From Policing to Design: A Qualitative Multisite Study of Generative Artificial Intelligence and SDG 4 in Higher Education

Generative artificial intelligence (AI) is now embedded in the everyday practice of higher education. This qualitative, multisite study examines how university faculty perceive where generative AI advances or threatens Sustainable Development Goal (SDG) 4, which commits education systems to inclusive, equitable, high-quality learning across the lifespan. We conducted semi-structured interviews and focus groups with 36 academics across three universities, complemented by document and artefact analysis. Guided by critical pedagogy, sociomateriality, and technological pedagogical content knowledge (TPACK), we used reflexive thematic analysis to identify five cross-cutting themes. Faculty reported inclusion gains through rapid accessibility work, multilingual support, and differentiated feedback, alongside risks that undermine SDG 4, including bias, expansion of surveillance, unreliable outputs, paywalled access advantages, and work intensification. Assessment emerged as the decisive site of tension: staff rejected detection-led policing and favoured designs that reward process, critique, and provenance. We offer a practical framework, aligned to SDG 4 targets, that translates these insights into commitments, indicators, and a 12-month programme plan. The sector should move beyond bans and hype. Responsible adoption requires equity by design, assessment redesign, institutionally guaranteed access, transparent evaluation, and protected time for teacher development.