Can AI Prove Understanding? Rethinking Assessment in the Age of Generative AI

As generative AI reshapes how learners produce answers, traditional approaches to assessment are increasingly under question.

This session examines how understanding can be more meaningfully evidenced through application in unfamiliar contexts. Drawing on research in transfer, productive failure, and motivation, alongside a recent pilot, we explore how AI-mediated simulations and role-plays create conditions where learners must act, decide, and justify under uncertainty.

• The difference between performance and understanding in AI-supported learning

• How transfer and productive failure contribute to deeper evidence of learning

• The role of uncertainty and novel contexts in strengthening assessment

• Designing AI-supported assessment that is ethical, motivating, and psychologically safe

Cohen Ambrose is Course Director at the Digital Learning Institute and an educational developer, learning experience designer, and researcher with over 15 years’ experience across higher education and professional learning.

His work focuses on digital pedagogy, learning theory, and the evolving role of AI in education. He has led the design and delivery of blended and online programmes, supported educators in developing effective learning experiences, and contributed to curriculum and programme reform across both academic and workplace contexts.

Cohen’s research and practice centre on how learning design can better support performance, transfer, and learning in the flow of work.