Stop Talking about AI-Proofing Courses
As a new semester approaches, I am reviewing applications to my advanced writing course while trying to plan a campus-wide summit on the challenges generative AI poses to pedagogy. Several of these applications were clearly generated by AI. Meanwhile, I’m reading articles and having conversations with colleagues about how to AI-proof the classroom. (There’s actually a great discussion by Shannon Brenner and Jordon O’Connell about AI-Proofing where even they admit the limits.) Faculty recommend interviewing students and video comments in forums, for example, or collecting writing samples at the start of a term. I even had ChatGPT generate a list of common attributes in essays written by AI (which it did without irony). Is there syllabus language that can mitigate the use of AI? Are attempts at AI-proofing even worth our time or are we just ZeroGPT-ing into the wind?
The AI Student
Jeremy Douglass, Professor of English at UC Santa Barbara, has a solution I like that does seems to avoid the cat and mouse game (see this writeup by Lillian Mina at U Alabama at Birmingham on AI detection tools). The process follows an idea he had suggested at our Future of Writing Symposium in 2023, where he suggested having an “AI Student” in your course, doing all the assignments.
The idea is first to see what a minimally prompted AI can do:
- Put all your assignment prompts into one or more LLMs.
- Publish the results for the class to see.
- Ask students to review it, evaluate it, comment on it.
Then, of course, have a discussion about what the AI did. That also gives the class a sense of what the base-line answer to this assignment is.
Next, announce that is output is considered D/F level work. If literally anyone can produce this content by simply inputting the assignment prompt without even taking the class, then students who wish to demonstrate learning the material and making an original contribution must to do better than (or something different from) this.
That exercise reminds us of another important goal: Talk with students. Let them know what’s important about the assignment? What’s its goal? What are your values? Why do you want them to do their own work? I think we assume (to our students’ loss) that students know why some of us value the writing process so much. And we should share our struggles and the joy of that struggle.
Above all, be empathetic: As we did with plagiarism cases, ask why are they doing it? Are they just shirking the work? Or are they being practical? Or are they being thoroughly modern? Do they have trouble getting started? Do they have insufficient models of what a genre of writing should look like? One of the things I dislike about the antagonist style of pedagogy is that we assume “cheaters are bad” rather than writers who need help.
Listen, I’m okay with you trying to respond to various uses of AI in your classroom. Maybe we settle for AI-resistant. But if you’re thinking you’re AI-proofing your course, absent the unplugged Enemy of the State-style clean-room, you’re most likely, well, hallucinating.
