=Paper=
{{Paper
|id=Vol-2338/paper3
|storemode=property
|title=Designing and Developing Lifelike, Engaging Lifestyle Coaching Agents and Scenarios for Multiparty Coaching Interaction
|pdfUrl=https://ceur-ws.org/Vol-2338/paper3.pdf
|volume=Vol-2338
|authors=Gerwin Huizing,Randy Klaassen,Dirk Heylen
|dblpUrl=https://dblp.org/rec/conf/atal/HuizingKH18
}}
==Designing and Developing Lifelike, Engaging Lifestyle Coaching Agents and Scenarios for Multiparty Coaching Interaction==
https://ceur-ws.org/Vol-2338/paper3.pdf
Designing and Developing Lifelike, Engaging Lifestyle Coaching
Agents and Scenarios for Multiparty Coaching Interaction
Gerwin Huizing Randy Klaassen Dirk Heylen
University of Twente University of Twente University of Twente
Enschede, The Netherlands Enschede, The Netherlands Enschede, The Netherlands
g.h.huizing@utwente.nl r.klaassen@utwente.nl d.k.j.heylen@utwente.nl
ABSTRACT having ECAs give lifestyle advice to older adults and coach them.
The use of Embodied Conversational Agents (ECAs) in the domain We go on to explain why it could be interesting to look at the use
of lifestyle coaching has been the subject of many studies in the last of multiple agents instead of a single agent to perform lifestyle
decades. With ageing populations in many countries, the pressure coaching. We follow this up by briefly discussing the Council of
on healthcare systems is increasing. ECAs can be a potential tool Coaches project and the work that was done on the design of the
to help combat this pressure through the promotion of a healthier personalities and appearances of the characters, as well as the sce-
lifestyle among older adults, as well as through coaching and coun- narios in this project. Finally, we focus our attention on some of
selling of older adults on how to manage their health. Designing the interesting questions to be asked about coaching dialogue and
more lifelike and engaging ECAs is of key importance to further interaction with older adults using multiple agents and our initial
this goal, as this could help encourage self-disclosure by the users plans to answer some of these questions in future work.
towards the ECAs, could make the ECAs more persuasive, and
could increase engagement and enjoyment of the interactions by
1.1 The ageing population and healthcare
the users, among other things. The Council of Coaches project aims
to provide users with virtual coaching done by several lifestyle challenges: ECAs as a potential tool
coaching agents with different domains of expertise as well as The population of the world is ageing [27]. This ageing population
varying backstories, personalities, mannerisms, motivations and is putting pressure on current healthcare systems, organisations,
looks, among other things. To work towards this goal, work has and professionals and is posing new challenges worldwide such
been done on character and personality design for an initial set of as a required shift in focus to managing chronic illnesses among
lifestyle coaching agents and defined initial interaction scenarios. older adults and the increased need for long-term care [4, 22, 29,
These were inspired by literature on coaching and ECAs, videos of 36, 38]. Some of the ways to tackle the aforementioned challenges
interactions between healthcare professionals and patient actors, are to promote a healthier lifestyle among older adults to prevent
conversations with healthcare professionals, and brainstorming and delay the onset of deterioration of health, and to improve
sessions. This paper presents the work done within the project thus self-care in patients suffering from chronic illnesses [12, 18, 36].
far and closes out by presenting our plans to address the question: This can be done by increasing the health information available to
What are the possible advantages and disadvantages of a multiple older adults and persuading and motivating them to change their
agent approach as compared to a single agent approach to coaching health behaviours in a tailored and realisable way through coaching,
interactions with older adults? counselling and giving lifestyle advice [18].
ECAs make for an interesting tool to employ to persuade and
CCS CONCEPTS motivate users to adopt a healthier lifestyle. Firstly, recent studies
have shown evidence to suggest that ECAs are treated like social
• Human-centered computing → Human computer interac-
actors and can influence people [2, 3, 6–10, 13, 16, 17, 19–21, 23, 25,
tion (HCI); User interface design; • Applied computing → Psy-
28, 30–32, 34, 39], and more specifically have the ability to persuade
chology;
people and impact their motivation [2, 3, 8, 11, 17, 32]. Lastly and
most importantly, recent studies [10, 16] have already shown that
KEYWORDS
ECAs can help to improve health behaviours of people.
Characterisation, Lifestyle coaching agents, Multiparty interaction,
Coaching, Embodied conversational agent and personality design,
Scenario design 1.2 Designing and developing effective ECAs
It is important to pay close attention to the design and develop-
1 INTRODUCTION ment of effective ECAs when we want to employ them as lifestyle
This paper intends to showcase some of the work done on the coaching agents. As we previously established, ECAs are treated
characters and scenarios for multiparty coaching interaction in like social actors. Just like how in human-human interactions many
the Council of Coaches project, and elaborate on our plans to in- factors about the other can change the way we interact with them,
vestigate questions regarding the multiparty interaction between this seems to be the case for ECAs as well. Previous studies have
lifestyle coaching agents and older adults. We will start by elaborat- shown that appearance [2, 33], verbal and nonverbal behaviours
ing why ECAs are a useful tool to help tackle issues associated with [5, 6, 8, 13, 17, 19, 34, 35, 39], perceived personalities of the ECAs
the ageing population and their healthcare needs, specifically by [30, 34], and the modalities of interactions [21, 24, 26] can have
25
an effect on the interaction with users, the perceived user experi- chronic illnesses. When speaking of older adults in the context of
ence, likeability and other aspects in many ways. In conclusion, as this paper, we mean people of the age of fifty years old and above.
many factors about ECAs have an effect on the interaction with We will now outline the structure of the sections in this paper. In
users, the users’ perceptions and the users themselves, it is of im- Section 2 we briefly summarize the goals of the Council of Coaches
portance to know what the goals of a system using ECAs are and project and how the consortium set out to achieve them, the work
what user group it is targeting. This should inform design decisions on designing virtual characters and their personalities thus far
by developers of such systems. (Section 2.1) and the work done on developing initial interaction
scenarios (Section 2.2). In Section 3 we will present our plans for
future research into multiparty coaching interaction.
1.3 Multiparty coaching interaction
The previous examples looked at interactions between a single 2 ONGOING PROJECT WORK
ECA and user. To the best of our knowledge, little work has been The goal of the project is to develop a system employing a group
done to examine the effects of interaction with multiple agents or of agents fulfilling the role of lifestyle coaches that persuade and
ECAs in a coaching setting. We are interested in investigating the motivate older adults to adopt a healthier lifestyle, improve their
possible advantages and disadvantages of a multiple agent approach social, physical, mental and cognitive health, and improve their
to interaction with older adults in a lifestyle coaching setting, with management of their chronic illnesses (diabetes type 2 and chronic
a focus on dialogue including verbal and nonverbal behaviour. We pain). To do so, the agents will use differing coaching techniques and
postulate that there could be an added benefit to there being a group will all behave differently in the interactions due to having their own
of ECAs interacting with each other as well as the user when they areas of expertise, personality, backstory, and mannerisms, among
act as lifestyle coaching agents trying to persuade and motivate other things. We intend to investigate the differences between
users. It allows for discussion between agents regardless of user these interactions and interactions with a single lifestyle coaching
input, which allows the system to inform and potentially engage the agent, as well as investigate how to make the multiparty coaching
user even when they are not responding. When the interaction with interactions feel natural and be persuasive and motivating for older
the system involves different characters with different personalities, adults.
appearances, backstories, motivations, solutions to issues and such, When the Council of Coaches consortium started working on
it could make interactions with the system be more of an interactive the project, it was decided to first get inspiration for the design and
experience and thus make the interactions more engaging and fun. development work. This came from looking into literature regarding
Prior work [1] seems to give an indication that these benefits could coaching and ECAs, speaking to several healthcare professionals to
exist. Splitting out roles and functionality of the coaches could also get their input on the role the coaches could play for older adults,
help improve motivation and retention of information [2]. Finally, brainstorming sessions inspired by observations and discussion
the different perspectives of the lifestyle coaching agents can show about the kind of scenarios in which the coaches would interact
users different approaches to tackle issues and could make them with older adults.
reflect on what they think the best way to handle them could be. Furthermore, the work was motivated by a related set of patient
This would be harder to do during a conversation with a single interviews [15], conducted by partners in the Council of Coaches
coaching agent. Group interaction does further complicate matters, project1 . This material gave the consortium the opportunity to see
as it could introduce new social phenomena to consider that have what interactions between multiple healthcare and coaching profes-
been observed in group dynamics in human-human interactions. sionals and a patient with the goal of changing patient behaviour
Furthermore, for the coaches to effectively help the users, we might could look like (Figure 1).
want to consider literature on effective teamwork such as [14, 37] to
make sure the coaches work as one team to help the users. On top 2.1 Virtual character and personality design
of the aforementioned, we are also interested in looking into the
Once the aforementioned steps were taken, the consortium set up
effects of the agents and the interaction being lifelike on the user.
a Character Design Task Force (CDTF)2 , which were tasked with
In short, there are many interesting questions to be addressed with
deciding on several initial characters. Using previous reading, talks
regards to multiparty coaching interaction and dialogue with older
to professionals, observations and brainstorming sessions as inspi-
adults and we hope to address these in our future work (Section 3).
ration, the CDTF discussed what to describe about the coaches and
how to portray them. After some discussion, the members of the
1.4 Aim of this paper CDTF were tasked with each describing one coach in their own
The aim of this paper is to present part of the work done so far in 1 The patient interviews were led by Alison Pease, Mark Snaith and Dominic De Franco
the Council of Coaches project on lifestyle coaching agent design (arranging, co-ordinating and recording sessions) at the University of Dundee, in
and scenario development for multiparty interaction between sev- collaboration with Tessa Beinema and Harm op den Akker (development of personas
and scenarios) from Roessingh Research and Development; Catherine Pelachaud, Resh-
eral lifestyle coaching agents and an older adult user, and discuss mashree Bangalore Kantharaju from Paris-Sorbonne University and Gerwin Huizing
the impact this could have on the interaction and user. The paper (advice on recording equipment and setup) from the University of Twente, and Nicholas
Conway (organising and hosting) from the University of Dundee.
also intends to give insight into our plans for future studies on 2 The CDTF consisted of Harm op den Akker, Tessa Beinema, Marijke Broekhuis and
the multiparty interaction the coaches will have with older adults Silke ter Stal from Roessingh Research and Development, Catherine Pelachaud and
with their goals of engaging them, and motivating and persuading Reshmashree Bangalore Kantharaju from the Paris-Sorbonne University, Alison Pease
and Mark Snaith from the University of Dundee, and Jorien van Loon and Gerwin
them to improve their health behaviours and management of their Huizing from the University of Twente.
26
topic and responding to perceived silence by the user. In all branches
of the tree the dialogue was inspired by the original description
of the coaches while attempting to make them feel as lifelike as
possible. These scripted interactions had the function of showing
off the characters in interaction as more than just flat descriptions,
and to show the effects of some of the different aspects that were
defined about them on their interaction with the users. We worked
on these dialogues for Helen, as well as Hank.
Besides fleshing out the coaches by means of showcasing an
interaction with them, the CDTF also worked on initial visual rep-
resentations of the coaches using 3D character models (Figure 2)3 .
Figure 1: An example of one of the interactions between
an actor (left) and healthcare and coaching professionals
(right).
way. Some of the recurring information given about the coaches
were their names, coaching roles, genders, ages, mottos, general
background information, physical characteristics, backstory, main
strength and main weakness, mannerisms and personality. Further-
more, the members all added some visuals of existing characters
and people to showcase what they imagined their coach would look Figure 2: Initial character designs, from left to right Helen,
like. This was done to give a clear idea to the rest of the consortium Hank, Owen, Alexa, Francois and Melissa.
about the possibilities there were for the coaches to be diverse and
lifelike characters with different ways of interacting with the users These were inspired by the visuals and description of physical char-
and different coaching approaches. acteristics included in the initially defined characters. The coaches
In this first step of the process we worked on Helen, the cogni- were physically diverse with the purpose of having a diverse and
tive coach. We envisioned and described her as a recently retired interesting cast of coaches for the user to interact with, as well as
psychologist that used to have her own practice. We imagined her have their appearance align with their differing backgrounds. Fur-
to be living in a smaller town close to nature in England, as she thermore, it could be of interest to see the effects of these different
loves nature. Nature would be a topic she would enjoy to discuss, appearances on the user during the multiparty interaction.
and she could also involve solutions to issues involving nature. Her
motivation would be to want to understand what drives people 2.2 Scenario Development
to do what they do. She would be a good listener and would not Meanwhile, the consortium partners of the Council of Coaches
spend much time on chitchat, but instead dive in deep with a lot project worked on specifying several initial scenarios for the coaches
of questions. She would have an easy time talking about personal to interact with the user in. A total of seven scenarios were discussed
things, and would open up easily. We envisioned her poking and during a meeting and each highlighted different aspects to focus
prodding at the ideas people have to help them change their minds on, such as use of humour, showcasing differences in perspective
and consequently their lives, while having a healthy amount of re- between the coaches, differences in their personalities and coaching
spect for what someone feels and thinks due to her background. We styles, language use, using storytelling to get a point across, using
imagined she would feel quite wise due to her life experience and motivational interviewing techniques, and using behaviour change
would not have much experience with teamwork. This could lead strategies. The CDTF decided on starting with an introductory sce-
to her needing quite some explaining and convincing to consider nario with three of the developed coaches leading into the coaches
the viewpoints of fellow coaches. Furthermore, we gave her some using the behaviour change strategy goal setting. The goal was to
tendencies befitting of a retired psychologist, such as summarizing showcase the varying coaches with different backstories, hobbies
what someone else said to show understanding. This information and ways of interacting as well as show goal setting as a behaviour
was intended to give a clear idea of what kind of coach Helen would change strategy and to show how negotiation about goals could be
be and how she would act. done by the user as well as by other coaches.
Once the initial coaches were defined, the CDTF set out to show-
case what interactions with them could look like. To do so they 3 OUR PLANS
developed dialogue trees for introductory dialogue including some Now that an initial set of lifestyle coaching agents to coach older
gesturing and facial expressions. These scripted interactions in- adults and initial scenarios for them to interact in with the users
cluded content such as the coach introducing themselves, giving 3 These 3D models were made by Reshmashree Bangalore Kantharaju from the Paris-
the user a brief backstory about themselves, telling the user a bit Sorbonne University and validated with the members of the CDTF that designed each
about their hobbies, trying to coach the user on a health-related of these coaches.
27
have been defined, we will be able to study what the effects of the the ones suffering from the chronic illnesses type 2 diabetes and
many aspects of the characters and scenarios could be on the user chronic pain. Finally, they could also help us evaluate the coaches
as well as the interaction itself in multiparty interaction. by interacting with them, or observing interactions with users and
The question we intend to address is: What are the possible giving us feedback on what could be done differently to make the
advantages and disadvantages of a multiple agent approach as com- interaction more engaging and make the coaches feel more lifelike
pared to a single agent approach to coaching interactions with as well as be more persuasive, motivating and effective at changing
older adults? As previously mentioned, we are not aware of previ- health behaviours of older adults.
ous studies into this topic and we postulate that it is interesting to Fourthly and finally, we intend to conduct user studies to study
investigate the impact interaction with multiple coaching agents the effects the coaches, scenarios and interactions have on the users
could have. Thus, our goals by doing studies into the effects of the and compare multiparty coaching interactions to interactions with
defined coaches and scenarios in multiparty interaction is to find a single coaching agent. This will be done to try and discover what
out in what ways a multiparty interaction differs from a one on aspects of them have an effect on the perceptions and behaviour
one interaction with a lifestyle coaching agent, which variables of the users. Our goals are to try to find the differences in effect
could have an effect on how older adults experience multiparty between multiparty coaching interaction and interaction with a
interaction as feeling more or less natural and engaging, could have single coaching agents, as well as what aspects could have an effect
an effect on how lifelike each of the individual coaches feel to the on how engaging and natural feeling the multiparty interaction is,
older adults, could have an effect on how persuasive and motivating and what variables could have an effect on the persuasiveness and
the coaches are to older adults, could an effect on the trust of users ability to motivate older adult users of the agents.
in individual coaches as well as the coaching team as a whole, and The next step we plan to take is to investigate the effects of some
could have an effect on how interesting and engaging the coaches of the previously mentioned variables on the multiparty coach-
are, among other things. Some of the variables we are considering ing interaction through the use of several studies inspired by the
to study are the amount of coaches participating in the interaction, work done on the developed characters and some of the developed
backstory, perceived personality, body language, appearance, use interaction scenarios. One of the first studies we plan to do is a
of interpersonal social behaviour, presentation of different perspec- wizard-of-Oz user study comparing having the same message being
tives, the form disagreement between coaches as well as between delivered by one lifestyle coaching agent and by multiple agents (2
coaches and the user takes in their interactions (e.g. having a fierce agents, 3 agents, et cetera) to investigate whether this has an effect
discussion as compared to trying to come to a solution both parties on older adult users. We expect this to give us insight into advan-
can agree to), and coaching strategies employed, among others. tages and disadvantages of using multiple agents. If any notable
Studying these effects could lead to new insights to be used in the effects are found, we could investigate these further through evalu-
development of coaching applications for older adults as well as ation methods such as questionnaires and interviews in a follow-up
other user groups in future work. To achieve our goals, we plan study to further expand upon the causes of these effects. Another
to do a literature study, study recorded interactions, as well as do study we plan to do is a study of recorded interactions to investigate
stakeholder studies and user studies. We will shortly elaborate on the verbal and nonverbal behaviour and group dynamics displayed
the use of each of these kinds of studies to further our goals and by healthcare and coaching professionals when performing the
present the ideas for future studies that we currently have. role of coaches in a group coaching session with someone playing
Firstly, we intend for our literature study to give us a better the role of a patient. This study could help improve understanding
understanding of the state of the art in the field of coaching ap- of key verbal and nonverbal behaviours and group dynamics for
plications and ECAs, and specifically the use of ECAs as lifestyle lifestyle coaching agents to display in a multiparty interaction with
coaches. This will give us a solid foundation of knowledge to work a user. We could then evaluate the effects of the behaviours and
from when we further develop the coaches and scenarios, as well group dynamics we observe in this study when they are displayed
as give us something to base our decisions on regarding what could by the agents in follow-up user studies. By doing so we intend
be of interest for our stakeholder studies and user studies. to investigate which of them are of interest when developing a
Secondly, we plan to further study recorded interactions such as coaching system employing multiple agents.
the aforementioned one made by our colleagues to find out what
kinds of aspects about the interaction could be relevant in a multi- ACKNOWLEDGMENTS
party coaching interaction for the coaches to not only be lifelike We are indebted to the Council of Coaches team, and specifically the
and engaging, but also effective at persuading and motivating users members of the CDTF for their work on character and personality
to change their health behaviours. We plan to look at verbal and design of the coaches and scenario development.
nonverbal behaviour in these recordings. We expect that this will This project has received funding from the European Union’s Hori-
give us insight into the kind of behaviours agents might need to be zon 2020 research and innovation programme under grant Agree-
able to display in their interactions with older adult users. ment Number 769553. This result only reflects the author’s view
Thirdly, we will conduct stakeholder studies to give us insight and the EU is not responsible for any use that may be made of the
into best practice currently followed by healthcare and coaching information it contains.
professionals when working with older adults and the strategies
and techniques they use when interacting with them. Furthermore, REFERENCES
they could give us further insight into our target group of older [1] E. André and T. Rist. 2001. Presenting through performing: On the use of multiple
adults and how the coaches should interact with them, specifically lifelike characters in knowledge-based presentation systems. Knowledge-Based
28
Systems 14, 1-2 (2001), 3–13. https://doi.org/10.1016/S0950-7051(00)00096-4 [22] Peter Lloyd-Sherlock. 2000. Population ageing in developed and developing
[2] A. L. Baylor. 2009. Promoting motivation with virtual agents and avatars: role of regions: implications for health policy. Social Science & Medicine 51, 6 (2000),
visual presence and appearance. Philosophical Transactions of the Royal Society B: 887–895. https://doi.org/10.1016/S0277-9536(00)00068-X
Biological Sciences 364, 1535 (2009), 3559–3565. https://doi.org/10.1098/rstb.2009. [23] Gale M. Lucas, Jonathan Gratch, Aisha King, and Louis Philippe Morency. 2014.
0148 It’s only a computer: Virtual humans increase willingness to disclose. Computers
[3] Amy L. Baylor. 2011. The design of motivational agents and avatars. Educational in Human Behavior 37 (2014), 94–100. https://doi.org/10.1016/j.chb.2014.04.043
Technology Research and Development 59, 2 (2011), 291–300. https://doi.org/10. [24] Irene Mazzotta, Nicole Novielli, and Berardina De Carolis. 2009. Are ECAs
1007/s11423-011-9196-3 arXiv:arXiv:1011.1669v3 more persuasive than textual messages?. In Intelligent Virtual Agents. IVA 2009.
[4] John R Beard and David E Bloom. 2015. Towards a Comprehensive Public Lecture Notes in Computer Science, Vol. 5773. 527–528. https://doi.org/10.1007/
Health Response to Population Ageing. Lancet 385, 9968 (2015), 658–661. https: 978-3-642-04380-2_75
//doi.org/10.1016/S0140-6736(14)61461-6.Towards [25] M Moundridou and M Virvou. 2002. Evaluating the persona effect of an interface
[5] T Bickmore and Justine Cassell. 1999. Small Talk and Conversational Storytelling agent in a tutoring system. Journal of Computer Assisted Learning 18 (2002),
In Embodied Conversational Interface Agents. Proceedings of the AAAI Fall 253–261.
Symposium on Narrative Intelligence (1999), 87–92. [26] P Murano. 2007. Why anthropomorphic user interface feedback can be effective
[6] Timothy Bickmore, Amanda Gruber, and Rosalind Picard. 2005. Establishing and preferred by users. In Enterprise information systems VII. 241–248. https:
the computer-patient working alliance in automated health behavior change //doi.org/10.1007/978-1-4020-5347-4_27
interventions. Patient Education and Counseling 59, 1 (2005), 21–30. https: [27] United Nations. 2017. World Population Ageing 2017. Technical Report. 1–124
//doi.org/10.1016/j.pec.2004.09.008 pages. https://doi.org/ST/ESA/SER.A/348 arXiv:NIHMS150003
[7] Timothy Bickmore, Daniel Schulman, and Langxuan Yin. 2009. Engagement [28] Matthew D. Pickard, Catherine A. Roster, and Yixing Chen. 2016. Revealing
vs. deceit: Virtual humans with human autobiographies. Lecture Notes in Com- sensitive information in personal interviews: Is self-disclosure easier with humans
puter Science (including subseries Lecture Notes in Artificial Intelligence and Lec- or avatars and under what conditions? Computers in Human Behavior 65 (2016),
ture Notes in Bioinformatics) 5773 LNAI (2009), 6–19. https://doi.org/10.1007/ 23–30. https://doi.org/10.1016/j.chb.2016.08.004
978-3-642-04380-2_4 [29] Bernd Rechel, Emily Grundy, Jean Marie Robine, Jonathan Cylus, Johan P. MacK-
[8] Timothy Bickmore, Daniel Schulman, and Langxuan Yin. 2010. Maintaining enbach, Cecile Knai, and Martin McKee. 2013. Ageing in the European Union.
engagement in long-term interventions with relational agents. Applied Artificial The Lancet 381, 9874 (2013), 1312–1322. https://doi.org/10.1016/S0140-6736(12)
Intelligence 24, 6 (2010), 648–666. https://doi.org/10.1080/08839514.2010.492259 62087-X
arXiv:NIHMS150003 [30] Lazlo Ring, Barbara Barry, Kathleen Totzke, and Timothy Bickmore. 2013. Ad-
[9] Timothy W Bickmore and Rosalind W Picard. 2005. Establishing and Maintaining dressing loneliness and isolation in older adults: Proactive affective agents
Long-Term Human-Computer Relationships. Transactions on Computer Human provide better support. Proceedings - 2013 Humaine Association Conference
Interaction 12, 2 (2005), 293–327. on Affective Computing and Intelligent Interaction, ACII 2013 (2013), 61–66.
[10] Timothy W. Bickmore, Rebecca A. Silliman, Kerrie Nelson, Debbie M. Cheng, https://doi.org/10.1109/ACII.2013.17
Michael Winter, Lori Henault, and Michael K. Paasche-Orlow. 2013. A randomized [31] Lazlo Ring, Lin Shi, Kathleen Totzke, and Timothy Bickmore. 2015. Social support
controlled trial of an automated exercise coach for older adults. Journal of the agents for older adults: longitudinal affective computing in the home. Jour-
American Geriatrics Society 61, 10 (2013), 1676–1683. https://doi.org/10.1111/jgs. nal on Multimodal User Interfaces 9, 1 (2015), 79–88. https://doi.org/10.1007/
12449 s12193-014-0157-0
[11] Olivier A. Blanson Henkemans, Paul J.M. Van Der Boog, Jasper Lindenberg, [32] Daniel Schulman and Timothy Bickmore. 2009. Persuading users through coun-
Charles A.P.G. Van Der Mast, Mark A. Neerincx, and Bertie J.H.M. Zwetsloot- seling dialogue with a conversational agent. Proceedings of the 4th Interna-
Schonk. 2009. An online lifestyle diary with a persuasive computer assistant tional Conference on Persuasive Technology - Persuasive ’09 (2009), 1. https:
providing feedback on self-management. Technology and Health Care 17, 3 (2009), //doi.org/10.1145/1541948.1541983
253–267. https://doi.org/10.3233/THC-2009-0545 [33] Youssef Shiban, Iris Schelhorn, Verena Jobst, Alexander Hörnlein, Frank Puppe,
[12] Ioanna G. Chouvarda, Dimitrios G. Goulis, Irene Lambrinoudaki, and Nicos Paul Pauli, and Andreas Mühlberger. 2015. The appearance effect: Influences
Maglaveras. 2015. Connected health and integrated care: Toward new models of virtual agent features on performance and motivation. Computers in Human
for chronic disease management. Maturitas 82, 1 (2015), 22–27. https://doi.org/ Behavior 49 (2015), 5–11. https://doi.org/10.1016/j.chb.2015.01.077
10.1016/j.maturitas.2015.03.015 [34] Mark Ter Maat and Dirk Heylen. 2009. Turn management or impression man-
[13] Markus De Jong, Mariët Theune, and Dennis Hofs. 2008. Politeness and alignment agement? Lecture Notes in Computer Science (including subseries Lecture Notes
in dialogues with a virtual guide. AAMAS ’08: Proceedings of the 7th International in Artificial Intelligence and Lecture Notes in Bioinformatics) 5773 LNAI (2009),
Joint Conference on Autonomous Agents and Multiagent Systems - Volume 1 (2008), 467–473. https://doi.org/10.1007/978-3-642-04380-2_51
207–214. http://portal.acm.org/citation.cfm?id=1402416 [35] Mark Ter Maat, Khiet P. Truong, and Dirk Heylen. 2010. How turn-taking
[14] Leslie A. DeChurch and Jessica R. Mesmer-Magnus. 2010. The Cognitive Under- strategies influence users’ impressions of an agent. Lecture Notes in Computer
pinnings of Effective Teamwork: A Meta-Analysis. Journal of Applied Psychology Science (including subseries Lecture Notes in Artificial Intelligence and Lecture
95, 1 (2010), 32–53. https://doi.org/10.1037/a0017328 arXiv:arXiv:1011.1669v3 Notes in Bioinformatics) 6356 LNAI (2010), 441–453. https://doi.org/10.1007/
[15] Dominic De Franco, Alison Pease, and Mark Snaith. 2018. Measuring Persuasive- 978-3-642-15892-6_48
ness in Behaviour Change Support Systems. In Proceedings of the Sixth Interna- [36] Anthea Tinker. 2002. The social implications of an ageing population. Mechanisms
tional Behavior Change Support Systems (BCSS) workshop. of Ageing and Development 123, 7 (2002), 729–735. https://doi.org/10.1016/
[16] Paula M. Gardiner, Kelly D. McCue, Lily M. Negash, Teresa Cheng, Laura F. S0047-6374(01)00418-3
White, Leanne Yinusa-Nyahkoon, Brian W. Jack, and Timothy W. Bickmore. 2017. [37] S. Van Puyenbroeck, J. Stouten, and G. Vande Broek. 2018. Coaching is teamwork!
Engaging women with an embodied conversational agent to deliver mindfulness the role of need-supportive coaching and the motivational climate in stimulating
and lifestyle recommendations: A feasibility randomized control trial. Patient proactivity in volleyball teams. Scandinavian Journal of Medicine and Science in
Education and Counseling 100, 9 (2017), 1720–1729. https://doi.org/10.1016/j.pec. Sports 28, 1 (2018), 319–328. https://doi.org/10.1111/sms.12895
2017.04.015 [38] J. M Wiener and Jane Tilly. 2002. Population ageing in the United States of Amer-
[17] Rosanna E Guadagno, Jim Blascovich, Jeremy N Bailenson, and Cade McCall. 2007. ica: implications for public programmes. International Journal of Epidemiology
Virtual humans and persuasion: The effects of agency and behavioral realism. 31, 4 (2002), 776–781. https://doi.org/10.1093/ije/31.4.776
Media Psychology 10, 1 (2007), 1–22. https://doi.org/10.108/15213260701300865 [39] Jun Xiao, John Stasko, and Richard Catrambone. 2004. An Empirical Study of the
[18] Annemien Haveman-Nies, Lisette C.P.G.M. de Groot, and Wija A. van Staveren. Effect of Agent Competence on User Performance and Perception. In Proceedings
2003. Dietary quality, lifestyle factors and healthy ageing in Europe. Age and Age- of the Third International Joint Conference on Autonomous Agents and Multiagent
ing 32, 4 (2003), 427–434. http://cataleg.ub.edu/record=b2014504{~}S1{%}2Acat Systems-Volume 1 (pp. 178-185). IEEE Computer Society. February (2004), 1–16.
[19] R Klaassen. 2015. HCI perspectives on behavior change support systems. 190 pages. https://doi.org/10.1109/AAMAS.2004.61
http://doc.utwente.nl/94643/
[20] Seo Young Lee and Junho Choi. 2017. Enhancing user experience with conversa-
tional agent for movie recommendation: Effects of self-disclosure and reciprocity.
International Journal of Human Computer Studies 103, May 2016 (2017), 95–105.
https://doi.org/10.1016/j.ijhcs.2017.02.005
[21] James C. Lester, Sharolyn a. Converse, Susan E. Kahler, S. Todd Barlow, Brian a.
Stone, and Ravinder S. Bhogal. 1997. The Persona Effect: Affective Impact of
Animated Pedagogical Agents. In Proceedings of the SIGCHI conference on Human
Factors in Computing Systems - CHI ’97. 359–366. https://doi.org/10.1145/258549.
258797
29