This is a very favorable study design (the doctors wrote clean scenarios and recruited participants to play along). Real patients will mention irrelevant/distractor/somatic symptoms, and may use language related to conspiracy theories. Chatbots will tend to reinforce those. doi.org/10.1038/s415...

Products that lull regulators into complacency, assuming that tests that would be valid for assessing correctness of conventional devices and people are also valid for products based on fitting data without a mechanism. A system that seems trustworthy, then mistakes the carotid artery during surgery

Thread is absolutely correct. The "fooled you" dollars are not good and will only get worse as the MLM scam spreads. It won't displace established authors, but is a denial of service attack on new authors finding an audience, their new gem in a sea of slop as readers stick to known authors.

The main conceit of "AI" (namely chatbots and "agents") are the twin promises of control and productivity. The mechanism of control is that you can be replaced more effectively by a machine, or someone who wields the machine more adroitly than you.

I do want to shout-out my fellow sociologists, who have collectively created a discipline so woke that not a single one of our introductory textbooks can make it past Florida's censors. Great work everyone.

No joke: I got angry hate mail today for writing an obituary of a Black woman scientist—because the person felt she did didn’t deserve the recognition. Which just makes me want to share it again: www.nature.com/articles/d41...

Internal search launched in Dec and closed in Jan with no shared governance or faculty representation in the committee, despite the fact that the current chancellor's term doesn't end until Jun 2027. This isn't a good-faith search, but a shady ploy to shut the university community out of the process

Caught with fingerprints all over the Rule 11 Violation Machine and the Rule 11 Violation Obfuscator. 14 of 60 cited cases do not exist and other cases were misquoted, but this man thought he could talk his way out of sanctions in 97 pages of testimony under oath. Get his ass, Judge Failla.

The Trump administration is sending a loud and clear message that they will not stop schools from discriminating. This is a necessary step on the path to resegregating schools. Which has been their openly stated goal all along.

The asymmetry of bullshit: You can make an anti-vax conspiracy bot (with incidental factual statements), but you can't make an accurate-science bot using LLM methods (they don't represent facts and will reinforce false claims even if all training data were strictly factual in narrative voice).

The Wayback is now so load-bearing we should be protecting it with our actual lives

Also very true of universities: from the HUAC and McCarthy era to today, ordinary faculty, staff, and students are far more committed to the university charter than are upper administrators and regents.

1. I am tracking graphics and figures in academic journals made gen AI here: bit.ly/academic-slop 2. Keeping gen AI of the sort described here should be a completely solvable problem with run of the mill peer review and editorial oversight. 3. Change research assessment incentives.

@biblioracle.bsky.social has a great interview with @mattseybold.bsky.social on techno-feudalism, mass surveillance, Ed Tech, and what this means for academic freedom, and higher ED more generally. A chilling read. @aaup.org academicfreedomontheline.substack.com/p/the-techno...

Thank you, Victor. I spent a few weeks on this. As usual, it’s always a negotiation with length. So a few other bon mots: 1. The admin has issued DOJ guidance to challenge state laws that would regulate AI www.wired.com/story/ai-sup...

This is overstating the decision. Alsup only addressed training, not uses of the trained model. Using it for search is fair use, but it remains to be adjudicated whether distributing the model weights or output is fair use when those include near-verbatim reproduction of books.

EdTech, it doesn't work... How long until we can just accept the science on this? www.felienne.nl/2026-05/#edt...

As these teens describe, AI can diminish human relationships; devalue art; threaten the environment; lead to laziness; give unreliable results; pose privacy concerns; and be misused. So, please, stop with the narratives of inevitability and let's embrace a pedagogy and politics of refusal.

University administrators across the country are holding this "compact", drafted by one Epstein associate to give another Epstein associate leverage over their institutions, hoping that silence will protect them from public scrutiny. "No" is a complete sentence and faculty were correct to demand it.

The question should never be "is fraud profitable" or "is plagiarism good for my h-index". A call for dispassionate study fails to confront power structures that got us here, cedes the framing, and ignores the fact that decision-makers won't fund and won't listen to results of critical studies.

Faculty did not ask for this. The chancellor did this while claiming austerity and collecting a million dollars in salary and benefits. While TAMU is closing women's and gender studies programs out of pointed cruelty, the CSU is doing the same out of willful starving of resources and layoffs.

A lot of institutions overrepresented in the Epstein files should be thinking about whether this is the reputation they want. And if not, what they can do now to change culture and hire people of such character that won't be in the Epstein files of 2050. MeToo barely scratched the surface.

beyond words to express how much i hate this... looks like "AI summary" not the author's written abstract is displayed by default for all ACM digital library publications. i, for one, didn't ask for it and hate it to my core

Imagine if we relied on oil companies to publish evidence that CO2 emissions cause climate change. This statement against interest by Anthropic illustrates the epistemic vulnerability in which funding agencies and universities have uncritically accepted vibes-based claims. arxiv.org/pdf/2601.20245

“Most of the reports the charity receives involve wearable health technology like smartwatches and rings, but she said abusers are also using smart locks, heating technology, and even fertility trackers to manipulate and control victims.”

The administrators and this company they hired have conspired to commit academic and journalistic misconduct more flagrant than press-release plagiarism (@plagiarismtoday.com). If faculty governance fails to drag the administrators for this, their code of professional conduct will mean nothing.

With this flagrant violation of academic freedom, TAMU administrators are defying shared governance by unilaterally choosing not to be a university. This is a betrayal of students, faculty, and the public.

I was interviewed for two documentaries about "AI" that both premiered at Sundance this year: @ghostdoc2026.bsky.social ("Ghost in the Machine") and "The AI Doc: How I Became an Apocaloptimist". As a pair, these films are a study in contrasts. Another long 🧵>>

Cloudflare beclowns themselves publishing a blog on what purports to be a Matrix impl that "runs in production today" on Cloudflare workers, but it's just vibe-coded slop that doesn't do any such thing. University administrators want to give students the hubris to do this. tech.lgbt/@JadedBlueEy...

I just got done watching a truly horrible movie at Sundance ("The AI Doc: Or How I Became an Apocaloptimist"). It was introduced by the Sundance programming rep as having "a strong sense of informational integrity", but it's hard to imagine a characterization more off the mark. A long thread 🧵>>

Coercive tactics like these, which have been reported since at least summer 2025, are widespread enough that internal/self-reported data are not reliable measures of sentiment or impact. When internal dissent/critique is so chilled, external levers (regulation, audits) are more important than ever.

Last summer, as OpenAI sought to transition from nonprofit to for-profit, civil society groups pressed the company for accountability. Tyler Johnston, the founder of an AI watchdog, was among them. Then OpenAI showed up with a subpoena at his door. My latest for @moreperfectunion.bsky.social.

Some tools for thinking about AI in a time when the companies making them are trying to destroy teaching and learning itself... 1/n 🧵

Confessing to a denial of service attack: "We're flooding the zone." LLMs do not perform the metacognitive processes that are necessary to comply with 5 U.S.C. 3331. (h/t Judge F Kay Behm) Oath of office: "I will well and faithfully discharge the duties of the office on which I am about to enter"

The biggest risk in this moment is something sociologists call “symbolic compliance.” That’s when an institution that is violating civil rights gives the public just enough symbolic victories that accountability efforts lose steam before there has been any meaningful change.

K-12 is awash with products like IXL, and you're supposed to take it on faith that the mirage of personalization aids learning. AFT leadership and universities are lending credibility to the marketing/inevitability based on vibes and FOMO, mostly from administrators and without validated pedagogy.

🚨🚨🚨 "At least 63 times since 2007, data from some of the 28 human genomic repositories that the N.I.H. controls was improperly released to researchers, used for unapproved purposes or made vulnerable to theft..." (Gift Link) www.nytimes.com/2026/01/24/u...

All of this. Disrupt the “violence comes from emotions running too high” narrative whenever you can. Patriarchal violence doesn’t come from someone losing control. It comes from someone using a very specific set of tactics to seek more control than they already have.

Again, when people say trust in higher ed is down, I genuinely want to know whose trust hasn't been harmed. Why would anyone assume I feel great about my sector when it did things like this?

It's wild how strong the cognitive biases are. There is no excuse in 2026 for journalists to portray chatbot output as a product of a reasoning process (it's linguistic statistics, not ideas), much less to suggest that chatbots can perform the metocognitive processes to explain its own outputs.

Yes! ✊ Never forget who the plaintiffs are (AAUP and MESA) and are not (any universities). The moral conviction of university counsel and administration flows like water, indifferent to the university mission and governance necessary for higher ed to serve society. Support orgs that actually fight.

"However, my case reveals a fundamental weakness: these tools were not developed with academic standards of reliability and accountability in mind." Erm, no. We already knew that. It is a hyped black box built on plagiarised content + labour exploitation. The arrogance & lack of reflection 🙄

See how cryptogyny works. Emmy Noether was, as @olivia.science would say, a gigachad. My daughter's middle name is in her honor.

For context, this "base case" estimate (left) is somewhat above the top of the range from the 2024 LBNL report (middle). escholarship.org/uc/item/32d6... It is also well above McKinsey's estimate from that time (right). www.mckinsey.com/industries/p...

Last week, Nathan Evans, Assoc VC for Academic Affairs at CSU, was a panelist at the U of Colorado Board of Regents meeting, along with several other "AI" boosters busy destroying their institutions. It was a 75-min unchecked Gish gallop. Please sign in solidarity to counter this toxic info bubble.

Reminder that doomer and booster are two sides of the same coin, seeking investor money because the product isn't good for most legal/consensual uses and the illegal/nonconsensual (CSAM, plausible deniability for racial profiling and military strikes on civilians) aren't enough for the valuations.

I wrote about how concentration camp regimes evolve and why it's not too late to stop this, even though we are deep in it now.

There's so much wrong here, but the Alps are a battery that surpasses the combined utility-scale battery and pumped-storage hydro in North America, and European countries have been planning and building this for many decades. en.wikipedia.org/wiki/Pumped-... www.hydropower.org/publications...

“But what deeply troubles me now is that for all the steps we've taken toward integration, I've come to believe that we are integrating into a burning house."  Dr. King said this famous quote on March 27th, 1968, mere days before his murder. But the context around the quote is interesting...

These examples hammer home that the deployment of this tech is intended to be a Plausible Deniability Generator, something that creates pretext and paper trail to cover things that governments and corporations wanted to do to us anyway.