Stop Chasing AI. Start Fixing Higher Ed’s Broken Assessment.
By Autar Kaw
For the past two years, universities have been getting completely bent out of shape about AI and assessment, pouring enormous energy into a never-ending cat-and-mouse game. New tools show up, new detection schemes follow, and the cycle repeats. Meanwhile, we are losing the plot: assessment should drive learning, not policing.
AI is not the root cause of higher education’s crisis; it has merely held up a mirror to a 35-year-old rot we have refused to address.
The public is questioning the value of a degree because universities operate as if they can continue business as usual. In typical large courses, 500 students sit in a room staring at PowerPoint slides. They are evaluated three or four times a semester on ill-formed multiple-choice exams, often unproctored and online, that make up their entire grade. What critical thinking takes place here?
Source: Gemeni Generated
Even from a measurement standpoint, the system is defenseless. Question banks are recycled without basic checks for reliability, validity, difficulty, or discrimination index. When our primary mode of instruction is passive listening and our main evaluation is unproctored recall, we are testing compliance, not mastery. AI didn’t break this system; it just made the cracks impossible to ignore.
We don’t need complex, esoteric solutions like surveillance software or prompt policies. We need alignment, priority, and courage.
Here is a pragmatic reset: The 90/90 Remedy.
-
Allocate 90% of grades to in-class, proctored assessments in 90% of courses. Bring back monitored exams and blue books. Put students in environments where they demonstrate thinking in real time. Crucially, align these exams with higher-order learning objectives, not procedural recall.
-
Decouple homework from grades. Keep assigned practice, but remove high-stakes grading. Learning thrives in low-stakes environments; heavy grading invites gaming the system.
-
Move terminal content online to reclaim class time. Saying “this takes too much class time” is an excuse, not a constraint. Identify topics with minimal impact on future courses and move them online via focused videos, worked examples, and formative practice. Reclaim live hours for authentic assessment.
-
Audit test metrics. Faculty must do the basic psychometric work to ensure exams actually differentiate deep understanding from wild guessing.
This approach centers learning over compliance and removes the incentive to cheat.
Why isn’t this widely adopted? Because it disrupts comfort, challenges administrative insistence on mass-lecture economics, and dispels the illusion that complexity equals rigor. It is easier to chase detection tools than to rethink course design.
Stop cramming content and chasing AI misuse forever. Curate your courses, move what you can online, and assess what truly matters—in class, with integrity.