Student GenAI use under changing institutional policies: a mixed-methods case study in undergraduate physics

The rise of generative artificial intelligence (GenAI) in higher education and the resulting institutional responses have prompted research to support evidence-based policy decisions. As policies change to adjust to this disruptive technology, it demands agility from educators. The impacts of these policy changes on student behaviours and outcomes are currently poorly understood. This study examines how students used GenAI across three assessments in an undergraduate physics unit after assessment redesign forced by institutional policy that permitted purpose-specific and open use. Using a mixed-methods approach, student declarations of GenAI use were coded into four functional categories and mapped to the Substitution, Augmentation, Modification, and Redefinition (SAMR) framework. Learning and Formatting dominated early use and aligned with substitution and augmentation for a presentation assessment type. Later assessments showed a marked growth in Feedback uses, which spanned all SAMR levels and became the most common pattern. ‘Doing Task’ uses appeared in smaller numbers but illustrated modification and redefinition when GenAI generated step-by-step solutions for unfamiliar problems. Across all assessments, GenAI users achieved slightly higher scores, although differences were not statistically significant, so this trend should be interpreted as a tentative observation. Findings suggest that assessment design, which rewards understanding and requires student judgement, aligns with the theoretical markers of evaluative judgement. While this study did not directly measure internal cognitive states, the observed change in functional behaviour, specifically the transition toward iterative feedback loops, suggests that GenAI could serve as a catalyst for learning. This study provides empirical evidence on how assessment redesign based on policy shifts influences GenAI use patterns and proposes design principles for authentic assessment in science, technology, engineering, and mathematics (STEM) contexts. It highlights authentic assessment and constructive alignment as guiding principles for sustainable policy and pedagogy in the GenAI era, ensuring coherence among learning outcomes, activities, and assessment.