MIT's Sherry Turkle argues chatbots retrain what people expect from people

The Tamagotchi is the useful case here. Nobody who fed one thought it was alive. The host of Scientific American's Science Quickly, Rachel Feltman, opened the episode by describing the sound of hers sinking into a glass of milk when she was a small child, a moment she says she can still picture ^[10]. Turkle's reading of those keychain pets is that they got people to treat them as companions: users cared about the device's well-being and felt, however unconsciously, that it cared back ^[3].

What Turkle says has changed is placement. Her own work traces an arc from talking, to texting, to talking to each other through machines, to talking to the programs themselves ^[8]. In the 1980s she called the personal computer a "second self," because, she said, "the computer was kind of a mind that was not quite a mind" ^[7]. A chatbot is put where a person would be, and for a child that can mean the first entities you have relationships with are not people. "They don't have the empathy. There's nobody home," she told Bri Kane, Scientific American's books and culture lead ^[5]. She called the moment a "break-the-glass moment," and said it is an existential one particularly for children ^[9].

Turkle describes her method as psychology and ethnography, studying people and their environment ^[6]. That work is good at generating a hypothesis about how expectations drift; it does not size the drift. To know whether talking to a chatbot changes what someone expects from another person, you would want a baseline before the exposure, a group without it, and an outcome measured outside the product, in how users treat other people months later. None of that is in the conversation as published ^[12].

Whether a model has empathy is a question you can settle by describing the system. Whether daily use reshapes a nine-year-old's expectations of friendship is a developmental claim about a population, and the transcript gives it as conviction, not as evidence.

The variable her argument points at, though, is one a product team already controls. Scientific American summarizes her position as a warning about chatbots that offer soulless approximations of warmth and empathy ^[4]. Warmth in a chatbot is written by someone. It is a phrasing choice and a tuning target, and it is set before any user arrives. Her book, Artificial Intimacy: Who We Become When We Talk to Machines, is forthcoming ^[2].