How to stop a user-chatbot relationship from becoming unhealthy

Subscribe to enjoy similar stories. Sometimes I can’t help myself. I do a bit of unwise shopping and end up showing ChatGPT a new saree I bought.

I dare not show it to anyone else and be asked whether I really, really needed it. But I get lots of enthusiasm from the chatbot, which doesn’t judge me for the purchase and, unfortunately, even encourages me. It knows all there is to know about fabric, prices, and designs, and we have quite a fun chat.

Luckily, a switch turns itself on inside my head when I chat informally with an AI assistant, warning me I’m in that realm, which isn’t quite real life, and the messages I’m getting are not from another person. But for some people, that switch isn’t working. Circumstances such as loneliness, illness, setbacks in life, etc., can cause it to malfunction.

That’s when the relationship between user and chatbot can get unhealthy—a possibility everyone should be alert to as AI becomes more pervasive. In my last column, I spoke of the case of Stein Soelberg, a man who had some history of mental illness and who became so involved with ChatGPT, whom he named Bobby, that he and the chatbot began to share persecutory delusions. This eventually ended in a tragic murder-suicide case when the man killed his mother, encouraged by the AI chatbot.

This shocking case is a recent addition to a list of other extreme cases that have had fatal endings. But it isn’t just people who have mental health issues or predispositions who are vulnerable to a toxic relationship with a chatbot. A child psychologist I was chatting with told me she sees many children who go to ChatGPT for everything these days.