Policy Ambiguity and Academic Integrity: Student Experiences of AI Governance in Higher Education

This study examines undergraduate students’ experiences of academic integrity governance in AI-mediated higher education, focusing on how institutional policies are interpreted and enacted in everyday contexts. Using qualitative data from a student focus group and analysis informed by policy enactment and sociomaterial perspectives, the study investigates how learners make sense of artificial intelligence (AI) rules in routine academic work. The findings show that integrity challenges associated with AI are driven less by intentional misconduct than by policy ambiguity, abstraction, and uneven enactment across modules and instructors. These patterns were consistent across themes, where students expressed a strong desire to act ethically, yet struggled to translate high-level policy statements into actionable guidance for common tasks such as drafting, language support, and revision. In the absence of assessment-level clarity, academic integrity was frequently reframed as risk management. Decisions were shaped by perceived detectability, instructor tolerance, peer norms, and potential consequences. These conditions encouraged adaptive behaviours including strategic concealment, selective disclosure, avoidance of AI tools, and heightened anxiety and emotional fatigue. Some students also avoided AI despite recognising potential learning benefits. The study conceptualises integrity as a situated, enacted practice produced through interactions among policy texts, assessment design, technological affordances, and student sense-making. It shows how policy ambiguity mediates between institutional governance and student enactment, redistributing ethical labour and generating unintended behavioural and emotional outcomes. The paper argues for human-centred AI governance that prioritises assessment-level guidance, learning-oriented integrity framing, and integrated AI literacy to support confident, transparent, and ethically grounded engagement with AI.