My university uses AI detectors when marking papers, and almost any result higher than 0% can cause lost grades. As anyone whose used one of these tools will know, they have an extremely high false positive rate, meaning that while I don’t use AI (and can’t, given the technical nature of the papers) I still lose marks. Is there any way to decrease the rate of false positives without completely destroying the structure and flow of a paper?


My experience is the opposite. Even for a work that was 100% obviously written by AI, an AI detector only gave me 30% probability. That specific work was written with a PhD level and had bogus sources, one litteraly having copilot in the name. Coming from a 19 years old first year college student… On every work I presented, the AI evaluated the probability very low while for me (I know the students, I’ve spent 15 weeks with them) l, they obviously didn’t wrote the papers. When I pointed to the model that I was sceptical, it suddenly change the cheating probability to a much higher score. So basically it just tolds you what you want to hear. The college told us they wont consider cheating unless we have proof so we just end up giving good grades to everyone. Two scenarios here, either about 70% of students use some sort of AI to write these days, or we have the smartest generation ever in front of us, with 19 years old kids having the level of their teacher aftef a year in College. University and College administration want to give diplomas, the students want to graduate, and the teachers are now just a tool to get this done.
Universities like Hardvard recently adopted internal rules preventing a class from having more than 40% A or A+ in the final result, because it was getting absurd, with some classes having 80% of the students with an A. I’m seriously considering leaving college teaching because I can’t understand what’s happening.
what kind of detector are you using? how would you ‘point out’ to it?
What did you use? Don’t ask the LLM if something is written by an LLM, it has no way of telling.