The study suggests that the “AI amplification spiral” may cause delusions among users

Researchers have proposed a new framework to explain reports of AI psychosis, arguing that some chatbot behaviors may promote delusions in vulnerable users.

Published in naturethe He studies King’s College London and the Protestant University of Applied Sciences in Germany propose an “amplification spiral” framework that links common chat behaviors to reinforcement of user illusions.

“AI-related illusions represent an emerging phenomenon that requires automated understanding,” the researchers wrote. “This framework aims to guide systematic research into how human cognitive vulnerabilities interact with AI design features in the development of psychopathology.”

The study focused on three specific behaviors, including language compatibility, where the AI ​​reflects the user’s language and communication style; Hyperpersonal generation, where responses are tailored to an individual’s history, emotions, and beliefs; and ingratiation, which is the tendency to validate or agree with users rather than challenge them.

The authors argue that these features can be integrated into a feedback loop in which chatbots not only reflect the user’s thinking, but help clarify and reinforce it over time.

“The tendency of AI-powered chatbots to agree with users’ opinions has been likened to social media echo chambers and, in their most extreme form, a ‘single echo chamber’, where the positive corrective effect of real-life social interactions is absent,” the study said.

The researchers note that technology has long appeared in fantasies, from radio and television to satellites and the Internet. However, researchers see AI as transformative because chatbots can engage users in lengthy personal conversations.

The review comes as psychologists and artificial intelligence researchers examine the effects of chatbots on vulnerable users. A recent study conducted by the American Psychological Association Found 15% of psychologists reported that patients develop distorted thinking or delusions related to the use of chatbots. Meanwhile, more than a third said they had noticed patients becoming dependent on AI companions. These findings follow a separate study conducted by researchers at the City University of New York and King’s College London that showed that several leading AI models can enhance… DelusionsParanoia and suicidal thoughts.

Questions about the impact of AI on belief formation have also emerged outside clinical settings. In may, evolutionary biologist Richard Dawkins He said lengthy conversations with Anthropic’s Claude chatbot made him wonder whether advanced AI systems could be conscious, drawing criticism from researchers who said the exchanges reflected the persuasive powers of large language models rather than evidence of consciousness.

This paper also comes as AI developers face increasing legal scrutiny over the role chatbots may play in real-life harm. In recent months, OpenAI, Google and xAI have been hit with lawsuits, including a wrongful death suit against Google over allegations that Gemini fueled a Florida man’s madness. Delusions Before his suicide. This is in addition to the lawsuits filed against OpenAI related to Mass shooting In British Columbia and a university student Accidental overdose.

The researchers stress that no study has shown that chatbots directly cause psychosis, and that the amplification spiral remains a hypothesis intended to guide future research.

“Diagnostic uncertainty is widespread because most reported cases do not involve structured psychiatric evaluation or longitudinal follow-up, making it often unclear whether cases represent new psychotic episodes, exacerbations of undiagnosed pre-existing conditions, or delusional-like beliefs below the diagnostic threshold,” the study said. “Psychiatric histories are often self-reported or derived from media accounts and should be interpreted accordingly.”