From Expectations to Measured Pragmatism: A Pre- and Post-Experience Study of Student Engagement in AI-Supported Academic Exams

Generative AI (GenAI) is transforming higher education assessments, yet empirical research on students’ lived experiences with GenAI during graded, time-constrained classroom assessments remains scarce. This study investigates how direct experience with GenAI in examinations shapes student perceptions of learning, metacognition, and engagement. Drawing on self-regulated learning research and cognitive load theory, we employed a retrospective pre–post design to analyze qualitative reflections and quantitative data from 90 undergraduate computer science and engineering students. Our qualitative analysis suggests a complex recalibration from idealized expectations of efficiency toward what may be described as a state of measured pragmatism. Interpretive analysis of Post-experience reflections indicates that direct practical engagement appeared to make students more conscious of the need for metacognitive engagement, with a focus on real-time output verification and the restrictive role of time pressure. Concerns regarding assessment authenticity and fairness emerged only after direct engagement. Quantitative results show that although 68.5% preferred the GenAI format, this preference did not correlate significantly with academic performance (r = 0.014, p = 0.89). Those findings suggest that student engagement is driven by pedagogical and professional relevance rather than grade improvement alone. Overall, the findings underscore the need for assessment designs that balance cognitive support with active student monitoring and responsibility.