Understanding students’ learning attitudes and behaviors toward generative AI: an integrated social cognitive theory and diffusion of innovation perspective

Purpose This study aims to examine how psychosocial factors and innovation characteristics shape students’ learning attitudes and behaviours toward generative artificial intelligence (GenAI) in higher education by integrating social cognitive theory (SCT) and diffusion of innovation (DoI) perspectives. Design/methodology/approach A cross-sectional survey design was used. Data were collected from 403 undergraduate and graduate students and analysed using partial least squares structural equation modelling (PLS-SEM) to test the proposed structural and mediation models. Findings The findings indicate that self-efficacy, observational learning, and relative advantage have significant positive effects on students’ attitudes toward Gen AI, whereas complexity and outcome expectancy exhibit more limited influence. Furthermore, students’ attitudes partially mediated several relationships between SCT and DoI factors and GenAI-based learning behaviour. These results suggest that students’ engagement with GenAI is driven by internal evaluative processes involving confidence, social learning experiences, and perceived benefits rather than the mere availability of AI technologies. Practical implications This study highlights the importance of strengthening students’ self-efficacy through practice-based training and early mentoring, supported by clear models of GenAI use. Higher education institutions should prioritise tangible academic benefits, including improved conceptual understanding and learning efficiency, when implementing AI-supported learning strategies. Originality/value This study contributes to the GenAI adoption literature by integrating SCT and DoI to explain how psychosocial factors and innovation characteristics shape students’ learning behaviours through attitude. It highlights attitude as an evaluative mechanism and identifies self-efficacy, observational learning, and relative advantage as the key drivers of GenAI-based learning behavior.