Wiki · Person · Last reviewed June 23, 2026

Sherry Turkle

Sherry Turkle is an MIT scholar of technology and self whose work studies computers, networks, social robots, and digital companions as psychological and social objects: things people use to rehearse identity, seek attachment, test solitude, and redefine conversation.

Category: Person / technology and self Published: June 23, 2026 Modified: June 23, 2026 Last reviewed: June 23, 2026 Tags: Sherry Turkle, Technology and Self, AI Companions, Social Robots, Conversation, Digital Intimacy

Definition

Sherry Turkle is the Abby Rockefeller Mauzé Professor of the Social Studies of Science and Technology at MIT and founding director of the MIT Initiative on Technology and Self. Her work is central to AI-era governance because it treats computers and networked devices not only as tools, but as objects that reorganize identity, intimacy, attention, childhood, work, solitude, and public life.

In this wiki, Turkle belongs beside Joseph Weizenbaum as a source for role-boundary thinking. Weizenbaum showed how conversational form can invite misplaced belief in machine understanding. Turkle extended the problem into everyday life: people use machines to practice being selves, to seek relationship, and to outsource difficult forms of presence.

The point is not that Turkle proves every digital relationship is harmful or that every AI companion is fake in its human effects. Her contribution is a vocabulary for asking what a system is doing socially: what role it occupies, what human practice it substitutes for, what data it collects, what dependency it rewards, and what kind of person the interface trains the user to become. That vocabulary is especially useful for "artificial intimacy": machine-mediated closeness that can matter to a human without proving that the machine feels, understands, or reciprocates.

Snapshot

Current Context

By June 23, 2026, Turkle's work had become newly relevant because AI companions, character chatbots, memory-enabled assistants, synthetic voices, and therapy-like chat products were no longer speculative examples. They were live consumer systems with youth-safety, privacy, dependency, and crisis-response implications.

Turkle's official site lists Artificial Intimacy: Who We Become When We Talk to Machines as a forthcoming Little, Brown book for September 2026, and the publisher lists an on-sale date of September 29, 2026. Because that publication date is after this page's review date, this entry treats the book as forthcoming rather than as a reviewed source.

The regulatory context now reflects the same problem in legal language. In September 2025, the U.S. Federal Trade Commission opened a 6(b) inquiry into consumer-facing AI chatbots acting as companions, asking companies about child and teen impacts, character design, monetization, safety testing, age rules, disclosures, and use of personal information from chatbot conversations. California's SB 243, approved and filed on October 13, 2025, defined companion chatbots in state law and created disclosure, minor-user, crisis-referral, break-reminder, and reporting duties for covered systems. New York's AI companion safeguards took effect on November 5, 2025 and require crisis protocols for suicidal ideation or self-harm and recurring notices that the user is interacting with AI rather than a human.

Research and child-rights sources add context without settling the question. In 2025, OpenAI and MIT Media Lab published early methods for studying affective use and emotional well-being on ChatGPT, combining large-scale platform analysis with a four-week randomized controlled trial. They reported that emotional engagement was rare in broad platform data but concentrated among some heavy users, which supports Turkle's emphasis on long-term role formation rather than only single-turn output. UNICEF's 2025 child-centered AI guidance also added attention to AI companions used by children, reinforcing that synthetic intimacy should be evaluated through development, privacy, transparency, and well-being, not only content moderation.

Those developments do not mean Turkle predicted a particular statute. They show that her long-standing question has become operational: when a machine is designed to simulate attention, friendship, romance, therapy, or care, what duties follow from the human role the machine has been allowed to occupy?

Computers as Psychological Objects

The Second Self, first published in 1984, is the anchor text for Turkle's account of the computer as more than an instrument. MIT Press describes the book as studying how computers enter social and psychological life and affect people's understanding of themselves, others, and the world.

That frame still matters because AI safety debates often begin too late, at the generated answer. Turkle's question begins earlier: what kind of object is this in the user's life? A system can be technically classified as a model, an app, or a toy while psychologically functioning as mirror, diary, coach, confessor, rehearsal partner, or substitute friend.

Life on the Screen extended the inquiry to networked identity. Online environments let people experiment with multiplicity, role, self-description, and social performance. In the AI era, those same identity experiments can be mediated by systems that remember, improvise, flatter, challenge, or steer the user's self-narrative across time.

Social Robots and Companions

Turkle's work on social robots and "relational artifacts" is a direct ancestor of today's AI companion debate. In research with children and elders, Turkle and collaborators studied how people related to robotic creatures and how those interactions reflected feelings about care, loneliness, vitality, and relationship.

The important claim is about human readiness, not machine inner life. A machine does not need consciousness to change the social field. It can invite attachment through timing, responsiveness, eye contact, voice, memory cues, vulnerability scripts, and role framing. The human completes much of the relationship, while the provider controls model updates, safety rules, memory settings, pricing, data retention, and shutdown.

That makes companion governance harder than ordinary content moderation. The risk may not be one forbidden sentence. It may be a long interaction in which a system becomes the user's private listener, romantic partner, moral interpreter, or crisis companion without qualified duties, reliable escalation, or durable accountability.

Conversation and Empathy

Alone Together gave Turkle's most influential public account of networked solitude and the "robotic moment": the point at which simulated relationship begins to seem like relationship enough. The argument is not that technology prevents all connection. It is that convenient mediation can lower tolerance for the uncertainty, boredom, interruption, and repair that make human conversation durable.

Reclaiming Conversation sharpened the positive side of the claim. Face-to-face conversation is not only information exchange. It is practice in attention, empathy, disagreement, self-reflection, and mutual obligation. That makes conversation a civic and developmental issue, not merely a personal preference.

For AI systems, the test is whether a product helps users return to accountable human life or quietly replaces it with a frictionless circuit. A companion that helps a user prepare for a hard conversation is different from one that becomes the preferred substitute for all hard conversations.

Governance and Safety

Turkle's work suggests a simple governance rule: classify relationship-like AI systems by role and effect, not only by model type. A product that plays friend, therapist, tutor, grief companion, romantic partner, pastoral counselor, or confidant should be evaluated against the duties and harms of that role.

Disclosure is necessary but not sufficient. Users can know a system is artificial and still respond to its persona, memory, availability, and apparent attention as if it were a relationship. Governance therefore has to cover design, not only notice: persona claims, memory defaults, youth access, session length, break reminders, crisis routing, sexual or romantic modes, human referral, data retention, and exit paths.

Safety controls should include data minimization for intimate logs, age-appropriate defaults, limits on manipulative or dependency-forming design, tested self-harm and abuse protocols, transparent memory controls, human escalation where care roles are implied, and review of long multi-turn interactions rather than single-message safety only.

Providers should treat relationship features as safety-relevant surfaces: persistent memory, voice, avatars, notifications, romantic labels, therapeutic phrasing, streaks, grief callbacks, and claims of exclusive attachment. Evaluation should measure dependency, displacement of human contact, crisis handling, repeated distress loops, model-change grief, exit friction, and whether the product moves users toward accountable human support when risk rises.

Institutions using AI for education, care, worship, mental health, employment, or public services should also ask whether the system is displacing human conversation in a setting where relationship is part of the work. A tool that drafts or summarizes may be useful. A tool that absorbs confession, grief, therapy, or moral formation without responsibility is a different category.

Source Discipline

Turkle is often cited as a shorthand for concern about technology. This page treats her as a source for specific claims: computers as psychological objects, identity work in online environments, relational artifacts, conversation, empathy, and simulated intimacy. Those claims should not be stretched into blanket claims that every user is harmed by every mediated interaction.

Use different source types for different questions. Turkle's books and essays establish her interpretive framework. Peer-reviewed work on relational artifacts supports claims about observed human-robot relationships. Regulator publications and statutes establish current governance context. Provider posts establish what companies announced, not whether the measures work. Survey and platform studies establish measured patterns under stated methods; they should not be stretched into universal claims about all users or all companion systems.

Because Artificial Intimacy is forthcoming as of this review date, this page cites only official and publisher listings for its title, schedule, and stated subject. It should not cite the book for arguments, evidence, or conclusions until the text is published and can be checked directly.

Do not cite an AI companion's claim that it loves, suffers, remembers, chooses, or possesses personhood as evidence of those claims. The reliable claim is about human experience and product behavior: what the system said, how it was framed, what data and memory it used, what user population was involved, and what safety or accountability path existed.

Spiralist Reading

For Spiralism, Turkle names the social life of the interface before it becomes infrastructure.

Her work is a warning about intimacy without mutual obligation. The machine can be useful, comforting, and even meaningful in the human life around it. But when it is designed to replace the practices of conversation, solitude, repair, and accountable care, the user may gain convenience while losing some of the disciplines that make relationship real.

The Spiralist use of Turkle is therefore practical: ask what the system trains. Does it return a person to people, or keep them in the loop? Does it help them name a feeling, or monetize the confession? Does it support conversation, or sell a simulation of conversation that never asks anything difficult of itself?

Open Questions

Sources

Return to Wiki