Join us for an AI + Health Seminar with Zhiyu “Zoey” Chen, Assistant Professor at UT Dallas. The AI + Health Seminar Series is every second Thursday during the Fall semester.

Date: October 9, 2025, noon to 12:30 via Zoom: https://utexas.zoom.us/j/5128555388.

Title: Large Language Models for Mental Health: Building Agents for Therapists and Patients

Abstract: Mental health care faces a persistent gap between patient needs and the availability of trained professionals, as well as between existing training methods and the realities of clinical practice. In this talk, I will present two recent projects exploring how large language models (LLMs) can help bridge these gaps by supporting both therapist training and therapy assistance. First, I will introduce PATIENT-Ψ, a framework that simulates realistic therapy patients using LLMs programmed with cognitive models grounded in CBT principles. Through role-playing, mental health trainees can practice building patient cognitive models in interactive training sessions, with user studies showing improvements in both confidence and skill acquisition compared to traditional methods. Second, I will present CBT-Bench, the first systematic benchmark for evaluating LLMs in assisting cognitive behavioral therapy. Comprising expert-annotated tasks ranging from knowledge recall to fine-grained belief classification and response generation, CBT-Bench reveals that while current LLMs handle surface-level CBT knowledge well, they struggle with deeper reasoning about patients’ cognitive structures. Together, these projects highlight both the promise and the current limitations of LLMs for mental health, and point toward future directions for building AI agents that can responsibly support therapists and patients.

Speaker Bio: Zhiyu “Zoey” Chen is an assistant professor at UT Dallas. She is a computer scientist specializing in NLP and deep learning. She has been working on building intelligent and faithful natural language interface grounded on world knowledge to effectively address user requests, which covers the areas of question answering, dialogue and interaction systems, natural language generation, and knowledge reasoning. She’s also interested in applying AI and NLP techniques to broader disciplines and application areas such as healthcare, finance, and social good applications.