AI Literacy and Self-Perceived Cognitive Learning Outcomes Among University Students in AI-Integrated Courses: Associations with Instructor Feedback and AI Use Indicators

Abstract

Artificial intelligence (AI) is rapidly being integrated into university curricula, yet quantitative indicators of AI use reveal little about how learners use AI as a learning resource or what educational outcomes follow. This cross-sectional survey study of 212 university students enrolled in AI-integrated courses examined the associations of AI literacy, instructor feedback, and two single-item AI use indicators—the proportion of in-class AI use and total weekly AI use time—with self-perceived cognitive learning outcomes, measured across the six cognitive processes of the revised Bloom’s taxonomy. Confirmatory factor analyses supported multidimensional and higher-order structures, but the cognitive domains overlapped substantially (interfactor correlations up to 0.943; HTMT up to 0.946), so domain-level distinctions should be interpreted with caution. A regression model with the four predictors explained 46.3% of the variance in overall self-perceived cognitive learning outcomes (R2 = 0.463, adjusted R2 = 0.452). When all predictors were considered simultaneously, only AI literacy showed a significant positive association (B = 0.615, β = 0.593, 95% CI [0.475, 0.754], p < 0.001); the data did not provide evidence for independent associations of instructor feedback or the two AI use indicators, whose weaker associations may partly reflect their single-item measurement. AI literacy remained significantly associated with all six cognitive domains after Benjamini–Hochberg correction. These findings suggest—within the limits of a cross-sectional, self-report design—that the quantity of AI use and learners’ competency to understand, evaluate, and self-regulate AI use are empirically distinct indicators that universities should measure separately.