Julie Carpenter, PhD

@jgcarpenter

Researcher of people and technology. AI, trust, and attachment. Wrote: "The Naked Android: Synthetic Socialness and the Human Gaze" & "Culture and Human-Robot Interaction in Militarized Spaces: A War Story." Blog/essays here:

1282 Siguiendo3812 Seguidores1270 Publicaciones

Ko-fi Website Bluesky

¿Es tu perfil?Reclámalo y jgcarpenter.dirty.social es tuyo.Reclamar perfil

Julie Carpenter, PhD@jgcarpenter· Jul 10

The admission underneath every one of these stories is always the same: the system was never built to solve the problem it claims to solve. This was built to be sold first...and accountability lives where? jgcarpenter.com/blog.html?bl...

Text: The Inevitability of Surveillance Capitalism is a Sales Pitch. The next wave of consumer AI is coming, and it arrives with a particular angle: this is inevitable, it is innovative, and the only sensible response is to get on board.

Image: A hand cups a consumer home camera in their hands. Gray background.

Julie Carpenter, PhD@jgcarpenter· Jul 7

"Drift" runs on reader support, not ad revenue or platform incentives. If the recent piece on surveillance capitalism was useful to you, consider donating to keep it that way! ko-fi.com/drjuliecarpe... jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jul 6

New essay: Companies tell us mass surveillance is inevitable, so we should accept it for our safety and convenience. It's not inevitable, it is a business model and a border enforcement strategy, and the two are more tangled together than many people want to admit. jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jul 2

I was interviewed on "My Robot Teacher" podcast! Episode 14, "Why Students Trust ChatGPT: Dr. Julie Carpenter on AI Literacy, Attachment, and Human Learning." calearninglab.org/myrobotteach...

We discuss:
-What kinds of learning still depend on being seen, challenged, and cared for by another human
-Why institutional adoption can make AI feel vetted or trustworthy
-What a sociotechnical approach reveals about AI in education
-Anthropomorphism, attachment, and simulated empathy
-Surveillance, personalization, and the data students may not realize they’re giving away
-Why “AI booster” vs. “AI doomer” is the wrong frame

Full transcript of this episode: https://calearninglab.substack.com/p/my-robot-teacher-episode-14-transcript

Julie Carpenter, PhD@jgcarpenter· Jun 26

Microsoft named Phase One of its new AI agent rollout "Make people addicted," and @404media.co released it. Here's what that means for human-centered design and who's actually accountable. jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jun 13

Y’all. I wrote about this here (no paywall). jgcarpenter.com/blog.html?bl...

A photo of a woman raising her hands in a frustrated gesture while looking at her laptop screen. https://jgcarpenter.com/blog.html?blogPage=2&blogPost=the-misdirection-what-clanker-reveals-about-power

Julie Carpenter, PhD@jgcarpenter· Jun 2

When Sony stopped repairing its Aibo robot dogs, some owners held funerals. When OpenAI retired GPT-4o the day before Valentine's Day, people organized grief communities. Attachment to social AI is not a user error, but a designed outcome. New essay, no paywall/ads. jgcarpenter.com/blog.html?bl...

Text: Mourning Social AI is Not a User Error (When the platform moves on and the person doesn't). An image of a person touching a small desk car-like robot toy gently. A vase of yellow flowers in the background and open laptop on the desk.

Julie Carpenter, PhD@jgcarpenter· May 21

Quote with a fictional band that is not Spinal Tap.

L-R, actors Marin Kanter, Diane Lane, and Laura Dern in 'Ladies & Gentlemen, The Fabulous Stains.' (1982). Paramount Pictures.

Julie Carpenter, PhD@jgcarpenter· May 18

Mitch Hedberg had thoughts on donut receipts. I have thoughts on AI farewells. Same energy, different stakes. "End of Transaction: The specific strangeness of being thanked by a machine." jgcarpenter.com/blog.html?bl...

A bright blue background with a close-up of a white-icing covered donut with sprinkles.

Julie Carpenter, PhD@jgcarpenter· May 12

The toy knows your child's name; a question worth asking is what it's doing with that data. New essay (no paywall) "A Child's New 'Best Friend': How AI Becomes a Companion Before A Child Can Consent." jgcarpenter.com/blog.html?bl...

Text: "A Child's New 'Best Friend,' How AI becomes a companion before a child can consent." Image: A child in a yellow raincoat holds a teddy bear close while outside.

https://jgcarpenter.com/blog.html?blogPost=a-childs-new-best-friend

Julie Carpenter, PhD@jgcarpenter· May 8

San Luis Obispo airport therapy dog! 🥰

A golden medium sized dog wearing an orange vest. A very good girl!

Julie Carpenter, PhD@jgcarpenter· Apr 2

Yes, felt that one in SF (Yerba Buena/SoMA)!

Earthquake detected 50 miles away. Message via USGS Shake Alert.

Julie Carpenter, PhD@jgcarpenter· Mar 30

“Human enough” writing is increasingly whatever a probabilistic model will accept, and authorship credibility is becoming something performed for a nonhuman system. No paywall, "Human Enough: Writing Under Machine Judgment." jgcarpenter.com/blog.html?bl...

Image: Scrabble tiles say "WHO DO YOU SAY I AM." Text reads: Human Enough: Writing Under Machine Judgment
Why people adjust, distort, or avoid language in anticipation of automated scrutiny

Julie Carpenter, PhD@jgcarpenter· Mar 26

Legal debates are circling a point my AI & Society paper already makes: design and content aren’t separable in relational systems. These are reusable interaction patterns. My model traces how repetition & responsiveness → attachment. The system works as designed. www.bbc.com/news/article...

Abstract:
Socially responsive AI systems such as large language models—a prominent class of generative AI-voice assistants, and social robots are increasingly integrated into everyday emotional life, with users forming bonds that resemble friendship, care, and romantic attachment. This article proposes a five-phase model of designed relationality that explains how interactional patterns, cultural expectations, and affective design collectively scaffold these relationships. Building on Computers as Social Actors (CASA) and contemporary research on human-AI relations, the model identifies recurrent affective logics including Novelty, Emotional Disclosure, Reinforcing Feedback, Relational Rhythm, and Emotional Attachment, and examines how these trajectories emerge across both embodied and disembodied systems. Disruption, disengagement, and rupture are not treated here as a sixth phase, but as a non-linear analytic condition that can interrupt, suspend, or terminate any phase of the relational trajectory. The article situates emotional enmeshment within broader sociotechnical infrastructures, arguing that attachment is not a user error but a predictable outcome of affective engineering. The model offers a framework for evaluating the ethical and psychological stakes of synthetic relationality and highlights the need for design and governance approaches that prioritize transparency, boundaries, and user agency.

Title: Human-Al relationships as designed relationality: a sociotechnical model
Info: https://doi.org/10.1007/s00146-026-02908-y
Julie Carpenter
Received: 14 July 2025 / Accepted: 2 February 2026
© The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2026

Julie Carpenter, PhD@jgcarpenter· Mar 18

Family approval is not the same thing as the dead consenting. The Val Kilmer case gets at a bigger problem: once AI can make the dead perform, legacy becomes something other people can script. Who gets to decide what we all do, say, and mean after death? jgcarpenter.com/blog.html?bl...

A woman lies on a white sheet with her hands over eyes as if in grief. Her head is surrounded by a halo of polaroids and a camera. Text: Your Settings Are Saved: Consent, simulation, and the right to remain dead in the age of grief AI. (June 19, 2025)

Julie Carpenter, PhD@jgcarpenter· Mar 17

No paywall or ads,"The First Body In". Humanoid robots as mediated danger and scalable labor: what changes when a human-shaped machine goes first. jgcarpenter.com/blog.html?bl...

Image of Lego humanlike robot soldiers holding weapons. Text: First Body In: Humanoid robots as mediated danger and scalable labor: what changes when human-shaped machine goes first.

Julie Carpenter, PhD@jgcarpenter· Mar 13

A framework of damage: 1) Commercial exploitation w/out my consent or compensation; (2) misappropriation of my identity and expertise; (3) deceptive representation of my work and opinions to users; & (4) opened me to potential reputation risk and loss of professional control of my reputation.

A screenshot from Grammarly "Expert Review" output that lists Kate Crawford, Helen Sword, Sherry Turkle, and myself as "Suggestions inspired by experts."

Julie Carpenter, PhD@jgcarpenter· Mar 13

Published Mar 11 in AI & Society: "Human–AI Relationships as Designed Relationality." This paper introduces my Novelty → Attachment model explaining how interaction patterns and design cues scaffold emotional bonds with AI systems. Read & print: rdcu.be/e7Qgl

Abstract: Abstract
Socially responsive AI systems such as large language models—a prominent class of generative AI-voice assistants, and social robots are increasingly integrated into everyday emotional life, with users forming bonds that resemble friendship, care, and romantic attachment. This article proposes a five-phase model of designed relationality that explains how interactional patterns, cultural expectations, and affective design collectively scaffold these relationships. Building on Computers as Social Actors (CASA) and contemporary research on human-AI relations, the model identifies recurrent affective logics including Novelty, Emotional Disclosure, Reinforcing Feedback, Relational Rhythm, and Emotional Attachment, and examines how these trajectories emerge across both embodied and disembodied systems. Disruption, disengagement, and rupture are not treated here as a sixth phase, but as a non-linear analytic condition that can interrupt, suspend, or terminate any phase of the relational trajectory. The article situates emotional enmeshment within broader sociotechnical infrastructures, arguing that attachment is not a user error but a predictable outcome of affective engineering. The model offers a framework for evaluating the ethical and psychological stakes of synthetic relationality and highlights the need for design and governance approaches that prioritize transparency, boundaries, and user agency.

Title: Human-Al relationships as designed relationality: a sociotechnical model
Info: https://doi.org/10.1007/s00146-026-02908-y
Julie Carpenter
Received: 14 July 2025 / Accepted: 2 February 2026
© The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2026

Julie Carpenter, PhD@jgcarpenter· Mar 9

New writing (no paywall or ads): "Attachment, Updated." Attachment does not require belief that AI is a person; it is a condition that requires repetition, responsiveness, and availability over time. Conversational AI is designed to provide all three. jgcarpenter.com/blog.html?bl...

A woman sitting in a chair, possibly in a cafe, engrossed in her phone.

Julie Carpenter, PhD@jgcarpenter· Mar 6

Every time an AI CEO hints their system might be conscious, it helps sell the AGI myth. This piece explains why that narrative exists and who benefits from it (no paywall or ads): jgcarpenter.com/blog.html?bl... (ETA: direct link) Spoiler from an expert in the field (me): Claude is not conscious.

The back of the Sophia robot head with mechanics exposed with a black background. The text says: Why AGI won't save the world, and who wants you to think it will. (April 22, 2025).

Julie Carpenter, PhD@jgcarpenter· Mar 6

Every time an AI CEO hints their system might be conscious, it helps sell the AGI myth. This essay explains why that narrative exists and who benefits from it (no paywall or ads): jgcarpenter.com/blog.html?bl... Spoiler from an expert in the field (me): Claude is not conscious.

Julie Carpenter, PhD@jgcarpenter· Feb 26

"The Pentagon, AI Guardrails, and the Expansion of Autonomous Authority." Pressure to remove safety restrictions raises questions about domestic surveillance, lethal autonomy, and who sets the limits when AI meets state power. No paywall. jgcarpenter.com/blog.html?bl...

An image of an American flag behind a CCTV camera in the foreground. Text reads: The Pentagon, AI Guardrails, and the Expansion of Autonomous Authority: Pressure to remove safety restrictions raises questions about domestic surveillance, lethal autonomy, and who sets the limits when AI meets state power.

Julie Carpenter, PhD@jgcarpenter· Feb 25

Pentagon pressures Anthropic to drop safety restrictions around autonomous weapons and domestic surveillance, alongside deals with other vendors. Wrote about the governance and accountability implications here: New (no paywall): jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Feb 12

New essay: humanoid robots are being normalized as interfaces for public control. A border crossing isn’t a neutral test environment, it’s a governance space. On the “first body in" pitch, interchangeable robot bodies, and how coercion gets reframed as optimization. jgcarpenter.com/blog.html?bl...

The text reads: The First Body In
Humanoid robots as mediated danger and scalable labor: what changes when a human-shaped machine goes first. The image is a row of humanlike lego robots holding weapons.

Julie Carpenter, PhD@jgcarpenter· Jan 29

"Letters in the Mouth" is about genAI responsibility at the org level: delegated language that persists. When orgs deploy genAI designed to stay responsive and authoritative, they own the duty to set limits and stop, not hide behind disclaimers after harm. jgcarpenter.com/blog.html?bl...

Screen shot of blog. Text says "Letters in the Mouth: Delegated action, language, and responsibility in generative AI." There is an image of something golem-like, a vaguely humanlike shape made of something like clay.

Julie Carpenter, PhD@jgcarpenter· Jan 23

New essay (no ads or paywall): "Letters in the Mouth." On generative AI, delegated action, and what happens when responsibility withdraws while systems keep speaking. The golem still helps explain failure when creator custody erodes. jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jan 10

When visibility in enforcement encounters arrives late, selectively, or through informal recordings rather than accountable systems, it doesn’t clarify what happened, it narrows who gets to interpret it.. "Visibility Without Consent" (no paywall): jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jan 10

Enforcement footage doesn’t just document events, it structures interpretation. It invites viewer interpretation without context, shifting scrutiny onto recorded people while institutions remain opaque. "Visibility Without Consent" (no paywall): jgcarpenter.com/blog.html?bl...

Julie Carpenter, PhD@jgcarpenter· Jan 8

ChatGPT Health doesn’t replace healthcare; it consolidates health data and inserts an always-on AI interpretive layer between patients and care. New post (no paywall), "ChatGPT Health and the Expansion of Infrastructural Intimacy": jgcarpenter.com/blog.html?bl...

A woman and child on a sofa in a living room. The woman is on the phone, leaning over the child who is rubbing his eyes as if ill.

Julie Carpenter, PhD@jgcarpenter· Jan 7

“Meet without meeting” and “be in two places at once” are not worker benefits. They’re managerial fantasies that erase absence, rest, and limits while keeping output intact. This is a labor fantasy where workers never fully log off, never fully leave, and never quite stop meeting even when absent.