Knowledge Structures for Decolonial AI

This essay is an exposition of approaches that collectively are resisting the status quo. Researchers, AI developers, legal scholars, activists and communities working on and from the African continent are decolonizing and crafting new narratives for African AI and the underlying knowledge structures. My artifacts present concrete examples of tokenistic participation of Africans in AI value chains, extractive data collection practices funded by big tech and philanthropy, and the perpetuation of Western knowledge hierarchies through Large Language Models (LLMs) trained predominantly on Western languages and contexts, with African representation slapped on at the end. Rather than dwelling extensively on the problems therein already covered by other scholars or diving deeper into theoretical frameworks of decolonization, the essay and artifacts I present offer insight into a decolonial moment. This is an optimistic collection of artifacts that showcase infrastructures and approaches that are being developed and an imaginary of what AI could look like when it prioritizes equitable distribution of resources and opportunities, technology transfer that builds local capacity, funding for local AI research and development in underrepresented regions, and non-extractive access to AI data and tools.