Unpacking Learner Agency and Epistemic Justice in AI-Augmented Open and Distance Learning for Marginalised Students

Abstract

Intelligent learning technologies are reshaping university-based Open and Distance Learning (ODL), raising questions about how digitally mediated pedagogies redistribute knowledge, power, and participation for students from structurally marginalised contexts. Moving beyond issues of access, efficiency, and automation, this paper examines the epistemological implications of AI-augmented ODL by analysing learner agency and epistemic justice. Drawing on critical pedagogy and theories of epistemic injustice, we employ a theory-synthesis design to integrate conceptual and empirical work on ODL, learning analytics, and algorithmic personalisation. The analysis demonstrates how predictive models, recommendation engines, and automated moderation can both scaffold and constrain autonomy, participation, and recognition; it identifies pathways of testimonial and hermeneutical injustice when systems misinterpret non-dominant discourse, proxy sensitive attributes, or inhibit contestation. In response, we propose a critical framework that redefines learner agency as relational and contextually situated, organised around three dimensions—recognition, voice, and power—and operationalised through three design checkpoints—credibility, comprehensibility, and control (CCC). We outline practical applications for course and assessment design, analytics pipelines, student data rights, and continuous improvement, while specifying policy requirements for participatory governance, impact assessment, and repair. The framework is intended to be applicable across regions globally where similar platform logics circulate, while remaining adaptable to local conditions. Ultimately, the paper offers a justice-oriented lens for reimagining ODL as a space where student dignity, cultural relevance, and equitable participation in knowledge construction are central. The study contributes to knowledge by delivering a coherent, testable model and CCC criteria that translate epistemic justice theory into actionable design and governance for AI-mediated ODL.