Die Charts sprechen für sich. Die allgemeinen Fähigkeiten von LLMs machen es viel einfacher, Hausaufgaben zu machen, aber wenn Sie nicht aufpassen, werden Sie am Ende weniger gebildet und sachkundig sein.
https://www.economist.com/graphic-detail/2026/08/18/does-ai-stop-children-from-learning
Von dostre
27 Kommentare
Not to be rude but, no shit?
Interesting how the AI-usage group stops being a symmetrical normal distribution and has a long tail. I wonder why that is.
Nothing new. Well, they AI used to be the older sibling but the result were the same.
Did the author not label the y axis because he used AI too much?
Lots of words to say students who take homework and study seriously perform better than those who don’t
Kids who work on their work do better in tests. More news at 11.
Lots of “does a bear shit in the woods?” research, lately
so the lesson here is to do your homework yourself and then at the very end have AI correct it.
Having something do the work for you doesn’t help you learn. Surprise, surprise.
I think theres 2 things that are interesting. 1 is that AI use will heavily correlate with „AI misuse“ and by that I mean „claude solve these 10 questions for me“. It will do a good job at that, it will not teach you anything. While you can use AI to learn and do so effectively, like with any studying, its active learning and the AI doesnt phase the knoweldge into your brain. 2 is that test outcome doesnt necessarily correlate with knowledge learned or retained beyond that test, tests have been a bit of a contentious topic in pedagogy
This isn’t really about AI though, this is more about not doing your own homework and therefore not learning anything. No different than having another person do your homework for you
That’s some shitty graphics
So to dodge AI warping grades, maybe a teacher should put more grade weight on exams?
I’m pretty sure if you looked at Chegg users back in the day you’d see the same result.
I ton of people I went to college with (I went for Electrical Engineering) basically Chegg’d the answers to all their homework anyway. Almost all of those people were clueless when it came to the actual subject matter and constantly got destroyed come exam time.
I almost always found some way to get a teachers edition or otherwise get the answers to the homework problems too, but I’d always try it on my own first, then look at the answer and try to understand what I did wrong, if I got it wrong.
You can definitely use an LLM to help you learn, they’re actually great at it. But you need to stay mindful of the point of homework and actually do the work yourself. Have the LLM check your work and explain things to do, but be sure you put in the work yourself.
So AI use for school work (i.e, not doing the work yourself) means that kids are not retaining information in order to learn and thus cannot retrive information for testing..
No Duh.
Who would have thought that you get a higher score on your homework after basically cheating on it
I’m old enough to remember when there was similar concern about web search.
I would have expected at least some people to see improved exam scores by using AI. Even if it’s not usable during the exam itself, it would be a great tool for developing study guides, practice tests, and engaging in things like Socratic method for the exam topic. But maybe AI comes along with a drop in overall effort to understand the material.
Isn’t this the same as those kids who used to get their mates to do the coursework and then struggle at exam time?
This is how we devolve into the eloi.
Can we have genuine discussion here? As technology get more complicated, we have to operate at a higher level of abstraction.
I had to learn to write complex SQL queries by hand on paper when I was at university. Now, AI can just give me the answer. But even back then, I didn’t know the full detail of how an OS or computer hardware operates.
Is it enough to learn the fundamentals, let AI do the heavy lifting and focus on higher level reasoning rather than understanding the logic under the hood.
It’s a shame because AI can be astonishingly good at teaching – but there’s obviously a temptation to skip that part and just get it to do the work on your behalf instead.
As a kid who had undiagnosed ADHD, I did great on tests but often didn’t turn in homework. I’m like the polar opposite of this lol.
If you’re not doing the processing in your brain, you’re not developing the neural pathways.
The job market will be wide open to people who did the work and developed their brain.
the paper itself notes students who use LLMs but spend just as much time on the homework (likely using the LLM as a guide rather than a crutch) paid little penalty. no one here discussing that.
these findings are parallel to students before LLMs. cheat on your homework and get great scores, fail the tests later.
This just in: students who don’t do their homework get worse test scores. More news at 8.
What’s frustrating is that people use this to argue that any AI use is automatically harming you. Obviously outsourcing your brain is bad, but working with AI to brainstorm is one of the most mentally engaging things you can do.
Do exam results correspond to job performance? Did they ever?