Nowcasting for hunger relief: a study of promise and perils

Pitched as an aid to better development decision-making, the website HungerMap LIVE presents composite data on, and machine-learning-derived predictions of, food insecurity in 90 countries. Of its current version, this article asks the following questions: What work is HungerMap LIVE called upon to do in ICT for development (ICT4D) practice? How well is it set up to do that work? Combining technical (both computer science and statistical) and social analysis, this article employs a close reading method drawn from humanities and legal research not usually directed at digital platforms or websites in combination with interview-based techniques. By this means, it scrutinizes HungerMap LIVE’s potential to guide or mislead users and canvasses some elaborations that could enhance its usability. It argues that interdisciplinary research of this kind can counter both the historical and technological determinism troubling the ICT4D field and better position decision-makers to employ machine learning in history- and context-attentive ways.