WHOSE KNOWLEDGE DOES AI TEACH? DECOLONIZING EDUCATION, ALGORITHMS AND EPISTEMIC AUTHORITY

Abstract

The rapid incorporation of artificial intelligence (AI), particularly generative AI, into education is transforming how knowledge is produced, accessed, evaluated and communicated. Although AI is frequently presented as a neutral technology capable of democratizing educational opportunities, its knowledge architectures are shaped by particular languages, datasets, institutions, commercial interests and cultural assumptions. This raises a fundamental educational question: whose knowledge does AI teach? This conceptual paper examines the relationship between artificial intelligence, education, coloniality and epistemic authority through a decolonial perspective. Drawing on recent scholarship on algorithmic colonialism, data colonialism, epistemic injustice, decolonial AI and AI in education, the paper argues that AI-mediated education can reproduce existing hierarchies of knowledge when Western-centric datasets, languages and epistemological assumptions become embedded within algorithmic systems. The paper conceptualizes this process as algorithmic epistemic authority, whereby AI systems increasingly participate in determining which knowledge appears credible, relevant, visible and pedagogically valuable. Against this tendency, the paper proposes a Decolonial AI-Education Framework based on five principles: epistemic plurality, knowledge sovereignty, algorithmic reflexivity, participatory governance and contextual pedagogical agency. The paper argues that decolonizing AI in education does not require rejecting artificial intelligence. Rather, it requires transforming the conditions under which AI is designed, trained, deployed and evaluated so that diverse knowledge systems, languages, communities and educational purposes can participate in determining what counts as knowledge.