Linking perceived AI integration quality, pedagogical innovation, and subsequent student engagement across disciplinary course groups

Amid AI use in higher education, this study examines whether perceived pedagogical innovation predicts subsequent engagement beyond prior engagement, whether perceived AI integration quality remains associated with later engagement, and whether these associations vary across course groups. A multi-wave field study involved undergraduates enrolled in STEM, business and management, and humanities and social sciences courses. Engagement was assessed longitudinally, while perceived AI integration quality and pedagogical innovation were assessed during the study. Structural equation modeling incorporated autoregressive controls, bootstrapped indirect-association analysis, cluster-robust analyses, and descriptive multi-group comparisons. Perceived pedagogical innovation remained positively associated with subsequent engagement after baseline and proximal engagement were controlled. AI integration quality had no direct association with later engagement but showed a significant indirect association through pedagogical innovation. Because both constructs were assessed concurrently, this finding represents a theory-guided statistical indirect association rather than causal mediation. Group-specific estimates suggested that the relationship between AI integration quality and pedagogical innovation was stronger in STEM, whereas the relationship between pedagogical innovation and engagement was stronger in the other course groups. These patterns indicate preliminary contextual variation rather than established disciplinary differences. Overall, the findings provide a conservative longitudinal account of how AI-supported pedagogy relates to subsequent engagement.