The moderating effect of academic self-efficacy on artificial intelligence-assisted learning and academic achievement
Artificial intelligence (AI) is increasingly reshaping higher education, yet the psychological mechanisms that determine when and for whom AI-assisted learning translates into academic success remain insufficiently understood, particularly within developing educational contexts. This study examined the relationship between artificial intelligence-assisted learning and academic achievement among university students, with academic self-efficacy tested as a moderating variable, drawing on Bandura's Social Cognitive Theory. A quantitative, cross-sectional survey design was employed, and data were collected from 477 undergraduate and postgraduate students enrolled in public and private universities across Punjab and Sindh, Pakistan, using a convenience sampling technique. A structured questionnaire comprising validated scales measuring artificial intelligence-assisted learning, academic self-efficacy, and academic achievement was administered via Google Forms, and data were analyzed using IBM SPSS Version 27. Pearson correlation analysis revealed a strong, statistically significant positive relationship between artificial intelligence-assisted learning and academic achievement (r = .684, p < .001). Multiple linear regression confirmed that artificial intelligence-assisted learning significantly predicted academic achievement, accounting for 49.9% of the variance (R² = .499, F(1, 475) = 470.836, p < .001). An independent samples t-test showed that students with high academic self-efficacy achieved significantly higher academic outcomes than those with low academic self-efficacy (t(475) = -14.126, p < .001). Most notably, hierarchical multiple regression analysis demonstrated that academic self-efficacy significantly moderated the relationship between artificial intelligence-assisted learning and academic achievement (ΔR² = .029, ΔF(1, 473) = 12.87, p < .001), with the interaction term emerging as a significant predictor (B = .156, β = .142, p < .001), indicating that the positive effect of artificial intelligence-assisted learning on academic achievement was stronger among students with higher academic self-efficacy. These findings extend Social Cognitive Theory into AI-mediated educational settings and suggest that the effectiveness of AI-assisted learning technologies is conditional upon students' belief in their own academic capabilities. The study offers practical implications for higher education institutions seeking to maximize the benefits of AI-based learning tools by simultaneously strengthening students' academic self-efficacy.