Teaching Thursday: AI and Cheating

In the lead up to the new semester, there is a lot of buzz about dealing with AIs in the classroom and the predictable gaggle of well-meaning, but angst-ridden punditry about how ubiquitous access to AI will “change everything” and how unprepared “we” (faculty?) is for the change that’s about to happen. 

This all seems a bit breathless to me, but I get that breathlessness sells page views and subscriptions and that is the way of the world.

Here is my warm take on AI in the history classroom.

1. Cheaters Gonna Cheat. Some of the most breathless anxieties about LLM AIs in the classroom is that it’s going to make it easier for students to cheat on writing assignments. This may well be true, but what’s also true is that almost all forms of cheating are, by definition, easier than doing the work to learn how to read and write. In other words, cheating is easy by design. In my experience, students who cheat do so because they either think that they have no choice or because they simply don’t want to do the work. LLM AIs aren’t going to change this. 

The other issue, of course, is that there is this feeling that it’ll be harder for us to “catch” students using AIs. This may be the case, but I think most faculty have given up on “cop shit” in their classroom a long time ago. My goal is not to “catch” my students cheating, but to create situations where cheating isn’t seen as a viable recipe for success in the class. 

2. Clearly communicate expectations and course goals. First, I stress to my students that none of the work is this course should require the use of AI. In other words, using AI to help with writing or to produce the writing required in this class is neither a necessary nor acceptable way to achieve one’s learning goals. Second, I make explicit how the assignments achieve the learning goals. For example, I tend to stress how writing and reading are interconnected and how relying on a LMM AI to summarize complex texts or write challenging papers is unlikely to produce the kind of stronger reading and writing skills that we have designed the class to build.

Will this stop a student committed to cheating from using an AI to read and write? I doubt it, but it will ensure that students committed to the class understand why using AI isn’t a good way to achieve its goals.

3. Creating good assignments. Finally, there are practical ways to make sure that the learning goals are transparent in assignments. In the bad old days, course assignments served as either sorting to discern who best understood course content or checks to keep students on track with attendance, participation, and outside of class work. These aren’t necessarily bad things, but they’re often seen as routine or secondary to the work of learning the material.

The growth of LMM AIs will encourage us to create assignments that are scaffolded to clearly represent the course goals. For writing, this means outlines, attentions to citational practices, revisions and drafts, and reflective work. These things are difficult for an AI to simulate successfully or require such an investment in working with AI that it’ll deter casual cheating. But more importantly, these skills are central to becoming a confident, polished writer.