From Digital Hoarding to Open Archives: Using AI Agents to Unlock West African Newspaper Collections

Arlette Farge's celebrated model of archival research — the slow, tactile encounter with documents in a silent reading room — presupposes well-ordered European institutions. For historians of West Africa, the reality is different: newspaper archives scattered across countries and continents, poorly catalogued, physically deteriorating, and sometimes easier to find in Berlin or Washington than in their countries of origin. Fieldwork often becomes a race to photograph as much as possible, leaving thousands of images unprocessed on hard drives for years. The gap between digitisation (capturing documents) and datafication (transforming them into searchable, structured data) is one of the most significant unacknowledged obstacles in contemporary historical research. This presentation draws on the Islam West Africa Collection (IWAC), an open-access database of over 14,500 newspaper articles, to show how AI can address this bottleneck. It discusses multimodal OCR for complex layouts, named entity recognition adapted to West African contexts, and corpus-level triage through sentiment analysis. It then introduces agentic AI pipelines — systems where specialised sub-agents autonomously handle image processing, text extraction, and metadata enrichment under human supervision — as a way to make these methods accessible to researchers without programming expertise. The talk concludes by reflecting on the stakes of using Western-built tools to make non-European knowledge visible.