Generative AI and moral deskilling in vocational education: A dialectical agency perspective

Abstract

Purpose – This study develops a Dialectical Agency framework to theorise conditions under which artificial intelligence (AI) may augment or erode judgment in vocational education. It addresses moral deskilling when AI-mediated activity displaces, rather than scaffolds, practices through which judgment, ethical discernment, and responsibility are cultivated. Methods – The study employs a literature-based analysis integrating Ellul’s account of technique, Ihde’s postphenomenology of human–technology relations, Lund and Vestøl’s relational account of structure and practice, and Vallor’s account of moral skill. Studies from vocational and professional contexts are examined according to evidential function and domain proximity. Findings – The framework conceptualises AI-mediated skill formation through two poles: S1, the structural configuration of AI use, and S2, the learner’s mode of engagement. Their interaction constitutes a tension–negation–transformation mechanism through which AI may operate as a scaffold for judgment when verification and reflection are institutionally supported, or as a substitute when AI outputs are accepted without evaluation. Four propositions are derived; the two configurations in which structural pressure and engagement mode diverge remain empirically underdetermined. Research implications – The framework calls for research assessing practical judgment and ethical discernment in vocational settings, while pointing to curriculum, assessment, AI-literacy, and institutional-governance levers for shaping structural conditions. Originality – The study advances relational approaches to AI-mediated learning by specifying a mechanism linking structural configurations and human–technology relations to moral deskilling across diverse vocational education contexts without presuming AI inherently empowering or deskilling. It identifies institutional conditions for sustaining AI as decision support rather than decision replacement.