AI, Cheating, and the Illusion of Learning

I have noticed a troubling trend recently when my students cheat with AI.

At the beginning, my undergraduates used to simply cut and paste whatever the AI gave them. Then they learned to put that output through other AIs to “humanize” the writing. (Thanks Grammarly!) The tech-savvy students would create their own prompts (“write like a smart student who is rushed and makes occasional syntactical errors”); the paranoid ones would go through the essay and make lots of minor changes. 

But now, a lot of my students have turned to what I call the “split screen” strategy: They’ll input the professor’s directions into AI (maybe on their phones or a second screen) and then retype what the AI tells them, making minor changes (sometimes every paragraph, sometimes every sentence) as they type. 

From a detection perspective, it becomes almost impossible for even the best AI detectors to identify such writing as AI-generated. Moreover, going into the basic back-end analytics (whether in Google Docs or MS Word) shows you that, yep, the student did in fact take a bunch of time to write the document (since they spent the time typing it up). And student after student using this “split screen” strategy has initially proclaimed to me, “I wrote it myself!” 

Technically, they’re right.