This qualitative case study examines undergraduate students’ engagement with generative artificial intelligence (GenAI) in academic learning in an English-medium university setting. This study employs self-determination theory (SDT) as the primary interpretive framework, treating technology acceptance perceptions (e.g., usefulness and ease of use) as descriptive cues rather than explanatory constructs. Data from 23 semi-structured interviews were analyzed using reflexive thematic analysis, complemented by epistemic network analysis (ENA) to examine the structural relationships among themes in students’ discourse, with automated coding validated against a manually coded subset. Students frequently described GenAI as supporting efficiency and conceptual understanding; however, their accounts revealed persistent tensions concerning creativity, trust, and academic integrity. The ENA results showed that these concerns were systematically interconnected: discussions of learning support consistently co-occurred with verification practices, reflecting a “trust-but-verify” repertoire through which students calibrated their reliance on AI while maintaining epistemic control. Beyond instrumental evaluations, students’ narratives highlighted broader value- and norm-related considerations, including algorithmic bias, environmental sustainability, and the positioning of AI within human–teacher learning networks. Overall, the findings suggest that students’ engagement with GenAI is best understood as a motivated and socially situated learning practice shaped by the negotiation of competence, autonomy, and relatedness. Pedagogically, the results support a shift from prohibition-oriented responses to transparent institutional guidance, autonomy-supportive scaffolding of verification practices and AI literacy, and process-oriented assessment designs that make students’ reasoning visible. Learning analytics approaches, such as ENA, may further assist educators in examining how these practices become integrated into students’ learning processes.
Isaeva et al. (Fri,) studied this question.