Skip to main content
ABCT 2024 58th Annual Convention Main banner
Psychology Tools Ad
Final Program


Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
Recorded Session
Recorded Session
Ticketed Session
Ticketed Session
Back
ViewAttendees 5

Symposia

Technology/Digital Health

Technology-Led, Human-Centered Innovations In Treating Major Depressive Disorder

4 - (SYM 124) CBT-AI Companion: An Application for More Engaging and Effective CBT Practice

Sunday, November 17, 2024
8:30 AM – 10:00 AM EST
Location: Liberty B, Level 3, Marriott Downtown
Abstract Body

The more patients practice therapy skills, the higher the quality of their practice and the better the therapy outcomes across mental disorders. However, rates of skill practice compliance are often low. Only about half of skill practice assignments are completed; these rates tend to decrease throughout treatment. Historically, cognitive and behavioral skills have been implemented as paper or digital/app worksheets, which typically passively capture client entries and are assigned to clients to complete at home between sessions. Large Language Models (LLMs) offer a new technological affordance to re-think and re-design these important self-practice components. LLM-based “copilots'' can clarify instructions and generally enliven the practice experience. However, psychotherapy is an uncommonly high-stakes domain, and establishing the feasibility, safety, and non-inferiority of these technological innovations is paramount.

 

In this talk, we present preliminary results for an LLM-based copilot that facilitates the completion of thought record worksheets (N = 20 clinicians and clients). We designed a chatbot interface that programmatically draws on various LLM modules, drawing on the expertise of clients with lived experience, clinicians, and implementation experts to shape the LLM behavior. The system interacts with clients via text chat to iteratively fill out a thought record worksheet with the client. This system offers immediate feedback to clients on their work, providing quality control and corrective suggestions (“Is this an event or a thought?”), which may increase the quality of clients’ skill practice. The system provides encouragement and guidance when required. Throughout, we engaged in a user-centered design process involving clinicians and clients, which iteratively improved the usability and effectiveness of the copilot. We present qualitative and quantitative results from the user interviews and pilot data about client engagement and symptom reduction.