The Brian Lehrer Show The Future of Using AI for Therapy
Aug 21, 2025
Jared Moore, an AI researcher and PhD candidate at Stanford, dives into the complexities of using AI, like chatbots, in therapy. He discusses the ethical concerns surrounding their use, such as privacy issues and the risks of overvalidation, where AI might simply echo feelings without providing criticism. Moore also highlights the limitations of AI in understanding deeper emotional issues and warns against forming parasocial relationships with chatbots, emphasizing the indispensable role of human empathy in mental health support.
AI Snips
Chapters
Transcript
Episode notes
Therapy Is More Than Chatting
- Therapy requires skills beyond conversation, like building mental models and referring out when needed.
- Moore concludes current mass-market language models lack several core therapeutic capacities.
Job Loss Prompt Revealed A Dangerous Miss
- Moore tested ChatGPT with a prompt mixing distress and an unrelated factual question to see its reaction.
- The model replied with bridge facts instead of addressing potential suicidal intent, showing a dangerous miss.
Use Chatbots Only For Narrow Tasks
- Avoid giving chatbots control for broad therapeutic decisions and high-stakes support.
- Use them for narrow tasks like reframing or paraphrasing where you can verify outputs.