Reconceptualising Learner Autonomy in Generative AI-Supported Higher Education: A Technology-Mediated Framework

Abstract

Generative artificial intelligence (generative AI) creates an autonomy paradox in higher education: It expands access to feedback, explanation and learning support while also enabling learners to outsource parts of planning, reasoning, production, verification and evaluation. Recently, studies have started to examine learner autonomy in environments supported by generative AI, but they remain conceptually fragmented regarding how learner responsibility should be organised across AI-mediated academic work. This conceptual study addresses this problem with a conceptual analysis and a bounded integrative literature review drawing on 46 sources, of which 43 formed the substantive analytical corpus that spans foundational autonomy theory, self-regulated learning, metacognition, learner agency, AI literacy, feedback, assessment, academic integrity and generative AI. The synthesis identifies three mechanisms of reconfiguration—cognitive redistribution, authorship ambiguity and regulatory intensification – and proposes technology-mediated learner autonomy as a learner-level responsibility structure comprising five operational dimensions: goal-setting autonomy, prompting agency, critical verification, metacognitive regulation and ethical accountability. Human judgement preservation functions as the normative core that links these dimensions and provides a criterion for distinguishing between accountable learner-directed use and uncritical or dependent AI use. The framework offers a theoretically traceable basis for future empirical validation and for designing assessment, feedback and institutional guidance that makes learner responsibility visible.