Iris van Rooij 💭

More than 1000 French academics from universities countrywide have signed a manifesto declaring their conscientious objection to the deployment of degenerative AI in their institutions. #ResistAI ✊🏽

Figure 3 from "The Tractable Cognition Thesis (van Rooij, 2008): "According to the Tractable Cognition thesis, the set of functions describing possible cognitive capacities is a subset of the set of tractable functions. The dotted line indicates that the set of tractable functions remains to be formalized."

🧵Thread overview of ALL my publications related to "Cognition and Intractability" in chronological order: 🌟

See this thread on the bayesian point bsky.app/profile/iris...

And to the catchy phrase about correlation

bsky.app/profile/oliv...

7/n

And ultimately relates to why models such as these are insufficient to say the least as models of cognition

bsky.app/profile/oliv...

6/n

This also relates to the problem with guardrails 5/n

bsky.app/profile/oliv...

And this one 4/n

bsky.app/profile/iris...

Especially here bsky.app/profile/iris...

3/n

If you're interested in this see: @irisvanrooij.bsky.social's other work too...

2/n

bsky.app/profile/iris...

Simply? No. Neither. Abduction for example, isn't based on statistics. But statistics itself — given you've granted cognition is (or could be) computational by asking if it's statistics — isn't far up the Chomsky hierarchy. So it's already a dead end.

Useful talk: bsky.app/profile/iris...

1/n

🔥

In the summer, together with @olivia.science @irisvanrooij.bsky.social and other scientists and scholaes, we started an open letter calling Dutch institutions of higher education to Stop the Uncritical Adoption of AI Technologies in Academia:
openletter.earth/open-letter-...

Yes, this is also why I don't think we should do all the nonsense I see some of my colleagues doing, which is basically abusive of their junior staff and students in the name of avoiding any AI cheating. More thoughts here if useful: bsky.app/profile/oliv...

Or better yet: make GenAI about an assault on civics. Because if you are selling a product that necessarily attempts to con one person into believing they are engaging with another person when they are not, you’re not just ruining education. You’re dismantling society’s foundations in social trust.

So the next time someone tries to draw you into a conversation about that piece they saw on 60 Minutes or read in the NYT about how educators are coping with/ combatting/whatever students using GenAI, don’t accept that framing. Make it about the industry that is impoverishing their education.

It's a difficult needle to thread because a bully (AI industry) has no moral compass nor will stop bullying if we acquiesce. But that doesn't also mean we don't fight back. We try to make this clear from the abstract here too: "AI cannot be allowed to drive change."

doi.org/10.5281/zeno...

1/n

Dear @olivia.science : Thank you for the work you & others do.

It was refreshing to read someone clearly delineate larger issues in the study of human development that I’d only seen play out in a clinical context with individuals.

I didn’t have the words to describe what I saw. Now I do.

we must be critical when we see statements such as: “It’s worth noting that AI can pass the clinical social worker exam — even without seeing the questions. But that shows the weaknesses of exams more than the strength of AI” (Caldwell, 2024, n.p.). An AI product passing an exam means nothing about the “weak- nesses” of exams in the abstract. Anybody can ‘win’ a marathon if driven by car. Does that make it a weakness of the marathon race as an endurance event? Or does it reflect a deeper category error on our behalf, like with the misapplication of statistical tests, that an assumption has been violated? The race as the exam assumes an unaided human performs it. Any digital computer can sur- pass a human at many feats, e.g. calculations per second, but that grants no humanity to the machine. 8 Do not Embrac

The point isn't assessment standards...

doi.org/10.31234/osf...

Framing GenAI as a battle between teachers and students is a red herring. Students and educators are on the same side. The real opposition are the data extraction firms and brokerages and their allies among the managerial class.

Any they want to further erode relationships between individuals so that, eventually, we simply decide it’s just easier to cede decision-making to models, which they can sell us.

A disrespect for me as an educator means that a basic piece of sociality and trust has been broken and so I don’t owe anyone generous feedback. But I don’t need to be punitive either. That’s what those who are selling these products want. They want to exhaust us into resignation.

It is a waste of educators’ labor to worry about policing students’ work or jury rigging AI-proof assignments.There are desultory ways of going about responding to such papers. But that doesn’t mean acquiescence. That means working towards and demanding structural changes at our institutions.

I’ve been thinking about the intellectual and affective labor involved in reading AI Generated student papers. I’m forecasting a little of the proposal that @cnygren.bsky.social and I advance in a piece that’ll be out soon but here’s where I’ve personally landed (o speak for myself not for us both):

This is important, without any doubt: the link between slowness - timelessness - and healthy scientific practice.

The challenge is people find deep thinking uncomfortable. AI doesn't just replace thinking, it soothes that discomfort. It's less a tool and more a cognitive sedative. A comfort blanket for the brain. Feels great to externalise all that thinking. Too great! That’ll be a hard habit to kick for many.

On a transmis ce manifeste dans la liste de diffusion générale de mon laboratoire (le LIS) et c’était une signature instantanée de ma part. Bravo. 👏

❤️‍🔥

👏🏼

Brilliant

D’ailleurs, au delà du manifeste en soi, n’oubliez pas de lire l‘excellente annexe « “Oui, mais l’IAg….” Réponses à quelques arguments courants en faveur de l’intelligence artificielle générative ». doi.org/10.58079/158mj