, 2015).
In technology-based intervention studies, researchers must establish not only the cause of behavior
change, but also the mechanism of change (Dallery et al, 2015). That is, they must provide
evidence through what processes the intervention, previously demonstrated to be effective,
achieved its effects. Often, this requires identifying the variables that mediate the effect of an
intervention to the desired outcome. Fortunately, in recent years, integrated theoretical
frameworks in travel behavior change – for example, the Stage Model of Self-Regulated Behavior
Change (Bamberg, 2013a) and Comprehensive Action Determination Model (Klöckner and
Blöbaum, 2010) have been proposed, which can give insights on the possible mediators in the
causal chain of travel behavior change.
In this chapter, we focus on the mechanism of change induced by a technology-based intervention
for travel behavior change, called “Blaze”. The name “Blaze” comes from “trailblazers”,
individuals who show the way or new behavior to a larger population. In Blaze, social traces or
patterns of behavior of trailblazers are provided as social information to others in order to induce
behavior change. In the development of the technology-based interventions within Blaze, the
Stage Model of Self-Regulated Behavioral Change (SSBC) is used. Analysis shows that Blaze is
effective in inducing a positive change across a wide range of indicators, as specified by the stage
model: car use reduction, stage membership, intentions and socio-cognitive determinants.
Research on mechanisms in travel behavior change in non-technological context has been done
(e.g. Thøgersen, 2009); but, to the best of our knowledge, there is none yet on the technological
context. Dallery et al (2015) and Baraldi et al (2015) argue there may exist significant differences
in mechanisms of change between non-technological and technological contexts. Our work thus
represents one of the first studies on the mechanisms associated with TBIs for travel behavior
change. In contrast, studies on the nature of mechanisms of behavior change in the health domain,
including reviews, have been extensively carried out (e.g. Schwarzer et al, 2011; Dallery et al,
2015).
139
Understanding the mechanisms of change from intervention to outcome typically involves
mediation analysis. In this analysis, two theoretical components are clarified: the action theory
and the conceptual theory. The first component, action theory, represents a theory about the
relationship between the intervention and the mediating variable. The second component,
conceptual theory, represents the relation between the mediating variable and the outcome (Baraldi
et al, 2015).
Both components are based on prior research about the relationships among the interventions,
potential mediators and the outcome of interest. Recently, Bamberg (2013a) proposed a
comprehensive theoretical framework of travel behavior change – the Stage Model of Self-
Regulated Behavioral Change (SSBC). The theory provides a blueprint of relationships of how
the outcome may be influenced by determinants (conceptual theory), and how the determinants
may be activated by different intervention types (action theory) (Bamberg et al, 2011; Bamberg,
2013a; Bamberg, 2013b). In our work, the SSBC is used to guide the systematic development of
interventions of Blaze. Using SSBC as a blueprint, we develop interventions targeting and
activating the determinants in the model.
After this Introduction, we discuss the pathway of behavior change from intervention to outcome
via the mediators in Section 9.2. We discuss the results in Section 9.3. In Section 9.4, we present
the limitations and implications of our results, and discuss areas for future work.
9.2. Pathway from Interventions to Behavior Change via Change in Mediators
In this section, we consider the specific mechanism by which the intervention influences the
mediating variables, which in turn influences the outcome. In this regard, we construct two models.
In the first model, the influence of the antecedents, the previous behavior and the intervention on
current behavior is analyzed. In the second model, we analyze whether behavior change is
influenced by and whether behavioral impacts of the intervention are mediated through changes in
the antecedents (c.f. Thøgersen, 2009; Dallery et al, 2015). The model structure is simply the base
model (see Chapter 8, Figure 8-5).
140
In Model 1, we model the effect of intervention on car use at T1, using cross-sectional data at T1
and entering the car use at T0 as covariate. In the model we only draw paths from the intervention
to intentions and behavior. We ignore the paths from interventions to the socio-cognitive variables.
The fit indices show a good fit: χ2 (df = 30, n = 241) = 41.906, p =0.073; RMSEA = 0.041; CFI =
0.991; TLI = 0.984; SRMR = 0.014. The R2 values are also high, indicating that our model can
explain well the variances in the intentions and behavior. We present the results of our estimates
in Table 9-1.
Table 9-1 Parameter Estimates of Model 1
Path B SE p β R2
gi0 → gi* .112 .048 .019 .112
.656
att → gi*** .388 .049 <.001 .379
reff → gi*** .227 .048 <.001 .284
posem → gi .085 .060 .153 .072
pnorm → gi** .137 .046 .003 .146
pbc → gi .025 .046 .585 .031
intervention → gi .194 .204 .340 .037
bi0 → bi*** .221 .046 <.001 .199
.722 gi → bi** .172 .063 .006 .141
att → bi** .169 .060 .005 .135
ap → bi*** .496 .042 <.001 .540
intervention → bi .281 .222 .205 .044
ii0 → ii*** .141 .034 <.001 .134
.858
bi → ii*** .206 .045 <.001 .194
cp → ii .009 .049 .861 .008
ap → ii** .137 .052 .008 .140
reff → ii*** .574 .046 <.001 .551
intervention → ii** .486 .171 .005 .071
beh0 → beh*** .568 .045 <.001 .578
.743
ii → beh* -.220 .095 .021 -.170
gi → beh -.001 .079 .992 -.000
pbc → beh .029 .065 .662 .021
reff → beh** -.286 .109 .009 -.213
intervention → beha -.500 .299 .095 -.057
Significance codes: Bold ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘a’ 0.1
Relative to the base model, some paths in Model 1 are not anymore significant: pbc→ gi; posem
→ gi; cp → ii; gi → beh; pbc → beh. We note that the base model is obtained by combining the
data from both time-points (baseline and 4 weeks later), with bigger sample size (N=482). Model
1, however, is constructed using data at a single time-point (4 weeks later), with smaller sample
size (N=241). Nonetheless, the acceptability of the indices of Model 1 indicates that the model
141
structure essentially remains invariant, but the invariance of parameter estimates cannot be claimed,
as previously pointed out.
From the same table, we observe that previous intentions have strong effect on present intentions
(the paths gi0 → gi, bi0 → bi, and ii0 → ii are all significant at different levels). Moreover, we find
that car use behavior at T0 (beh0) is a significant covariate. This path is significant at 99.9% level.
Thus, past behavior is a stronger predictor of current behavior than any of the previously mentioned
antecedents/determinants. At the same time, car use is also a conscious or volitional decision
(implementation intention and recovery efficacy, which are the volitional factors, are significant
predictors of behavior in the model, although their influence is relatively weaker).
We now check if the intervention succeeds in influencing the intentions. Goal and behavioral
intentions are not significantly influenced by the intervention (the paths intervention → gi and
intervention → bi are not significant). Nonetheless, the intervention has a significant effect on
implementation intention. This may imply that our technology-based intervention, or any other
technology tool for that matter, is only a useful aid in putting plan into practice, and therefore may
only affect the proximal intention (i.e. implementation intention) and not the distal ones (goal and
behavioral intentions). Furthermore, we observe that the intervention has a direct on behavior as
well (the path intervention → beh is significant at 90% level).
The preceding results imply that the intervention has mediated effect (via implementation
intention) on car use. We also test if the intervention has mediated effect via recovery efficacy but
we find that the effect of intervention on recovery efficacy is not significant.
Next, we construct Model 2, referred here as change model shown in Figure 1 (c.f. Dallery et al,
2015). Since the pathway may be stage-dependent, we fit two sets of data into our model: the first
dataset comes from the early stages (combined predecision and preaction), and the second dataset
is obtained from the late stages (combined action and postaction).3 This is done because changes
3 Ideally, we should fit five datasets, corresponding to five stages we identified in our stage model. However, due to
limited sample sizes for some stages, we combine some stages together and simply distinguish two general
stages/phases: “motivational” (i.e. predecision and preaction) and “volitional” (i.e. action and two post-actions). For
theoretical background on the distinction between motivational and volitional phase, kindly refer for example to
Schwarzer, Lippke & Luszczynska (2011).
142
in the endogenous variables due to changes in their determinants are mediated by stages. For
example, change in the implementation intention is assumed to be significant predictor of change
in behavior only in the latter stages.
In our (change) model, we examine if the intervention induces a change in the determinants and in
the intentions, which then cause the behavior to change. For both data sets, the model fit is slightly
lower, though acceptable. This confirms that our base model structure (see Chapter 8, Figure 8-5)
has acceptable predictive power not only for cross-sectional data but also for change data. In other
words, it can predict reasonably well change in intentions and weekly car use.
Figure 9-1 Model 2, the change model. Path coefficients are reported early stages/late stages. Significant path
coefficients at 95% level are in bold, and at 90% level are underlined. Fit indices of early stages: χ2 (df = 18, n =
115) = 19.921, p=.337; RMSEA = 0.031; CFI = 0.990; TLI = 0.978; SRMR = 0.037. Fit indices of late stages: χ2
(df = 18, n = 126) = 31.223, p=0.027; RMSEA = 0.076; CFI = 0.946; TLI = 0.885; SRMR = 0.041.
We highlight following observations from Figure 9-1. First, we note that the intervention is able
to cause a change in the behavior (intervention→ Δbeh), regardless whether the individual is at
the early or late stages (at 90% significance level). This validates the effectiveness of our
intervention in inducing a change in actual behavior. Our follow-up question is then how the
intervention is able to achieve its effects.
143
First, we consider the results corresponding to the late stages. Our change model using data from
the late stages shows that the paths (Δgi → Δbi; Δbi → Δii; Δii → Δbeh) are all significant. This
means, by causing changes to the goal intention, we can change the behavioral intention, which in
turn can induce change in implementation intention, possibly leading to a change in behavior.
However, our intervention is not able to cause a change in goal and behavioral intentions among
those in the late stages. The paths (intervention → Δgi; intervention → Δbi) are insignificant.
However, we observe that the path (intervention → Δii) is significant. This indicates that a
pathway exists from intervention to behavior change via the mediation of, among other factors, a
change in implementation intention (intervention → Δii → Δbeh).
Looking at the results corresponding to the early stages, we observe that the intervention likewise
succeeds in inducing a significant change in the implementation intention. The path (intervention
→ Δii) is significant. Considering that the individual is in the early stages, the intervention is thus
able to make the habit-driven student (note that habit influences current behavior) reflect and
conscious about his behavior, which is seen in the change in implementation intention.
Nonetheless, the change in implementation intention does not have a significant effect on the
change in car use. This makes sense because since the student is in the early stages, a change in
his implementation intention does not necessarily translate to a change in car use behavior. This
is confirmed by the path (Δii → Δbeh), which is insignificant and almost negligible. Therefore,
the path (intervention → Δii → Δbeh) cannot be claimed as a possible pathway of intervention.
Examining if the other candidate path, via the change in recovery efficacy, can be considered a
pathway, we find that the link (intervention → Δreff) is significant (p<0.01). Hence, we can
suppose that the intervention is able to cause a change in behavior via, among other factors, a
change in recovery efficacy (intervention → Δreff → Δbeh).
What our results show is that our technology-based intervention succeeds in causing change in
behavior via the mediation of implementation intention and self-efficacy variables (i.e. recovery
144
efficacy). Although we identified a number of possible mediators drawn from the stage model of
behavior change, in the end mediation occurs only through the implementation intention and self-
efficacy variables. Our findings are somehow in agreement with Dallery et al (2015), who in their
review of technology-based interventions for tobacco, alcohol and substance abuse, enumerated a
number of studies that identified the main mechanism of behavior change via self-efficacy.
9.3. Discussion
In this section, we discuss the mechanism of change based on our results. We discuss the
mechanism separately for conceptual and action components. The conceptual component tackles
the relationship between the mediating variables and the outcome of interest, while the action
component deals with the relationship between the intervention and the mediating variables.
9.3.1. Conceptual component
In summary, we claim the following about the mechanism of change in the reduction of car use.
Previously, in Chapter 7, we thoroughly discussed the results of our validation of the stage model
of self-regulated behavior change. In this section, we simply summarize briefly our previous
discussion.
Decrease in car use is associated with progression through a sequence of temporal stages. We
observe that there is a corresponding reduction in car use among those who transition to advanced
stages in the behavior change process. Transition to later stages is also associated with an increase
in three intention types: goal, behavioral and implementation intentions. We also find some
evidence that transition to a particular stage is especially associated with a specific intention type,
not only increasing, but also crossing a certain threshold. Behavior is not only determined by
implementation and goal intentions, but also by self-efficacy or ease of behavior (recovery efficacy
and perceived behavioral control). Recovery efficacy has the largest influence on behavior,
followed by implementation intention. Implementation intention is predicted by behavioral
intention, action planning, coping planning and recovery efficacy. Behavioral intention is formed
by goal intention, action plan and attitude towards alternatives. Goal intention is activated by
145
personal norm, attitude, anticipated positive emotions and self-efficacy (recovery efficacy and
perceived behavioral control).
9.3.2. Action component
Our main conclusions regarding the effect of the intervention, Blaze, on the mediators are as
follows. The intervention is able to induce significant changes in behavior. For individuals
belonging to the early stages in the behavior change process, the intervention also succeeds in
making them become more conscious about their behavior, which is reflected in a change in
implementation intention. However, change in implementation intention does not have a
significant effect on the change in car use. Instead the intervention is able to induce a change in
behavior via inducing a change in sense of recovery efficacy. For individuals in the late stages,
the intervention induces behavior change through the mediation of changes in the implementation
intention.
9.4. Limitations, Implications and Future Direction
We acknowledge a number of weaknesses. First, although we claim our study has a longitudinal
character, we are only limited by two data points (before-after the intervention). Our study design
is able to capture the immediate effect of the technology intervention, but we fail to understand its
(potential) long-term effect. Other studies (e.g. Thøgersen, 2009) conducted data collection in
three waves (before and after the intervention, and 6 months later). Such studies are able to assess
the long-term impact (if any) of the intervention. Klöckner (2014) also employed repeated
measures for two months. Studies like these, with frequent data collection, are able to capture, not
only between-group dynamics, but also the intra-person (within-person) dynamics over time.
Second, we have a small sample size, especially for some stages. We observe that many of our
respondents belong to the predecision and late post-action but very few of them belong to certain
stages, such as action and early post-action. This makes it necessary to consider with caution our
analysis regarding stage progression. Third, we employed quasi-experimental design, and not a
randomized controlled trial (RCT) as our experimental design, a standard requirement for studying
mechanism of effect (Kazdin, 2007). In our study design, we used convenience sampling primarily
146
due to resource constraints, but in RCT, samples must be drawn randomly from the population.
This limits our ability to draw robust inferences between intervention and change, as well as to
generalize our results to general population. However, as Bamberg and Rees (2017) recently noted,
in transportation research, quasi-experimental design may count already as sufficient evidence,
depending on the objectives of the evaluation. Finally, though we were able to demonstrate the
effectiveness of our technology-based intervention as a whole, we were not able to identify the
specific elements, components or aspects of Blaze, our TBI (technology-based intervention), that
produced the change (c.f. Ritterband et al, 2009). Identifying the elements within the intervention
that effect the change is essential in informing and guiding future intervention developments4.
Nonetheless, we suggest our results have important implications for intervention development
involving technology. First, we have seen that a TBI is able to induce changes in the mediating
variables and also car use behavior. Technology must be then further exploited in delivering large-
scale interventions aimed at promoting sustainable travel behavior.
Second, the fact that the mechanisms of change are stage-dependent highlights that one size does
not fit all. Abraham (2008) identifies two possible tailoring approaches: stage-matched and menu-
based behavior-change interventions. Stage-matched interventions are those tailored to match the
needs of distinct, stage-defined groups (e.g. Bamberg, 2013b). Menu-based interventions are
generic interventions that allow individuals to self-tailor and select from a “menu of interventions”
any material they find most relevant using self-assessments. In Blaze, we combine both approaches.
Third, our technology-based intervention is able to effect changes only in the mediating variables
that are most proximate to behavior (e.g. implementation intention and recovery efficacy). This
shows our intervention affects mainly the latter “volitional phase” (goal-pursuit), and not the
earlier “motivational phase” (goal-setting), of the behavior change process. It appears then that
our intervention – and perhaps, any technology-based intervention – is effective because it
4 We did administer a mid-term survey two weeks after first contact, asking participants to rate the usefulness of some
features of Blaze. Ideally, if the sample is large enough, we can identify the elements or components within Blaze
that caused the changes in the mediators, as specified by the stage-based model. We obtained N=125 responses.
However, the number of samples that are matched with responses from other time points, and therefore can be used
for subsequent analysis, is only N=43; this sample size is too limited to detect any significant changes that will enable
us to draw out conclusions.
147
provides technical support such as technology-enabled implementation plans (e.g. trip/travel plans
with a simulation of potential outcomes) to individuals so they can be helped to translate their
intention into actual behavior and also to break any unwanted (car-use) habit that hinders the
initiation of the new wanted behavior (Adriannse et al, 2011). If the initiation of this new behavior
can be sustained over long term with the aid of technology, then habits can be formed.
However, our experiments are not sufficient to rule out the potential of technology-based
interventions in influencing behavior change among those who have deeply-entrenched
orientations. This can be an area for future work. Although our TBI is able to change only the
proximate implementation intention and not the more distal ones (e.g. goal and behavioral
intentions), we also imagine that appropriate TBIs can be designed that can influence the more
distal determinants, changing even ingrained beliefs, attitudes and lifestyles. If this is achieved, it
will have important implications for soft transport policy measures. In mobility management,
interventions aimed at changing attitudes and tendencies are mostly carried out through campaigns
and personal dialogue by human agents (e.g. Davies, 2012). Furthermore, in mobility management,
the short-term effect of interventions is easily demonstrable, but persistence of their effects on
behavior change is largely unproven and area of further work as noted in two studies in Bamberg
et al (2011) and Richter et al (2011). In conclusion, we suggest that technical surrogates can
support or replace human agents and therefore broaden the impact range of mobility management
both in terms of population reached as well as possibly the time period of the intervention and its
effect.
148
CHAPTER 10
Conclusion
10.1. Summary
The main objective of this dissertation is to design, develop and evaluate a mobility behavior
change support system (mBCSS) as a travel demand management measure in the university setting
for developing world cases. In the design, development and evaluation of the system, the stage
model of self-regulated behavior change theory is used as a blueprint and theoretical framework.
To identify gaps in knowledge and therefore to position this study within existing literature, we
undertook a review of the literature.
In Chapter 2, we reviewed some travel demand management (TDM) measures implemented in
universities or schools around the world. Two main issues in the literature emerge from this
review: the dominance of hard TDM measures and the transferability of documented measures to
cases in the developing world.
In Chapter 3, we carried out a review of existing mobility behavior change systems for promoting
sustainable travel behavior. We extracted the persuasive features embedded in these systems and
evaluate their persuasive potential by using the persuasive systems design (PSD) model. Our
evaluation reveals that some features crucial for successful travel behavior change, such as
tunneling, rehearsal and social facilitation, are missing. Furthermore, we assessed studies
conducted to evaluate the effectiveness of these BCSSs in changing behavior and find indications
that effect sizes are mostly small though methodologically robust studies are largely missing and
hence no definitive conclusion yet can be derived. Based on these findings as well as literature
related to public health where BCSSs appear to be further developed, we then derived three
important suggestions: explicitly grounding the BCSS on a behavior change theory; provision of
trip/travel plan which the users can use to rehearse their behavior; and making use of social
interaction.
149
In Chapter 4, we discussed how behavior change theories can be used to inform the development
of mobility behavior change systems. Specifically, we discussed how stage models (namely, the
transtheoretical model and the stage model of self-regulated behavior change) can be used to
systematically inform the development of travel behavior change interventions at large scale. A
review of intervention studies applying the stage models reveals little guidance from existing
literature on the systematic application of the theory in the design of interventions. Stage-tailoring,
a design approach that matches the interventions to the stages, is the most prominent strategy
among the few studies that do systematically apply the theories, but an alternative approach, called
menu-based, is also introduced, which is grounded on the theoretical conceptualization that stages
represent an underlying continuum of action readiness. As illustration on how menu-based
approach may be applied, we presented Blaze mobility behavior change system.
From these three reviews, we then described in Chapter 5 how the Stage Model of Self-Regulated
Behavioral Change (SSBC) is used for the first time in the development of Blaze, a mobility
behavior change support system, consisting of Smartphone and web application. SSBC posits that
behavioral change is achieved by a transition through a temporal sequence of different stages.
Using SSBC, we systematically developed theory-based interventions.
In Chapter 6, we provided the study context. Blaze was deployed among the students of the Ateneo
de Manila University, a tertiary educational institution located in Metro Manila, Philippines, a
developing country. We described the details of the field experiment conducted in the university,
the survey instrument used and its reliability.
In Chapter 7, we evaluated the effectiveness of Blaze. Our analysis shows that Blaze is able to
induce a positive change across a wide range of indicators, as specified by the theory: car use,
stage membership, intentions and socio-cognitive determinants. Nonetheless, the effect size of the
intervention in reducing car use is small, a result that is expected given that the setting of the field
experiment is a university in a developing country, where good infrastructures supporting
sustainable commuting are deficient. Furthermore, we conducted a usefulness evaluation to assess
the relevance of the features within Blaze and associate it with stage membership. We discussed
150
the implications of our results, especially on the development of interventions and transport
policies for cases in developing countries.
In Chapter 8, our aim was to validate and extend the stage model of self-regulated behavior change.
To do so, we characterized the stages of change in terms of car use level, strength of the three
intention types, and stability. Our objective is to describe the relationships of these variables with
each other. We also described how the three intention types mark the transition points between
stages. We then constructed a model specifying the determinants or predictors of the three
intention types. We ended by discussing how our results confirm, validate and extend further the
stage model of self-regulated behavior change (SSBC). Our main findings suggest that, in
agreement with SSBC, travel behavior change is achieved by a transition through a temporal
sequence of four stages: predecision, pre-action, action and post-action. In an extension from
SSBC, we further distinguished post-action depending on whether the behavior is on initiation or
under maintenance. We observed that the former (initiation) is characterized by instability (either
relapse or progress), while the latter (maintenance) by stability.
Finally, in Chapter 9, we described how our mobility behavior change support system is able to
achieve its effects in changing the travel behavior of university students. We identified a causal
pathway linking the effect of the technology intervention to its behavioral outcome through the
mediation of a number of variables. The validated stage model of self-regulated behavioral change
(SSBC), described previously, is used as a theoretical framework to understand how the outcome
may be influenced by determinants (conceptual theory), and how the determinants may be
activated by different intervention types (action theory). We showed that the action and conceptual
components of the Blaze impact pathway on travel behavior change differ by stages: individuals
belonging to later stages change their behavior via the mediation of a change in implementation
intention, while those in the early stages undergo behavior change through the mediation of a
change in self-efficacy. We discussed how these results may aid in understanding the potential
role of technology interventions in mobility management.
151
10.2. Main Findings and Contribution
The first contribution of this study is the development of behavior change interventions based
explicitly on a theory, the stage model of self-regulated behavior change (SSBC). Existing
literature suggests that stage models propose what is called the “stage-tailoring approach” that
matches interventions to stages (Bamberg, 2013b; Bamberg et al, 2015; Friman et al, 2017). This
means developing different modules for different stages. Nonetheless, examining the content of
interventions in studies applying the stage models reveals that there is a content overlap among
multiple stages, suggesting that developing different modules may not be necessary. We therefore
suggest a “menu-based intervention approach” (Abraham, 2008). In this approach, a generic
intervention is given to all, regardless of their stage memberships, but using self-assessments
participants can self-tailor and select items from the menu that are deemed relevant. In this study,
we provide a theoretical grounding for this approach and also provide evidence that it may work.
Furthermore, even with a theoretical framework that can be used as a blueprint, there is little
guidance on how to develop theory-based interventions (Michie et al, 2008). Thus, the second
contribution of our work is that it represents one of the first attempts in the literature to
operationalize interventions explicitly based on SSBC. In particular, we suggest developing multi-
target menu-based interventions. We also discussed ten possible interventions based on SSBC
(e.g. Blaze map, ride plan, summary, etc).
In addition, our work demonstrates – and this is the third contribution – useful insights that can be
obtained about the travel behavior of students by using the stage model of self-regulated behavior
change (SSBC) theory. Travel behavior of university students is not well understood yet (c.f. Danaf
et al, 2014). Using the SSBC as the framework, we gain more insights about the behavior of the
university students. From SSBC, a specific pattern can be postulated about the car use and
intentions (goal, behavioral, implementation), depending on the stage memberships of the students.
The fourth contribution is the validation and extension of the stage model of self-regulated
behavior change. Bamberg (2013a) acknowledged at least two weaknesses of his study: (a) mere
correlation of variables using cross-section data, and; (b) failure to assess some constructs such as
152
coping/action planning, maintenance- and recovery self-efficacy. Furthermore, related researches
in the health domain (e.g. Schwarzer, 2008) suggest a further division of the postaction stage into
initiative and maintenance – a distinction not considered in the current stage model of self-
regulated behavior change. In the present work, longitudinal data measuring the model variables,
including coping/action planning, maintenance- and recovery- self-efficacy, obtained at two time
points were used to confirm and validate the core ideas of the current stage model. We also extend
the stage model by presenting evidence for the need to distinguish initiation and maintenance. This
work thus represents an independent validation and extension of the stage model using longitudinal
data.
In this study, we are also able to show that our mobility behavior change support system is able to
induce a positive change across a wide range of indicators, as specified by the stage model of self-
regulated behavior change theory: car use, stage membership and intentions. Hence, the fifth
contribution is the demonstration of effectiveness of our system. However, due to resource
constraints, only a convenience sample of university students is used, and thus we cannot
generalize our results to the general population of the university. Nonetheless, since our
methodology includes an adequate control group, this is considered to be of higher methodological
quality than previous evaluation studies of mobility behavior change support systems, all of which
only employ before-after comparison without adequate control groups.
Finally, we identify a causal pathway linking the effect of the technology intervention to its
behavioral outcome through the mediation of a number of variables. This is the sixth contribution.
We are able to show that the action and conceptual components of the Blaze impact pathway on
travel behavior change differ by stages: individuals belonging to later stages change their behavior
via the mediation of a change in implementation intention, while those in the early stages undergo
behavior change through the mediation of a change in self-efficacy. In the literature, research on
mechanisms in travel behavior change in non-technological context has been done; but, to the best
of our knowledge, there is none yet on the technological context. There may exist significant
differences in mechanisms of change between non-technological and technological contexts. Our
work thus represents one of the first studies on the mechanisms associated with technology
intervention for travel behavior change.
153
10.3. Implications
The results of this study have significant implications for transport policy. First, we have seen our
technology-based intervention is able to induce changes in the mediating variables and also car
use behavior. In the literature, there are many technology-based interventions developed in the
health domain, but these are sorely lacking in the travel behavior domain. Technology must be
then further exploited in delivering large-scale interventions aimed at promoting sustainable travel
behavior.
Second, we support the idea of explicitly grounding change programs in a behavior change theory,
not only for greater insight into understanding travel behavior, but also for greater effectiveness.
Indeed, it is claimed that a theoretical grounding is missing in many of the behavior change
programs. Moreover, many of the mobility BCSSs that have been developed so far put little effort
on grounding the BCSSs in an explicit behavior change theory.
Third, travel demand management measures implemented in the university setting are dominated
by hard measures. Nonetheless, these measures may not be appropriate for cases of universities in
the developing countries. An example of a soft measure implemented in the university setting,
TravelSmart, may also not be appropriate as it is costly due to the need for travel counselors. Our
mobility behavior change support system, which is demonstrated to have caused shifts in attitudes
and mode use among university students, is an appropriate soft measure. Universities in the
developing world are confronted by resource constraints. Hence, a highly scalable and cost-
effective approach is very much preferred. Furthermore, the rise in smartphone penetration rate
and internet usage in many developing countries (c.f. Poushter, 2016) poses the possibility of
deploying mBCSSs among larger segments of the population, especially among young people.
Majority of the members of the university community are the youth, who are known to have high
Smartphone ownership and usage. In addition, “public transportation infrastructure may be more
facilitative of commuting trips to the university than to other destinations in the city, making the
switch to alternative transport for commuting trips an easier or more convenient choice than for
non-commuting trips” (Kormos et al, 2014). This is an important consideration since the
154
effectiveness of mBCSSs implemented in the universities will be strengthened if combined with
hard transport policy measures or if there exist good infrastructures supporting sustainable travel
behavior. In this study, we think that the small effect size of our mBCSS (=0.041) can be further
increased if there exist more viable options or infrastructure supporting commuting by alternatives.
Fourth, our technology-based intervention is able to effect changes only in the mediating variables
that are most proximate to behavior (e.g. implementation intention and recovery efficacy). This
shows our intervention affects mainly the latter “volitional phase” (goal-pursuit), and not the
earlier “motivational phase” (goal-setting), of the behavior change process. It appears then that
our intervention – and perhaps, any technology-based intervention – is effective because it
provides technical support such as technology-enabled implementation plans (e.g. trip/travel plans
with a simulation of potential outcomes) to individuals so they can be helped to translate their
intention into actual behavior and also to break any unwanted (car-use) habit that hinders the
initiation of the new wanted behavior (Adriannse et al, 2011). If the initiation of this new behavior
can be sustained over long term with the aid of technology, then habits can be formed.
10.4. Limitations
We note some of the limitations of the present study. First, we only investigate the home-school
commuting behavior of students, which is only one part of overall student travel behavior.
Although home-school commuting is the “single largest impact” that university has on the
environment, a more complete travel data will be able to better capture the full travel behavior of
the students, especially since a number of them frequently go out of campus between classes.
Second, although we claim our study has a longitudinal character, we are only limited by two data
points (before-after the intervention). Our study design is able to capture the immediate effect of
the technology intervention, but we fail to understand its (potential) long-term effect. Other studies
(e.g. Thøgersen, 2009) conducted data collection in three waves (before and after the intervention,
and 6 months later). Such studies are able to assess the long-term impact (if any) of the
intervention. Klöckner (2014) also employed repeated measures for two months. Studies like these,
155
with frequent data collection, are able to capture, not only between-group dynamics, but also the
intra-person (within-person) dynamics over time.
Third, we have a small sample size, especially for some stages. We observe that many of our
respondents belong to the predecision and late post-action but very few of them belong to certain
stages, such as action and early post-action. This makes it necessary to consider with caution our
analysis regarding stage progression.
Fourth, we used single-item measures only. Multiple-item measures are generally recommended,
but caution must be taken on the length of questionnaire so as not to unnecessarily burden the
respondents. Perhaps, two-item measures per construct are acceptable.
Fifth, this dissertation only tackles behavior outcomes, but adherence of the students to the
intervention is not discussed or investigated. Adherence is considered critical to intervention
efficacy, since for interventions where participation is voluntary, users will stay in the study and
receive treatment or dose of intervention that is proportional to their level of adherence (Cugelman
et al, 2011). In this work, in order to encourage participation and adherence, we provided
incentives in the form of class points and a chance of winning cash in a lottery. We nonetheless
did not investigate the relationship among incentive, adherence, and intervention efficacy.
Sixth, we employed quasi-experimental design, and not a randomized controlled trial (RCT) as
our experimental design, a standard requirement for studying mechanism of effect (Kazdin, 2007).
In RCT, investigators randomly allocate individuals to an intervention or control group, while in
quasi-experimental designs, matched but not randomised control groups are used (c.f. Graham-
Rowe et al, 2011, for methodological discussion). The use of quasi-experimental design limits our
ability to draw robust inferences regarding the causal pathway between intervention and change.
However, as Bamberg and Rees (2017) recently noted, in transportation research, quasi-
experimental design may count already as sufficient evidence, depending on the objectives of the
evaluation.
156
Finally, though we were able to demonstrate the effectiveness of our technology-based
intervention as a whole, we were not able to identify the specific elements, components or aspects
of Blaze, our TBI (technology-based intervention), that produced the change (c.f. Ritterband et al,
2009). Identifying the elements within the intervention that effect the change is essential in
informing and guiding future intervention developments
10.5. Recommendation: Towards a Computer-Tailored, Theory-Driven Development of
Mobility Behavior Change Support System
In this section, we present a conceptualization of a mobility behavior change support system or
mBCSS that uses a computational model based explicitly on the stage model of self-regulated
behavior change theory to generate computer-tailored interventions. In the health domain, a
system like this, called eMate, has already been implemented (Klein et al, 2014), and this chapter
draws ideas from it. In what follows, we present a recommended improvement of our current
mobility behavior change support system, which we call Blaze 2.0.
The improved version performs the following tasks which will be described further in the next
subsections:
1. determining the current stage membership and cognition of the user;
2. generating automated travel diary and monitoring the level of car use;
3. reasoning about required changes;
4. suggesting relevant interventions in the menu for use
10.5.1. Determining the current stage membership and cognition
When the user starts using Blaze 2.0, he first answers a brief stage questionnaire, very similar to
the stage instrument used in this study. Aside from this, he is also directed to answer a
questionnaire on thoughts on car use consisting of questions, measuring on a 11-point scale (for
example) the socio-cognitive constructs in the SSBC model. To ensure construct validity, the
questions can be derived from Schwarzer (2008), Bamberg (2013b), Klöckner (2014) and
Klöckner (2017). The number of items per construct depends on the level of response burden that
the user can handle. In original version of Blaze, we used single-item measures per construct.
157
10.5.2. Generating automated travel diary and monitoring the level of car use
IPET (Meloni et al, 2014) and QT (Jariyasunant et al, 2015) have described the possible technical
implementation of how an automated activity-travel diary can be generated. The travel diary
includes details like origin, destination, mode, start time, duration, distance, carbon emission, cost,
calories burned, etc. A map showing the trips made can also be included in the visualization. A
graph showing an overview of aggregated data (e.g. car use or carbon emission) can also be
supported.
10.5.3. Reasoning about required changes
Blaze 2.0 will then intelligently determine the level of car use, car use reduction level, or carbon
emission level that is appropriate for the user. The prescribed value can be based, for instance, on
the level set by the user himself, by an expert or authority, or by local/national averages. Moreover,
it can also be stage-dependent. For instance, those who are already in the post-actional stage can
be encouraged to further lower their carbon emission, perhaps even lower than the local average.
Those in the actional stage can be asked to aim for a car use level that is 10% lower than their
previous week’s average. In any case, this is freely set by the designer, depending on his own
good judgment. The system will then help the user behave in compliance with the prescribed level
of car use or carbon emission. If the monitoring reveals non-compliance or sub-optimality, the
system will determine some possible causes. It does so by identifying the bottlenecks, those
constructs in the model that prevent the individual from complying. Bottlenecks are defined as
“the construct in the [path] that prevents a [person] from progressing from one stage of change to
another” (Klein et al, 2014). A path, moreover, is a sequence of edges linking a sequence of
constructs from one stage to the next stage. For instance, if in the individual is in the actional stage
and has difficulty transitioning to the next post-actional stage, or reaching the set target, we can
investigate which of the proximate – or even distal – determinants along the path from actional to
post-actional stage that are problematic. These bottlenecks can be identified by examining the
responses of the individual to the questionnaire described in 5.1. An algorithm can be developed
that is able to realistically determine these bottlenecks. If the user, however, proves resistant to
158
changes in outward behavior, the system can target latent changes such as progression through
stages, even though the same (high) level of car use is maintained. Once the bottlenecks are
identified, the system can pose a few questions to validate these potential bottlenecks. If they
prove to be indeed problematic, they can then be the target determinants of the interventions.
Alternatively, instead of identifying bottlenecks, one can identify the relevance likelihood of the
determinants. A determinant or construct is considered relevant if it is in the path from one stage
to the next.
10.5.4. Suggesting relevant interventions in the menu for use
The next step is to propose some suggestions to the user. These suggestions consist of identifying
the pages or features in the website that the user have to use. For example, if in the previous step,
we identified the lack of implementation plan to be the bottleneck that prevents the user with high
intention to finally succeed in reducing his car use to the set level, then we can ask him to focus
on using the trip planning or ridesharing page of the website. In IPET, a personalized trip plan
(PTP) is automatically generated for the user. Alternatively, we can ask the user to try
implementing this automatically generated PTP, without him having to prepare one. Other pages
in the menu of interventions can provide additional support depending on their relevance. In other
words, the PTP, as the main intervention, constitutes the central route to persuasion, while the
other relevant pages are the peripheral routes (c.f. Petty & Cacioppo, 1986).
Other functionalities can be added such as monitoring adherence to suggestions, dynamic updating
of beliefs about the user, changing the user’s perception, etc (c.f. Klein et al, 2014), but we believe
that the four tasks described above are the essential ones.
159
References
Abou-Zeid, M. & Ben-Akiva, M. (2011). The effect of social comparisons on commute well-being.
Transportation Research Part A: Policy and Practice 45.4, 345-361.
Abraham, C. (2008). Beyond Stages of Change: Multi‐Determinant Continuum Models of Action
Readiness and Menu‐Based Interventions. Applied Psychology, 57(1), 30-41.
Adriaanse, M. A., Gollwitzer, P. M., De Ridder, D. T., De Wit, J. B., & Kroese, F. M. (2011).
Breaking habits with implementation intentions: A test of underlying processes. Personality and
Social Psychology Bulletin, 37(4), 502-513.
Agatz, N., Erera, A., Savelsbergh, M., & Wang, X. (2012). Optimization for dynamic ride-sharing:
A review. European Journal of Operational Research, 223(2), 295-303.
Ajzen, I. (1991). The theory of planned behavior. Organizational behavior and human decision
processes, 50(2), 179-211.
Ampt, E. (2003). Voluntary household travel behaviour change–theory and practice. In 10th
International Association of Travel Behaviour Research Conference, Lucerne, Switzerland,
August.
Antypas, K., & Wangberg, S. C. (2014). Combining users’ needs with health behavior models in
designing an internet-and mobile-based intervention for physical activity in cardiac rehabilitation.
JMIR research protocols, 3(1), e4.
Aoun, A., Abou-Zeid, M., Kaysi, I., & Myntti, C. (2013). Reducing parking demand and traffic
congestion at the American University of Beirut. Transport Policy, 25, 52-60.
Armitage, C. J., & Arden, M. A. (2002). Exploring discontinuity patterns in the transtheoretical
model: An application of the theory of planned behaviour. British journal of health psychology,
7(1), 89-103.
Arnott, B., Rehackova, L., Errington, L., Sniehotta, F. F., Roberts, J. R., & Araujo-Soares, V.
(2014). Efficacy of behavioural interventions for transport behaviour change: systematic review,
meta-analysis and intervention coding. International journal of behavioral nutrition and physical
activity, 11(1), 133.
Ateneo de Manila University (2014). Towards Efficient Mobility in a Sustainable Campus.
(2014, July). Retrieved June 15, 2017, from
http://www.ateneo.edu/sites/default/files/The%20Ateneo%20Sustainability%20Report%202014.
Avineri, E. & Waygood, E. O. D. (2013). Applying valence framing to enhance the effect of
information on transport-related carbon dioxide emissions. Transportation Research Part A:
Policy and Practice, 48, 31-38.
160
Axsen, J., & Kurani, K. S. (2012). "Interpersonal influence within car buyers' social networks:
applying five perspectives to plug-in hybrid vehicle drivers." Environment and Planning-Part A
44.5, 1047.
Balsas, C. J. (2003). Sustainable transportation planning on college campuses. Transport
Policy, 10(1), 35-49.
Bamberg, S. (2007). Is a stage model a useful approach to explain car drivers' willingness to use
public transportation?. Journal of Applied Social Psychology, 37(8), 1757-1783.
Bamberg, S., Fujii, S., Friman, M., & Gärling, T. (2011). Behaviour theory and soft transport
policy measures. Transport policy, 18(1), 228-235.
Bamberg, S. (2013a). Changing environmentally harmful behaviors: A stage model of self-
regulated behavioral change. Journal of Environmental Psychology, 34, 151-159.
Bamberg, S. (2013b). Applying the stage model of self-regulated behavioral change in a car use
reduction intervention. Journal of Environmental Psychology, 33, 68-75.
Bamberg, S., Behrens, G., Bergmeyer, M., Brewitt, K., Papendick, M., Rees, J., & Zielinski, J.
(2015). “Development of a Theory-Driven, Web-based Behavioral Change Support System for
Environmentally Friendly Behavior”. In Evans, D. (Ed.). Social Marketing: Global Perspectives,
Strategies and Effects on Consumer Behavior (109-120). NY: Nova Science Publisher.
Bamberg, S., & Rees, J. (2017). The impact of voluntary travel behavior change measures–A meta-
analytical comparison of quasi-experimental and experimental evidence. Transportation Research
Part A: Policy and Practice, 100, 16-26.
Baraldi, A.N., Wurpts, I.C., Mackinnon, D.P., & Lockhart, G. (2014). Evaluating mechanisms of
behavior change to inform and evaluate technology-based interventions. Behavioral Healthcare
and Technology: Using Science-Based Innovations to Transform Practice, 187.
Barata, E., Cruz, L., & Ferreira, J. P. (2011). Parking at the UC campus: Problems and
solutions. Cities, 28(5), 406-413.
Baumer, E. P., Katz, S. J., Freeman, J. E., Adams, P., Gonzales, A. L., Pollak, J., Retelny, D.,
Niederdeppe, J., Olson, C.M. & Gay, G. K. (2012). Prescriptive persuasion and open-ended social
awareness: expanding the design space of mobile health. In Proceedings of the ACM 2012
conference on Computer Supported Cooperative Work, ACM, February, 475-484.
Bélanger-Gravel, A., Godin, G., & Amireault, S. (2013). A meta-analytic review of the effect of
implementation intentions on physical activity. Health Psychology Review, 7(1), 23-54.
Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus single-item
measures of the same constructs. Journal of marketing research, 44(2), 175-184.
161
Bhattacharjee, D., Haider, S. W., Tanaboriboon, Y., & Sinha, K. C. (1997). Commuters' attitudes
towards travel demand management in Bangkok. Transport Policy, 4(3), 161-170.
Bird, J., & Rogers, Y. (2010). The pulse of tidy street: Measuring and publicly displaying domestic
electricity consumption. In workshop on energy awareness and conservation through pervasive
applications (Pervasive 2010).
Bond, A., & Steiner, R. (2006). Sustainable campus transportation through transit partnership and
transportation demand management: A case study from the University of Florida. Berkeley
Planning Journal, 19(1).
Bonham, J., & Koth, B. (2010). Universities and the cycling culture. Transportation research part
D: transport and environment, 15(2), 94-102.
Brendryen, H., Drozd, F., & Kraft, P. (2008). A digital smoking cessation program delivered
through internet and cell phone without nicotine replacement (happy ending): randomized
controlled trial. Journal of Medical Internet Research, 10(5), e51.
Brög, W., Erl, E., & Mense, N. (2002, December). Individualised marketing changing travel
behaviour for a better environment. In Paper presented at the OECD Workshop: Environmentally
Sustainable Transport (Vol. 5, pp. 06-12).
Broll, G., Cao, H., Ebben, P., Holleis, P., Jacobs, K., Koolwaaij, J., Luther, S. & Souville, B.
(2012). Tripzoom: an app to improve your mobility behavior. In Proceedings of the 11th
International Conference on Mobile and Ubiquitous Multimedia, ACM, December, 57.
Brown, J., Hess, D. B., & Shoup, D. (2003). Fare-free public transit at universities: An
evaluation. Journal of Planning Education and Research, 23(1), 69-82.
Brynjarsdottir, H., Håkansson, M., Pierce, J., Baumer, E., DiSalvo, C., & Sengers, P. (2012).
Sustainably unpersuaded: how persuasion narrows our vision of sustainability. In Proceedings of
the SIGCHI Conference on Human Factors in Computing Systems, ACM, May, 947-956.
Busch, M., Schrammel, J., FLU, M. A., Kruijff, E., & Tscheligi, M. (2012). Persuasive Strategies
Report. CURE–Center for Usability Research & Engineering.
Cairns, S., Sloman, L., Newson, C., Anable, J., Kirkbride, A., & Goodwin, P. (2008). Smarter
choices: assessing the potential to achieve traffic reduction using ‘soft measures’. Transport
Reviews, 28(5), 593-618.
Carreras, I., Gabrielli, S., Miorandi, D., Tamilin, A., Cartolano, F., Jakob, M., & Marzorati, S.
(2012). SUPERHUB: a user-centric perspective on sustainable urban mobility. In Proceedings of
the 6th ACM workshop on Next generation mobile computing for dynamic personalised travel
planning, ACM, June, 9-10.
162
Chan, N. D., & Shaheen, S.A. (2012). Ridesharing in north america: Past, present, and
future." Transport Reviews 32.1, 93-112.
Chatterjee, K. (2009). A comparative evaluation of large-scale personal travel planning projects in
England. Transport Policy, 16(6), 293-305.
Christophersen, T., & Konradt, U. (2011). Reliability, validity, and sensitivity of a single-item
measure of online store usability. International Journal of Human-Computer Studies, 69(4), 269-
280.
Cohen-Blankshtain, G., & Rotem-Mindali, O. (2016). Key research themes on ICT and sustainable
urban mobility. International Journal of Sustainable Transportation, 10(1), 9-17.
Conroy, D. E., Yang, C. H., & Maher, J. P. (2014). Behavior change techniques in top-ranked
mobile apps for physical activity. American journal of preventive medicine, 46(6), 649-652.
Cooper, C. (2007). Successfully changing individual travel behavior: Applying community-based
social marketing to travel choice. Transportation Research Record: Journal of the Transportation
Research Board, (2021), 89-99.
Cowan, L. T., Van Wagenen, S. A., Brown, B. A., Hedin, R. J., Seino-Stephan, Y., Hall, P. C., &
West, J. H. (2012). Apps of steel: are exercise apps providing consumers with realistic
expectations? A content analysis of exercise apps for presence of behavior change theory. Health
Education & Behavior, 1090198112452126.
Cugelman, B., Thelwall, M., & Dawes, P. (2011). Online interventions for social marketing health
behavior change campaigns: a meta-analysis of psychological architectures and adherence
factors. Journal of medical Internet research, 13(1).
Dallery, J., Jarvis, B., Marsch, L., & Xie, H. (2015). Mechanisms of change associated with
technology-based interventions for substance use. Drug and alcohol dependence, 150, 14-23.
Danaf, M., Abou-Zeid, M., & Kaysi, I. (2014). Modeling travel choices of students at a private,
urban university: insights and policy implications. Case Studies on Transport Policy, 2(3), 142-
152.
Davies, N. (2012). What are the ingredients of successful travel behavioural change
campaigns?. Transport Policy 24, 19-29.
Demissie, M. G., Correia, G., & Bento, C. (2015). Analysis of the pattern and intensity of urban
activities through aggregate cellphone usage. Transportmetrica A: Transport Science, 11(6), 502-
524.
Diniz, I. M., Duarte, M. D. F. S., Peres, K. G., de Oliveira, E. S., & Berndt, A. (2015). Active
commuting by bicycle: results of an educational intervention study. Journal of physical activity
and health, 12(6), 801-807.
163
DiSalvo, C., Sengers, P., & Brynjarsdóttir, H. (2010, April). Mapping the landscape of sustainable
HCI. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, April,
ACM, 1975-1984.
Dorsey, B. (2005). Mass transit trends and the role of unlimited access in transportation demand
management. Journal of Transport Geography, 13(3), 235-246.
Delmelle, E. M., & Delmelle, E. C. (2012). Exploring spatio-temporal commuting patterns in a
university environment. Transport Policy, 21, 1-9.
Enoch, M., & Zhang, L. (2008). Travel plans. In The Implementation and Effectiveness of
Transport Demand Management Measures: An International Perspective. Ashgate London.
Ettema, D., Theo Arentze, and Harry Timmermans. "Social influences on household location,
mobility and activity choice in integrated micro-simulation models." Transportation Research
Part A: Policy and Practice 45.4 (2011): 283-295.
Faulkner, L. (2003). Beyond the five-user assumption: Benefits of increased sample sizes in
usability testing. Behavior Research Methods, 35(3), 379-383.
Fitzsimons, G. M., & Finkel, E. J. (2010). Interpersonal influences on self-regulation. Current
Directions in Psychological Science, 19(2), 101-105.
Fogg, B. J. (2002). Persuasive technology: using computers to change what we think and do.
Ubiquity, December, 5.
Fogg, B. J. and Eckles, D. (2007). Mobile Persuasion: 20 Perspectives on the Future of Behavior
Change. Stanford Captology Media, Stanford.
Friman, M., Huck, J., & Olsson, L. E. (2017). Transtheoretical model of change during travel
behavior interventions: an integrative review. International journal of environmental research and
public health, 14(6), 581.
Froehlich, J., Dillahunt, T., Klasnja, P., Mankoff, J., Consolvo, S., Harrison, B., & Landay, J. A.
(2009, April). UbiGreen: investigating a mobile tool for tracking and supporting green
transportation habits. In Proceedings of the SIGCHI Conference on Human Factors in Computing
Systems (pp. 1043-1052). ACM.
Fujii, S., & Taniguchi, A. (2005). Reducing family car-use by providing travel advice or requesting
behavioral plans: An experimental analysis of travel feedback programs. Transportation Research
Part D: Transport and Environment, 10(5), 385-393.
Fujii, S., & Taniguchi, A. (2006). Determinants of the effectiveness of travel feedback programs—
a review of communicative mobility management measures for changing travel behaviour in
Japan. Transport policy, 13(5), 339-348.
164
Fujii, S., Bamberg, S., Friman, M., & Gärling, T. (2009). Are effects of travel feedback programs
correctly assessed?. Transportmetrica, 5(1), 43-57.
Furuhata, M., Dessouky, M., Ordóñez, F., Brunet, M. E., Wang, X., & Koenig, S. (2013).
Ridesharing: The state-of-the-art and future directions. Transportation Research Part B:
Methodological, 57, 28-46.
Gaker, D., Vautin, D., Vij, A., & Walker, J. L. (2011). The power and value of green in promoting
sustainable transport behavior. Environmental Research Letters, 6(3), 034010.
Gärling, T. and Fujii, S. (2002). Structural equation modeling of determinants of
planning. Scandinavian Journal of Psychology 43.1, 1-8.
Gärling, T., & Schuitema, G. (2007). Travel demand management targeting reduced private car
use: effectiveness, public acceptability and political feasibility. Journal of Social Issues, 63(1),
139-153.
Gatersleben, B., & Appleton, K. M. (2007). Contemplating cycling to work: Attitudes and
perceptions in different stages of change. Transportation Research Part A: Policy and
Practice, 41(4), 302-312.
Goetzke, F., & Rave, T. (2010). Bicycle use in Germany: explaining differences between
municipalities with social network effects. Urban studies.
Gollwitzer, P. M. (1999). Implementation intentions: strong effects of simple plans. American
psychologist, 54(7), 493.
Gonzalo-Orden, H., Rojo, M., Velasco, L., & Linares, A. (2012, December). Mobility surveys and
sustainable policies in universities. In Proceedings of the Institution of Civil Engineers-Municipal
Engineer (Vol. 165, No. 4, pp. 219-229). Thomas Telford Ltd.
Graham-Rowe, E., Skippon, S., Gardner, B., & Abraham, C. (2011). Can we reduce car use and,
if so, how? A review of available evidence. Transportation Research Part A: Policy and
Practice, 45(5), 401-418.
Guo, Z., Derian, A., & Zhao, J. (2015). Smart devices and travel time use by bus passengers in
Vancouver, Canada. International Journal of Sustainable Transportation, 9(5), 335-347.
Heckhausen, H., & Gollwitzer, P. M. (1987). Thought contents and cognitive functioning in
motivational versus volitional states of mind. Motivation and emotion, 11(2), 101-120.
Hedeker, D., & Mermelstein, R. J. (1998). A multilevel thresholds of change model for analysis
of stages of change data. Multivariate Behavioral Research, 33(4), 427-455.
Hedeker, D., Mermelstein, R. J., & Weeks, K. A. (1999). The thresholds of change model: an
approach to analyzing stages of change data. Annals of Behavioral Medicine, 21(1), 61-70.
165
Hemmingsson, E., Udden, J., Neovius, M., Ekelund, U., & Rössner, S. (2009). Increased physical
activity in abdominally obese women through support for changed commuting habits: a
randomized clinical trial. International journal of obesity, 33(6), 645.
Hong, J., & Thakuriah, P. (2016). Relationship between motorized travel and time spent online for
nonwork purposes: An examination of location impact. International Journal of Sustainable
Transportation, 10(7), 617-626.
Hooper, D., Coughlan, J., & Mullen, M. (2008). Structural equation modelling: Guidelines for
determining model fit. Articles, 2.
Hsieh, H. S., Kanda, Y., & Fujii, S. (2017). Reducing car use by volitional strategy of action and
coping planning enhancement. Transportation research part F: traffic psychology and behaviour,
47, 163-175.
Hwang, W., & Salvendy, G. (2010). Number of people required for usability evaluation: the 10±2
rule. Communications of the ACM, 53(5), 130-133.
Javid, M. A., Okamura, T., Nakamura, F., Tanaka, S., & Wang, R. (2013). Factors Influencing the
Acceptability of Travel Demand Management (TDM) Measures in Lahore: Application of
Behavioral Theories. In Proceedings of the Eastern Asia Society for Transportation Studies (Vol.
9).
Jariyasunant, J., Abou-Zeid, M., Carrel, A., Ekambaram, V., Gaker, D., Sengupta, R., & Walker,
J. L. (2015). Quantified traveler: Travel feedback meets the cloud to change behavior. Journal of
Intelligent Transportation Systems, 19(2), 109-124.
Julsrud, T.E. & Denstadli, J.M. (2017). Smartphones, travel time-use and attitudes to public
transport services. Insights from an explorative study of urban dwellers in two Norwegian cities.
International Journal of Sustainable Transportation. DOI: 10.1080/15568318.2017.1292373.
Jylhä, A., Nurmi, P., Sirén, M., Hemminki, S., & Jacucci, G. (2013). Matkahupi: a persuasive
mobile application for sustainable mobility. In Proceedings of the 2013 ACM conference on
Pervasive and ubiquitous computing adjunct publication, ACM, 227-230.
Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research. Annu. Rev.
Clin. Psychol., 3, 1-27.
Khattak, A., Wang, X., Son, S., & Agnello, P. (2011). Travel by university students in Virginia: Is
this travel different from travel by the general population?. Transportation Research Record:
Journal of the Transportation Research Board, (2255), 137-145.
Kelly, J. A., & Fu, M. (2014). Sustainable school commuting–understanding choices and
identifying opportunities: A case study in Dublin, Ireland. Journal of Transport Geography, 34,
221-230.
166
Klein, M., Mogles, N., & Van Wissen, A. (2011). Why won’t you do what’s good for you? Using
intelligent support for behavior change. In Human Behavior Understanding, Springer Berlin
Heidelberg, 104-115.
Klein, M., Mogles, N., & van Wissen, A. (2014). Intelligent mobile support for therapy adherence
and behavior change. Journal of biomedical informatics, 51, 137-151.
Klöckner, C. A., & Blöbaum, A. (2010). A comprehensive action determination model: Toward a
broader understanding of ecological behaviour using the example of travel mode choice. Journal
of Environmental Psychology, 30(4), 574-586.
Klöckner, C. A. (2014). The dynamics of purchasing an electric vehicle–A prospective
longitudinal study of the decision-making process. Transportation Research Part F: Traffic
Psychology and Behaviour, 24, 103-116.
Klöckner, C. A. (2017). A stage model as an analysis framework for studying voluntary change in
food choices–The case of beef consumption reduction in Norway. Appetite, 108, 434-449.
Kormos, C., Gifford, R., & Brown, E. (2014). The influence of descriptive social norm information
on sustainable transportation behavior: a field experiment. Environment and Behavior,
0013916513520416.
Kwasnicka, D., Dombrowski, S. U., White, M., & Sniehotta, F. (2016). Theoretical explanations
for maintenance of behaviour change: a systematic review of behaviour theories. Health
psychology review, 10(3), 277-296.
Lane, N. D., Mohammod, M., Lin, M., Yang, X., Lu, H., Ali, S., Doryab, A., Berke, E., Choudhury,
T & Campbell, A. (2011). Bewell: A smartphone application to monitor, model and promote
wellbeing. In 5th international ICST conference on pervasive computing technologies for
healthcare, May, 23-26.
Langrial, S., Lehto, T. & Oinas-Kukkonen, H., Harjumaa,M. & Karppinen, P. (2012). Native
Mobile Applications For Personal Well-Being: A Persuasive Systems Design Evaluation.
In PACIS, 93.
Lathia, N., Pejovic, V., Rachuri, K. K., Mascolo, C., Musolesi, M., & Rentfrow, P. J. (2013).
Smartphones for large-scale behavior change interventions. IEEE Pervasive Computing, (3), 66-
73.
Lehto, T. & Oinas-Kukkonen, H. (2010). Persuasive features in six weight loss websites: A
qualitative evaluation." Persuasive technology. Springer Berlin Heidelberg, 2010. 162-173.
Lehto, T. & Oinas-Kukkonen, H. (2011). Persuasive features in web-based alcohol and smoking
interventions: a systematic review of the literature. Journal of Medical Internet research 13.3, e46.
167
Limanond, T., Butsingkorn, T., & Chermkhunthod, C. (2011). Travel behavior of university
students who live on campus: A case study of a rural university in Asia. Transport policy, 18(1),
163-171.
Lippke, S., Ziegelmann, J. P., & Schwarzer, R. (2005). Stage-specific adoption and maintenance
of physical activity: Testing a three-stage model. Psychology of Sport and Exercise, 6(5), 585-603.
Lovejoy, K., & Handy, S. L. (2011). Mixed Methods of Bike Counting for Better Cycling
Statistics: The Example of Bicycle Use, Abandonment, and Theft on UC Davis Campus.
In Transportation Research Board 90th Annual Meeting (No. 11-3795).
Lozano, R. (2006). Incorporation and institutionalization of SD into universities: breaking through
barriers to change. Journal of cleaner production, 14(9), 787-796.
Ludden, G., Hekkert, P. (2014). Design for healthy behavior: design interventions and stages of
change. Paper presented at the 9th International Conference on Design & Emotion: Colors of Care,
Bogota, Colombia, October 6-9.
Lustria, M. L. A., Cortese, J., Noar, S. M., & Glueckauf, R. L. (2009). Computer-tailored health
interventions delivered over the Web: review and analysis of key components. Patient education
and counseling, 74(2), 156-173.
Luszczynska, A., Mazurkiewicz, M., Ziegelmann, J. P., & Schwarzer, R. (2007). Recovery self-
efficacy and intention as predictors of running or jogging behavior: A cross-lagged panel analysis
over a two-year period. Psychology of Sport and Exercise, 8(2), 247-260.
Magliocchetti, D., Gielow, M., De Vigili, F., Conti, G., & De Amicis, R. (2011). A personal
mobility assistant based on ambient intelligence to promote sustainable travel choices. Procedia
Computer Science, 5, 892-899.
McKee, R., Mutrie, N., Crawford, F., & Green, B. (2007). Promoting walking to school: results of
a quasi-experimental trial. Journal of Epidemiology & Community Health, 61(9), 818-823.
Meloni, I., Sanjust, B., Sottile, E., & Cherchi, E. (2013). Propensity for voluntary travel behavior
changes: An experimental analysis. Procedia-Social and Behavioral Sciences, 87, 31-43.
Meloni, I., Sanjust, B., Delogu, G., & Sottile, E. (2014). Development of a Technological Platform
for Implementing VTBC Programs. Transportation Research Procedia, 3, 129-138.
Meloni, I., & di Teulada, B. S. (2015). I-Pet Individual Persuasive Eco-travel Technology: A tool
for VTBC program implementation. Transportation Research Procedia, 11, 422-433.
Meloni, I., di Teulada, B. S., & Spissu, E. (2016). Lessons learned from a personalized travel
planning (PTP) research program to reduce car dependence. Transportation, 1-18.
168
Michie, S., Johnston, M., Francis, J., Hardeman, W., & Eccles, M. (2008). From theory to
intervention: mapping theoretically derived behavioural determinants to behaviour change
techniques. Applied psychology, 57(4), 660-680.
Mintz, J., Branch, C., March, C., & Lerman, S. (2012). Key factors mediating the use of a mobile
technology tool designed to develop social and life skills in children with Autistic Spectrum
Disorders. Computers & Education, 58(1), 53-62.
Mintz, J., & Aagaard, M. (2012). The application of persuasive technology to educational settings.
Educational Technology Research and Development, 60(3), 483-499.
Miralles-Guasch, C., & Domene, E. (2010). Sustainable transport challenges in a suburban
university: The case of the Autonomous University of Barcelona. Transport policy, 17(6), 454-
463.
Mohammed, A. A., & Shakir, A. A. (2013). Factors that affect transport mode preference for
graduate students in the national university of Malaysia by logit method. Journal of Engineering
Science and Technology, 8(3), 351-363.
Molina-García, J., Castillo, I., Queralt, A., & Sallis, J. F. (2013). Bicycling to university:
evaluation of a bicycle-sharing program in Spain. Health promotion international, 30(2), 350-358.
Möser, G., & Bamberg, S. (2008). The effectiveness of soft transport policy measures: A critical
assessment and meta-analysis of empirical evidence. Journal of Environmental Psychology, 28(1),
10-26.
Mun, M., Reddy, S., Shilton, K., Yau, N., Burke, J., Estrin, D., Hansen, M., Howard, E., West, R.
& Boda, P. (2009). PEIR, the personal environmental impact report, as a platform for participatory
sensing systems research. In Proceedings of the 7th international conference on Mobile systems,
applications, and services, ACM., June, 55-68.
Mundorf, N., Redding, C., Paiva, A., Brick, L., Prochaska, J., & Fu, T. (2013). Promoting
sustainable transportation across campus communities using the transtheoretical model of change.
In Conference on Communication and the Environment, Uppsala, Sweden.
Mutrie, N., Carney, C., Blamey, A., Crawford, F., Aitchison, T., & Whitelaw, A. (2002). “Walk
in to Work Out”: a randomised controlled trial of a self help intervention to promote active
commuting. Journal of Epidemiology & Community Health, 56(6), 407-412.
National Economic and Development Authority (2014). Roadmap for Transport Infrastructure
Development for Metro Manila and its Surrounding Areas. (2014, March). Retrieved June 15,
2017, from http://www.neda.gov.ph/roadmap-transport-infrastructure-development-metro-
manila-surrounding-areas-region-iii-region-iv/
Nielsen Global Survey of Automative Demand (2014). Rising Middle Class Will Drive Global
Automotive Demand in the Coming Two Years. (2014, April). Retrieved June 29, 2017. from
169
http://www.nielsen.com/my/en/press-room/2014/rising-middle-class-will-drive-global-
automotive-demand.html
Oinas-Kukkonen, H. (2010). Behavior change support systems: A research model and agenda. In
Persuasive Technology, Springer Berlin Heidelberg, 4-14.
Oinas-Kukkonen, H. (2013). A foundation for the study of behavior change support systems.
Personal and ubiquitous computing, 17(6), 1223-1235.
Oinas-Kukkonen, H., & Harjumaa, M. (2009). Persuasive systems design: Key issues, process
model, and system features. Communications of the Association for Information Systems, 24(1),
28.
Patterson, Z., & Fitzsimmons, K. (2016). DataMobile: Smartphone Travel Survey
Experiment. Transportation Research Record: Journal of the Transportation Research Board,
(2594), 35-43.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances
in experimental social psychology, 19, 123-205.
Ploderer, B., Reitberger, W., Oinas-Kukkonen, H., & van Gemert-Pijnen, J. (2014). Social
interaction and reflection for behaviour change. Personal and ubiquitous computing, 18(7), 1667-
1676.
Poushter, J. (2016). Smartphone ownership and internet usage continues to climb in emerging
economies. Pew Research Center, 22.
Prochaska, J. O., & Velicer, W. F. (1997). The transtheoretical model of health behavior
change. American journal of health promotion, 12(1), 38-48.
Prochaska, J. O., DiClemente, C. C., & Norcross, J. C. (1992). In search of how people change:
Applications to addictive behaviors. American psychologist, 47(9), 1102.
Richter, J., Friman, M., & Gärling, T. (2011). Soft transport policy measures: Gaps in
knowledge. International journal of sustainable transportation, 5(4), 199-215.
Riley, W. T., Rivera, D. E., Atienza, A. A., Nilsen, W., Allison, S. M., & Mermelstein, R. (2011).
Health behavior models in the age of mobile interventions: are our theories up to the task?.
Translational behavioral medicine, 1(1), 53-71.
Rimer, B. K., & Kreuter, M. W. (2006). Advancing tailored health communication: A persuasion
and message effects perspective. Journal of Communication, 56(s1).
Ripplinger, D., Hough, J. A., Brandt-Sargent, B., Urban, S., & Center, R. T. (2009). The Changing
Attitudes and Behaviors of University Students Toward Public Transportation (No. DP-222).
Upper Great Plains Transportation Institute, North Dakota State University.
170
Rissel, C. E., New, C., Wen, L. M., Merom, D., Bauman, A. E., & Garrard, J. (2010). The
effectiveness of community-based cycling promotion: findings from the Cycling Connecting
Communities project in Sydney, Australia. International journal of behavioral nutrition and
physical activity, 7(1), 8.
Ritterband, L. M., Thorndike, F. P., Cox, D. J., Kovatchev, B. P., & Gonder-Frederick, L. A.
(2009). A behavior change model for internet interventions. Annals of Behavioral Medicine, 38(1),
18.
Rojas IV, M. B., Sadeghvaziri, E., & Jin, X. (2016). Comprehensive Review of Travel Behavior
and Mobility Pattern Studies That Used Mobile Phone Data. Transportation Research Record:
Journal of the Transportation Research Board, (2563), 71-79.
Rose, G., & Marfurt, H. (2007). Travel behaviour change impacts of a major ride to work day
event. Transportation Research Part A: Policy and Practice, 41(4), 351-364.
Rose, G. (2008). Encouraging sustainable campus travel: self-reported impacts of a university
TravelSmart initiative. Journal of Public Transportation, 11(1), 5.
Rotaris, L., & Danielis, R. (2014). The impact of transportation demand management policies on
commuting to college facilities: A case study at the University of Trieste, Italy. Transportation
Research Part A: Policy and Practice, 67, 127-140.
Rotaris, L., & Danielis, R. (2015). Commuting to college: The effectiveness and social efficiency
of transportation demand management policies. Transport Policy, 44, 158-168.
Rothman, A. J., Baldwin, A. S., Hertel, A. W., & Fuglestad, P. T. (2004). Disentangling behavioral
initiation and behavioral maintenance. Handbook of Self-Regulation. Research, Theory, and
Applications; Guilford Press: New York, NY, USA, 130-148.
Sadeghvaziri, E., Rojas IV, M. B., & Jin, X. (2016). Exploring the Potential of Mobile Phone Data
in Travel Pattern Analysis. Transportation Research Record: Journal of the Transportation
Research Board, (2594), 27-34.
Schmettow, M. (2012). Sample size in usability studies. Communications of the ACM, 55(4), 64-
70.
Schrammel, J., Busch, M., & Tscheligi, M. (2013). Peacox-Persuasive Advisor for CO2-Reducing
Cross-Modal Trip Planning. In PERSUASIVE (Adjunct Proceedings).
Schwarzer, R. (2008). Modeling health behavior change: How to predict and modify the adoption
and maintenance of health behaviors. Applied Psychology, 57(1), 1-29.
171
Schwarzer, R., Lippke, S., & Luszczynska, A. (2011). Mechanisms of health behavior change in
persons with chronic illness or disability: the Health Action Process Approach
(HAPA). Rehabilitation psychology, 56(3), 161.
Seethaler, R., & Rose, G. (2006). Using the six principles of persuasion to promote travel
behaviour change: Findings of a TravelSmart pilot test. Road & Transport Research: A Journal of
Australian and New Zealand Research and Practice, 15(2), 94.
Shannon, T., Giles-Corti, B., Pikora, T., Bulsara, M., Shilton, T., & Bull, F. (2006). Active
commuting in a university setting: assessing commuting habits and potential for modal
change. Transport Policy, 13(3), 240-253.
Siddiqi, Z. & Buliung, R. (2013). Dynamic ridesharing and information and communications
technology: past, present and future prospects. Transportation Planning and Technology 36.6,
479-498.
Tang, L., & Thakuriah, P. V. (2012). Ridership effects of real-time bus information system: A case
study in the City of Chicago. Transportation Research Part C: Emerging Technologies, 22, 146-
161.
Thøgersen, J. (2006). Understanding repetitive travel mode choices in a stable context: A panel
study approach. Transportation Research Part A: Policy and Practice, 40(8), 621-638.
Thøgersen, J. (2009). Promoting public transport as a subscription service: Effects of a free month
travel card. Transport Policy, 16(6), 335-343.
Tolley, R. (1996). Green campuses: cutting the environmental cost of commuting. Journal of
Transport Geography, 4(3), 213-217.
Vaz, S., Falkmer, T., Passmore, A. E., Parsons, R., & Andreou, P. (2013). The case for using the
repeatability coefficient when calculating test–retest reliability. PLoS One, 8(9), e73990.
Vij, A., Carrel, A., & Walker, J. L. (2013). Incorporating the influence of latent modal preferences
on travel mode choice behavior. Transportation Research Part A: Policy and Practice, 54, 164-
178.
Virzi, R. A. (1992). Refining the test phase of usability evaluation: How many subjects is enough?.
Human factors, 34(4), 457-468.
Voils, C. I., Gierisch, J. M., Yancy Jr, W. S., Sandelowski, M., Smith, R., Bolton, J., & Strauss, J.
L. (2014). Differentiating behavior initiation and maintenance: Theoretical framework and proof
of concept. Health Education & Behavior, 41(3), 325-336.
Wang, J., Wang, Y., Wei, C., Yao, N., Yuan, A., Shan, Y., & Yuan, C. (2014). Smartphone
interventions for long-term health management of chronic diseases: an integrative review.
Telemedicine and e-Health, 20(6), 570-583.
172
Webb, T., Joseph, J., Yardley, L., & Michie, S. (2010). Using the internet to promote health
behavior change: a systematic review and meta-analysis of the impact of theoretical basis, use of
behavior change techniques, and mode of delivery on efficacy. Journal of medical Internet
research, 12(1), e4.
Weiss, M., Staake, T., Mattern, F., & Fleisch, E. (2012). PowerPedia: changing energy usage with
the help of a community-based smartphone application. Personal and Ubiquitous Computing,
16(6), 655-664.
Wen, L. M., Orr, N., Bindon, J., & Rissel, C. (2005). Promoting active transport in a workplace
setting: evaluation of a pilot study in Australia. Health promotion international, 20(2), 123-133.
Whalen, K. E., Páez, A., & Carrasco, J. A. (2013). Mode choice of university students commuting
to school and the role of active travel. Journal of Transport Geography, 31, 132-142.
Wilson, D., Bopp, M., Colgan, J., Sims, D., Matthews, S., Rovniak, L., & Poole, E. (2016). A
social media campaign for promoting active travel to a university campus. Journal of Healthcare
Communications.
Yang, C. H., Maher, J. P., & Conroy, D. E. (2015). Implementation of behavior change techniques
in mobile applications for physical activity. American journal of preventive medicine, 48(4), 452-
455.
Zhang, D., Schmöcker, J. D., Fujii, S., & Yang, X. (2015). Social norms and public transport usage:
empirical study from Shanghai. Transportation, 1-20.
Zhou, J. (2012). Sustainable commute in a car-dominant city: Factors affecting alternative mode
choices among university students. Transportation research part A: policy and practice, 46(7),
1013-1029.
