Sherry Turkle’s research associate used ChatGPT to simulate her late mother, drawing on Turkle’s memoir. The bot recalled family details, but its comforting answer about the afterlife sounded to Turkle like sentimentality and clichés, a striking glimpse of both AI’s reach and its limits.
Watch Desk analysis
What happened
Turkle, an MIT sociologist and clinical psychologist who has studied technology’s effects on people for five decades, describes the encounter in her new book, Artificial Intimacy: Who We Become When We Talk to Machines. A review in The New Republic recounts how her associate supplied the chatbot with her memoir, The Empathy Diaries, and prompted it to play her mother, Harriet.
The chatbot recalled details from their shared life. When Turkle asked whether her mother was safe in the afterlife, it described a warm, quiet morning by the sea. Turkle recognised that nobody was speaking to her from beyond the grave, yet kept asking personal questions. The encounter offered a simulation of a deeply intimate conversation, not contact with the person she had lost.
Read the review in The New Republic.
Why it matters
This is not a benchmark for chatbot capability or evidence that a model understands grief. It is a vivid example of how personal material and fluent replies can make an AI interaction feel intimate, even when the person using it knows what the system is.
That tension sits at the heart of Turkle’s wider argument about people turning to chatbots for companionship, therapy or emotional support. The technology can offer responsiveness on demand. It cannot make the relationship reciprocal, though the conversation may still feel meaningful to the person having it. That distinction is easy to state and rather harder to live with.
Our read
The most revealing detail is not that ChatGPT could retrieve facts from a memoir. It is that Turkle knew the exchange was artificial and still found herself drawn into it. That deserves more attention than either “the machine brought her mother back” or “it was only autocomplete”. Neither captures what the experience felt like to the person in front of the screen.
Read this as a personal account and a critic’s argument, not a clinical study. Its value is in making the emotional stakes concrete: what happens when a system can echo a person’s memories and supply a convincing reply, but cannot be that person?
What to watch
- How AI companion and therapy products describe their emotional role to users.
- Whether people treat these systems as a supplement to human support or a substitute for it.
- How designers handle personal information used to make chatbot interactions feel more familiar.
Discussion spark: If you knew a chatbot was simulating someone you had lost, would an emotionally convincing conversation be comforting, unsettling, or both?
Sources and evidence
- Source update (20 November 2026, 11:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.