A two-stage SEM–ANN analysis of digital attitude, cognitive engagement, and digital self-efficacy as determinants of learning effectiveness in the AI era

Although the rapid advancement of Artificial Intelligence (AI) has a great impact on education, students' learning effectiveness is still the ultimate goal of learners as well as educators. In this research, we aim to investigate the impact of the core drivers of learning effectiveness in the AI era. In particular, we explore the interplay among digital self-efficacy, digital attitude, and cognitive engagement, and how they contribute to learning effectiveness in a digitized learning process. In this empirical study, a dual analytical procedure incorporating structural equation modelling (SEM) and artificial neural network (ANN) is applied to examine the associations between the drivers and to analyze the predictive power of the proposed model. The results confirmed the significance of all direct and indirect associations between variables. In particular, the mediating role of cognitive engagement on the relationship between digital attitude toward educational AI technologies and learning effectiveness, as well as the moderating effect of AI self-efficacy on the impact of attitude on learning engagement were demonstrated. The predictive power of the proposed model was analyzed by integrating ANN to take into account the model's nonlinearity and complexity. Not only this approach enables us to assess the predictive power, but also rank the importance of the independent variables' predictive effects on the dependent variable. The results found digital self-efficacy as the most crucial driver of students' perceived learning effectiveness, followed by cognitive engagement and digital attitude toward educational AI technologies.