Ubuntu, generative AI and machine-mediated relationships in South Africa

Abstract

Generative artificial intelligence (AI) systems, especially large language models (LLMs) deployed as chatbots and digital assistants, are increasingly presented as conversational partners that listen, respond and even console. In South Africa, where the ethic of ubuntu has played a central role in public discourse and philosophical reflection, this raises a pressing question: what happens to a relational conception of personhood when many everyday interactions are mediated by machines that simulate but do not share in human social life? Drawing on contemporary debates about ubuntu as a moral theory and political ideal, I argue that machine-mediated relationships risk habituating users to a one-sided form of interaction that is structurally at odds with ubuntu’s emphasis on mutual recognition, shared vulnerability and relational accountability. At the same time, the deployment of AI in customer service, mental health support and other care-adjacent domains threatens to entrench a new form of “relational apartheid” in which those with fewer resources are increasingly addressed by machines rather than persons. The article critically examines proposals to “encode” ubuntu in AI, suggesting that while systems can be designed to support communities in ubuntu-compatible ways, the attempt to treat AI itself as a bearer of ubuntu misunderstands both the ontology of persons and the socio-historical character of the ubuntu tradition. I conclude by sketching some normative conditions under which AI in South Africa might augment rather than displace human relationships, and by arguing that an ubuntu-informed critique can help to reorient AI policy and design towards relational justice rather than mere efficiency.