White Paper March 2026
Are we assessing learning – or just grading production?
Why assessment design, more than AI policy, decides whether students learn or just produce.
This is relevant to you because:
- You suspect your assessments reward polished output more than demonstrated understanding of the material.
- You've written an AI policy and found it confuses honest students more than it deters others.
- You want evidence that assessment design shapes AI use more than any policy does.
- You're weighing whether verification moments like short oral checks are worth the operational cost.
Abstract
Assessment defines the curriculum students actually study. This paper argues that when assessments reward production rather than understanding, students optimise for production, and that generative AI has made the gap impossible to ignore.
Drawing on Wonkhe research finding that 38% of students had submitted work they could not fully explain, it argues design matters more than policy. Students who expect to discuss their work use AI to test their reasoning; those submitting a one-off artefact use it on autopilot. The paper sets out four responses: build in verification moments, replace blanket bans with stage-specific guidance, protect equity for students who rely on AI as assistive technology, and close feedback loops.