Generative AI has dramatically accelerated assessment content creation. However, producing a plausible question is not the same as crafting a valid, fair, and effective assessment item.
For assessment professionals, the core challenge is moving from AI-generated questions to trustworthy quality. AI-produced items can appear fluent and convincing while hiding issues with accuracy, ambiguity, distractor quality, learning-objective alignment, accessibility, or bias.
In this session, Miguel Prieto, VP of Corporate Strategy at Open Assessment Technologies (OAT), and Renzo Degiovanni, Senior Research Scientist at the Luxembourg Institute of Science and Technology (LIST*), will present a research-informed approach to AI-assisted validation. They will demonstrate how independent AI validators can evaluate generated items against structured educational quality criteria before human review, creating a robust layer of quality assurance rather than relying on a single model to both generate and evaluate its own content.
The discussion will examine what assessment quality truly means in an AI-enabled workflow, how validation supports key areas such as educational alignment, question quality, effectiveness, accessibility, and fairness, and why human expertise remains central to final decisions.
The ultimate goal goes beyond faster authoring. It is achieving trustworthy assessment by design.
