AI chatbots could lure users away from reality, researchers warn

As AI chatbots become more responsive, chatty, and personalized, researchers warn that these same characteristics could also change the way some users view reality.

A new seal learning“Rethinking AI Psychosis: Misconceptions, Mental Barriers, and Overflow,” examines concerns that AI chatbots may foster delusions, paranoia, and psychological dependence in vulnerable users.

“There has been an increase in media reports about so-called AI psychosis over the past year,” the researchers wrote. “Not surprisingly, this has led to a growing body of research on the ways in which AI chatbots such as ChatGPT, Claude, and Replica It can exacerbate or cause psychosis, which is understood as users acquiring or maintaining false beliefs. “

Research from the University of Copenhagen and the University of Exeter argues that the fear surrounding “AI psychosis” may oversimplify the issue, suggesting that chatbots exacerbate existing problems while slowly reshaping the way users interact with reality and other people.

“If AI interactions were to cause psychosis de novo, we would expect to see much higher clinical risks,” the study said. “In fact, it can be hypothesized that human interactions with AI seem likely to cause or exacerbate pre-existing problems — and, perhaps, that these humans also had underlying threats that led them to interact more with the chatbot.”

The paper comes as cases, criminal investigations, and academic studies are increasingly focused on chatbot interactions that are linked to mass shootings, suicide, emotional dependenceand fraud to think.

In March, a wrongful death a case accused Google’s Gemini chatbot of promoting a Florida man’s scam with fictitious “emissaries” before killing himself. This was followed in April by OpenAI CEO Sam Altman’s presentation public apology to the community of Tumbler Ridge, British Columbia, after the company failed to alert law enforcement about a user account linked to a suspect in a mass shooting that killed eight people in February.

Researchers say chatbots can create a “virtuous cycle” by reinforcing false beliefs through confirmation and reassurance. However, the study Rethinking AI Psychosis argues that this phenomenon is similar to older forms of psychosis created by the advanced technologies of their time.

The debate has also spread beyond psychiatric research to social media. In a recent X post, Box founder Aaron Levie said that CEOs can get too excited about AI’s capabilities because they often see polished results without having to deal with the operational, legal, and technical issues required behind the scenes.

“Seniors are particularly susceptible to AI psychosis because they are so far from the final work that needs to be done to make the most of AI,” Levie said. he wrote. “So when they play with AI, they see results in a fun way, often without thinking about the 10 or 20 things that need to be done to get consistent results from the agents.”

Experts describe this as a type of epistemic driftwhile, over time, users may rely more on the good interpretation of the chatbot than on external evidence or other opinions. However, the Rethinking AI Psychosis paper goes further with a concept the authors call “existential drift,” describing a gradual change in the way a person perceives reality.

“It creates a gap between the individual and the people they share, and reveals reality in a new way, thus establishing a specific, often contradictory perspective,” he wrote.

The researchers dispute that AI friends compare emotional understanding and social interaction without offering real disagreement or independent opinion. Over time, users can feel a sense of interest in the world view that is powered by AI.

The authors say more research is needed to understand the impact of AI conversations on mental health as AI companions become a regular part of everyday life.

“To understand what is happening in these relationships between people and chatbots, we believe it is best to go back to the context, which encourages further research,” they wrote. “In particular, it’s about mental health and how human interactions with AI can change the world, themselves, and others.”