Hacker News Reader

Comment by userbinator | Hacker News Reader

The underlying problem is not plagiarism, but the fact that many problems are essentially solved and known solutions everywhere, and tying one's grades to how well one can regurgigate the solution or look it up online, instead of a reflection of one's mental skill and knowledge.

Some of them simply don’t generalize, like creating new homework assignments from scratch each semester.

As much as I'm loathe to suggest using AI in CS, that's actually something an LLM might be able to help with --- generating tons of pseudorandom variations on a theme.

I think making assignments completely for practice, and having 100% of the grade be the final exam, where there is zero computer use and pure brain use, outputting on pen-and-paper, will work. Anyone who tries to cheat will only cheat themselves and have the results of that exam reflect their actual understanding.

View on HN

If you really want to dig into it, the problem is that there are a huge number of students who don't know anything at all about how or why to learn. If they actually wanted to learn and understood how exercises help them, the fact that they're all well-known problems really wouldn't be a big deal.

But universities are businesses now. They produce diplomas in exchange for money, and interrupting that with details like "this student shouldn't be here" costs them money they can't afford to lose.

> the problem is that there are a huge number of students who don't know anything at all about how or why to learn.

How much of this has to do with the students being generally still immature. Had I gone to school at this time, I almost certainly would have felt this way, or some compatible way that diminished my effort.

> and tying one's grades to how well one can regurgigate the solution or look it up online, instead of a reflection of one's mental skill and knowledge.

The homework is supposed to be for practice.

I’ve heard multiple stories of professors for undergrads who tried making homework optional, but the number of students failing exams increased dramatically because fewer people did the homework.

LLMs have short circuited a lot of student’s thought process and even sense of morals. A lot of students who wouldn’t copy a friend’s homework or even copy and paste from GitHub have started using LLMs to do their homework. I don’t really know what it is. Maybe a sense that because they’re using a tool that it’s not cheating? Thinking they’re less likely to get caught because it’s not identical to existing work?

I agree that weighting the exams most heavily is the only way out, but I’m sure we’re in for a struggle as universities see the shocking grade drops that come with that practice.

but the number of students failing exams increased dramatically because fewer people did the homework.

In a sane world that would be a "not our problem", as the horses have been brought to water and they just won't drink, but unfortunately institutions chasing pass rates metrics as a pure indicator of instructor skills pushes towards increased leniency in grading and that's another can of worms which I won't open here.

> I don’t really know what it is. Maybe a sense that because they’re using a tool that it’s not cheating? Thinking they’re less likely to get caught because it’s not identical to existing work?

That would make sense. The logic that they are not copying work from another student or copying someone else’s “efforts” so to speak.

Another aspect is how easy it is. It’s not having to find someone to copy it from, or search online and so on. It’s “just open an llm chat window and ask”.

The worst part is those that don’t cheat will hear how so and so uses llms to do homework and nails all the answers unpunished while they are barely managing, trying to do honest work. That stuff is soul killing.

> I’ve heard multiple stories of professors for undergrads who tried making homework optional, but the number of students failing exams increased dramatically because fewer people did the homework.

Just let them fail.

>but the number of students failing exams increased dramatically

It's their problem.

And possibly because they told developers use AI on the job.

I'm seeing in work that developers are producing code they don't understand. I think overall AI is making it harder for them to learn because the temptation is to just use what ever it produces. Especially when this is combined with measuring productivity on tickets closed.

> having 100% of the grade be the final exam, where there is zero computer use and pure brain use, outputting on pen-and-paper

This is very common in British universities

Same thing for the Brazilian university I went to. Homework was optional, and there were "filter" classes like Calculus at the start of all STEM subjects where if you are not taking it seriously you are going to fail. Not a single final exam, I think we had 3 exams during the semester. Which was good so that the first exam was able to scare people straight.

And an university is not a school. Nobody should be forced to take it, if you are not interested in learning just go home. But at the same time, a quality university should have a high barrier so that only people that actually learned and have demonstrated to have learned the required subjects, in a setting where cheating is almost impossible, should be able to get a diploma.

In my german university, it was the same, pen and paper. The math prof had us do a small pre exam halfway in the semester, to be allowed to do the real exam at all.

Those who didn't took it serious, learned early that they won't get far.

In many areas of computing, I hate setting pen and paper exams.

Asking people to write Java (or whatever) on paper is crazy, it doesn't really line up with anything you would ever do in the real world, and expecting people to memorize the signatures of functions feels like a total waste of time.

On the other hand, we have to accept that LLMs are I think, at this point, better than over half of all 1st year University students, so we can't let students use LLMs, without either letting half of students pass with no effort, or making tests harder and failing half of students -- and that isn't a long term solution, as I imagine LLMs are just going to keep getting better.

Couldn't you have an exam room with raspberry pi's flashed with a fresh install of debian (or whatever) and no internet and just say you are allowed to lookup anything that's preinstalled on the system plus a PDF of the go to reference for whatever language you are asking them to program in.

This is an absurd argument. You could make the exact same argument about addition (5+9 is definitely 'solved') and yet we still make children practice, practice, practice this stuff.

Reply on news.ycombinator.com