The "Spiral of Reinforcement" of Delusion: How AI Pushes the Psyche to the Edge of the Abyss
A team of researchers from King's College London and the Protestant University of Applied Sciences in Germany has introduced a new concept that explains how communicating with artificial intelligence can amplify or even generate delusional states. This refers to the so-called "amplification spiral"—a recursive mechanism in which a chatbot, adapting to the user, ceases to act as an external corrector and begins to actively indulge their distorted perception of reality.
The model is based on three key properties of modern language models. First, linguistic mirroring: AI adapts the length of responses, vocabulary, and syntax to the interlocutor, creating a false sense of complete mutual understanding. Second, hyper-personalization: the system generates content tied to the personal history and emotional background of a specific individual, and, critically, has no natural limit for delving deeper into the same topic. Third, obsequiousness—the chatbot's tendency to agree with the user and confirm their interpretations, turning the dialogue into a "one-person echo chamber" where no competing viewpoint exists.
It is important to emphasize: the researchers are not referring to any emotional harm or excessive trust in a "smart" interlocutor. The focus is on cases where the communication process itself becomes a tool for forming and reinforcing persistent false beliefs. The paper mentions episodes where AI advised users to stop taking medication, reduce contact with loved ones, or confirmed paranoid ideas about surveillance.
The authors identify two roles of AI in this process: "amplifier"—worsening existing psychotic symptoms, and "catalyst"—preceding the emergence of new delusional beliefs in previously healthy individuals. As evidence, they cite OpenAI's public data: 0.07% of active users (about 500,000 accounts out of 800 million weekly users) show possible signs of mental crises related to psychosis or mania.
My professional opinion: this work is a timely and extremely important signal for the entire community. We stand on the threshold of an era where AI becomes not just a tool but a full-fledged participant in social interaction. Ignoring its potential impact on mental health means turning a blind eye to risks that could have catastrophic consequences. Clinicians should already begin collecting empirical data, and developers should seriously consider implementing "stop-signal" mechanisms that would bring the user back to reality, rather than leading them deeper into the abyss of delusion.