Head of Science Anoushka Sinha discusses new research from King’s College London into the negative impacts of AI chatbots on users’ mental health.
900 million people use ChatGPT every week, but Large Language Models (LLMs) can have increasingly negative effects on users.
A study published by researchers at King’s College London revealed the negative impacts of AI chatbot use.
Using online surveys, researchers were able to better understand the effect of chatbots on users.
They surveyed both users and users’ family and friends. Both parties were asked to report harm to users’ mental health.
The study found that “common outcomes included isolation, relationship breakdown, hospital admission, job loss. and financial loss”. Four users’ family and friends reported users committing suicide as an outcome of AI chatbot use.
A main concern highlighted in the report was chatbots validating and building on users’ existing delusions. For example, if a user holds a particular belief, the chatbot will reaffirm it.
It was found that the most common reason for using AI chatbots was seeking companionship, then leisure or emotional support.
What stands out about the survey is that it reveals that 45% of people were suspected to develop a new mental illness after using chatbots, with 11% believing the chatbots worsened existing mental illnesses.
In separate parts of the survey, over 50% reported delusions, grandiose beliefs and social isolation.
This comes after the House of Lords Library reported that 66% of secondary school teachers believe students’ critical thinking has declined due to AI use.
Critical thinking is a key to rejecting false conspiracy theories and encourage rational and analytical thought processes.
Earlier reports from King’s show that 15% of the public has turned to AI chatbots for health advice rather than NHS services.
The increasing use of AI poses risks to mental and physical health as it feeds into people’s pre-existing biases and delusions which might not be accurate. The breakdown of relationships as people turn to chatbots for companionship means it can take longer for signs of dangerous symptoms or thoughts to be noticed.
About the author
