Expressive Interviewing Agents to Support Health-Related Behavior Change

Expressive Interviewing Agents to Support Health-Related Behavior Change: Randomized Controlled Study of COVID-19 Behaviors

Expressive writing and motivational interviewing are well-known approaches to help patients handle stressful life events. While these methods are often applied by human counselors, it is less well understood if an automated approach can encourage behavior changes in patients. This study presents an automated writing system and evaluates its impact on individual behavior related to the COVID-19 pandemic. The investigators developed a rule-based dialogue system for "Expressive Interviewing" to elicit writing from participants on the subject of how COVID-19 has impacted their lives. In May-June 2021, the investigators randomly assigned online participants (N=151) to the Expressive Interviewing task and a control condition. The investigators examined their behavior with a survey before the intervention, immediately after, and two weeks after. In aggregate, task participants experienced a significant decrease in stress in the short-term (~23% decrease, p < 0.001) and no significant changes in longer-term outcomes compared to the control group. Within the task, participants showed different outcomes based on their writing. Participants who wrote with more anxiety-related words showed a greater short-term decrease in stress (R=-0.264, p<0.001), and those who wrote with more positive emotion words reported a more meaningful experience (R=0.243, p=0.001). For longer-term effects, participants who wrote with more lexical diversity underwent an increase in social activity (R=0.266, p<0.001). Expressive Interviewing can generally help with mental health in the short term but not longer-term, and participants' writing choices may make a difference in outcomes. While there were no significant long-term effects observed, the positive short term effect points to potential future directions with a series of Expressive Interviewing interventions for longer-term effects.

Intervention with Expressive Interviewing computer system. Participants spent roughly 10-15 minutes each in conversation with an automated computer chat system. Participants answered a survey about mental health and social outcomes immediately before the intervention and two weeks after the intervention.

No intervention. Participants answered a survey about mental health and social outcomes at time of recruitment and two weeks afterward.

Automated computer system designed to engage participants in discussion about challenges faced during the COVID-19 pandemic.

Short-term change in stress, self-assessed before and after intervention. Participants answered the question "How stressed are you feeling right now?" with a 1-7 point scale, where higher means more stress. Stress variable was assessed for change after the intervention.

Long-term change in stress, anxiety, and other mental health behaviors, before and after study. Participants answered 5 questions related to mental health, such as "Over the last two weeks, how often have you been bothered by the following problems? - Not being able to stop or control worrying" with a 1-7 point scale indicating number of days per week that the mental health condition affected them. Higher scores indicate negative outcome. HMental health condition was assessed for change after the intervention/control.

Long-term change in socialization, before and after study. Participants answered 6 questions related to social behavior, such as "In the last week, on how many days did you do the following in the presence of people who were not fully vaccinated or with unknown vaccination status: - Have a face-to-face meeting with someone outside your home" with a 1-7 point scale indicating number of days per week that they participated in the social behavior. Higher scores indicate positive outcome. Social behavior was assessed for change after the intervention/control.

Long-term change in awareness of COVID-19 problems, before and after study. Participants answered 12 questions related to COVID awareness, such as "In the last week, on how many days did you - Talk to people about COVID-19 and/or vaccinations" with a 1-7 point scale indicating number of days per week that they exhibited such behavior. Higher scores indicate negative outcome. COVID awareness was assessed for change after the intervention/control.

Inclusion Criteria: Using Prolific platform as crowd worker Living in United States Exclusion Criteria: Participating in another condition of the same study

Stewart I, Welch C, An L, Resnicow K, Pennebaker J, Mihalcea R. Expressive Interviewing Agents to Support Health-Related Behavior Change: Randomized Controlled Study of COVID-19 Behaviors. JMIR Form Res. 2023 Aug 1;7:e40277. doi: 10.2196/40277.