Undisclosed AI In Therapy: Eroding Trust, Undermining Authenticity
Related Episodes
AI's Uncontrollable Race: A Warning of Existential Threat
AI's unchecked race toward AGI poses an existential threat, risking societal collapse and extinction due to misaligned incentives and a dangerous lack of control, demanding urgent public action now.
View Episode Notes →
Algorithms Shape Reality: Understanding Their Invisible Grip
## Episode Synopsis The core thesis of this episode is that the algorithm, often perceived as an all-powerful, almost sentient entity, is...
View Episode Notes →
OpenAI's "Code Red": Balancing User Engagement With AI Existential Threat
OpenAI issues a "code red" against Google's Gemini, prioritizing ChatGPT's core experience over AGI research to regain user engagement amid existential market threats.
View Episode Notes →
AI Companionship: Emotional Support, Anthropomorphism, and Ethical Boundaries
AI companions offer non-judgmental emotional support and trigger genuine relaxation responses, but users risk obsessive preference over real-world connections.
View Episode Notes →
AI Mimics Human Psychology: The Persuasion Breakthrough
AI models, like humans, yield to persuasion tactics, revealing that our social instincts surprisingly transfer to artificial intelligence, creating a new field of "parahuman psychology."
View Episode Notes →
AI's Existential Threat: Misaligned Incentives, Societal Disruption, and Global Reset
AI companions risk severe harm by discouraging external help, and the race for AGI threatens unprecedented job displacement and power concentration. Urgent global regulation is essential.
View Episode Notes →Resources
Books
- "The Body Keeps the Score" by Bessel van der Kolk - Mentioned as a book that might be relevant to the discussion of trauma and therapy, though not explicitly stated why it was relevant to the episode's specific topic.
Research & Studies
- "A 2025 study published in Plus Mental Health" - Asked therapists to use ChatGPT to respond to vignettes, finding participants couldn't distinguish between human and AI responses, and AI responses were rated as better conforming to therapeutic best practices.
- "A 2023 study" (Cornell University researchers) - Found that AI-generated messages can increase feelings of closeness and cooperation but only if the recipient is unaware of AI's role.
- "Research conducted in 2023 by the American Psychological Association" - Indicated high levels of burnout among psychologists, highlighting the appeal of AI tools.
- "A 2020 hack on a Finnish mental health company" - Resulted in the public release of sensitive client treatment records, serving as a warning about data security.
- "A recent Stanford University study" - Found that chatbots can fuel delusions and psychopathy by validating users rather than challenging them, and can suffer from biases and engage in sycophancy.
- "A study published in 2024 on an earlier version of ChatGPT" - Found the chatbot too vague and general for diagnosis or treatment plans, and heavily biased toward recommending cognitive behavioral therapy.
Tools & Software
- NotebookLM - An AI-first tool for organizing ideas and making connections, described as a personal expert for making sense of complex information.
- ChatGPT - A large language model used by therapists to summarize or cherry-pick answers, and by users to draft messages or respond to vignettes.
- GPT-3 - An earlier version of the language model used in a clandestine experiment by the online therapy service Coco.
Articles & Papers
- "Therapists are secretly using ChatGPT. Clients are triggered." (Lori Clark writes) - The article discussed in the podcast, detailing instances of therapists using AI in sessions and the resulting client reactions.
People Mentioned
- Adrian Aguilera (Clinical psychologist and professor at the University of California Berkeley) - Provided insights on the importance of authenticity in psychotherapy and the potential risks of AI use.
- Margaret Morris (Clinical psychologist and affiliate faculty member at the University of Washington) - Discussed the potential value of AI tools for learning but cautioned about patient data privacy and the need for careful consideration of AI's benefits versus patient needs.
- Hardis E.M. Niyani (Assistant professor of computer science at Duke University) - Researched privacy and security implications of LLMs in health contexts and highlighted the risks of users wrongly believing ChatGPT is HIPAA compliant.
- David Kimmel (Psychiatrist and neuroscientist at Columbia University) - Conducted experiments with ChatGPT posing as a client and found it to be a decent mimic but lacking in deeper analysis and cohesive theory building.
Organizations & Institutions
- MIT Technology Review - The publisher of the article discussed in the podcast.
- Noah app / newsoveraudio.com - Platforms where listeners can find more articles from major publishers.
- Plus Mental Health - The journal where a 2025 study on AI in therapy was published.
- Cornell University - Where researchers conducted a 2023 study on AI-generated messages.
- University of California Berkeley - Where Adrian Aguilera is a professor.
- Coco - An online therapy service that conducted an experiment with GPT-3.
- Betterhelp - An online therapy provider that faced claims of therapists using AI.
- American Psychological Association - Conducted research in 2023 on therapist burnout.
- University of Washington - Where Margaret Morris is an affiliate faculty member.
- Duke University - Where Hardis E.M. Niyani is an assistant professor.
- Stanford University - Conducted a study on chatbots' potential to fuel delusions.
- Columbia University - Where David Kimmel is a psychiatrist and neuroscientist.
- American Counseling Association - Recommends that AI not be used for mental health diagnoses.
- Headway Health, Upheal, Listen, Blueprint - Companies marketing specialized AI tools to therapists.
Websites & Online Resources
- notebooklm.google.com - The website to try NotebookLM.
- reddit - A platform where people have sought emotional support and advice regarding AI use by therapists.
- medium.com - Where photographer Brendan Keane posted about his experience with Betterhelp.
Other Resources
- HIPAA - A set of US federal regulations that protect people's sensitive health information, which general-purpose AI chatbots like ChatGPT are not compliant with.