“After just “10 minutes of AI-assisted problem-solving, people who lost access to the AI performed worse and gave up more frequently than those who never used it”.
An MIT study showed that test groups who used ChatGPT had the lowest brain activity and “consistently underperformed at neural, linguistic, and behavioural levels”. We undergo not just cognitive offloading, but cognitive surrender: we forget what was possible.”
The youth who are using ai slop or have parents using ai slop are all losing hope. They aren't being allowed to actually live.
@RosannaSibora
As Euclid put it, there's 'no royal road to Geometry'...
But now there's a slip-and-slide in the opposite direction.
@RosannaSibora@fosstodon.org
No surprise.
Aug 8, 2026, 9:17 PM@RosannaSibora I just don't understand how this is the gazillionth example of something that's forced into the world, we all know it's bad for us, our leaders know it too, and yet...
Honestly, wtf are we doing?
The "leaders" are doddering dotards.
@thisislieven @RosannaSibora The money and potential for profit without consequence was too great a lure... of course it isn't really "we", it is things that happen to us and are left to deal with.
Aug 9, 2026, 9:57 AM@nini @RosannaSibora Well, sort of. At the end of the day in democracies, even when they don't function optimally, we still choose our leaders and somehow most of us keep picking people who won't act in our best interest.
There's of course a much larger story here in degraded public knowledge, education, normalcy bias, decreased standards of living that make us pick other priorities and whatever else.
But the people in general are not without blame. And if a product wasn't adopted voluntarily by the public at large as well things would look different. It might not fully disappear but it would be less central in our lives. Much as we may hate AI, millions of people use it on the daily.
@thisislieven @RosannaSibora Eh, it's a confluence of bad actors making sure electable candidates are still very much in the pockets of those not acting in anyone's best interests but their own and a mind-sapping technology people are forced into using by employers that appeases their need to work less but is a net negative on every scale.
You are right, people do bear some responsibility but they're also encouraged to do things against their interests all the time because the consequences are worse than accepting the clear badness. Some day people will hopefully fight back en masse but until then they're still being offered a shit sandwich and told to be happy that it's not worse.
Aug 9, 2026, 11:21 AM@nini @RosannaSibora I don't know where you are in the world as there's a big difference in how much big money is able to buy candidates from country to country.
In the US the situation is really bad (though there have been some promising recent primaries and elections - not yet enough to make a dent though).
In the EU (my home) and wider Europe there's some of this as well but nowhere near the level as there is in the US. Most of Europe also has a multi-party system making it more difficult for vested interests to capture politicians.
Elsewhere in the world I find difficult to judge.
@thisislieven @RosannaSibora All true and places with majority rule are especially vulnerable to it. All democracies though are at risk and the rise of the far-right in many democracies funded by bad actors, some corporate and others foreign, is part of that. You don't need to capture the politicians if your Russian funded party ballooned by misinformation campaigns now has some influence.
@RosannaSibora @sjpurrs This describes some of my coworkers to a tee.
the guardian says that an MIT study found such and such
Is the guardian correct ?
Is the study valid ?
even if Valid, does it apply outside of some narrow range of circumstances ?
who know; who cares, this is social media, confirmation bias is good !!
@RosannaSibora is that causation or just a correlation
far as I can tell, there is no link to the MIT study, so that is a red flag
I think this is the stuudy
https://arxiv.org/pdf/2506.08872v1
but maybe it is this which is not yet peer reviewed and it is not from MIT (solely)
@failedLyndonLaRouchite @RosannaSibora it doesn’t seem far fetched. I’ve heard similar studies about paying people to solve problems found that people would give up more easily when paid or when they were formerly paid for solving problems. The conclusion may be questionable, even if the study results are rigorous. There are different thresholds, rewards, and motivations. I enjoyed hearing about a novel application of this experiment.
Thank you for the study links. 🙂
I seem to recall the 1980s, when it was calculators that were ruining kids' brains.
@tuban_muzuru @RosannaSibora yes and coders who wouldn’t use C, only ASM.
AI can be a tool used discriminately and effectively like a calculator or compiler OR you can “to someone with a hammer, everything looks like a nail” things.
@ozeng @tuban_muzuru @RosannaSibora the difference is calculators usually give the correct answer.
@cinebox @ozeng @RosannaSibora
There is no difference: they're both machines. You push the buttons, the current sloshes around in there.
@tuban_muzuru But calculators give the right answer consistently, it's not working with a unnecessarily huge blob of data and plucking out an answer based on algorithmic estimations that may or may not be correct. They might both be machines but one is using clear logic, the other is electronic dice rolling.
Aug 10, 2026, 7:19 AM@nini @tuban_muzuru @cinebox @ozeng @RosannaSibora
More importantly, you can build a mental model of how a calculator works and learn to use it correctly. They offload a specific step in a calculation.
And using calculators in class does harm children’s ability to do mental arithmetic, which is why they’re not introduced into classrooms until children are past the stage where mental arithmetic is important.
They also cause the ability to do long division, multiplication, and so on to atrophy. That normally doesn’t matter because these are not actually useful skills, the ability to understand when to use them is. Using a calculator helps learning once you get to algebra because it’s useful only for a handful of steps and those are not the important ones for what you’re learning. Using a symbolic algebra system would let the machine do all of the steps, but only up to a certain limit and you’re acquiring a skill that includes knowing how to formulate the equation, so using a symbolic algebra system will increase your abilities later.
And, at that point, the systems aren’t entirely deterministic. They have tactics that will try to solve equations in different ways and they will select between them based on heuristics. But to really get the most out of these systems you need to understand how the tactics work and when they will be useful, which is what you learned solving equations by hand. And, once you understand that, then they behave deterministically: if the heuristics don’t work, you can apply steps manually (still much faster than doing it by hand).
There is no equivalent for an LLM. It will generate a sequence of tokens that is a statistically plausible continuation of the sequence that you provide as a prompt. But that can vary wildly based on the random seed. It will also vary wildly based on which model you use. In contrast, any skill you learn on something like Matlab will transfer directly to something like GNU Octave.
The impossibility of theory building is what really sets LLMs apart. Humans recognise patterns even when there are no patterns and so people come up with a load of things that they throw into their prompts because they made things better in a few previous attempts and build religions around them but they aren’t mentally modelling how the LLMs work.
Aug 10, 2026, 7:21 AM@nini @tuban_muzuru @cinebox @ozeng @RosannaSibora I wouldn't engage with them, the signs of their cognitive collapse are showing.
@paul @nini @cinebox @ozeng @RosannaSibora
Well, as it says in my sig...two kinds of monkeys
Aug 9, 2026, 12:09 AM@RosannaSibora You know... this makes me wonder...
I'm two years post stroke. I still have trouble with my right hand, especially my index finger. I keep defaulting to this awkward left-handed mouse, and I awkwardly type with only six fingers unless I REALLY force myself.
Therapists tell me 'use it or lose it' is a real thing I need to be aware of.
I wonder if this is a similar phenomenon.
@Longspeak @RosannaSibora I believe it 100% apt. AI shouldn't be used as a mental crutch or substitute for higher-level of any level of thinking, but I think those who approach it this way are going to experience mental atrophy much like the physical atrophy undoubtedly you are trying your best to overcome. Bless you and keep it up
Aug 9, 2026, 1:34 AM@pinoycelt @RosannaSibora Thank you. I'm trying. It's insidious. I'll just fall into the pattern without question.
Still better than the alternative, so I'll take it.
https://www.desmog.com/2025/12/11/the-koch-network-is-pushing-trump-to-accelerate-ai-documents-show/
https://medium.com/@richard.brody/trumps-war-on-education-366c82569c65
https://jacobin.com/2018/06/public-education-privatization-koch-brothers-teachers
https://gisme.georgetown.edu/publications/trump-won-because-voters-are-ignorant-literally/
https://jacobin.com/2018/06/public-education-privatization-koch-brothers-teachers
Koch Network spent a lot of money to create an illiterate peasantry.
https://nymag.com/intelligencer/article/texas-private-school-vouchers-josh-cowen-research.html
https://michiganadvance.com/2024/09/26/a-religious-separatist-movement-in-american-education/
https://michiganadvance.com/2024/09/27/a-religious-separatist-movement-in-american-education-2/
https://www.economist.com/united-states/2025/02/26/edward-blum-has-new-allies-in-his-fight-against-affirmative-action
https://archive.is/fkhbD
1/
2/
Average American literacy is down to 6th grade after years of attacks on teachers & deficient school funding
https://en.wikipedia.org/wiki/Literacy_in_the_United_States
https://ballsandstrikes.org/law-politics/ed-blum-cbc-scholarship-lawsuit/
https://slate.com/news-and-politics/2022/10/supreme-court-edward-blum-unc-harvard-myth.html
The six dynasties behind Projecf 2025 are terrified of an educated electorate.
https://theintercept.com/2022/08/25/student-loans-debt-reagan/
https://www.washingtonpost.com/made-by-history/2022/12/29/history-student-loan-debt/
https://en.wikipedia.org/wiki/Roger_A._Freeman_(economist)
Their oligarchic "forever wealth" depends on a frying planet & the uneducated.
https://www.desmog.com/2024/08/14/project-2025-billionaire-donor-heritage-foundation-donald-trump-jd-vance-charles-koch-peter-coors/
https://www.thenationalliteracyinstitute.com/2024-2025-literacy-statistics
@RosannaSibora
Does anyone know the details of the MIT study mentioned in the article? I couldn’t see a reference & I would like to read it for myself. Thanks for any help! 🙏🏻
@RosannaSibora I mean, that’s what they said about social media. And the internet. And personal computers. And pocket calculators.
I’m not saying people should learn AI first, not by a long shot. But the “it makes things too easy” is an argument decades old about every new technology.
There is plenty to complain about with AI without resorting to the simple “hate it because it’s new.”
@AnonymousFreak @RosannaSibora Asbestos has great fire-proofing attributes but the secondary issues with it made it unusable because it was that unsafe, to me LLMs have that exact same issue of being this miracle material on paper but has severe issues where the risks outweigh the gains.
@RosannaSibora Education is about transforming, and being transformed, for the sake of understanding.
It is ridiculously wrong to treat the required personal, transformational work as something to be relegated in the name of efficiency while dropping everything personal and transformational.
Would it become more obvious if one proposed to send an AI proxy to college or university instead of the actual student?
@RosannaSibora While I recognise many of the referenced facts from studies, unfortunately (but somewhat understandably) the article does not cite them.
If it did, it would have been quite an awesome resource.
@RosannaSibora @imaginaryuncle AI users and lovers are in total denial.
@RosannaSibora anecdotal,but I feel it myself after just a short period of LLM "programming".
Hate that I have to do it, that colleagues don't care and management is so all-in on "AI" that it is half of all sentences spoken for all tasks.
Aug 9, 2026, 2:12 PM@RosannaSibora Now we need a study about those who have to spend time reporting about AI, I feel like there would be fatigue too…
@RosannaSibora well, duh.
@RosannaSibora does anyone have the study?
@RosannaSibora It would be interesting to see whether the same harm results if the users understand that they are not in fact interacting with another consciousness, but obtaining guesses based on training data.