Harnessing responsible and sustainable AI adoption in higher education: Empirical evidence from the Netherlands

Artificial intelligence can advance the Sustainable Development Goals, yet its societal impact depends on equitable, responsible, context-sensitive adoption. Across higher education, fragmented or restrictive policies, limited stakeholder engagement, and unclear guidance can suppress legitimate use and widen inequities. This article synthesises two empirical studies of ChatGPT adoption among Dutch higher-education students. Study 1, an eight-month longitudinal panel (n = 222), shows that declining use is predicted by reduced trust, lower perceived behavioural control, and heightened emotional creepiness. Study 2, a cross-sectional survey (n = 355) extending meta-UTAUT, finds that anthropomorphism, trust, and design novelty shape attitudes and intentions, while institutional policy negatively moderates the intention-use link, indicating that restrictive governance can deter engagement. The evidence supports SDG 4 (Quality Education), SDG 10 (Reduced Inequalities), and SDG 16 (Strong Institutions) and yields actionable recommendations to implement transparent, proportionate rules, build targeted AI literacy, minimise emotional discomfort, and embed safeguards that limit misuse. Grounded in Responsible Research and Innovation, the article positions universities as settings for testing governance models that balance innovation with educational integrity and inclusion, and unveils an impact pathway from institutional levers to short-, medium-, and long-term societal outcomes, informing policymakers and institutional leaders on responsible AI integration.