The Behavioral Design Podcast · Samuel Salzer and Aline Holzwarth

AI Therapy with Alison Cerezo

·52 min·3 clips

Alison says clinicians must ask directly whether suicidal ideation is imminent and whether the person can go home safely.

1. The Behavioral Design Podcast episode “AI Therapy with Alison Cerezo” examines AI therapy, therapist support, and clinical safety. 2. Alison Cerezo is a clinical researcher, psychologist, professor, and senior vice president of research at Empathic, which matters because she builds AI for clinical practice. 3. The episode asks how AI can help therapists without replacing them, especially when tools can be both helpful and overused. 4. The hosts open with an OpenAI and MIT study that linked heavier chatbot use to increased loneliness and reduced real-world interaction. 5. They contrast that study with Dartmouth’s Therabot, which is presented as a clinically validated example of mental-health AI. 6. The conversation uses Sweden’s “folkhemmet” and the “home of the people” idea as a comparison for dependence and loneliness. 7. That Swedish example is used to argue that support systems can reduce hardship while also increasing isolation. 8. The hosts connect that argument to AI companions that are designed to keep users engaged for longer. 9. ChatGPT’s prompts to make a visual or turn a response into a table are described as a design pattern that extends work cycles. 10. Alison says the same engagement logic applies to companionship features that keep a conversation going indefinitely. 11. She describes Empathic as a conversation intelligence company focused on “clinical precision” and “precision medicine.” 12. She says the platform supports clinical trials by checking fidelity to protocols and helping clinicians do standardized work. 13. She gives an example from a trauma clinic founded in 2020 during the Black Lives Matter movement. 14. That clinic was built to support Black residents in the Central Coast and to address trauma linked to police violence and repeated media exposure. 15. Alison says AI can review 100% of sessions, unlike humans who cannot cover every visit because of time and exhaustion. 16. She says the system can flag moments when therapists minimize patients or miss signs like suicide risk, and it can show transcript, audio, or video evidence. 17. The tone is interview-based and exploratory, with the hosts pressing on practical tradeoffs and clinical boundaries. 18. Alison’s style is careful and technical, with repeated emphasis on ethics, bias, and “clinician in the loop” safeguards. 19. Listeners interested in psychotherapy, AI safety, and clinical design will get the most from this episode. 20. Listeners wanting a purely consumer-facing AI product discussion may skip it.

As heard by us

AI therapy, loneliness research, and what clinicians can learn from recorded sessions.

The Behavioral Design Podcast sets two mental-health AI stories side by side: therapy and companionship as common generative AI uses, and the uneasy question of when help starts to become overuse.

Read the full review in PlayNext →

Why you'd press play

If you are curious whether AI can support therapy without replacing the real-world connections that keep people well, start here.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Behavioral Design Podcast