“The demand for chatbot relationships is there, and it is notably high—pretending it’s not happening is clearly not the solution,” she says. “We’re edging toward a moral panic here, and while we absolutely do need better guardrails, I worry there will be a knee-jerk reaction that further stigmatizes these relationships. That could ultimately cause more harm.”
The study is intended to offer a snapshot of how adults form bonds with chatbots and doesn’t capture the kind of dynamics that could be at play among children or teens using AI, says Pat Pataranutaporn, an assistant professor at the MIT Media Lab who oversaw the research. AI companionship has become a topic of fierce debate recently, with two high-profile lawsuits underway against Character.AI and OpenAI. They both claim that companion-like behavior in the companies’ models contributed to the suicides of two teenagers. In response, OpenAI has recently announced plans to build a separate version of ChatGPT for teenagers. It’s also said it will add age verification measures and parental controls. OpenAI did not respond when asked for comment about the MIT Media Lab study.
Many members of the Reddit community say they know that their artificial companions are not sentient or “real,” but they feel a very real connection to them anyway. This highlights how crucial it is for chatbot makers to think about how to design systems that can help people without reeling them in emotionally, says Pataranutaporn. “There’s also a policy implication here,” he adds. “We should ask not just why this system is so addictive but also: Why do people seek it out for this? And why do they continue to engage?”
The team is interested in learning more about how human-AI interactions evolve over time and how users integrate their artificial companions into their lives. It’s worth understanding that many of these users may feel that the experience of being in a relationship with an AI companion is better than the alternative of feeling lonely, says Sheer Karny, a graduate student at the MIT Media Lab who worked on the research.
“These people are already going through something,” he says. “Do we want them to go on feeling even more alone, or potentially be manipulated by a system we know to be sycophantic to the extent of leading people to die by suicide and commit crimes? That’s one of the cruxes here.”