A study out of Mount Saint Vincent University last week put a sentence to something I've heard in every faculty room I've been in for two years. A participant told the researchers, "I feel like a detective, not a teacher." The study surveyed 53 faculty and ran focus groups with a dozen more, and its central finding is that instructors are now assessing student work through suspicion rather than through evidence.
I recognize the position. You read a paragraph that's competent and slightly airless, and the question that arrives isn't "is this good" but "did they write it." Those are different questions and only one of them is teaching. Once the second question is in your head it doesn't leave, and it changes how you read everything else in the pile, including the honest work.
The scale of the shift is documented. The Digital Education Council's global survey in July drew more than 45,000 responses across 35 countries, including 27,000 students and 18,000 faculty. Eighty-eight percent of students use AI in their learning. Seventy-seven percent of faculty use it in teaching, up 16 points in a year. So the tool is on both sides of the desk, which is worth stating because a lot of the public conversation still imagines otherwise.
The assessment findings are the ones I'd put in front of a curriculum committee. Fifty-seven percent of students say assessments lack adequate guidance on AI use. Only 28 percent feel assessments reflect the AI-enabled skills they'll need at work. Twenty-four percent report having no permission to use AI in assessments at all, and in the US and Canada that figure rises to 38 percent.
The regional numbers are the part I didn't expect. Eighty-one percent of North American students worry AI is making their learning too shallow, against 66 percent globally. Seventy-three percent worry about classmates misusing it, against 60 percent globally. And faculty intent to use AI in the US and Canada fell nine points, from 76 to 67 percent, while the global trend moved the other way. Students here are more anxious about depth than students elsewhere, and their instructors are pulling back.
I don't think those two facts are unrelated. A student who's told nothing about what's permitted, in a course where the instructor is quietly withdrawing from the tool, will assume the worst and use it anyway. That's not a discipline problem. It's a design problem, and the design is ours.
The institutional response is moving faster than I expected. AACSB's framework for AI in business education went from 26 participating schools last July to 84 this July. The July update's emphasis is on AI literacy across all pathways rather than in a specialist track, and on responsible use as infrastructure rather than a standalone ethics module. That's the right shape. Whether it reaches a 300-person introductory course is a separate matter.
Which is the honest constraint. Louis Volante made the argument in Policy Options in June that assessment redesign in Canada is blocked by resources rather than by pedagogy. Oral exams and portfolios work in a graduate seminar of fifteen. They don't work in a first-year lecture of five hundred, and no amount of principled writing about authentic assessment changes the arithmetic. I'd add that the people asked to redesign are usually the ones with the heaviest teaching loads.
What I've settled on for my own courses is narrow. I say what's permitted, in writing, per assignment, in language a nineteen-year-old can act on. I assess more of the thinking in the room, where I can see it happen. And I've stopped trying to detect, because detection is a race I lose while spending my attention on the wrong thing. If a student hands me work that isn't theirs, the cost lands on them in a conversation later, and it lands harder than a flagged report would.
I'm not confident this is right. I am confident that reading a stack of essays as evidence to be examined rather than work to be responded to will make anyone worse at the job, and will make students feel it before the instructor notices.
Sources
- Academics lacking clear AI policy guidance for classrooms (HRD Canada, on the Mount Saint Vincent University study, August 2026): hcamag.com
- Global AI in Higher Education Survey 2026 (Digital Education Council, July 2026): edtechinnovationhub.com
- A Framework for Artificial Intelligence in Business Education, July 2026 update (AACSB): aacsb.edu
- Louis Volante on AI and university assessment (Policy Options, June 2026): policyoptions.irpp.org