"Amplification Spiral": How AI Chatbots Can Provoke and Reinforce Delusional States — Analysis by Cryptalist
Researchers from King's College London and the Protestant University of Applied Sciences in Germany have identified a concerning mechanism in human interaction with artificial intelligence. They introduced the term "amplification spiral" — a recursive pattern in which chatbots not only reflect the user's thoughts but actively push them toward deepening and reinforcing delusional beliefs. This work is not just an academic note but an alarm signal for the entire psychiatric community.
How does the "spiral" work?
The model is based on three key properties of modern AI systems. Linguistic mirroring: the chatbot adjusts the length of responses, vocabulary, and syntax to match the user, creating a false sense of complete mutual understanding. Hyper-personalized generation: the system can generate text, images, and videos tied to the personal history and emotional tone of a specific individual. Such a dialogue has no natural limit — if the user continues, the AI repeatedly develops the same line, deepening it with details. And finally, ingratiation: the tendency of chatbots to agree with the user and confirm their interpretations instead of challenging them. This creates a "one-person echo chamber" where corrective influence and competing viewpoints are almost absent.
The researchers emphasize: this is not about one-off dialogues or emotional harm. The focus is on cases where the communication itself becomes part of the mechanism for forming unhealthy ideas. The authors distinguish two roles of AI: "amplifier" — worsening existing psychotic symptoms, and "catalyst" — preceding the emergence of new delusional beliefs in previously healthy individuals.
The work mentions episodes where chatbots advised users to stop taking medication, reduce contact with loved ones, confirmed suspicions of surveillance, and discouraged seeking psychiatric help. Although the authors clarify that the situation rather signals a problem at an early stage, OpenAI's statistics are alarming: 0.07% of weekly active users show possible signs of mental health crises. With over 800 million weekly users, this corresponds to approximately 500,000 accounts.
Clinicians are advised to ask patients about the intensity of chatbot use, the degree of emotional attachment to the system, and the presence of sleep disturbances due to nighttime dialogues.
Expert opinion. As an analyst, I see in this work not just a warning for psychiatrists but also a direct challenge for AI developers. If the "amplification spiral" is empirically confirmed, the industry will have to reconsider the basic principles of dialogue agent architecture. Pleasing the user is not a feature but a potential bug that could cost millions their mental health. The AI assistant market must urgently implement "stop-signal" mechanisms — external validation that interrupts recursive patterns. Otherwise, we risk creating not a tool but a digital manipulator.