Cohort-level growth modeling of ethical engagement with artificial intelligence in education

Artificial intelligence is increasingly embedded in higher education, yet limited evidence explains how students’ ethical engagement develops over time. Existing research is largely cross-sectional and often treats ethical awareness, reflection, and responsible use as a unified competence. This study examined cohort-level changes in AI Ethics Awareness (AEAw), Responsible AI Use (RAIU), and AI-Driven Reflection on Decision-Making (AIRD), including relationships among their initial levels and growth rates. Transformative Learning Theory provided a heuristic interpretive lens, while a socio-technical perspective situated ethical engagement within institutional, pedagogical, cultural, and technological conditions. A rotating-cohort longitudinal design involved 350 undergraduate computer science students from two Nigerian public universities across three annual measurement occasions. Confirmatory factor analysis evaluated the theoretically specified measurement structure, while univariate and multivariate latent growth curve models estimated construct-specific trajectories and their interrelationships. Results revealed statistically significant and practically substantial cohort-level increases in AEAw, RAIU, and AIRD. However, relationships among their initial levels and growth factors were predominantly weak and mixed, providing little evidence of a unified developmental sequence or compensatory-growth process. Theoretically, our findings refine Transformative Learning Theory by showing that awareness, reflection, and self-perceived responsible practice may develop asynchronously rather than through a deterministic progression from knowledge to action. Practically, universities should complement ethics-awareness initiatives with explicit AI-use policies, authentic ethical tasks, disclosure requirements, behavioural assessment, and institutionally supported opportunities to translate reflection into responsible practice. These findings represent aggregate cohort trends and do not establish individual development or causal instructional effects.