Generative AI, social normalisation, and the governance gap: a cross-disciplinary study of AI adoption and sustainability in UK medical education
TL;DR
Findings tentatively suggest that GenAI use is becoming socially normalised and potentially habitual among students, with TPB providing only partial explanatory power.
Abstract
Generative artificial intelligence (GenAI) has become embedded in higher education, yet its adoption has outpaced institutional governance and critical reflection particularly in medical and health science programmes. This exploratory study addresses two underexamined dimensions: the psychological and social factors shaping student adoption, and the environmental footprint of AI-intensive learning cultures. A cross-sectional survey ( N = 80) examined medical and health science students at the University of Birmingham using the Theory of Planned Behaviour (TPB) and Geels' Multi-Level Perspective (MLP). A supplementary wave (Wave 2) introduced environmental awareness and habit measures. A focused document analysis examined AI governance policies at five UK universities. Literature was identified through searches of PubMed, ERIC, Web of Science, and Google Scholar (2019–2026) using terms including 'generative AI', 'medical education', 'AI governance', and 'sustainability'. GenAI use was near-universal (96.3%), yet only 40.0% had received formal training. Significant associations were identified between programme of study and disclosure behaviour ( p = .003) and between training and trust ( p = .004). Negative correlations were found between frequency of use and perceived behavioural control ( r = − .409, p < .001) and behavioural intention (rs = − .514, p < .001); these findings are exploratory given the sample size. No student raised environmental concerns unprompted, and systematic document review found no AI policy at any of the five institutions referenced sustainability considerations. Findings tentatively suggest that GenAI use is becoming socially normalised and potentially habitual among students, with TPB providing only partial explanatory power. A governance gap was identified: AI policies address academic integrity but not environmental sustainability, despite separate institutional sustainability commitments existing at all five institutions. Structured AI literacy curricula and sustainability-aware governance are recommended. Findings should be interpreted cautiously given the exploratory, single-institution design.