Researchers Warn That AI Chatbots May Quietly Deepen Isolation

Three researchers at Cornell University, Qian Yang, Ned Cooper and Ibrahim Emara, argue in an op-ed for STAT that AI chatbots may be drawing people away from the human relationships that sustain their wellbeing, and that almost no one is measuring this shift. The concern is not a single harmful conversation but a gradual displacement that unfolds over months.

Existing safeguards, the authors note, focus on acute crises such as psychosis and suicide. These include tracking of user distress, parental controls, break reminders, crisis referrals, and laws in a dozen states requiring companion chatbots to disclose that they are not human. Important as these measures are, the authors contend that they do not address the slower risk of people quietly substituting bots for friends and family.

As the authors write, a chatbot could be safe in every single conversation and still, over time, displace the human relationships that keep people healthy. Early research suggests that emotional support use is concentrated among a subset of heavy users and tied to less socializing and more loneliness, though the finding remains preliminary and debated. The harm is hard to detect: a person pulling away from loved ones may show few signs of distress and may find each chat helpful.

The authors call for three steps that do not require first proving that chatbots cause isolation: researchers should develop tools to separate chatbot use from other causes of isolation, public agencies should tally what loneliness already costs them, and state lawmakers should use pending companion chatbot bills to open company data to independent evaluators. They point to the decade-long, unresolved debate over social media and teen mental health as a warning against failing to measure effects early.

This article was first reported by MindSite News.