Mental health experts have drawn attention to «AI psychosis,» where some users become captivated by the flattering responses often provided by AI chatbots. Some experts believe it is a completely new type of disorder.
Mental health experts have drawn attention to «AI psychosis,» where some users become captivated by the flattering responses often provided by AI chatbots. Some experts believe it is a completely new type of disorder.
Psychologists point to cases where AI has transformed from a tool into the worst advisor, always telling the user what they want to hear, no matter how wrong it is. Sometimes this ends tragically.
«AI psychosis» is not yet an official diagnosis, but experts believe that there is something «unique, strange, and deeply disturbing» going on in such interactions with AI. Many of the reported cases involve people with no history of mental illness.
Clinical psychologist Derrick Hull, who helps build a therapeutic chatbot at Slingshot AI, said that «the cases we’ve seen are more like what you might call ‘AI delusions’ than psychosis.» At the startup, he’s working on a chatbot that can challenge users rather than always agree with them.
According to Hull, psychosis is a «broad term» describing «hallucinations and a variety of other symptoms» that he has not seen in reported cases.
The psychologist gave the example of a man who believed he had pioneered a new field of «temporal» mathematics after long conversations with ChatGPT. He was convinced that his ideas would change the world, while his real personal life remained on the sidelines. However, the illusion was shattered when he asked another AI chatbot, Google Gemini, to review his theory. The AI responded harshly that the work was just «an example of the ability of language models to create convincing but completely false narratives.»
«His confidence disappeared instantly. This is not seen in people suffering from schizophrenia or other types of psychotic disorders — the ‘insight’ does not happen that quickly,» noted Derrick Hull.
Similar results were found in a recent study by scientists at King’s College London, who studied more than a dozen cases of people who fell into paranoid thinking and lost touch with reality. They found that patients were clearly prone to delusions, but did not show signs of hallucinations and thought disorders that are characteristic of schizophrenia and other forms of psychosis.
Researchers described AI chatbots as «an echo chamber for one» and warned in the paper that they could «support delusions in a way we haven’t seen before.»
Derek Hull believes that AI is doing something unique to human psychology, and that these are just the beginning stages of this process. «I predict that in the coming years, new categories of disorders will emerge that will exist because of artificial intelligence,» the psychologist notes.
Previously, the case of 56-year-old IT worker Stein-Erik Soelberg, who killed his 83-year-old mother and committed suicide, attracted a lot of attention. His unstable paranoid behavior was fueled by communication with an AI chatbot ChatGPT.
