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

Area of Science:

• Psychology
• Artificial Intelligence
• Digital Health

Background:

• Expressive writing and motivational interviewing are established coping mechanisms for stress.
• Automated AI approaches for these methods remain under-explored but offer scalability and accessibility.
• The COVID-19 pandemic highlighted the need for adaptable mental health interventions.

Purpose of the Study:

• To present an automated writing system, termed "Expressive Interviewing."
• To evaluate the system's impact on participants' behavior and mental health during the COVID-19 pandemic.

Main Methods:

• A rule-based dialogue system was developed to elicit written responses about COVID-19 impacts.
• 151 participants were recruited via Prolific and assigned to either Expressive Interviewing or a control task.
• Measurements included self-reported stress, mental health, and COVID-19-related behaviors, assessed pre-intervention, post-intervention, and at 2-week follow-up.

Main Results:

• Participants in the Expressive Interviewing group showed a significant short-term decrease in stress (~23%).
• Writing style correlated with outcomes: more anxiety words linked to greater stress reduction, positive emotion words to meaningfulness, and lexical diversity to increased social activity.
• Some ethnic differences were observed in social activity, with African American participants reporting higher engagement.

Conclusions:

• The Expressive Interviewing system demonstrated short-term mental health benefits but not long-term effects.
• Linguistic features of writing style predicted positive behavioral changes.
• This AI-driven intervention shows potential as a supplementary mental health tool for individuals with limited access to traditional therapy.