• The activity was developed to address two interconnected challenges: the growing use of generative AI in take-home assignments, and students producing correct answers without genuinely understanding the reasoning behind them. Rather than preventing AI use, it changes what is being assessed entirely.
• The design turns assessment into a cycle of explaining → evaluating → receiving feedback → reflecting, shifting the emphasis from the final written product to the student's mathematical sense-making and communication.
• Grounded in Generative Learning Theory and the robust self-explanation effect documented in the science of learning, the activity prompts students to generate explanations of why each mathematical step is valid, rather than simply stating what it is, thereby supporting the formation of conceptual knowledge and more long-lasting learning.
• Peer assessment exposes students to the diversity of ways in which the same problem can be approached and communicated, with students reporting that they particularly valued seeing different approaches to the same mathematical problem.
• Students may use AI to investigate the mathematics, but must independently explain their reasoning on video — making genuine understanding impossible to fake.
• The approach has produced encouraging results: a 16% increase in pass rate, a drop in exam absenteeism from over 15% to 9%, and an increase in A and B grades.
• The core design is transferable beyond mathematics — any discipline where students can be asked to make their thinking visible and evaluate how others think can adopt the same assessment architecture.

‍