How University Students Use Claude

(www.anthropic.com)

Show context

dtnewman ◴[09 Apr 25 16:23 UTC] No.43633873[source]▶

> A common question is: “how much are students using AI to cheat?” That’s hard to answer, especially as we don’t know the specific educational context where each of Claude’s responses is being used.

I built a popular product that helps teachers with this problem.

Yes, it's "hard to answer", but let's be honest... it's a very very widespread problem. I've talked to hundreds of teachers about this and it's a ubiquitous issue. For many students, it's literally "let me paste the assignment into ChatGPT and see what it spits out, change a few words and submit that".

I think the issue is that it's so tempting to lean on AI. I remember long nights struggling to implement complex data structures in CS classes. I'd work on something for an hour before I'd have an epiphany and figure out what was wrong. But that struggling was ultimately necessary to really learn the concepts. With AI, I can simply copy/paste my code and say "hey, what's wrong with this code?" and it'll often spot it (nevermind the fact that I can just ask ChatGPT "create a b-tree in C" and it'll do it). That's amazing in a sense, but also hurts the learning process.

replies(34): >>43633957 #>>43634006 #>>43634053 #>>43634075 #>>43634251 #>>43634294 #>>43634327 #>>43634339 #>>43634343 #>>43634407 #>>43634559 #>>43634566 #>>43634616 #>>43634842 #>>43635388 #>>43635498 #>>43635830 #>>43636831 #>>43638149 #>>43638980 #>>43639096 #>>43639628 #>>43639904 #>>43640528 #>>43640853 #>>43642243 #>>43642367 #>>43643255 #>>43645561 #>>43645638 #>>43646665 #>>43646725 #>>43647078 #>>43654777 #

1. bboygravity ◴[09 Apr 25 18:42 UTC] No.43635830[source]▶

>>43633873 #

I don't get this reasoning. Without LLMs I would learn how to write sub-optimal code that is somewhat functional. With LLMs instantly see "how it's done" for my exact problem case which makes me learn way faster. On top of that it always makes dumb mistakes which forces you to actually understand what it's spitting out to get it to work properly. Again: that helps with learning.

The fact that you can ask it for a solution for exactly the context you're interested in is amazing and traditional learning doesn't come close in terms of efficiency IMO.

replies(3): >>43635984 #>>43640019 #>>43640050 #

2. dingnuts ◴[09 Apr 25 18:49 UTC] No.43635984[source]▶

>>43635830 (TP) #

> With LLMs instantly see "how it's done" for my exact problem case which makes me learn way faster.

No, you see a plausible set of tokens that appear similar to how it's done, and as a beginner, you're not able to tell the difference between a good example and something that is subtly wrong.

So you learn something, but it's wrong. You internalize it. Later, it comes back to bite you. But OpenAI keeps the money for the tokens. You pay whether the LLM is right or not. Sam likes that.

replies(2): >>43638780 #>>43639717 #

3. Spivak ◴[09 Apr 25 22:37 UTC] No.43638780[source]▶

>>43635984 #

This makes for a good sound bite but it's just not true. The use case of "show me what is a customary solution to <problem>" plays exactly into LLMs strength as a funny kind of search engine. I used to (and still do) search public code for this use case to get a sense of the style and idioms common in a new language/library and the plausible set of tokens is doing exactly that.

4. margalabargala ◴[10 Apr 25 01:19 UTC] No.43639717[source]▶

>>43635984 #

> So you learn something, [...] You internalize it.

Or they don't.

5. layer8 ◴[10 Apr 25 02:19 UTC] No.43640019[source]▶

>>43635830 (TP) #

It’s more like looking up the solution to the math problem you’re supposed to solve on your own. It can be helpful in some situations, but in general you don’t learn the problem-solving skills if you don’t do it yourself.

replies(1): >>43642722 #

6. hackable_sand ◴[10 Apr 25 02:26 UTC] No.43640050[source]▶

>>43635830 (TP) #

I would recommend programming, and designing your system, on a piece of paper instead.

It's the most efficient few-shot which beats the odds on any SotA model.

7. Ekaros ◴[10 Apr 25 11:08 UTC] No.43642722[source]▶

>>43640019 #

Exactly. For vast majority of students myself included just looking at ready solution is actually very poor way to study. And LLMs are exactly this. Ready solution generators. Doing with things like math and programming is learning.

And same goes for art. You do not become master in art by looking at art or even someone drawing...

↑