AI-Era Assessment Must Certify Human Judgement

Generative AI has made one educational mistake impossible to ignore: a polished submission is not the same thing as demonstrated learning. A learner can now produce fluent prose, a plausible lesson plan, a sermon outline, computer code or a policy brief without possessing all the judgement that the finished object appears to display. That does not make assessment impossible. It makes the purpose of assessment clearer.

Assessment exists to produce valid evidence about what a person knows, can do, can judge and can be trusted to carry into new situations. It should support learning, guide feedback, protect standards and certify competence. If an assessment records only the final artefact, it may certify the combined capability of learner, model, template and editor while saying little about the learner's own formation.

The answer is not universal surveillance, unreliable detection or a nostalgic return to one examination format. It is assessment that makes consequential human judgement visible: what the learner can explain, perform, verify, defend, transfer and repair. AI may be permitted within that design, but the institution must remain able to distinguish independent competence, assisted competence and governance competence.

On 28 October 2026, UNESCO reported on an international conference held in Almaty on 26–27 October. More than 70 policymakers, university leaders, experts and cultural professionals discussed higher education, qualification recognition, flexible pathways, arts, sustainability and AI. The…