AI-Adaptive Learning Systems and Student Engagement: A PLS-SEM Validation of the AI-Adaptive Learning Engagement Model (AALEM)

Artificial intelligence (AI)-adaptive learning platforms are increasingly embedded in higher education, yet the psychological mechanisms that translate platform quality into sustained student engagement and downstream academic performance remain incompletely modeled. This study proposes and validates the AI-Adaptive Learning Engagement Model (AALEM), a conceptual framework that integrates perceived usefulness and self-efficacy as sequential and interacting mechanisms linking AI-adaptive learning quality to student engagement and academic performance. Because no primary human-subjects data collection was undertaken, the model is validated using a calibrated synthetic dataset (N = 436) generated to instantiate plausible population-level structural relationships for demonstration and methodological illustration. Partial least squares structural equation modeling (PLS-SEM) confirms that AI-adaptive learning quality strongly predicts perceived usefulness (β = 0.49, p < .001), which in turn predicts self-efficacy (β = 0.40, p < .001) and, jointly with self-efficacy, predicts student engagement (β = 0.34 and β = 0.35 respectively, both p < .001), explaining 34.2% of its variance. A significant perceived usefulness × self-efficacy interaction (β = 0.11, p = .003) indicates that self-efficacy amplifies the usefulness–engagement relationship. Student engagement subsequently predicts academic performance (β = 0.40, p < .001). All constructs demonstrate acceptable reliability (Cronbach's α > 0.85, composite reliability > 0.90) and discriminant validity via the Fornell–Larcker criterion and heterotrait–monotrait ratio. Cross-checks of the measurement and structural estimates using regression procedures analogous to SPSS, JASP, and R corroborate the SmartPLS-style path estimates. The results offer a statistically coherent, reproducible template for how AI-adaptive learning quality cultivates engagement through usefulness perceptions and self-efficacy, with implications for instructional design, platform development, and future empirical validation with real student samples.