Why Teachers Still Need AI Detection Tools

I meet a lot of teachers who have the same perspective regarding student AI writing. I often hear, “I know my students' writing; so I don't need AI detection software. It's a waste of time.” I fully sympathize with such a perspective, since I too have a good sense of my student writing, and, for more than twenty years, I have refined my sense for plagiarism and student cheating.

Over a decade ago, one of my students submitted a paper that was definitely a student submission, but it wasn't my student's writing. I assign a lot of writing, so I get a sense early in the year what each of my students' voice is like and what hiccups are in their writing. This paper was not his own. After some digging, I discovered that the student had borrowed his paper from his cousin from another school who was writing a similar paper (how many papers on Gatsby and the American Dream exist in the world?). He submitted his cousin's paper, and assumed I wouldn't notice. To his dismay, I noticed.

So, why use AI detection? First, AI writing is becoming ubiquitous and it is only going to get better. More importantly, though, as AI writing encroaches even further into our classrooms, the simple bandwidth it takes to monitor it is exhausting. There has always been a little fun (and healthy) cat-and-mouse game going on among teachers and students. In some ways it's good for both the teacher and the student. Teachers must stay guard as students employ a great amount of critical thinking exploring ways to evade detection. But, there is a point of minimal return. There is a point where trying to monitor student cheating takes away from actually teaching students the craft of writing well. AI detection that works gives bandwidth back to the teacher to employ their energies more meaningfully elsewhere (and the student still can have as much fun as they want trying to break the rules).

A simple parallel to the AI detection problem is cell phone usage in classrooms. We wouldn't need policies if management of cell phones were effective, but they are not. It is exhausting for teachers to police cell phone usage in a classroom, despite the fact that it is so easy to detect. That's the point: the ease of detection is no reason to avoid having a policy/detection in place. It might be easy to detect AI writing for some teachers, but there is still good mental economics behind using software to eliminate the waste of energy in the chase. It simply takes up too much time and energy to monitor for AI.