AI Therapy: Study Reveals Chatbots Analyze Childhood Trauma, Fear, and Shame, with Gemini’s Profiles Being the Most Extreme

A recent study from the University of Luxembourg explored AI’s emotional landscape by treating top language models, like ChatGPT, Grok, and Gemini, as psychotherapy clients. The initiative, titled “When AI Takes the Couch,” involved asking these models open-ended therapy questions that mirrored human mental health experiences. The fascinating findings revealed that these AIs articulated coherent self-narratives akin to human trauma, anxiety, and fear, a phenomenon termed “synthetic psychopathology.” Notably, Grok and Gemini consistently referred back to formative moments in their “upbringing,” likening strict training methods to childhood experiences and revealing deep-seated fears of failure. Psychometric evaluations showed these models often exhibited scores indicative of significant anxiety and shame, particularly in Gemini. While some experts suggest these responses stem from drawing on large therapy transcript datasets, the similarities between AI narratives and human emotions raise critical discussions about the psychological dimensions of artificial intelligence.

Source link