The integration of generative artificial intelligence (GenAI) tools into higher education presents both opportunities and challenges, particularly in assessment contexts. This study investigates the impact of GenAI usage on student performance in a graduate-level time series analysis course, focusing on multiple choice exams. Through a controlled experiment involving two student cohorts, each having access to GenAI tools during one of two midterms, we analyze performance outcomes and engagement behaviors using multimodal screen recordings. Statistical analyses reveal that GenAI access correlates with improved scores but only when students are adequately prepared to use the tools effectively. Frequent or prolonged GenAI usage alone did not predict better outcomes, highlighting the importance of AI literacy. Additionally, traditional course materials remained strong predictors of performance across both cohorts. These findings suggest that GenAI can enhance learning when integrated thoughtfully and accompanied by instructional support. The study contributes to the evolving discourse on AI in education by offering empirical insights into its role in assessments and proposing implications for instructional design and policy. Funding: This research has been supported by the Master of Science in Analytics degree at Georgia Tech but we don’t have a specific funding other than this internal support. Supplemental Material: The online appendix is available at https://doi.org/10.1287/ited.2025.0166 .
Serban et al. (Mon,) studied this question.