What are medical students really doing with GenAI in their self-study? An epistemic entanglement framework approach

The rise of GenAI in health professions education has led to debates over its promise to support student learning and potential to encourage cognitive offloading. However, there is little empirical work examining how medical students actually use GenAI in their everyday self-study practices. To address this gap, this study draws on the Epistemic Entanglement Framework (EEF) to examine how knowledge is co-constructed between undergraduate medical students and GenAI chatbots.

We adopted a qualitative study design. Participants were clinical year medical students who commenced their medical training prior to the widespread availability of GenAI tools. Data were collected through semi-structured interviews, supplemented by participants' actual chat interactions with GenAI tools. Interviews and screen-based interactions were video-recorded. Data were analysed thematically, with iterative coding to identify patterns across participants' accounts. Coding analysis was abductive and informed by the EEF.

Twenty clinical year medical students participated in the study. Three key themes emerged. First, GenAI use was embedded within a learning network where students sometimes interacted with it as a human actor. Second, knowledge was co-constructed between and cognition was distributed across the learning network with varying levels of epistemic co-agency. Third, participants highlighted GenAI's unreliability as a learning resource especially in relation to accuracy and contextual understanding.

Medical students engaged with GenAI in complex and sophisticated ways within a dynamic learning network. The integration of GenAI into this network both gave rise to new learning practices, such as working with information that had been pre-processed by GenAI, while leading to the demise of others, such as synthesizing information from multiple sources. Thus, the GenAI usage has reshaped self-study behaviours by redirecting and redistributing cognitive effort, thereby shifting patterns of epistemic entanglement.