Incomplete Chronicles: Unveiling Data Bias in Maternal Health

Artificial intelligence (AI) has brought significant advancements in various domains, such as healthcare. However, the efficacy of these AI-driven technologies heavily depends on the quality and diversity of the data used to train them. In healthcare, where AI is increasingly leveraged for diagnostic and treatment purposes, the bias inherent in healthcare data sets can exacerbate existing health disparities, particularly in underrepresented regions such as Southern Africa (Zou & Schiebinger, 2021).

This report presents findings on how Dawa Health, a medical technology startup based in Lusaka, Zambia, has integrated AI into its maternal healthcare offerings and its DawaMom application. This ethnographic research examines the robustness and potential biases in the data sets and the technology behind the DawaMom app, including the influence that conventional categorisation and knowledge-storing approaches have on the dataset. Lastly, this research presents insights into how users interact with the DawaMom app, highlighting areas where the product's universalised design may only partially meet these expectant mothers' needs.