Theres a Tool to Catch Students Cheating with ChatGPT. OpenAI Hasn't Released It | Hacker News Reader

Theres a Tool to Catch Students Cheating with ChatGPT. OpenAI Hasn't Released It

wsj.com

12 pointsgrugagag10 comments

It’s basically an impossible situation. I sympathize with TAs that have to read through dozens of AI generated slop papers and want to put an end to it. But the fallout from a false accusation is serious and potentially life-changing. Schools don’t have to give you due process.

That’s why I was so upset about tools like “gpt zero” and the media’s subsequent fawning over it. They don’t care how accurate they are. They’ll add some fine print weasel words about how they only provide guidance but they know what they’re doing.

I used to teach high school. When a kid handed in writing that was not their own it was obvious. If anything, AI will make it more obvious and thus occur less often.

I've been taking some online classes these last two semesters, and I can cleanly group my classmates into three categories: the blatant LLM users, the rare students who actually invest themselves, and the majority whose contributions are so cursory that it hardly matters whether they wrote or generated it.

We built a simple but novel solution for that is far more reliable and works completely differently to gpt-zero and openai methods. I'm not posting a link as we're not ready for HN hug of death, but please PM if interested.

The saddest thing is that this project has been once of the most demoralizing I've ever taken on. Day to day we see so many students being failed by teachers and school leadership who care more about "adapting to AI" than real student outcomes today.

In practice, we've found teachers don't generally want to have the difficult conversations with students when the hard evidence of cheating is given to them.

And generally school/university/college leadership have no real tactics to implement their "AI strategy" other than train their own chat bots (wtf) and "adapting assessment to use AI".

Unfortunately, a simple non AI fix to the problem is definitely not as good for their careers.

IMHO without a change, we're creating a pretty bleak future for students of the next few years.

The problem is that a lot of students were already cheating, and there's no real benefit to teachers trying to punish the students for it.

The schools don't want to do it, because letting one or two students by is, to them, much better than {the public backlash of someone admitting to cheating, the loss of money from kicking paying college students out, ...}, and the teachers don't want to do it because the schools are going to tell them to shut up and stop making noise about it, and it might poison their career to not listen. Of course, if you do that, then they will keep doing it, and you'll get people graduating who poison your reputation eventually, but it turns out to take a while before that happens, if ever...

At least when I was in college, it was pretty obvious to the students which students Did Not Do The Work(tm), and it was eye-opening how bad it could get when you took the "for non-majors" versions of various classes.

ChatGPT et al are just bringing to the public eye how bad this already was by lowering the bar so much there's no barrier to entry, no "if you know you know" whisper network of people cheating for each other, no wink-and-a-nod "tutoring" service being paid for, just...press button receive paper.

Easy solution, allow the use of LLM's for work that can reasonably be completed with the aid of an LLM. Learning to use these tools to produce effective quality content is a skill worth teaching. Start marking based on slop value instead, LLM's still produce slop so marking against that will simultaneously reward students who didn't use LLM's while also encouraging students who used an LLM to read and edit what the LLM produced.

LLM's still aren't great (without reading produced text and iterative editing of prompt upon reflection etc) at producing good quality content first go. There's always things that are a bit 'off' so to speak.

https://archive.md/XeNGe

My middle school daughter wrote an essay and plugged it into an online tool assessing risk of AI utilization out of curiosity. It gave her an 80%, I said “good job!”

It’s not cheating

Exactly, they need to adapt evaluation system to actually evaluate the student, regardless of the tools they use.

Reply on news.ycombinator.com