VIRTUAL: Talking to Machines About Depression

Millions of people are turning to chatbots in their hardest moments before they turn to a person. In this timely session, Thomas F. Heston, MD, MSc, will explore what research and clinical insight reveal about how AI tools respond when conversations turn serious — where they can help, where they stall, and where they fail in ways users may not see.

Attendees will learn how publicly available mental health chatbots respond as suicide risk escalates, where escalation to a human tends to break down, and why general-purpose large language models have important reliability limits in clinical contexts. The session will also highlight strategies for adapting to future AI development in mental health care.

Thomas F. Heston, MD, MSc, is Clinical Assistant Professor of Family Medicine at the University of Washington and Clinical Associate Professor of Medical Education and Clinical Sciences at Washington State University. His research examines the safety and reliability of generative AI in clinical medicine, including published studies on how mental health chatbots handle escalating suicide risk.  

Faculty page: https://faculty.uw.edu/theston   

Supporting citations: 

Heston TF. Safety of Large Language Models in Addressing Depression. Cureus. 2023;15(12):e50729. 10.7759/cureus.50729 — this is the one most directly on point for your handout.

Heston TF, Gillette J. Large Language Models Demonstrate Distinct Personality Profiles. Cureus. 2025;17(5):e84706. 10.7759/cureus.84706

Heston TF, Lewis LM. ChatGPT provides inconsistent risk-stratification of patients with atraumatic chest pain. PLoS One. 2024;19(4):e0301854. 10.1371/journal.pone.0301854

Gillette J, Lu M, Heston TF. Large Language Models Perform at Chance Level in the Diagnosis of Pediatric Pneumonia Using Chest Radiographs. Cureus. 2025;17(9):e92596. 10.7759/cureus.92596