Symmetrical and asymmetrical insights into student adoption of AI-enhanced digital educational systems

Background Based on the stimulus–organism–response (SOR) model and the technology acceptance model (TAM), this study investigates how the functional and experiential features of artificial intelligence (AI)-enhanced educational systems influence users’ behavioral intention through perceived usefulness and perceived ease of use. Methods A multi-method approach integrating partial least squares structural equation modelling (PLS-SEM), artificial neural networks (ANN) and necessary condition analysis (NCA) was employed. Results The results of the PLS-SEM analysis indicate that both perceived usefulness and perceived ease of use have a significant positive impact on behavioral intention, with perceived ease of use exerting the strongest effect. Furthermore, the effects of functional and experiential features on perceived usefulness and perceived ease of use are asymmetric. ANN analysis confirms the robustness of the findings and identifies perceived ease of use as the most influential determinant of behavioral intention. In contrast, the NCA analysis indicates that perceived usefulness is the only statistically significant necessary condition for achieving higher levels of behavioral intention. Although this necessary relationship is statistically significant, its effect size is small, suggesting that its actual constraining effect is relatively limited. The bottleneck analysis further reveals the constraining characteristics of different stages in the formation of behavioral intention. Discussion By integrating symmetric, asymmetric and necessity-based analytical perspectives, this study provides a more comprehensive understanding of user acceptance behavior in AI-enhanced educational environments.