Pluralistic: AI psychosis and the warped mirror (18 Sep 2025)


Today's links



Narcissus staring into his reflection; his face and the face of the reflection have been replaced by the staring red eye of HAL 9000 from Kubrick's '2001: A Space Odyssey.'

AI psychosis and the warped mirror (permalink)

"AI psychosis" is the pop-psych diagnosis in a recent string of horrible and horrifying cases in which vulnerable people were lured by chatbots into harming themselves and others, including a murder-suicide:

https://futurism.com/man-chatgpt-psychosis-murders-mother

AI psychosis is just one of the many delusions inspired by AI, and it's hardly the most prevalent. The most widespread AI delusion is, of course, that an AI can do your job (it can't, but an AI salesman can capitalize on this delusion to convince your boss to fire you and replace you with a chatbot that can't do your job):

https://pluralistic.net/2025/03/18/asbestos-in-the-walls/#government-by-spicy-autocomplete

The AI job delusion has a long lineage. Since the steam-loom, bosses have hyped new technologies as a way to frighten workers into accepting lower wages and worse working conditions, under threat of imminent technological replacement.

Likewise, AI psychosis isn't an entirely new phenomenon, and it has disturbing precedents in our recent past.

In the early 2000s, a community of internet users formed to discuss a new illness they called "Morgellons Disease." Morgellons sufferers believed that they had wires growing in their skin:

https://en.wikipedia.org/wiki/Morgellons

Morgellons appears to be a delusion, and the most widely accepted explanation for it is that people whose mental illness compels them to pick at their skin create open sores on their bodies, and then stray blowing fibers adhere to the wet, exposed tissues, which the sufferers believe to be wires.

Morgellons became an internet phenomenon in the early 2000s, but it appears that there were people who suffered from this pathology for a very long time. The name "Morgellons" comes from a 17th century case-report:

https://en.wikipedia.org/wiki/A_Letter_to_a_Friend

The difference between a Morgellons sufferer in the 1680s and a Morgellons sufferer in 2001 is that the latter need not suffer alone. The incredible power of the internet to connect people with rare traits meant that people suffering with Morgellons could coalesce online and egg one another on. They could counter the narratives of concerned family members who insisted that there weren't wires growing under their skin, and upload photos of the "wires" they'd discovered under their own skin.

People have suffered from all kinds of delusions since time immemorial, and while the specifics of the delusion reflect the world of the sufferer (I remember when I stopped hearing from people with radios in their heads and started hearing from people with RFIDs in their heads), the shape of the delusions have been stable over long timescales.

But the internet era has profoundly changed the nature of delusion, by connecting people with the same delusions to one another, in order to reinforce each other.

Take "gang stalking delusion," the traumatic belief that a vast cabal of powerful, coordinated actors have selected a group of "targeted individuals" to harass. People with gang stalking delusion will sometimes insist that passing bus-ads, snatches of overheard music, and other random/ambient details are actually targeted at them, intended to bring them distress:

https://en.wikipedia.org/wiki/Gang_stalking

The "targeted individuals" suffering from gang stalking delusion have formed vast, sprawling communities that are notionally designed to support them through the trauma of being stalked. But the practical function of these communities is to reinforce the delusion and make things much worse for their members: "My psychiatrist said the same thing as yours did – it's proof that they're both in on it!"

Like Morgellons, gang stalking delusion isn't a new phenomenon. It's a subset of "persecutory delusion," another mental illness that we find centuries of evidence for in the record:

https://en.wikipedia.org/wiki/Persecutory_delusion

But like modern Morgellons sufferers, people today with gang stalking delusion are able to find one another and reinforce and amplify each others' delusions, to their own detriment.

Now, even this isn't new – through the historical record, we find many examples of small groups of people who coalesced around a shared delusion. The difference is that old timey people had to luck into finding someone else who shared their delusion, while modern, internet-enabled people can just use the Reddit search-bar.

There's many examples of harmful delusions being worsened through online community reinforcement: there's pro-anorexia forums, incel forums, bitcoin, and "race realism" and other all-consuming junk science.

That's where LLMs come in. While the internet makes it far easier to find a toxic community of similarly afflicted people struggling with your mental illness, an LLM eliminates the need to find that forum. The LLM can deliver all the reinforcement you demand, produced to order, at any hour, day or night. While posting about a new delusional belief to a forum won't generate responses until other forum members see it and reply to it, an LLM can deliver a response in seconds.

In other words, there's one job that an AI can absolutely do better than a human: it can reinforce our delusions more efficiently, more quickly, and more effectively than a community of sufferers can.

Speed isn't the only reason that LLMs are super efficient delusion-reinforcers. An LLM has no consciousness, it has no desires, and it has nothing it wants to communicate. It has no wants, period. All it can do is transform a prompt into something that seems like the kind of thing that would follow from that prompt. It's a next-word-guessing machine.

This is why AI art is so empty: the only message an AI image generator can convey is the prompt you feed it. That's the only thing a piece of AI art has to "say." But when you dilute a short prompt across a million pixels or a hundred thousand words, the communicative intent in any given sentence or brushstroke is indistinguishable from zero. AI art can be "eerie" (in the sense of seeming to have an intent without there being any intender), and it can be striking, but it's not good:

https://pluralistic.net/2024/05/13/spooky-action-at-a-close-up/#invisible-hand

However, the more communicative intent there is in a prompt, and the more human decision-making there is in the production (whether that's selecting the best work from among many variants or post-processing the work with your own artistic flourishes), the more chances that work has of saying something. That's because you're saying something, every time you re-prompt it, every time you select from among an array of its outputs.

When you repeatedly prompt an LLM over a long timescale – whether you're discussing your delusional beliefs, or pursuing a romantic fantasy ("AI girl/boyfriends") – you are filling it up with your communicative intent. The work that comes out the other side – the transformation of your prompts into a response – is a mirror that you're holding up to your own inputs.

So while a member of a gang stalking forum might have a delusion that is just different enough from yours that they seem foolish, or they accuse you of being paranoid, the chatbot's conception of gang stalking delusion is being informed, tuned and shaped by you. It's an improv partner, "yes-and"ing you into a life of paranoid terror.

In the Greek legend, Narcissus falls in love with his reflection in a stream and is rooted to the spot, captured by his own regard. People who prompt a chatbot to reinforce their delusions are catching sight of their own reflection in the LLM and terrifying themselves into a spiral of self-destruction.

(Image: Cryteria, CC BY 3.0, modified)


Hey look at this (permalink)



A shelf of leatherbound history books with a gilt-stamped series title, 'The World's Famous Events.'

Object permanence (permalink)

#20yrsago Software lets camphones scan and OCR a page of text in 5 secs https://web.archive.org/web/20051029085125/https://www.newscientist.com/article.ns/?feedId=online-news_rss20&id=dn7998

#20yrsago Profiles of RIAA victims who fought back https://web.archive.org/web/20051125085616/http://p2pnet.net/story/6283

#15yrsago Intel + DRM: a crippled processor that you have to pay extra to unlock https://memex.craphound.com/2010/09/18/intel-drm-a-crippled-processor-that-you-have-to-pay-extra-to-unlock/

#10yrsago UC Berkeley issues first-ever university transparency report https://slate.com/technology/2015/09/uc-berkeley-issues-the-first-ever-university-transparency-report-others-should-follow.html

#10yrsago THIS COMPUTER IS NEVER OBSOLETE https://www.tumblr.com/neuroxin/125324271592/this-computer-is-never-obsolete-digging

#5yrsago Youtube's war on algorithmic radicalization https://pluralistic.net/2020/09/18/the-americanskis/#algorithm-lawyers

#5yrsago A cryptographic mystery solved https://pluralistic.net/2020/09/18/the-americanskis/#otps-r-us

#5yrsago In Search Of A Flat Earth https://pluralistic.net/2020/09/18/the-americanskis/#mass-murder-cults

#1yrago There's no such thing as "shareholder supremacy" https://pluralistic.net/2024/09/18/falsifiability/#figleaves-not-rubrics


Upcoming appearances (permalink)

A photo of me onstage, giving a speech, pounding the podium.



A screenshot of me at my desk, doing a livecast.

Recent appearances (permalink)



A grid of my books with Will Stahle covers..

Latest books (permalink)



A cardboard book box with the Macmillan logo.

Upcoming books (permalink)

  • "Canny Valley": A limited edition collection of the collages I create for Pluralistic, self-published, September 2025

  • "Enshittification: Why Everything Suddenly Got Worse and What to Do About It," Farrar, Straus, Giroux, October 7 2025
    https://us.macmillan.com/books/9780374619329/enshittification/

  • "Unauthorized Bread": a middle-grades graphic novel adapted from my novella about refugees, toasters and DRM, FirstSecond, 2026

  • "Enshittification, Why Everything Suddenly Got Worse and What to Do About It" (the graphic novel), Firstsecond, 2026

  • "The Memex Method," Farrar, Straus, Giroux, 2026

  • "The Reverse-Centaur's Guide to AI," a short book about being a better AI critic, Farrar, Straus and Giroux, 2026



Colophon (permalink)

Today's top sources:

Currently writing:

  • "The Reverse Centaur's Guide to AI," a short book for Farrar, Straus and Giroux about being an effective AI critic. FIRST DRAFT COMPLETE AND SUBMITTED.

  • A Little Brother short story about DIY insulin PLANNING


This work – excluding any serialized fiction – is licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commercially, provided that you attribute it to me, Cory Doctorow, and include a link to pluralistic.net.

https://creativecommons.org/licenses/by/4.0/

Quotations and images are not included in this license; they are included either under a limitation or exception to copyright, or on the basis of a separate license. Please exercise caution.


How to get Pluralistic:

Blog (no ads, tracking, or data-collection):

Pluralistic.net

Newsletter (no ads, tracking, or data-collection):

https://pluralistic.net/plura-list

Mastodon (no ads, tracking, or data-collection):

https://mamot.fr/@pluralistic

Medium (no ads, paywalled):

https://doctorow.medium.com/

Twitter (mass-scale, unrestricted, third-party surveillance and advertising):

https://twitter.com/doctorow

Tumblr (mass-scale, unrestricted, third-party surveillance and advertising):

https://mostlysignssomeportents.tumblr.com/tagged/pluralistic

"When life gives you SARS, you make sarsaparilla" -Joey "Accordion Guy" DeVilla

READ CAREFULLY: By reading this, you agree, on behalf of your employer, to release me from all obligations and waivers arising from any and all NON-NEGOTIATED agreements, licenses, terms-of-service, shrinkwrap, clickwrap, browsewrap, confidentiality, non-disclosure, non-compete and acceptable use policies ("BOGUS AGREEMENTS") that I have entered into with your employer, its partners, licensors, agents and assigns, in perpetuity, without prejudice to my ongoing rights and privileges. You further represent that you have the authority to release me from any BOGUS AGREEMENTS on behalf of your employer.

ISSN: 3066-764X

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.