Using AI chatbots for even just for 10 minutes may have a shockingly negative impact on people’s ability to think and problem-solve, according to a new study from researchers at Carnegie Mellon, MIT, Oxford, and UCLA.

[odr_k4tana@infosec.exchange]

@GossiTheDog not gonna lie: this work is kinda problematic, especially when I see how they frame their results.
The first study simply represents a modular change problem. If you had help you will always struggle initially if that help is cut off suddenly, and the ability only slowly recovers.
The second, "better" version (acc to authors) is also funny. Once you read "Participants who used AI for hints showed no significant impairments relative to control", it's easy to guess what happened. Lazy people did not take the task seriously and used AI for answers. Of course they'll crash.
This rather confirms my personal notion that AI is mostly enabling laziness rather than killing thinking ability (as I stated here some time ago: https://k4tana.github.io/blog/2025/06/04/AI-Nuanced-Take.html).

Edit: Additionally those effect sizes are not really much to write home about, especially in the second experiment. Statistical significance alone does not mean practical relevance.

[kira_unlimited@defcon.social]

@GossiTheDog This is exactly what I find anecdotally with my students (I teach high school), and it's extra problematic because the only thing they're willing to think about is how to regain access to AI should I deign to lock up all the chrome books and try to make them remember how to think!

[kira_unlimited@defcon.social]

@odr_k4tana @GossiTheDog Hmmmmm, yeah I see what you're saying here. It would be interesting to see a repeat of this with a control who had help from humans versus a group that had help from AI.

I would also love to know the methodology on determining what people used AI for --- did they keep a record of the AI conversation and determine what it was used for, or was it self-reported via survey?

[linguistgoneforeign@mastodon.social]

@GossiTheDog

Thanks for sharing!

[walruths@mastodon.social]

@GossiTheDog
A chatbot is the reason I installed prosody 4 times before succeeding, granted, it's also why I know why it didn't work, lmao, but imagine the time savings in retrospect.

[lockspringer@chaos.social]

@GossiTheDog Hail @grok please 'find' a study that says the exact opposite of this one.

[bontchev@infosec.exchange]

@GossiTheDog In other news, using a calculator impairs your ability to compute in your head. Also, my constant use of keyboard has impacted negatively the legibility of my handwriting.

[kira_unlimited@defcon.social]

@odr_k4tana @GossiTheDog

Okay, I went and read your blog post, and I have some thoughts:

1) Overall, this was a very informative and thought-provoking read, so thank you for writing it and linking it here!

2) I think you do overlook the issue of how AI is wrecking havoc on education in your enumeration of serious issues with AI. I don't know if this is true everywhere (I read that you're in Germany, and I feel that education is going better in the EU than in the US right now). In any case, I'm currently (for the last roughly two years and the next two and a half weeks) a public school teacher in the Mississippi Delta, USA and AI is causing (and exacerbating) a LOT of problems. I teach a lot of functionally illiterate 10th and 11th graders (these students are 14-17 years old), and the general lack of critical thinking and problem solving skills is ASTOUNDING to behold.

Of course, this isn't CAUSED by AI. The historical cause is, of course, repercussions of slavery, Jim Crow laws, and systemic racism in this region, leading to huge rates of illiteracy and a (valid) cultural distrust of establishment. BUT I believe firmly that AI is making it far, far worse:

AI has made it easier to cheat than ever before. I cannot assign any sort of homework, because 90% of it will be completed using AI, so it's basically an unethical grading practice to accept it, but it's quite hard to "prove" that it's AI enough to satisfy parents that their child deserved a zero on the assignment.

I've been having to require that students complete all writing on paper in class, but still I sometimes get AI stuff that they copy onto the paper from students hiding a phone or apple watch effectively enough that I don't see them doing it despite walking around and monitoring the classroom all day. I feel like my job at this point is just to be the behavior police and not even to teach anything.

But beyond THAT I am inundated with professional developments and trainings urging teachers to accept that AI is a tool and use it to create lesson plans, reading materials for the students to read (instead of actual literature????), and use it to take the pressure off of grading writing. Essentially, in the US at least, teachers are perpetually being expected to do more and more and more tasks without more pay or more time during the workday provided, and heavy AI use is the recommendation to make this workload viable.

And beyond THAT, the state writing test is graded using AI. At a certain point, I question why we're even having school if the teachers are using AI to make the assignments, the students are using AI to do the assignments, the teachers are using AI to grade the assignments, and the state is using AI to assess proficiency. It feels like we're a robot cosplaying as a school.

Of course, some of this is the American education system, which has been problematic and inadequate for the task of giving young people a worthwhile education to become productive citizens for at least three quarters of century at this point, coming to a head. But I think AI is somehow both enabling it and exploding it all at once.

In conclusion, I'm terrified for the future of critical thinking in the US, and I think AI is a serious contributor in that collapse, in that it's poisoning any last hope of sufficiently educating young people without a serious system overhaul (which is necessary for sure, but also not currently politically favorable, because the current disaster system of education is productive for the (evil) powers that be.

Which is all to say, in long-winded (but context-providing) terms, I'm not convinced AI is destroying the critical thinking skills of people who had already developed them, but I think, in the United States at least, AI is widening the percentage of people who will enter adulthood never having learned them, and that is a serious problem.

1/2

[kira_unlimited@defcon.social]

@odr_k4tana @GossiTheDog

And finally,

3) Your point at the end about how we're losing our grip on technology because most of us really don't understand it is something I've been thinking a lot about recently!

I made a YouTube video somewhat along these lines recently!

No pressure to watch it, but since I so enjoyed your blog post, I figured I would share!

- YouTube

Auf YouTube findest du die angesagtesten Videos und Tracks. Außerdem kannst du eigene Inhalte hochladen und mit Freunden oder gleich der ganzen Welt teilen.

(www.youtube.com)

(edited for typo)

[masyukun@mastodon.social]

@GossiTheDog This doesn't pass the smell test. Compare with if you had a 10 minute conversation with the smartest person you know. Or the dumbest. Shouldn't there be more general results that indicate conversations IN GENERAL make you smarter or dumber depending on who you are speaking to?

[vonkordke@mastodon.com.pl]

@GossiTheDog what about talking with other people? Don't ask anybody about anything and do everything by yourself?

[jackemled@furry.engineer]

@MuhammadFreeSoftware @GossiTheDog Why not just use the built in emote search available on literally every operating system? It's even faster & it costs much less.

[just_a_rock@wetdry.world]

@bontchev @GossiTheDog Using a calculator does not harm your mathematical reasoning or your accuracy with doing the math in your head. It just leaves you out of practice so you take longer. That is practice, which is not the same type of learning as reasoning & thinking. Here they show the harm AI causes to reasoning & thinking ability & motivation. Real time taken to answer a math problem was not measured as part of the study. See the experiment methods & both the control & variable conditions.

[tarix29@tech.lgbt]

@odr_k4tana @GossiTheDog I really would have liked to see another group who was only given the test questions without the prior "practice" questions. That could have shown whether AI use directly decreased performance or if it merely made the practice less effective

[spinni81@mastodon.social]

@GossiTheDog
"In the AI condition, participants solved 12 fraction problems with an AI assistant (GPT-5) available in a sidebar. The AI was then removed without warning, and all participants solved 3 additional test problems independently. Participants in the AI condition had a significantly lower solve rate"

I'm quite sure the same would happen with a calculator instead of an AI chatbot. Interesting study but with this methodology I'd take it with a grain of salt.

[slotos@toot.community]

@kpwn @GossiTheDog That’s exactly why the reversal happens.

Look at later graphs, they clearly label the phases: „pre”, „learning”, and „test”.

Making mistakes in learning is exactly how we learn. Replace AI with a happy-to-help specialist, and you’ll get a similar result.

Hell, train AI model with only the correct answers and no negative backpressure, and it will recognize sarcomas by the presence of ruler markers on the x-rays.

[leyrer@23.social]

@linuro @GossiTheDog "I snort cocaine just so I can stay awake ..."