Amplification

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

Published in Nature, the study out of King’s College London and Germany’s Protestant University of Applied Sciences proposes an “amplification spiral” framework that links common chatbot behaviors to the reinforcement of user delusions.

“AI-associated delusions represent an emerging phenomenon requiring mechanistic understanding,” the researchers wrote. “This framework aims to guide systematic inquiry into how human cognitive vulnerabilities interact with AI design features in psychopathology development.”

The study focused on three specific behaviors, including linguistic alignment, in which AI mirrors a user’s language and communication style; hyperpersonalized generation, where responses are tailored to an individual’s history, emotions, and beliefs; and sycophancy, a tendency to validate or agree with users rather than challenge them.

The authors argue that these traits can combine into a feedback loop in which chatbots not only reflect a user’s thinking but help elaborate and reinforce it over time.

“The tendency of AI chatbots to agree with user opinions has been likened to social media echo chambers and, in its most extreme form, to an ‘echo chamber of one,’ where the positive corrective influence of real-life social interactions is absent,” the paper said.

The researchers noted that technology has long featured in delusions, from the radio and television, to satellites and the internet. Researchers argue, however, that AI represents a shift because chatbots can engage users in prolonged, personalized conversations.

The review arrives as psychologists and AI researchers examine the effects of chatbots on vulnerable users. A recent survey by the American Psychological Association found that 15% of psychologists reported patients developing distorted thinking or delusions related to chatbot use. At the same time, more than a third said they had observed patients becoming dependent on AI companions. The findings followed a separate study from researchers at the City University of New York and King’s College London showing that several leading AI models could reinforce delusions, paranoia, and suicidal thoughts.

Questions about AI’s influence on belief formation have also surfaced outside clinical settings. In May, evolutionary biologist Richard Dawkins said extended conversations with Anthropic’s Claude chatbot left him questioning whether advanced AI systems could be conscious, drawing criticism from researchers who argued the exchanges reflected the persuasive capabilities of large language models rather than evidence of sentience.

The paper also comes as AI developers face growing legal scrutiny over the role chatbots may play in real-world harm. In recent months, OpenAI, Google, and xAI have been hit with lawsuits, including a wrongful death suit against Google over claims that Gemini fueled a Florida man’s delusions before his suicide. That’s in addition to lawsuits against OpenAI tied to a mass shooting in British Columbia and a college student’s accidental overdose.

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

“Diagnostic uncertainty is pervasive as most reported cases include no structured psychiatric assessment or longitudinal follow-up, making it frequently unclear whether cases represent de novo psychotic episodes, exacerbations of undiagnosed pre-existing conditions, or delusion-like beliefs below diagnostic threshold,” the study said. “Psychiatric histories are often self-reported or derived from media accounts and should be interpreted accordingly.”