Virtual Agent-Based Communication Skills Training to Facilitate Health Persuasion Among Peers

TL;DR

This work tackles the challenge of enabling lay volunteers to influence peers on stigmatizing health topics by introducing a web-based virtual-agent system that teaches Motivational Interviewing (MI) skills and offers role-play practice with a vaccine-hesitant persona. It evaluates three training designs—Didactic, Role-Play, and Video—using standardized patients and independent judges to measure MI competence and satisfaction, revealing that interactive, agent-based training boosts learner satisfaction and confidence while producing trainees who meet basic MI benchmarks. The study also uncovers that while interactive training is preferred, MI skill improvements are not consistently superior across designs, highlighting the potential benefits of summative feedback and longitudinal practice. Overall, the findings support scalable, agent-based interpersonal skills training for health persuasion and provide design guidance for broader deployment in community health contexts.

Abstract

Many laypeople are motivated to improve the health behavior of their family or friends but do not know where to start, especially if the health behavior is potentially stigmatizing or controversial. We present an approach that uses virtual agents to coach community-based volunteers in health counseling techniques, such as motivational interviewing, and allows them to practice these skills in role-playing scenarios. We use this approach in a virtual agent-based system to increase COVID-19 vaccination by empowering users to influence their social network. In a between-subjects comparative design study, we test the effects of agent system interactivity and role-playing functionality on counseling outcomes, with participants evaluated by standardized patients and objective judges. We find that all versions are effective at producing peer counselors who score adequately on a standardized measure of counseling competence, and that participants were significantly more satisfied with interactive virtual agents compared to passive viewing of the training material. We discuss design implications for interpersonal skills training systems based on our findings.

Figures (4)

• Figure 1: A screenshot of a virtual agent on our web-based platform. This work is part of a larger project, focused on developing a health behavior change intervention for church communities in the US, which led to tailoring visual aspects of the design to these communities.
• Figure 2: Between-group comparison of Satisfaction with Training Media measure (p<0.05), as well as manipulation checks of Interactivity (p<0.05) and amount of Practice the training provided (p=0.07). Video condition was rated lower than both didactic and role-play in terms of trainee satisfaction, and these ratings were positively correlated with perceptions of interactivity and practice.
• Figure 3: Demonstration of the role-play dialogue flow, using the MI skill of active listening as an example. At each turn, the user is given a maximum of two utterance options that they can choose from, where one option is MI-adherent and one is not. The first blue block shows the omission of some utterances for brevity. These omitted utterances include the segments of the conversation representing the MI-adherent-path utterances from all prior skills, in this case 'starting a conversation' and 'asking open-ended questions'.
• Figure 4: The distributions of scores across all conditions in our study, with dashed lines representing their respective baselines. Measures with a meaningful neutral point in the Likert scale or a MITI-recommended threshold of counseling competency were tested using one-sample hypothesis tests. All Likert measures had means significantly higher than neutral. The mean for Global Relational scores was significantly higher, and the mean for R:Q Ratios was significantly lower than MITI threshold for counseling competency (see \ref{['tab:descriptivesandonesample']} for descriptives).