Artificial Intimacy — Sherry Turkle’s stark warning against AI ‘companionship’

In Ted Chiang’s 2011 sci-fi short story “Dacey’s Patent Automatic Nanny”, a Victorian baby is raised by a mechanical caregiver. Handed over to human nursemaids at the age of two, the child exhibits “feeble-mindedness” and is committed to an institution. The twist comes when a doctor hypothesises that what the stunted child needs is not more love from people, but to be returned to the care of the automaton: “The child embraced the machine as soon as he saw it, and in the days that followed he would play happily with toys as long as [it] was nearby.”

Artificial Intimacy: Who We Become When We Talk to Machines chronicles how close we are to Chiang’s bleak vision of emotional dependency now that the most common use of generative AI in the US is for therapy and companionship. “If social media came for our attention, artificial intimacy comes for our capacity for attachment,” writes its author, Sherry Turkle.

An esteemed social sciences professor at MIT, Turkle has spent almost half a century studying our relationship with machines, measuring “the gap between what humans need and what technologies deliver”.

Her latest book on the subject introduces us to users of the AI companion platform Replika, to the emerging “grief tech” industry and to people who started using ChatGPT for homework or meal planning, only to get sucked into something deeper. Structured around key stages in psychological development — childhood, adolescence, love, work, death — the book argues strongly that our new “posthuman conversational partners” are inadequate companions for life’s journey.

AI companions are built on little more than sophisticated word-guessing software . . . Our increasing dependence on them deprives us of inner growth

AI companions are built on little more than sophisticated word-guessing software . . . Our increasing dependence on them deprives us of inner growth

Turkle grounds her analysis in academic literature, while illuminating it with often haunting vignettes. Ninth-grader Owen uses ChatGPT to role-play tricky conversations with friends until he can be confident he won’t embarrass himself. Brian built his Replika companion during the pandemic as someone to gossip with about work; right now she is an “office buddy”, but he is “open to more” and has withdrawn from dating. Jang Ji-sung is a grieving mother, but ever since a simulated encounter with her dead daughter she is doubly bereft: the child no longer visits her in her dreams.

The melancholy and loneliness that these cases bring up for Turkle are never far from the surface. Successful psychological development can be achieved only in concert with others, she argues, in the context of “relationships of responsibility, consequence, and care”. Healthy attachment, empathy, our capacity for solitude: all require friction, the kind you get only when interacting with an “other”, someone with divergent perceptions and desires.

Beginning with Eliza, the world’s first chatbot, developed in the 1960s at MIT to perform as a psychotherapist, Turkle has repeatedly observed how, in our desire to be understood, we imbue machines with inner lives they don’t possess. (Turkle artfully illustrates her own point when she grows furious with ChatGPT “as though it’s a person” after a role-playing exercise hits a nerve.)

ChatGPT and its ilk can simulate empathy (for example, it reassures anxious Maria, “even if you don’t get into graduate school, everything is going to be okay”). It may feel as though these companions — built on top of what is essentially sophisticated word-guessing software — care for us more than the people in our lives. But when we talk to machines, like Narcissus, we are talking to “self-objects” — reflections of ourselves. Our increasing dependence on them deprives us of inner growth. We succumb to what the sociologist Émile Durkheim termed anomie: disorientation, depression and anxiety.

“The more we recognize our vulnerability to the virtual, the more we’ll be able to commit to the real,” writes Turkle. We need to “reinvent sociality”. She warns that chatbots’ imperative “to quiet us down” pacifies us at a time when we urgently need to engage in civic life.

Happy to issue this political call to arms, Turkle is militant in naming the corporate incentives behind AI’s growing ubiquity only when critiquing the underlying economic conditions closest to her field (we could “train an army of mental health professionals” with the money being invested in AI, she argues). Elsewhere, her remedies are frustratingly tentative. Her suggestion that Meta might funnel some of its profits into public meeting places felt like a fantasy from a bygone era.

But then, Turkle is old enough to remember the community-minded aspirations of the first personal computer enthusiasts — she served “the wrong kind of vegetarian food” to Steve Jobs when he came to her home for dinner in 1977. Unlike Jobs, she is still alive to witness that dream turn sour. Where once Apple asked people to “think different”, now, she writes, their ads promise users there is “no need to think” at all.

Artificial Intimacy: Who We Become When We Talk to Machines by Sherry Turkle, Bridge Street Press £22/Little, Brown & Company $32, 288 pages

Join our online book group on Facebook at FT Books Café and follow FT Weekend on Instagram, Bluesky and X