Reconceptualizing Student Agency in AI-Supported STEM Learning: A Qualitative Study of Autonomy, Regulation, and Dependency

This study explores student agency in Artificial Intelligence (AI)-supported STEM learning environments, addressing a critical gap in existing literature that predominantly focuses on learning outcomes rather than learner-centered processes. Drawing on an interpretive qualitative approach, this research investigates how students experience autonomy, self-regulation, and decision-making when interacting with AI technologies in STEM education. Data were collected from 15 participants engaged in AI-supported learning through in-depth semi-structured interviews, supported by observations and document analysis. The data were analyzed using thematic analysis to identify recurring patterns and meanings related to student agency. The findings reveal that student agency is a dynamic and multidimensional construct shaped by the interplay between technological affordances and learner engagement. Four major themes emerged: enhanced autonomy, development of self-regulated learning, negotiated decision-making, and ambivalent dependency on AI. While AI technologies provide adaptive support that empowers students to take control of their learning, they also introduce the risk of over-reliance, which may reduce cognitive engagement. This study contributes to the theoretical advancement of student agency by conceptualizing it as a spectrum rather than a fixed attribute, highlighting the dual role of AI as both an enabler and a constraint. The findings offer important pedagogical implications for designing AI-supported STEM learning environments that promote active, reflective, and responsible learning. Future research is recommended to explore this phenomenon across diverse contexts and through longitudinal designs.