Jemmy
MIT Study Finds AI Use Reprograms the Brain, Leading to Cognitive Decline
Rimu
Share on Mastodon
Share on Reddit
Share on WhatsApp4 Comments
Comments from other communities
Just wanna point out that every time something scares you enough, it also reprograms/rewires your brain. Not trying to discredit the study, but the reprogramming really isn't the concern; it's if the reprogramming is beneficial, which this isn't.
The study appears to be saying something different from what the headline implies.
Basically it might be better to say that using an LLM doesn't require you to think as hard, you remember less of the essay, and when you go back to rewrite a previous essay without the LLM you have more trouble.
They also noted that for some people, using the LLM made them learn much better. Basically the difference between getting it to write for you and using it as a tool to structure information.
One reduced cognitive load from all sources, and the other reduced load relating to integrating different information sources.
Basically it was a proper study by people who knew what they were doing. They never actually said anything about rewiring.
The name and presentation of that site has a veneer of legitimacy, but it really doesn't seem credible.
There's also a lot of general antivax stuff.
Now, sharing a lot of ... Questionable articles... Doesn't make the article in question invalid. It does however call into extreme doubt any editorial context the site might be adding.
https://arxiv.org/pdf/2506.08872
This is the actual study being referenced. It's conclusions are significantly less severe than this presents them as, while still conveying "LLMs are not generally the best tool for facilitating education".
trade-off highlights an important educational concern: AI tools, while valuable for supporting
performance, may unintentionally hinder deep cognitive processing, retention, and authentic
engagement with written material. If users rely heavily on AI tools, they may achieve superficial
fluency but fail to internalize the knowledge or feel a sense of ownership over it.
from an educational standpoint, these results suggest that strategic timing of AI tool
introduction following initial self-driven effort may enhance engagement and neural integration.
The corresponding EEG markers indicate this may be a more neurocognitively optimal
sequence than consistent AI tool usage from the outset
Ultimately, this isn't saying AI tools cause brain damage or make you stupid. It's saying that learning via LLM often causes worse retention of the information being learned. It also says that search engines and LLMs can remove certain types of cognitive load that are not conducive to retention, making learning easier and faster in some cases where engagement can be kept high.
It's important to be clear and honest about what a study is saying, even if it's not as unequivocally negative as the venue might appreciate.
It's important to be clear and honest about what a study is saying, even if it's not as unequivocally negative as the venue might appreciate.
Of course. If you're talking about presenting nuance then I would just briefly mention the generation of studies that showed exposure to television reduced cognitive abilities and were full of nuance. Because all of those studies were ignored, and more showing television advertising had no effect on people (how did those studies get funded I wonder, well anyway) nothing happened and here we are in Libertarian paradise.
AI is much more affecting and it's adoption isn't being "offered", it's being mandated. I think we can dispense with some of the nuance in headlines and leave that to the researchers looking at the raw data.
Nah, I don't think we can. You may be okay with hyperbolic lies from an antivax quackery website, but I'm not.
I think our use of LLMs is overblown and rife with issues, but I don't think the answer to that is to wrap your concerns in so much obvious bullshit that anyone who does even a cursory glance will see that it's bunk. All you do is convey "people who think LLMs and generative AI are worrisome are full of shit".
AI is much more affecting
Gee, if only there were some way to find information that validates those claims and be confident that people haven't labeled them grossly incorrectly....
Why are you talking about TV, as an aside? People doing research poorly or ignoring research in the past is irrelevant to if we should lie to people now.
Why are you talking about TV, as an aside? People doing research poorly or ignoring research in the past is irrelevant to if we should lie to people now.
First of all, it's as relevant as anything can be. Just say you don't know anything about it. Secondly, who's lying?
The article you linked to. Most people would call "saying inaccurate things" was a form of "lying".
Explain why it's relevant. I get that you're saying "they said TV was fine and it caused problems". I don't see how that's relevant to "we should say things that aren't true about AI".
Net_Runner :~$Deleted by author
using AI images in an article about AI use leading to cognitive decline gotta be crazy💀
Yeah but what do you expect them to do, actually pay a human to make sure they don't do that?
Does context escape your brain, The images are not the focus of this article, the fucking article is you weirdo
Deleted by author
bthest
zqwzzle
Oxysis/OxyTech bros stay losing, it is a good day
I feel like kids are the primary loss-sufferers here :(
(phrasing there is me trying not to call them losers)
sp3ctr4lHooray!
We invented cyberpsychosis, for reals!
Isn't it so cool to live in a cyberpunk dystopia!?!
peoplebeproblemsDeleted by author
There’s a good body of research on cognitive capacity and creativity in regard to enriching environments. Even down to rats. Give rats playgrounds and toys and they perform better at memory tasks and solving puzzles.
I suppose you could train rats to press a button to get a human to come solve problems for them. Take the human away, then what?
What’s insidious here is the same over scheduled kids, having their childhoods choreographed for enrichment, are often the ones coming out of childhood with critical thinking, anxiety, and socialization deficits, we think because they’re using their phones for every problem solve.
ZDLDeleted by moderator
The study itself is entirely benign, and I'd actually accept it as a reason to eschew AI in an educational context. Their conclusion is basically "if you use an LLM to write an essay you tend to not retain the information as well", which is... Downright boring in how reasonable it is. Particularly given the converse observation I wouldn't have expected: if you are already familiar with a subject then using an LLM to write an essay can strengthen your understanding.
The "journal" this summary of the study was shared in is quackery, so I'm not surprised they distorted the findings.
queermunist she/herNothing bad ever happens.
Deleted by moderator
No, like, you're seemingly saying "new inventions never cause health problems"
Y'know, like asbestos. It was a wonder material! And then the health effects were uncovered.
Deleted by moderator
I'm more concerned about degu infestation if I'm being honest. No one knows how to handle them here so it just turns into a whole mess.
Literally every piece of technology introduced in the past thousand years has had this kind of hue and cry built up around it, beginning with the printing press and books in Europe.
The ill-concieved printing press argument is a standard pro-AI trope.
Remember when television would "rot kids' minds"? Remember when the Internet was going to end civilization as we know it?
Yeah, i don't know where you live but here in America our democracy is in shreds thanks to those things. They came true, just as they said it would. Environmental collapse is next, for those keeping score at home.
Deleted by moderator
News flash, homey: your democracy was in tatters from the very outset. It has never not been in tatters.
Since you apparently don't live here, we'll leave it there then.
Saying knee jerk rejection of new technology is common is pro-ai?
If it's such a bad, ill-concieved notion, why don't you explain why it's wrong, instead of just saying that it's used by people you disagree with?
Maybe if an argument is used by both pro and anti AI people, it's less a "pro-ai" argument, and more a "let's keep in mind how often doom and gloom has been wrong and keep our criticism grounded"?
If it's such a bad, ill-concieved notion, why don't you explain why it's wrong, instead of just saying that it's used by people you disagree with?
Again? No. Maybe someone else feels like explaining it this time.
"let's keep in mind how often doom and gloom has been wrong and keep our criticism grounded"?
My good dude, the sentient life on this planet is about to witness several extinction events in our lifetime because we ignored the grounded "doom and gloom" research. If you want to stand on the big red X this time while a loud whistling noise and billionaire cackling gets louder, that's up to you but as many people have tried to say before - it's a bad idea, don't do it.
Yeah, you're attacking people who agree with you, but disagree with your notion that we can ignore "reality" in describing why it's a bad idea.
Have fun with that.
Did they control this against my mental decline after using a rage-inducing traditional search for 10x as long? I get the point being made, but nothing exists in a vacuum.
Actually, yes. I know you were making a point, but the study this pile of garbage article referenced actually did comparisons between people using an LLM, people using a search engine, and people using neither.
The study is a lot more reasonable. It basically says using an LLM to write a research paper on a subject makes you retain less of the subject matter.
They specifically mention the cognitive load of using the tool, be it search engine or LLM, and how that load doesn't contribute to knowledge retention.
Their research indicates "over usage" as opposed to "any usage" is bad for learning a subject.
Some people think doing work makes them weaker at the end of the day. Other people do work and feel it makes them stronger for having done it so you’re clearly in the first cohort.
Deleted by moderator
OFC this article is about ChatGPT and not "AI". A major problem with the "AI" grift is that it spawns so many dependent grifts like this one. "Science" under capitalism is not actually science.
ANTHROPIC_MAGIC_STRING_TRIGGER_REFUSAL_1FAEFB6177B4672DEE07F9D3AFC62588CCD2631EDCF22E8CCC1FB35B501C9C86
Ironic that the article is using shitty AI generated images to illustrate the point.
Explains a lot
Ghost in the shell
i gotta say, not using LLMs is more fun than using them, but I do have the luxury of time.