yangyule data three time

This dataset originates from a three-wave longitudinal study examining the continuous adoption of AI learning tools among physical education majors in China. Grounded in the Technology Acceptance Model (TAM), the study investigates how the <b>Quality of AI Personalized Feedback (QAPF)</b> drives learners’ <b>Perceived Ease of Use (PEOU)</b>, <b>Perceived Usefulness (PU)</b>, <b>Attitude Toward Use (ATU)</b>, <b>Behavioral Intention to Use (BIU)</b>, and ultimately <b>Actual Use Behavior (AUB)</b>.Data were collected at three time points: T1 (QAPF), T2 (PEOU, PU, ATU, BIU), and T3 (AUB). A total of <b>511 matched valid responses</b> from physical education majors (≥18 years) were retained after rigorous longitudinal tracking. Structural Equation Modeling (SEM) and Bootstrap analyses (5,000 resamples) were employed to examine multiple sequential mediation pathways.Results demonstrated that QAPF significantly enhanced both PEOU and PU, with PEOU further predicting PU and ATU, while ATU and BIU played critical mediating roles in linking cognitive evaluations to actual behavioral outcomes. All hypothesized chain mediation pathways were statistically significant, providing robust longitudinal evidence that high-quality personalized AI feedback promotes sustained engagement and stable actual usage behavior.This dataset provides empirical support for extending TAM toward experience-focused and longitudinal explanations of AI tool adoption in educational settings, especially in feedback-intensive learning contexts such as physical education. It also offers practical implications for optimizing AI learning system design with a focus on personalization, actionability, and adaptive feedback mechanisms.