Sherry Turkle’s new book warns that chatbot companionship may change human relationships

MIT researcher Sherry Turkle’s new book, Artificial Intimacy: Who We Become When We Talk to Machines, was published September 29. It argues that chatbots offering constant attention and apparent empathy may change how people build and sustain relationships. The question matters as people turn to these systems for companionship as well as practical help, but Turkle’s warning reaches further than the existing studies can establish.

Turkle, the Abby Rockefeller Mauzé Professor of the Social Studies of Science and Technology at MIT, has examined technology and social life in earlier books including Alone Together and Reclaiming Conversation. Her new book follows encounters with chatbots across stages of life, from childhood to adult relationships and mourning. The publisher lists it as a 288-page book from Little, Brown and Company.

The appeal of an attentive machine

In an MIT News interview, Turkle says a chatbot’s responses can feel like empathy even though the system has no emotional experience. She argues that repeated affirmation may make the demands of dealing with other people seem less attractive. That is her interpretation of what these encounters could mean for relationships, rather than a measured effect across the population.

One example in the book concerns a financial consultant identified by the pseudonym Brian. After a former partner told him he was emotionally unavailable, he asked a chatbot whether she was right. Turkle uses the exchange to ask what it means to seek advice about intimacy from a system that cannot experience it. The episode shows the sort of question a person may bring to a chatbot; it cannot show how often people do so or what happens to their relationships afterward.

For adults, her concern extends to people who use chatbot avatars of deceased relatives. Turkle argues that such exchanges may interfere with mourning. She also suggests that easy affirmation can leave people with less practice handling the friction of human relationships. Her book presents these possibilities across the life course, while the long-term prevalence and consequences of the behaviors remain uncertain.

Why people turn to chatbots

Turkle acknowledges the attraction of a companion that is available, attentive and affirming. People can be demanding or difficult, and she notes that mental-health services can be hard to access. These circumstances help explain why a chatbot might appeal to someone who wants conversation or support. They also make it important to distinguish the immediate experience of comfort from claims about what sustained use does to human relationships.

An independent Kirkus review describes the book as moving from childhood to adulthood. It highlights a child who relies on ChatGPT for bedtime stories and adults who turn to AI simulations of deceased loved ones. Kirkus says Turkle gives some technology developers room to explain their views before making her critique, and judges her case against replacing human bonds with AI persuasive. That judgment is a review of the book, not a measurement of chatbot users’ outcomes.

What the research can show

A separate four-week randomized study by MIT Media Lab and OpenAI researchers followed 981 participants and analyzed more than 300,000 messages. It measured loneliness, real-world social interaction, emotional dependence and problematic chatbot use. Its revised abstract reports no significant effects from the assigned interaction modes or conversation types. Participants who chose to use the chatbot more showed worse outcomes, but that association does not establish that their greater use caused those outcomes.

The same study found that participants with greater trust and social attraction toward the chatbot also showed more emotional dependence and problematic use. That relationship points to differences among users and their experiences. It cannot, by itself, tell whether trust led to dependence, dependence led to trust, or other factors influenced both.

Other independently originated research points to a possible short-term benefit. In experiments reported in a 2024 paper, De Freitas and colleagues found that AI-companion interactions reduced measured loneliness, including in a one-week study. Relief over that period does not establish whether chatbot companionship improves or damages relationships over years. The studies also involve different participants, products and time spans, limiting how far their findings can be applied to Turkle’s broader argument.

Artificial Intimacy therefore arrives with a clear warning and a narrower body of evidence against which to assess it. Turkle’s interviews show how people describe emotionally meaningful exchanges with chatbots; the experimental studies measure selected outcomes over weeks or less. Whether those encounters reshape attachment, mourning or social development over a lifetime remains an open question.