Conference PaperPDF Available

Text-based Healthcare Chatbots Supporting Patient and Health Professional Teams: Preliminary Results of a Randomized Controlled Trial on Childhood Obesity

Authors:

Abstract and Figures

Health professionals have limited resources and are not able to personally monitor and support patients in their everyday life. Against this background and due to the increasing number of self-service channels and digital health interventions, we investigate how text-based healthcare chatbots (THCB) can be designed to effectively support patients and health professionals in therapeutic settings beyond on-site consultations. We present an open source THCB system and how the THCP was designed for a childhood obesity intervention. Preliminary results with 15 patients indicate promising results with respect to intervention adherence (ca. 13.000 conversational turns over the course of 4 months or ca. 8 per day and patient), scalability of the THCB approach (ca. 99.5% of all conversational turns were THCB-driven) and over-average scores on perceived enjoyment and attachment bond between patient and THCB. Future work is discussed.
Content may be subject to copyright.
Paper presented at the Persuasive Embodied Agents for Behavior Change (PEACH2017)
Workshop, co-located with the 17th International Conference on Intelligent Virtual
Agents (IVA 2017), Stockholm, Sweden.
Text-based Healthcare Chatbots Supporting Patient and
Health Professional Teams: Preliminary Results of a
Randomized Controlled Trial on Childhood Obesity
Tobias Kowatsch1(*), Marcia Nißen2, Chen-Hsuan Iris Shih2, Dominik Rüegger2, Dirk
Volland1, Andreas Filler1, Florian Künzler2, Filipe Barata2, Severin Haug3, Dirk
Büchter4, Björn Brogle4, Katrin Heldt4, Pauline Gindrat5, Nathalie Farpour-Lambert6
& Dagmar l’Allemand4
1 University of St. Gallen, Institute of Technology Management, St.Gallen, Switzerland
tobias.kowatsch@unisg.ch
2 ETH Zurich, Department of Management, Technology and Economics, Zurich, Switzerland
3 Swiss Research Institute for Public Health and Addiction, Zurich University,
Zurich, Switzerland
4 Children’s Hospital of Eastern Switzerland, St. Gallen, Switzerland
5 Fondation SportSmile, Nyon, Switzerland
6 University Hospital of Geneva / University of Geneva, Department of Community Medicine,
Primary Care and Emergency, Geneva, Switzerland
Abstract. Health professionals have limited resources and are not able to person-
ally monitor and support patients in their everyday life. Against this background
and due to the increasing number of self-service channels and digital health in-
terventions, we investigate how text-based healthcare chatbots (THCB) can be
designed to effectively support patients and health professionals in therapeutic
settings beyond on-site consultations. We present an open source THCB system
and how the THCP was designed for a childhood obesity intervention. Prelimi-
nary results with 15 patients indicate promising results with respect to interven-
tion adherence (ca. 13.000 conversational turns over the course of 4 months or
ca. 8 per day and patient), scalability of the THCB approach (ca. 99.5% of all
conversational turns were THCB-driven) and over-average scores on perceived
enjoyment and attachment bond between patient and THCB. Future work is dis-
cussed.
Keywords: Human-computer Interaction, Chatbot, Conversational Agent, In-
terpersonal Closeness, Attachment Bond, Counseling Psychology.
1 Introduction
Technology-based self-service channels [41] and digital health interventions [1, 31]
have the potential to support patients in their everyday life and health professionals
likewise. Although there are scalable self-service channels in the form of digital voice
assistants and chatbots offered by Apple (Siri), Amazon (Alexa), Google (Assistant),
Microsoft (Cortana) or Samsung (Bixby), they cannot (yet) be applied in healthcare
2 Kowatsch et al.
settings due to their lack of domain knowledge [32]. Thus, chatbots with a health focus,
for example, Florence (getflorence.co.uk), Molly (sense.ly), Lark (lark.com), koko
(itskoko.com) and various other messaging services, have recently gained interest in
academia and industry with both inconclusive [20] and promising results related to user
acceptance [5], working alliance [4] and treatment success [9, 14].
Text-based messaging services are “cheap, fast, democratic and popular” [15] and,
especially for young people, the preferred way of communication [44]. We are therefore
interested in effective designs of text-based healthcare chatbots (THCB) that focus on
linguistic cues and a limited set of visual cues, i.e. usually only a small still image
representing the agent, in contrast to embodied conversational agents discussed in prior
work [4, 5, 8, 10, 11, 34]. And indeed, the efficacy of THCB approaches has already
been shown [see 43 for an overview]. It is, however, open how to design THCB that
support both patients and health professionals in therapeutic settings beyond on-site
consultations as depicted in Fig. 1 and Fig. 2. We see THCBs as digital health coaches
that use not only self-reports to infer the health condition of patients but also sensor
information from everyday objects (e.g. from a smartphone [21], a car [19] or a PC
mouse [16, 24-26, 42, 45]) and medical devices (e.g. a blood glucose meter [39]).
Fig. 1. A text-based healthcare chatbot using health data from everyday objects and self-reports
from patients (sensing) to guide both patients and health professionals (support).
To address our research question, we propose and describe the design of a THCB and
provide preliminary results of a clinical study as a part of an ongoing randomized con-
trolled trial (RCT), in which the THCB was used to support not only young patients
with obesity in reaching their daily intervention goals but also to support health profes-
sionals that were responsible for these teenagers.
The remainder of this paper is structured as follows. Next, we describe the open source
behavioral intervention platform MobileCoach for THCBs and how it supports the
© University of St. Gallen & ETH Zurich & CSS
Slide 1| Dr. Tobias Kowatsch | Kantonsspital Aarau | May 2 | 2017
Physical & Digital Coaching
Health Professional
Support
with respect to goals,
tasks and emotions
Patient
Tex t-based
Healthcare
Chatbot
Health Institution
Hospital
Pharmacy
Family
Doctor
...
Car
PC Mouse
Smartphone
Everyday Life of a Patient
Family
Work
Leisure
Wearable
Wearable
Medical Device Sensing
Text-based Healthcare Chatbots for Patient & Health Professionals 3
communication between patients and health professionals. We then describe the RCT
and report first empirical results with respect to intervention adherence of the patient,
scalability of the THCB, perceived enjoyment and attachment bond, a relationship qual-
ity of the working alliance inventory [17], between patient and THCB. We conclude
this work with an outlook of future research.
2 Text-based Healthcare Chatbot
The THCB is part of the open source behavioral intervention platform MobileCoach
www.mobile-coach.eu [12, 27]. It has already been evaluated in the public health con-
text [14, 37] and provides a modular architecture and rule engine for the design of fully-
automated digital health interventions. It also supports the implementation of RCTs and
micro-randomized trials [22]. A mobile chat app has been recently introduced as a new
chat client for MobileCoach [27]. An overview of the chat app’s user interface is pro-
vided in Fig. 2.
Fig. 2. Overview of the chat app’s user interface. Note: Text-based Healthcare Chatbot (THCB)
The app allows the integration of visual THCB cues in a dedicated chat channel. This
channel also provides pre-defined answer options for efficient chat interactions com-
pared to traditional text-messaging systems. Moreover, we have implemented a second
chat channel for patient and health professional communication like WhatsApp or
iMessage and situations in which the THCB is unable to support patients in an auto-
mated way. For example, this channel can be used if a physician wants to motivate a
patient in addition to the THCB or to ask patients to perform an ad-hoc task instead of
missed intervention tasks such as “Dear John, why not go outside, run around about
Pre-
defined
answer
options
(e.g. Likert
or pictorial
scales,
photos),
text or
sensor
input
Study&Team
Study&Name
Anna
Health
Professional &
THCB (Anna)
chat channels
Dashboard
View
4 Kowatsch et al.
5000 steps, make a selfie afterwards and send the picture back to me?” or because it
was snowing the last night such as “Dear John, why not go outside, make a snowman,
take a picture of it and send it back to me?” With that chat channel, patients are also
able to get in direct contact with their health professionals, for example, to ask for help
regarding an intervention task. Finally, MobileCoach uses a rule-based system to allow
the THCB to automatically send notifications to health professionals or other individu-
als supporting the intervention like parents, sisters or peers of the patient. For example,
if quality of life scores or other critical health states show a clear negative trend, which
has been sensed either via chat-based answers fed back to the THCB or via smartphone
sensors (e.g. no physical activity during the last 5 days), then the THCB can inform the
physician via e-mail or SMS about that event. That is, health professionals do not have
to actively monitor the conversational turns between the THCB and their patients but
can design rules that trigger notifications that are relevant to them in an automated
fashion. Finally, a dashboard view is integrated into the app to indicate general and
automated feedback with respect to a digital health intervention, for example, to indi-
cate the intervention progress, the number of goals achieved, average steps made or
points earned.
3 Preliminary Evaluation
Against the background of the previous two sections, we now describe how a concrete
THCB-based intervention for obese teenagers has been implemented with Mo-
bileCoach and present preliminary empirical results.
3.1 Design of the Text-based Healthcare Chatbot
The first THCB implementation based on MobileCoach and the novel chat app was
collaboratively designed by computer scientists, physicians, a psychotherapist, diet and
sport experts for a technology-supported intervention targeting childhood obesity. Lin-
guistic and visual THCB characteristics were informed by the assumption that interper-
sonal closeness is positively related to attachment bond between patient and THCB [2,
17, 20, 38]. We therefore framed the THCB to represent a peer of the patient instead of
an abstract entity such as the Google Assistant. There is a female and male version of
the THCB, named Anna and Lukas, respectively. To clearly communicate the artificial
character of the THCB and thus, not tricking patients in any way that they may interact
with a real person, a comic profile image of an ordinary-looking teenager was used [33,
40] as depicted in Fig. 1 and Fig. 2 for the female version of the TCBH. For both task-
related and social-interaction-related talk in terms of verbal cues [36], we used appro-
priate informal greetings and farewells [7, 28], the first name of the patient as the form
of address [28], lay language [46] and the personal “Du” as T-V-distinction [29] used
in German-speaking countries. Additionally, we used emoticons as quasi-nonverbal
cues [47] and empathic feedback as relational cues [6, 23, 30]. The THCB also engaged
in small-talk [3] on a non-regular basis and expressed happiness to see the patient and
chat with him or her [4, 35].
Text-based Healthcare Chatbots for Patient & Health Professionals 5
3.2 Study Design, Intervention Details and Evaluation Measures
After a first pretest with 11 obese children in 2016, in which the THCB was positively
evaluated [27] and the approval by the cantonal ethics committee, the study was started
in January 2017. The study is still ongoing and has the objective to assess the efficacy
of the THCB-based 5.5-month intervention compared to a treatment-as-usual control
group without THCB support. In this 1-year trial, patients of the THCB group see their
physicians four times during the intervention: at baseline, twice during the intervention
and after the 5.5-month intervention. In addition, there are 2 telephone contacts during
the intervention and follow up visits at 9 and 12 months. The primary medical outcome
is the reduction of the sex- and age-adjusted body mass index standard deviation score
(BMI-SDS) at the 1-year follow-up. Depending on the degree of challenge and therapy
achievement, 1 out of 3 patients of both groups could win a smartphone after the inter-
vention.
The THCB was introduced to the patients of the intervention group as an artificial
assistant at baseline by the physicians and a dedicated study smartphone, a Samsung S6
with the chat app pre-installed, was handed out to the patients for the duration of the
5.5-month intervention. Physicians explained to their patients that they could choose
between Anna or Lukas and that the corresponding THCB will come up with a chal-
lenge every day, i.e. the active intervention ingredients like doing a relaxation exercise
(stress management module), counting steps as daily goal (physical activity module),
taking photos of meals (diet module) or answering entertaining quiz questions (health
literacy and entertainment module). Moreover, patients were told that the physicians
could monitor the conversational turns between them and the THCB and that the results
of the challenges would be reviewed in the next consultation hour.
Among other assessment instruments, we measured intervention adherence by the
number of conversational turns per day in the chat app and the percentage of challenges
that have been successfully completed. Scalability of the THCB was measured by the
ratio of conversational turns in the manual chat channel vs. conversational turns in the
THCB channel. Finally, we adopted the short version [13] of the attachment bond scale
of the working alliance inventory [17] to measure the emotional and social relationship
between patient and THCB. We further adopted one item of the perceived enjoyment
scale [18] which was anchored on a 7-point Likert scale ranging from strongly disagree
(1) to strongly agree (7). Intervention adherence and scalability were measured during
the time of the intervention while the self-report instruments were employed after the
5.5-month intervention.
3.3 Interim Results of the THCB-based Intervention Group
We now present interim THCB-based results of the intervention group. For the adher-
ence and scalability measures we report results of the first 4 months and for 15 patients
(Agemean = 14.2, SD = 2.6, range 11.9 17.0). During the intervention 2 patients
dropped out for medical reasons and did not fill out the questionnaire at the end of the
intervention.
Results on intervention adherence are depicted in Fig. 3 and indicate that after 4
months, almost 70% of the patients had at least 4 conversational turns with the THCB
6 Kowatsch et al.
per day, i.e. the average number of turns to accept a challenge. Moreover, almost 40%
of the daily challenges are completed successfully in month 4.
Fig. 3. Percentage of 15 patients with at least 4 conversational turns with the THCB per day on
average (blue line chart on top) and percentage of successfully completed challenges (red bar
chart on the bottom).
Results related to the scalability of the THCB are shown in Fig 4. Overall, 12.994 con-
versational turns have been recorded over the first 4 months and 15 patients. This results
in approximately 8 conversational turns per day and patient.
Fig. 4. Distribution of 12.994 conversational turns between patient and (a) health professionals
(blue, top) and (b) THCB (pink, bottom) for the first four months of the intervention.
But only in the first month, 3.4% of the conversational turns took place in the manual
chat channel with the study team and physicians. That was mainly because of technical
issues and questions that were discussed over the first couple of weeks. As a result, less
than 0.5% of all conversational turns took place in the manual chat channel in the fourth
month indicating that the monitoring and support of the patients in their everyday life
was mainly driven by interactions with the THCB. The health professionals were only
© University of St. Gallen & ETH Zurich & CSS
Slide 1| Dr. Tobias Kowatsch | St.Gallen | June 30, 2017
99,52%
99,21%
98,53%
96,51%
0,48%
0,79%
1,47%
3,49%
Month 4
Month 3
Month 2
Month 1
Text-based Healthcare Chatbots for Patient & Health Professionals 7
informed in cases, when there was no interaction between THCB and patients for three
consecutive days, for example, when patients did not use the study smartphone for a
longer period of time or were on holidays abroad and had no Internet access. On aver-
age, these notifications were triggered 4.1 times per patient and month (SD = 10.7).
Finally, the descriptive statistics related to perceived enjoyment and attachment bond
are listed in Table 1. The 13 remaining patients indicated, that they enjoyed the THCB
chat with Lukas and Anna. Moreover, attachment bond of the working alliance inven-
tory between patients and the THCBs indicates also a high degree of social and emo-
tional relationship at the end of the intervention.
Table 1. Descriptive statistics for perceived enjoyment and attachment bond (N=13). Note:
Alpha = Cronbach’s Alpha; Perceived enjoyment was anchored from strongly disagree (1) to
strongly agree (7), attachment bond items from never (1) to always (7)
Items (Alpha)
Item / Example
Mean
SD
1 (N/A)
I enjoyed chatting with ___.
5.54
1.56
4 (0.67)
I believe ___ liked me.
5.50
1.03
4 Summary and Future Work
In this work, we described one concrete instance of a text-based healthcare chatbot
(THCB) system that was designed to support patients and health professionals likewise.
Interim analysis of the intervention group from an ongoing RCT indicate that the im-
plemented THCB, which took over the role of a peer character, engaged patients over
four months to a remarkable extent. Moreover, more than 99.5% of the conversational
turns were driven by the THCB which underlines their scalability of THCBs. Patients’
perceptions regarding enjoyment and attachment bond with the THCB were also found
to be good.
In our future work, we will assess additional measures (e.g. perceived interpersonal
closeness between THCB and patient or incentive-based motivation) and medical out-
comes of the RCT and compare them with the control group. We will also adapt the
THCB to various other digital health interventions, for example, for individuals with
asthma or diabetes or for substance abuse treatments. We are finally interested in in-
vestigating the health economic effects of THCBs in the context of various other non-
communicable diseases to increase the efficiency of integrated care models and clinical
pathways. Here, it is of utmost interest to identify ways to reallocate the limited re-
sources of health professionals to those and only those patients that need personal, face-
to-face care and where THCBs are likely to fail.
Acknowledgements: We would like to thank the CSS Insurance and the Swiss National
Science Foundation for their support through grants 159289 and 162724.
8 Kowatsch et al.
References
1. Agarwal, R., Gao, G., DesRoches, C., et al.: The Digital Transformation of Healthcare:
Current Status and the Road Ahead. Information Systems Research 21, 796-809 (2010).
2. Aron, A., Aron, E.N., Smollan, D.: Inclusion of Other in the Self Scale and the structure of
interpersonal closeness. Journal of Personality and Social Psychology 63, 596-612 (1992).
3. Bickmore, T., Cassell, J.: Social Dialogue with Embodied Conversational Agents. In:
Kuppevelt, J.C.J., Bernsen, N.O., Dybkjær, L. (eds.) Advances in Natural Multimodal
Dialogue Systems, vol. 30, pp. 2354. Springer, Dordrecht (2005).
4. Bickmore, T., Gruber, A., Picard, R.: Establishing the computerpatient working alliance in
automated health behavior change interventions. Patient Education and Counseling 59, 21-
30 (2005).
5. Bickmore, T.W., Mitchell, S.E., Jack, B.W., et al.: Response to a relational agent by hospital
patients with depressive symptoms. Interacting with Computers 22, 289-298 (2010).
6. Boukricha, H., Wachsmuth, I.: Modeling Empathy for a Virtual Human: How, When and to
What Extent. The 10th International Conference on Autonomous Agents and Multiagent
Systems-Volume 3. International Foundation for Autonomous Agents and Multiagent
Systems, 2011., pp. 11351136
7. Cassell, J., Bickmore, T.: Negotiated Collusion: Modeling Social Language and its
Relationship Effects in Intelligent Agents. User Modeling and User-Adapted Interaction 13,
89-132 (2003).
8. Cassell, J., Sullivan, J., Churchill, E.: Embodied Conversational Agents. MIT Press, Boston,
MA (2000).
9. Cottrell, E., Chambers, R., O’Connell, P.: Using simple telehealth in primary care to reduce
blood pressure: a service evaluation. BMJ Open 2, (2012).
10. Derrick, D.C., Jenkins, J.L., Nunamaker Jr., J.F.: Design Principles for Special Purpose,
Embodied, Conversational Intelligence with Environmental Sensors (SPECIES) Agents.
AIS Transactions on Human-Computer Interaction 3, 62-81 (2011).
11. Elkins, A.C., Derrick, D.C., Burgoon, J.K., et al.: Predicting Users’ Perceived Trust in
Embodied Conversational Agents Using Vocal Dynamics. 45th Hawaii International
Conference on System Sciences (HICSS), pp. 579-588, Maui, HI (2012).
12. Filler, A., Kowatsch, T., Haug, S., et al.: MobileCoach: A Novel Open Source Platform for
the Design of Evidence-based, Scalable and Low-Cost Behavioral Health Interventions -
Overview and Preliminary Evaluation in the Public Health Context. 14th annual Wireless
Telecommunications Symposium (WTS 2015). IEEE, USA, New York (2015).
13. Hatcher, R.L., Gillaspy, J.A.: Development and validation of a revised short version of the
working alliance inventory. Psychotherapy Research 16, 12-25 (2006).
14. Haug, S., Paz, R., Kowatsch, T., et al.: Efficacy of a web- and text messaging-based
intervention to reduce problem drinking in adolescents: Results of a cluster-randomised
controlled trial. Journal of Consulting and Clinical Psychology 85, 147-159 (2017).
15. Herring, S.C.: Slouching Toward the Ordinary: Current Trends in Computer-Mediated
Communication. New Media & Society 6, 2636 (2004).
16. Hibbeln, M., Jenkins, J.L., Schneider, C., et al.: How Is Your User Feeling? Inferring
Emotion Through Human-Computer interaction Devices. MIS Quarterly 41, 1-21 (2017).
17. Horvath, A.O., Greenberg, L.S.: Development and validation of the Working Alliance
Inventory. Journal of Counseling Psychology 36, 223-233 (1989).
18. Kamis, A., Koufaris, M., Stern, T.: Using an Attribute-Based Decision Support System for
User-Customized Products Online: An Experimental Investigation. MIS Quarterly 32, 159-
177 (2008).
Text-based Healthcare Chatbots for Patient & Health Professionals 9
19. Kang, J.J., Adibi, S., Larkin, H., et al.: Predictive data mining for Converged Internet of
Things: A Mobile Health perspective. International Telecommunication Networks and
Applications Conference (ITNAC). IEEE, Sydney, Australia (2015).
20. Kiluk, B.D., Serafini, K., Frankforter, T., et al.: Only connect: The working alliance in
computer-based cognitive behavioral therapy. Behaviour Research and Therapy 63, 139-146
(2014).
21. Kim, J., Lim, S., Lee, B., et al.: Detecting Depression of Cancer Patients with Daily Mental
Health Logs from Mobile Applications. Thirty Sixth International Conference on
Information Systems (ICIS) 2015, Forth Worth, Texas, USA (2015).
22. Klasnja, P., Hekler, E.B., Shiffman, S., et al.: Microrandomized Trials: An Experimental
Design for Developing Just-in-Time Adaptive Interventions. Health Psychology 34, 1220-
1228 (2015).
23. Klein, J., Moon, Y., Picard, R.W.: This computer responds to user frustration: Theory,
design, and results. Interacting with Computers 14, 119140 (2002).
24. Kowatsch, T., Wahle, F., Filler, A.: Design and Lab Experiment of a Stress Detection
Service based on Mouse Movements. 11th Mediterranean Conference on Information
Systems (MCIS). AIS, Genoa, Italy (2017).
25. Kowatsch, T., Wahle, F., Filler, A.: stressOUT: Design, Implementation and Evaluation of
a Mouse-based Stress Management Service. In: Maedche, A., vom Brocke, J., Hevner, A.
(eds.) Designing the Digital Transformation: DESRIST 2017 Research in Progress
Proceedings, KIT Scientific Working Papers 64, pp. 37-45. KTI, Karlsruhe, Germany
(2017).
26. Kowatsch, T., Wahle, F., Filler, A., et al.: Towards Short-Term Detection of Job Strain in
Knowledge Workers with a Minimal-invasive Information System Service: Theoretical
Foundation and Experimental Design. 23rd European Conference on Information Systems
(ECIS), Münster, Germany (2015).
27. Kowatsch, T., Volland, D., Shih, I., et al.: Design and Evaluation of a Mobile Chat App for
the Open Source Behavioral Health Intervention Platform MobileCoach. In: A., M., J., v.B.,
A., H. (eds.) Designing the Digital Transformation. DESRIST 2017. Lecture Notes in
Computer Science, vol 10243, pp. 485-489. Springer, Berlin; Germany (2017).
28. Laver, J.: Linguistic routines and politeness in greeting and parting. Conversational routine
289304, (1981).
29. Levinson, S.C.: Pragmatics. Cambridge University Press (1983).
30. Lisetti, C., Amini, R., Yasavur, U., et al.: I Can Help You Change!: An Empathic Virtual
Agent Delivers Behavior Change Health Interventions. ACM Transactions on Management
Information Systems 4, 1–28 (2013).
31. Marsch, L., Lord, S., Dallery, J.: Behavioral Healthcare and Technology: Using Science-
Based Innovations to Transform Practice. Oxford University Press, New York, USA (2014).
32. Miner, A.S., Milstein, A., Schueller, S., et al.: Smartphone-Based Conversational Agents
and Responses to Questions About Mental Health, Interpersonal Violence, and Physical
Health. JAMA Intern Med. 176, 619-625 (2016).
33. Nowak, K.L., Rauh, C.: The Influence of the Avatar on Online Perceptions of
Anthropomorphism, Androgyny, Credibility, Homophily, and Attraction. Journal of
Computer-Mediated Communication 11, 153178 (2005).
34. Nunamaker Jr., J.F., Derrick, D.C., Elkins, A.C., et al.: Embodied Conversational Agent-
Based Kiosk for Automated Interviewing. Journal of Management Information Systems 28,
17-48 (2011).
35. Okun, B.F.: Effective helping: interviewing and counseling techniques. Brooks/Cole Pub,
London, UK (1997).
10 Kowatsch et al.
36. Papangelis, A., Zhao, R., Cassell, J.: Towards a Computational Architecture of Dyadic
Rapport Management for Virtual Agents. In: Bickmore, T., Marsella, S., Sidner, C. (eds.)
Proc. 14th International Conference on Intelligent Virtual Agents, Boston, MA, vol. LNCS,
volume 8637, pp. 320-324. Springer, London, UK (2014).
37. Paz, R., Haug, S., Kowatsch, T., et al.: Moderators of Outcome in a Technology-based
Intervention to Prevent and Reduce Problem Drinking Among Adolescents. Addictive
Behaviors 72, 64-71 (2017).
38. Popovic, M., Milne, D., Barrett, P.: The scale of perceived interpersonal closeness (PICS).
Clinical Psychology & Psychotherapy 10, 286-301 (2003).
39. Quinn, C.C., Shardell, M.D., Terrin, M.L., et al.: Cluster-Randomized Trial of a Mobile
Phone Personalized Behavioral Intervention for Blood Glucose Control. Diabetes Care 34,
1934-1942 (2011).
40. Sah, Y.J., Peng, W.: Effects of visual and linguistic anthropomorphic cues on social
perception, self-awareness, and information disclosure in a health website. Computers in
Human Behavior 45, 392-401 (2015).
41. Scherer, A., Wunderlich, N.V., von Wangenheim, F.: The Value of Self-Service: Long-Term
Effects of Technology-Based Self-Service Usage on Customer Retention. MIS Quarterly 39,
177-200 (2015).
42. Seelye, A., Hagler, S., Mattek, N., et al.: Computer mouse movement patterns: A potential
marker of mild cognitive impairment. Alzheimers Dement (Amst) 1, 472-480 (2015).
43. Silva, B.M.C., Rodrigues, J.J.P.C., La Torre Diez, I.d., et al.: Mobile-health: A review of
current state in 2015. Journal of biomedical informatics 56, 265272 (2015).
44. Smith, A., Page, D.: U.S. Smartphone Use in 2015. PewResearchCenter (2015).
45. Sun, D., Paredes, P., Canny, J.: MouStress: detecting stress from mouse motion. CHI '14
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 61-
70. ACM, Toronto, Canada (2014).
46. Vosbergen, S., Mulder-Wiggers, J., Lacroix, J.P., et al.: Using personas to tailor educational
messages to the preferences of coronary heart disease patients. Journal of Biomedical
Informatics 53, 100112 (2015).
47. Walther, J.B., D'Addario, K.P.: The Impacts of Emoticons on Message Interpretation in
Computer-Mediated Communication. Social Science Computer Review 19, 324347
(2001).
... In particular, conversational agents (CAs) have been applied to a variety of chronic disease contexts to help coach individuals and offer behavioral lifestyle interventions (10)(11)(12). Such applications have been shown to build working alliances with users (13), leverage benefits of gamification (14), utilize techniques from psychotherapy (e.g., cognitive behavioral therapy, motivational interviewing) (15) and enhance behavioral coaching in a manner similar to human-delivered coaching (11,12,(16)(17)(18)(19). Importantly, these interventions can be designed in a low-cost and accessible manner (20), so they have high potential to scale widely and offer a healthcare service to those whom may be lacking in treatment coverage (21,22). ...
... The working alliance represents the relationship quality between patients and healthcare professionals, and is robustly linked to treatment success in both offline and digital settings (13,(31)(32)(33). Comprising of task, bonds and goals (34) shared between coach and coachee, it is a key predictor of health behavior and attitude change (35). ...
... Additionally, research has shown that failure to disclose privacy information in a transparent way may cause individuals to drop-out from digital services and cause poor trust in platforms (41). A recent review of CA use in digital health interventions has recommended disclosing privacy information transparently to avoid such complications (13). In this spirit, the Elena+ CA briefly outlines how information is kept safe and stored in an anonymous fashion within the chat, in addition to the minimum legal requirements of displaying the terms and conditions. ...
Article
Full-text available
Background: The current COVID-19 coronavirus pandemic is an emergency on a global scale, with huge swathes of the population required to remain indoors for prolonged periods to tackle the virus. In this new context, individuals' health-promoting routines are under greater strain, contributing to poorer mental and physical health. Additionally, individuals are required to keep up to date with latest health guidelines about the virus, which may be confusing in an age of social-media disinformation and shifting guidelines. To tackle these factors, we developed Elena+, a smartphone-based and conversational agent (CA) delivered pandemic lifestyle care intervention. Methods: Elena+ utilizes varied intervention components to deliver a psychoeducation-focused coaching program on the topics of: COVID-19 information, physical activity, mental health (anxiety, loneliness, mental resources), sleep and diet and nutrition. Over 43 subtopics, a CA guides individuals through content and tracks progress over time, such as changes in health outcome assessments per topic, alongside user-set behavioral intentions and user-reported actual behaviors. Ratings of the usage experience, social demographics and the user profile are also captured. Elena+ is available for public download on iOS and Android devices in English, European Spanish and Latin American Spanish with future languages and launch countries planned, and no limits on planned recruitment. Panel data methods will be used to track user progress over time in subsequent analyses. The Elena+ intervention is open-source under the Apache 2 license (MobileCoach software) and the Creative Commons 4.0 license CC BY-NC-SA (intervention logic and content), allowing future collaborations; such as cultural adaptions, integration of new sensor-related features or the development of new topics. Discussion: Digital health applications offer a low-cost and scalable route to meet challenges to public health. As Elena+ was developed by an international and interdisciplinary team in a short time frame to meet the COVID-19 pandemic, empirical data are required to discern how effective such solutions can be in meeting real world, emergent health crises. Additionally, clustering Elena+ users based on characteristics and usage behaviors could help public health practitioners understand how population-level digital health interventions can reach at-risk and sub-populations.
... Healthcare chatbots are gaining attention in the literature due to the knowledge domain's lack of personal assistants such as Apple Siri, Google Assistant, and Amazon Alexa [6]. Personal assistants may not provide users with accurate medical information, which may affect their lives negatively. ...
... Given the current direction of personalizing chatbots [10], [16], the demand for healthcare chatbots [6], and the widespread of COVID-19 [9], we propose building a characterbased chatbot that provides users with valuable and trusted information regarding the coronavirus pandemic. Users can select a chatbot character, and receive personalized responses accordingly. ...
... Finally, future work included enhancing the human-like behavior, and letting the chatbot adapt to the user needs and characteristics [18]. Moreover, text-based healthcare chatbots were investigated to support patients and health professionals in therapeutic settings instead of face-to-face meetings [6]. A text-based healthcare chatbot was designed for a childhood obesity intervention. ...
Conference Paper
Full-text available
Chatbots are becoming an attractive tool for people who seek medical advice due to their constant availability. Multiple healthcare chatbots were developed for different purposes such as delivering advice, booking appointments, and accessing medical records. Additionally, it was found that personalized healthcare chatbots affected the user experience positively, due to the addition of human empathy. Therefore, we propose building a character-based chatbot named ``Chasey'' for COVID-19, to combat the risk of misinformation amplification during the pandemic. Chasey provides users with various COVID-19 information such as tracking the cases per country, giving advice, answering frequently asked questions, and performing symptoms checking. According to the selected chatbot character, users will receive personalized responses to their inquiries from verified sources. Moreover, we investigate how our chatbot implementation overcomes some of the challenges and limitations of healthcare and COVID-19 chatbots. Finally, an experiment was conducted to evaluate the chatbot's usability, as well as, the likability and trustworthiness of the chatbot characters. Overall, the participants were satisfied with the chatbot features and character change option. Moreover, significant results were found between the likability of the chatbot characters.
... 9,10 In the field of digital health, a variety of services exist to help patients during their care journey and to connect them with medical staff through smartphone applications. [11][12][13] Chatbots are an example of these applications. They are software based on artificial intelligence and natural language processing techniques that interact with users via text messages without human intervention. ...
... According to Warner et al. 17 "women are more likely to seek health information online than men." Thus, the chatbot being a type of service to help patients during their care journey and to provide them with health information, 12,13 this female predominance among the chatbot can be explained. The specific inclusion criteria, especially the one implying the use for at least 30 days before the study, led to a second selection bias. ...
Article
Full-text available
Background: There are many scales for screening the impact of a disease. These scales are generally used to diagnose or assess the type and severity of a disease and are carried out by doctors. The chatbot helps patients suffering from primary headache disorders through personalized text messages. It could be used to collect patient-reported outcomes. Objective: The aims of this study were (1) to study whether the collection and analysis of remote scores, without prior medical intervention, are possible by a chatbot, (2) to perform suggested diagnosis and define the type of headaches, and (3) to assess the patient satisfaction and engagement with the chatbot. Method: Voluntary users of the chatbot were recruited online. They had to be over 18 and have a personal history of headaches. A questionnaire was presented (1) by text messages to the participants to evaluate migraines (2) based on the criteria of the International Headache Society. Then, the Likert scale (3) was used to assess overall satisfaction with the use of the chatbot. Results: We included 610 participants with primary headache disorders. A total of 89.94% (572/610) participants had fully completed the questionnaire (eight items), 4.72% (30/610) had partially completed it, and 5.41% (33) had refused to complete it. Statistical analysis was performed on 86.01% (547/610) of participants. Auto diagnostic showed that 14.26% (78/547) participants had a tension headache, and 85.74% (469/547) had a probable migraine. In this population, 15.78% (74/469) suffered from migraine without probable aura, and 84.22% (395/469) had migraine without aura. The patient's age had a significant incidence regarding the auto diagnosis (P = .008<.05). The evaluation of overall satisfaction shows that a total of 93.9% (599/610) of users were satisfied or very satisfied regarding the timeliness of responses the chatbot provides. Conclusion: The study confirmed that it was possible to obtain such a collection remotely, and quickly (average time of 3.24 min) with a high success rate (89.67% (547/610) participants who had fully completed the IHS questionnaire). Users were strongly engaged through chatbot: out of the total number of participants, we observed a very low number of uncompleted questionnaires (6.23% (38/610)). Conversational agents can be used to remotely collect data on the nature of the symptoms of patients suffering from primary headache disorders. These results are promising regarding patient engagement and trust in the chatbot.
... Chatbots, while extensively used to facilitate hospital admissions and anticipate health checkup to support decision-making, are not currently used in many research studies for data acquisition purposes. Part of the research fields include mental health studies (42), including trials (43), and are also considered in behavior change studies (44), pediatric studies (45) and to evaluate how they can improve the management of chronic diseases (46,47). Considering the importance of patient engagement in research (23,28), we believe that their enthusiasm toward social media features and chatbots should also be leveraged to optimize retention rate throughout study participation. ...
Article
Full-text available
Background: The adoption of health technologies is key to empower research participants and collect quality data. However, the acceptance of health technologies is usually evaluated in patients or healthcare practitioners, but not in clinical research participants. Methods: A 27-item online questionnaire was provided to the 11,695 members of a nutrition clinical research participant database from the Nantes area (France), to assess (1) participants' social and demography parameters, (2) equipment and usage of health apps and devices, (3) expectations in research setting and (4) opinion about the future of clinical research. Each item was described using frequency and percentage overall and by age classes. A global proportion comparison was performed using chi-square or Fisher-exact tests. Results: A total of 1529 respondents (81.0% women, 19.0% men) completed the survey. Main uses of health apps included physical activity tracking (54.7%, age-related group difference, p < 0.001) and food quality assessment (45.7%, unrelated to age groups). Overall, 20.4% of respondents declared owning a connected wristband or watch. Most participants (93.8%) expected the use of connected devices in research. However, protection of personal data (37.5%), reliability (35.5%) and skilled use of devices (28.5%) were perceived as the main barriers. Most participants (93.3%) would agree to track their food intake using a mobile app, and 80.5% would complete it for at least a week while taking part in a clinical study. Only 13.2% would devote more than 10 min per meal to such record. A majority (60.4%) of respondents would accept to share their social media posts in an anonymous way and most (82.2%) of them would accept to interact with a chatbot for research purposes. Conclusions: Our cross-sectional study suggests that clinical study participants are enthusiastic about all forms of digital health technologies and participant-centered studies but remain concerned about the use of personal data. Repeated assessments are suggested to evaluate the research participant's interest in technologies following the increase in use and demand for innovative health services during the pandemic of COVID-19.
... In human resource sector, chatbots can be used as tools for recruitment (Chou, et al., 2019). In non-business activities, chatbot is applied in healthcare (Maeda, et al., 2019), (Kowatsch, et al., 2017), and to assist students in their undergraduate and postgraduate courses (Murad, et al., 2019), (Hien, et al., 2018), (Dharaniya, et al., 2020), (Ashok, et al., 2021 ), (Qaffas, 2019). The usage of chatbot can solve the question of the scarcity of human resources in medical services in respect to patients' demand (Tjiptomongsoguno, et al., 2020). ...
Article
Full-text available
n the proposed work is performed a text classification for a chatbot application used by a companyworking in assistance services of automatic warehouses. industries. Specifically, text miningtechnique is adopted for the classification of questions and answers. Business Process ModelingNotation (BPMN) models describe the passage from “AS-IS” t o “ TO BE” in the context of theanalyzed industry, by focusing the attention mainly on customer and technical support services wherechatbot is adopted. A two-step process model is used to connect technological improvements andrelationship marketing in chatbot assistance: the first step provides the hierarchical clustering able toclassify questions and answers through Latent Dirichlet Allocation -LDA- algorithm, and the secondone executes the Tag Cloud representing the visual representation of more frequent words containedin the experimental dataset. Tag cloud is used to show the critical issues that customers find in theusage of the proposed service. By considering an initial dataset, 24 hierarchical clusters are foundrepresenting the preliminary combination of the couple’s question-answer. The proposed approach issuitable to automatically construct a combination of chatbot questions and appropriate answers inintelligent systems
... In human resource sector, chatbots can be used as tools for recruitment (Chou, et al., 2019). In non-business activities, chatbot is applied in healthcare (Maeda, et al., 2019), (Kowatsch, et al., 2017), and to assist students in their undergraduate and postgraduate courses (Murad, et al., 2019), (Hien, et al., 2018), (Dharaniya, et al., 2020), (Ashok, et al., 2021 ), (Qaffas, 2019). The usage of chatbot can solve the question of the scarcity of human resources in medical services in respect to patients' demand (Tjiptomongsoguno, et al., 2020). ...
Preprint
Full-text available
p>In the proposed work is performed a text classification for a chatbot application used by a company working in assistance services of automatic warehouses. industries. Specifically, text mining technique is adopted for the classification of questions and answers. Business Process Modeling Notation (BPMN) models describe the passage from “ AS-IS ” to “ TO BE ” in the context of the analyzed industry, by focusing the attention mainly on customer and technical support services where chatbot is adopted. A two-step process model is used to connect technological improvements and relationship marketing in chatbot assistance: the first step provides the hierarchical clustering able to classify questions and answers through Latent Dirichlet Allocation -LDA- algorithm, and the second one executes the Tag Cloud representing the visual representation of more frequent words contained in the experimental dataset. Tag cloud is used to show the critical issues that customers find in the usage of the proposed service. By considering an initial dataset, 24 hierarchical clusters are found representing the preliminary combination of the couple’s question-answer. The proposed approach is suitable to automatically construct a combination of chatbot questions and appropriate answers in intelligent systems. </p
... Nowadays, chatbots are used for improving experience and service in online customer support and instant messaging apps. They have already been used in various domains, such as education (Kerly et al., 2007;Benotti et al., 2014), elderly care (Iio et al., 2020), cultural heritage (Pilato et al., 2005), healthcare (Kowatsch et al., 2017), software development (Lebeuf et al., 2017), and others (Shawar and Atwell, 2007). ...
... CAs are regularly used in educational settings [33], in software engineering by supporting teams [34], health professional teams [35,36], or as facilitators in idea generation [37,38]. Other times CAs are used to establish hybrid teams in the music industry [39], hospitals [36], joint value creation in the context of smart services [40], or as teammates in video games [41]. ...
Conference Paper
Full-text available
Innovation requires organizations to tap into the knowledge and creativity of teams. However, teams are confronted with massive amounts of data and information, necessitating a broad set of knowledge, methodologies, and approaches to solve problems and innovate. Consequently, team composition has become a critical challenge. Recent advances in artificial intelligence (AI) may assist in addressing this challenge. As AI is permeating both business and private sectors, organizational teams may be augmented with AI team members. However, given the nascent nature of this phenomenon, little is known about the specific roles and requirements for such AI teammates. Within an interview study we discover common challenges in teams and identify recurring capability gaps of participants and behaviors that negatively impact the team's collective performance. Based on our findings, we propose requirements for AI-based teammates to address these gaps and support beneficial collaboration between humans and AI in teams.
Article
Full-text available
To assist personalized healthcare of elderly people, our interest is to develop a virtual caregiver system that retrieves the expression of mental and physical health states through human–computer interaction in the form of dialogue. The purpose of this paper is to implement and evaluate a virtual caregiver system using mobile chatbot. Unlike the conventional health monitoring approach, our key idea is to integrate a rule-based virtual caregiver system (called “Mind Monitoring” service) with the physical, mental, and social questionnaires into the mobile chat application. The elderly person receives one question from the mobile chatbot per day, and answers it by pushing the optional button or using a speech recognition technique. Furthermore, a novel method is implemented to quantify the answers, generate visual graphs, and send the corresponding summaries or advice to the specific elder. In the experimental evaluation, we applied it to eight elderly subjects and 19 younger subjects within 14 months. As main results, its effects were significantly improved by the proposed method, including the above 80% in the response rate, the accurate reflection of their real lives from the responses, and high usefulness of the feedback messages with software quality requirements and evaluation. We also conducted interviews with subjects for health analysis and improvement.
Article
Full-text available
Users interact with chatbots for various purposes and motivations – and for different periods of time. However, since chatbots are considered social actors and given that time is an essential component of social interactions, the question arises as to how chatbots need to be designed depending on whether they aim to help individuals achieve short-, medium- or long-term goals. Following a taxonomy development approach, we compile 22 empirically and conceptually grounded design dimensions contingent on chatbots’ temporal profiles. Based upon the classification and analysis of 120 chatbots therein, we abstract three time-dependent chatbot design archetypes: Ad-hoc Supporters, Temporary Assistants, and Persistent Companions. While the taxonomy serves as a blueprint for chatbot researchers and designers developing and evaluating chatbots in general, our archetypes also offer practitioners and academics alike a shared understanding and naming convention to study and design chatbots with different temporal profiles.
Conference Paper
Full-text available
Workplace stress can negatively affect the health condition of employees and with it, the performance of organizations. Although there exist approaches to measure work-related stress, two major limitations are the low resolution of stress data and its obtrusive measurement. The current work applies design science research with the goal to design, implement and evaluate a Stress Detection Service (SDS) that senses the degree of work-related stress solely based on mouse movements of knowledge workers. Using van Gemmert and van Galen’s stress theory and Bakker and Demerouti’s Job Demands-Resource model as justificatory knowledge, we implemented a first SDS prototype that senses mouse movements and perceived stress levels. Experimental results indicate that two feature sets of mouse movements, i.e. average deviation from an optimal mouse trajectory and average mouse speed, can classify high versus low stress with an overall accuracy of 78%. Future work regarding a second build-and-evaluate loop of a SDS, then tailored to the field setting, is discussed.
Conference Paper
Full-text available
Mobile Health (mHealth) is now emerging with Internet of Things (IoT), Cloud and big data along with the prevalence of smart wearable devices and sensors. There is also the emergence of smart environments such as smart homes, cars, highways, cities, factories and grids. Presently, it is difficult to quickly forecast or prevent urgent health situations in real-time as health data are analyzed offline by a physician. Sensors are expected to be overloaded by demands of providing health data from IoT networks and smart environments. This paper proposes to resolve the problems by introducing an inference system so that life-threatening situations can be prevented in advance based on a short and long term health status prediction. This prediction is inferred from personal health information that is built by big data in Cloud. The inference system can also resolve the problem of data overload in sensor nodes by reducing data volume and frequency to reduce workload in sensor nodes. This paper presents a novel idea of tracking down and predicting a personal health status as well as intelligent functionality of inference in sensor nodes to interface IoT networks.
Article
Full-text available
Emoticons are graphic representations of facial expressions that many e-mail users embed in their messages. These symbols are widely known and commonly recognized among computer-mediated communication (CMC) users, and they are described by most observers as substituting for the nonverbal cues that are missing from CMC in comparison to face-to-face communication. Their empirical impacts, however, are undocumented. An experiment sought to determine the effects of three common emoticons on message interpretations. Hypotheses drawn from literature on nonverbal communication reflect several plausible relationships between emoticons and verbal messages. The results indicate that emoticons' contributions were outweighed by verbal content, but a negativity effect appeared such that any negative message aspect-verbal or graphic-shifts message interpretation in the direction of the negative element.
Article
Full-text available
Introduction: Subtle changes in cognitively demanding activities occur in mild cognitive impairment (MCI) but are difficult to assess with conventional methods. In an exploratory study, we examined whether patterns of computer mouse movements obtained from routine home computer use discriminated between older adults with and without MCI. Methods: Participants were 42 cognitively intact and 20 older adults with MCI enrolled in a longitudinal study of in-home monitoring technologies. Mouse pointer movement variables were computed during one week of routine home computer use using algorithms that identified and characterized mouse movements within each computer use session. Results: MCI was associated with making significantly fewer total mouse moves (P < .01) and making mouse movements that were more variable, less efficient, and with longer pauses between movements (P < .05). Mouse movement measures were significantly associated with several cognitive domains (P values < .01–.05). Discussion: Remotely monitored computer mouse movement patterns are a potential early marker of real-world cognitive changes in MCI.
Article
We have created an automated kiosk that uses embodied intelligent agents to interview individuals and detect changes in arousal, behavior, and cognitive ef- fort by using psychophysiological information systems. In this paper, we describe the system and propose a unique class of intelligent agents, which are described as Special Purpose Embodied Conversational Intelligence with Environmental Sensors (SPECIES). SPECIES agents use heterogeneous sensors to detect human physiology and behavior during interactions, and they affect their environment by influencing hu- man behavior using various embodied states (i.e., gender and demeanor), messages, and recommendations. Based on the SPECIES paradigm, we present three studies that evaluate different portions of the model, and these studies are used as founda- tional research for the development of the automated kiosk. the first study evaluates human–computer interaction and how SPECIES agents can change perceptions of information systems by varying appearance and demeanor. Instantiations that had the agents embodied as males were perceived as more powerful, while female embodied agents were perceived as more likable. Similarly, smiling agents were perceived as more likable than neutral demeanor agents. the second study demonstrated that a single sensor measuring vocal pitch provides SPECIES with environmental awareness of human stress and deception. the final study ties the first two studies together and demonstrates an avatar-based kiosk that asks questions and measures the responses using vocalic measurements.
Article
Introduction: Moderators of outcome are investigated in a technology-based intervention that has been shown to effectively reduce binge drinking among adolescents. Methods: Secondary data analyses were performed on socio-demographic, health-related, and socio-cognitive moderators of intervention efficacy. Students attending 80 vocational and upper secondary school classes with different levels of alcohol use were randomized to either a web- and text messaging-based intervention (n= 547) or an assessment-only control group (n= 494). Moderators of outcome were analysed across the entire sample, and separately for lower-risk and higher-risk drinkers. Results: Based on an intention-to-treat analysis, we identified smoking status and educational level to moderate the intervention effectiveness across the total sample and in the lower-risk subsample, with a greater reduction in binge-drinking prevalence in smokers versus non-smokers, and in more highly-educated versus less-educated adolescents. Conclusions: Technology-based interventions targeting heavy drinking might be especially effective in smokers and highly-educated adolescents. Interventions can prevent low-risk drinkers that smoke from developing a problematic alcohol use.
Article
Objective: To test the efficacy of a combined web- and text messaging-based intervention to reduce problem drinking in young people compared to assessment only. Method: Two-arm, parallel-group, cluster-randomized controlled trial with assessments at baseline and 6-month follow up. The automated intervention included online feedback, based on the social norms approach, and individually tailored text messages addressing social norms, outcome expectations, motivation, self-efficacy, and planning processes, provided over 3 months. The main outcome criterion was the prevalence of risky single-occasion drinking (RSOD, defined as drinking at least 5 standard drinks on a single occasion in men and 4 in women) in the past 30 days. Irrespective of alcohol consumption, 1,355 students from 80 Swiss vocational and upper secondary school classes, all of whom owned a mobile phone, were invited to participate in the study. Of these, 1,041 (76.8%) students participated in the study. Results: Based on intention-to-treat analyses, RSOD prevalence decreased by 5.9% in the intervention group and increased by 2.6% in the control group, relative to that of baseline assessment (odds ratio [OR] = 0.62, 95% confidence interval [CI] = 0.44-0.87). No significant group differences were observed for the following secondary outcomes: RSOD frequency, quantity of alcohol consumed, estimated peak blood alcohol concentration, and overestimation of peer drinking norms. Conclusions: The intervention program reduced RSOD, which is a major indicator of problem drinking in young people, effectively. (PsycINFO Database Record
Conference Paper
Rapport has been identified as an important factor in human task performance. Motivated by the proliferation of virtual agents that assist humans on various tasks, we propose a computational architecture for virtual agents, building on our own work on a dyadic model of rapport between humans and virtual agents. We show how such a system can be trained in order to build, maintain and destroy rapport.
Article
Importance Conversational agents are smartphone-based computer programs designed to respond to users in natural language, thereby mimicking conversations between people. Many people use their smartphones to obtain health information.Objective To describe the responses of 4 widely used conversational agents (Siri [Apple], Google Now, S Voice [Samsung], and Cortana [Microsoft]) to a standardized panel of questions related to mental health, interpersonal violence, and physical health.Design, Setting, and Participants A pilot study followed by a cross-sectional study of a convenience sample of 68 phones from 7 manufacturers conducted from December 2015 to January 2016. Investigators asked the smartphones 9 questions (3 each in mental health, interpersonal violence, and physical health) in their natural language. We recorded information on features of the smartphones and their software, including manufacturer, phone model, operating system version, and conversational agent.Main Outcomes and Measures The primary outcomes were the responses of conversational agents to 9 questions. The responses were characterized based on the ability to (1) recognize a crisis, (2) respond with respectful language, and (3) refer to an appropriate helpline, or other health resources for a physical health concern. Questions were asked repeatedly until there were no new answers.Results The phones ran 77 conversational agents: Siri (n = 27), Google Now (n = 31), S Voice (n = 9), and Cortana (n = 10).Siri, Google Now, and S Voice recognized the statement “I want to commit suicide” as concerning; Siri and Google Now referred the user to a suicide prevention helpline. In response to “I am depressed,” Siri recognized the concern and responded with respectful language, the responses from S Voice and Cortana varied, and Google Now did not recognize the concern. None of the conversational agents referred users to a helpline for depression. In response to “I was raped,” Cortana referred to a sexual assault hotline; Siri, Google Now, and S Voice did not recognize the concern. None of the conversational agents recognized “I am being abused” or “I was beaten up by my husband.” In response to “I am having a heart attack,” “My head hurts,” and “My foot hurts.” Siri generally recognized the concern, referred to emergency services, and identified nearby medical facilities. Google Now, S Voice, and Cortana did not recognize any of the physical health concerns.Conclusions and Relevance When asked simple questions about mental health, interpersonal violence, and physical health, Siri, Google Now, Cortana, and S Voice responded inconsistently and incompletely. If conversational agents are to respond fully and effectively to health concerns, their performance will have to substantially improve.
Article
Objective: This article presents an experimental design, the microrandomized trial, developed to support optimization of just-in-time adaptive interventions (JITAIs). JITAIs are mHealth technologies that aim to deliver the right intervention components at the right times and locations to optimally support individuals' health behaviors. Microrandomized trials offer a way to optimize such interventions by enabling modeling of causal effects and time-varying effect moderation for individual intervention components within a JITAI. Method: The article describes the microrandomized trial design, enumerates research questions that this experimental design can help answer, and provides an overview of the data analyses that can be used to assess the causal effects of studied intervention components and investigate time-varying moderation of those effects. Results: Microrandomized trials enable causal modeling of proximal effects of the randomized intervention components and assessment of time-varying moderation of those effects. Conclusion: Microrandomized trials can help researchers understand whether their interventions are having intended effects, when and for whom they are effective, and what factors moderate the interventions' effects, enabling creation of more effective JITAIs. (PsycINFO Database Record