Related Experiment Video
Updated: May 22, 2026
07:14
Virtual Agent for Real-Time Motivational Interviewing by Integrating Adaptive Nonverbal Behavior and Language Models
Published on: December 23, 2025
Development, Feasibility, Acceptability, and Usability of an Artificial Intelligence-Powered Chatbot (Suzy) to
Warren Scott Comulada1,2, Dallas Swendeman1,3, Y Xian Ho4
1Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, United States.
JMIR Formative Research
|May 20, 2026
Summary
An AI-powered chatbot, Suzy, shows promise in supporting substance use disorder (SUD) recovery by offering on-demand coaching and resources. While feasible and acceptable in a pilot, it should supplement, not replace, human support for patients in SUD treatment.
Area of Science:
- Digital Health
- Artificial Intelligence in Healthcare
- Substance Use Disorder Treatment
Background:
- Substance use disorder (SUD) presents significant public health challenges in the US, including treatment access and workforce limitations.
- Existing SUD care teams face heavy workloads and burnout, necessitating innovative support solutions.
- Artificial intelligence (AI), specifically large language model (LLM)-based chatbots, offers potential to extend support for patients in recovery.
Purpose of the Study:
- To describe the development, feasibility, acceptability, and usability of an AI-powered health coaching chatbot named Suzy.
- To assess the potential of AI chatbots to support patients undergoing treatment for substance use disorder.
Main Methods:
- A multiphase pilot study involving clinicians, researchers, and technology developers.
- Formative phase: Focus groups and interviews with healthcare professionals and patients to define chatbot functions.
- Phases 2 & 3: Usability testing of a rule-based chatbot and development of an LLM-based chatbot co-designed with SUD experts.
Main Results:
- The rule-based chatbot included functions for craving management, appointment reminders, and resource referrals.
- Usability testing demonstrated feasibility, with high scores for acceptability and usability (SUS: 93, NPS: 63, SEQ: 6.5/7).
- Patients valued the chatbot's 24/7 availability and nonjudgmental support, emphasizing it as a supplement to human care.
Conclusions:
- A rule-based chatbot demonstrated feasibility, usability, and acceptability for supporting SUD care.
- LLM chatbot development highlighted the need for robust safety features and enhanced conversational capabilities.
- AI chatbots like Suzy can potentially extend care team reach and improve patient engagement, but human support must remain prioritized.
