I teach undergraduate courses at a mid-size university and we’ve been revising our academic integrity policy for the third time this year. Every time we think we have a workable framework, something changes.
Current state at my institution:
- Detection tools are “available as one input” but cannot be sole basis for misconduct charges
- Faculty can require process documentation (drafts, research notes)
- AI use for certain purposes (brainstorming, grammar) is permitted with disclosure
- AI use for generating submitted text is not permitted without explicit assignment design
The challenge is enforcement and consistency. Every faculty member interprets this differently. Some colleagues refuse to use any detection tools. Others rely on them heavily. Students get wildly different treatment depending on who teaches the section.
I’m curious how other institutions are approaching this. Is anyone getting it right? Or is everyone muddling through like we are?
Based on my conversations with colleagues across several institutions, everyone is muddling through. But some approaches work better than others.
The most effective policies I’ve seen:
- Focus on process over product. Require visible writing processes rather than trying to detect AI in final submissions.
- Redesign assessments. Oral defenses, in-class components, iterative drafts with feedback. These are harder to game with AI.
- Train faculty. The biggest inconsistency comes from uneven understanding of what the tools can and cannot do.
The institutions struggling most are the ones trying to use detection tools as the primary enforcement mechanism.
My university (im in the US) went through a similar evolution. they started with a blanket ban, then moved to “permitted with disclosure,” then added the process documentation requirement. each iteration was better than the last
what actually changed the conversation was when a senior professor’s published work got flagged by the university’s own detection tool. suddenly the false positive problem became very real to the administration
High school perspective: we’ve basically given up on detection tools and moved entirely to process-based assessment. more in-class writing, conferences about their work, portfolios showing development over time
its significantly more work for teachers but the outcomes are better for everyone. students who use ai to draft and then understand the material through revision are actually learning. students who copy-paste chatgpt cant explain their work in a conference. the assessment method IS the detection
@jonahHex99 The conference/oral defense approach is probably the most AI-resistant assessment there is. Hard to scale for large lecture courses though. I’ve been experimenting with peer review sessions where students must explain and defend their writing choices to classmates. Similar principle, more scalable.
@SophieB_92 The “professor got flagged” moment is so predictable and yet so effective at shifting institutional perspective. Nothing like personal experience to cut through abstract policy debates.