- Article
Artificial intelligence (AI) increasingly functions as an epistemic infrastructure that shapes how knowledge is accessed, produced, interpreted, validated, and circulated. In African contexts, critiques of AI have appropriately emphasized language bias, cultural inclusion, and the underrepresentation of African languages and knowledge traditions. While these concerns remain significant, this paper argues that representational and language-centered approaches are insufficient because expanding linguistic or cultural coverage does not necessarily transform how knowledge is interpreted, validated, or accorded authority within AI-mediated environments. Drawing on Ubuntu’s relational ethical foundations and African epistemological traditions, this paper proposes Mbuntu as an African relational epistemic framework for technologically mediated knowing. Mbuntu extends Ubuntu’s relational orientation from relational humanity toward relational knowing under technological mediation and is organized around four interrelated principles: relational intelligence, communal knowledge validation, moral accountability, and context sensitivity. Together, these principles provide conceptual heuristics for examining how computational representations acquire credibility and epistemic authority, while distinguishing the apparent accuracy of AI-mediated knowledge from the justification for treating it as authoritative within particular social and cultural contexts. This paper develops the implications of this distinction for education, AI design and evaluation, and governance, including questions of epistemic sovereignty. Rather than prescribing a universal technical model, Mbuntu provides a diagnostic and normative framework for moving beyond epistemic inclusion toward epistemic reorientation, in which the conditions through which AI mediates, validates, and authorizes knowledge themselves become objects of critical scrutiny.
20 November 2026


