University Honors

The rapid advancement of artificial intelligence (AI) has transformed how students access and use educational information. While AI-generated summaries and podcast-style explanations have become increasingly popular, less is known about how these formats influence immediate comprehension, delayed retention, engagement, and perceived understanding. This study compared AI-generated text summaries, AI-generated audio summaries, and human-written text summaries among undergraduate students enrolled in a business course at the University of California, Riverside. Participants were randomly assigned to one of the three summary-format conditions and completed an immediate comprehension assessment (T1), a delayed retention assessment approximately four hours later (T2), and post-learning measures of engagement, cognitive load, and confidence. Results indicated no significant differences in immediate comprehension across summary formats. The predicted overall decline from T1 to T2 was not observed; instead, delayed scores were slightly higher than immediate scores, and item-level comparisons suggested that some T2 items were easier than their T1 counterparts. However, a significant time-by-format interaction showed that changes in quiz performance differed across conditions, with the AI-generated text condition showing a more favorable delayed-retention pattern than the human-written condition. Engagement and cognitive load did not significantly predict delayed retention after controlling for initial comprehension and summary format. Exploratory correlational analyses indicated that confidence and performance were positively related across all conditions, although delayed metacognitive accuracy appeared strongest in the human-written condition. These findings suggest that AI-generated text may support short-term delayed recall under some conditions, but conclusions about overall retention should be interpreted cautiously because of quiz-version differences. As AI-assisted learning technologies continue to evolve, understanding how format shapes retention and self-monitoring may help educators and developers design more effective educational tools.