Universities are scrambling to fix assessment validity as AI use explodes in classrooms. Different schools have rolled out mixed AI policies—some ban generative AI outright, others allow it selectively, and some give free rein. The problem? Confused students and teachers. They don’t know what “appropriate use” even means. Cheating fears are skyrocketing.
The University of Leeds launched a “traffic light” system: red means no AI, orange allows some use, green lets students use AI freely. The University of Melbourne requires students to declare AI use on their assignments. These moves aim to protect "assessment validity" — making sure tests actually measure student skills, not AI skills or AI-paid-for work.
But a new peer-reviewed paper argues these rules aren’t enough. Clear instructions (called discursive changes) rely on students playing by the rules. That’s fragile because AI cheating is getting harder to detect. 
The researchers say schools need structural changes—redesigning assessments themselves. For example, swapping one-off essays for in-class progressive writing, or adding live discussions about student choices. These embed validity by design, not just by hoping students follow rules.
“If assessment is to retain its function as a meaningful claim about student capability, it must be rethought at the level of design.”
Discursive tweaks won’t cut it in an AI world, the paper warns. Structural redesign is the only way to keep assessments fair and valid. The challenge is big and will hit disciplines differently. But ignoring it risks turning university degrees into meaningless certificates.
Read the full paper here.
Also check out coverage on the rise of AI cheating by The Guardian and others.
