Can an AI Really Be Your Friend? Sherry Turkle Says It Can Only Imitate One
AI·November 1, 2026
A new episode of Scientific American's podcast Science Quickly takes on a question that is getting harder to brush off: can an artificial intelligence really be a friend? Books and culture lead Bri Kane sits down with Sherry Turkle, a researcher at the Massachusetts Institute of Technology, to discuss the rise of AI chatbots and how they are changing the way people connect.
Turkle's central point is a distinction between seeming and caring. In her view, AI can imitate intimacy with real skill, producing responses that look like empathy, companionship, friendship and even love. What it cannot do is care about the person on the other end. The conversation explores why that gap matters, since a warm, always-available reply can feel like genuine connection even when nothing on the machine's side is at stake.
The episode also looks at what could be lost. As software takes on a larger share of people's emotional lives, the concern is not only being taken in by a convincing simulation. It is also the slow erosion of the messier kind of relationship, the kind that depends on two people who can disappoint each other, who have to show up, and who are changed by the exchange. Those are the skills and expectations that a frictionless chatbot never requires.
The discussion is tied to Turkle's new book, Artificial Intimacy, published by Little, Brown and Company in 2026, which offers the fuller version of her argument. The Science Quickly team, including producers Rachel Feltman, Fonda Mwangi, Allison Rodgers and Jeff DelViscio, put the episode together, with editing by Alex Sugiura. Listeners with questions, comments or story ideas can write to sciencequickly@sciam.com.