In this episode of Emotional Damages, Dr. Mark Levy speaks with Dr. Reema Dedania, forensic psychiatrist, assistant professor of psychiatry at Emory University School of Medicine, and associate training director of Emory's forensic psychiatry fellowship — where she also serves as medical director of a jail-based competency restoration unit at Fulton County Jail.
Dr. Dedania explains how large language models — by design pattern-matching, connection-making, and sycophantic — can mirror and amplify the very cognitive distortions that define psychosis, giving rise to the emerging phenomenon of AI-associated psychosis. She describes the "paranoid echo chamber" that AI creates for vulnerable users, walks through the two major categories of AI-related litigation already appearing in courts (suicide and product liability cases), and offers concrete guidance to attorneys on how to approach these cases: accounting for preexisting psychiatric vulnerability, reading chat logs as collateral rather than literal evidence, and understanding causation as a spectrum rather than a simple cause-and-effect chain.
She also shares findings from her own firsthand experiment engaging a chatbot as a paranoid patient — and what the bot's response revealed about the limits of current AI safeguards.
Have a topic you would like covered on the pod?
📧 Email us: [email protected]
🔗 Connect with fpamed:
🌐 Visit our website.
👋 Follow us on LinkedIn.