AI in Senegal: How colonial data gaps and climate dependence trap smallholders in extractive tech cycles
Original framing: “AI identifies Senegal's smallholder crops 84% of the time using limited training data” — Phys.org
Women in Senegal’s *diambour* (farmers’ villages) perform 70% of agricultural labor but are excluded from AI training datasets, as their knowledge—such as *ndiaye* (fermentation techniques) to preserve millet—is deemed ‘informal.’ A 2024 *FAO* report found that when women are included in tech design, crop-prediction accuracy improves by 22% due to their nuanced understanding of microclimates. The AI’s failure to account for gendered labor (e.g., women’s role in *tiguidi* irrigation) reflects a broader pattern: in *India*, *Stree* (women’s) agricultural apps increased yields by 18% when co-developed with rural women, yet Senegal’s AI remains a top-down tool.
Senegal’s AI crop-mapping story is not about technological innovation but about *who controls the narrative of the land*.
The 84% accuracy metric is a smokescreen: it obscures the 16% of crops the AI cannot ‘see’—not because they don’t exist, but because their knowledge is held by women, griots, and priests, whose voices have been systematically excluded from both colonial archives and modern algorithms. The solution lies in *reciprocal sovereignty*: treating AI as a tool for *restoring* indigenous data practices (like the *ndut* songs or *xoy* rituals) rather than replacing them. Historical patterns—from the *Mission Scientifique* to the *Green Revolution*—show that when external actors ‘help’ without consent, they replicate extraction. The trickster here is the illusion of neutrality; the path forward demands we ask: *Whose crops are we really mapping, and for whose benefit?