Responsible use of artificial intelligence and machine learning for food security early warning systems

Paper by Weston Anderson et al: “Artificial intelligence (AI) and machine learning (ML) methods offer substantial promise for monitoring and predicting acute food insecurity when paired with domain experts as part of a trusted and accountable system. However, using AI/ML-based methods may cause costly, dangerous mistakes if implemented uncritically. Funding for humanitarian aid has been drastically cut, putting tremendous pressure on food security early warning systems to use AI as a means of cutting costs. In this Comment, we outline where AI/ML methods offer promise to make early warning systems more adaptable and effective as well as where the use of AI/ML is ill advised. We recommend that AI/ML be used to improve monitoring and forecast models in the data-rich portions of food security early warning systems, such as those that rely on climate models and remote sensing. Where data are irregular and sources are varied, AI/ML should instead be used to improve the accessibility and timeliness of socioeconomic data collation. AI can augment food security analyst capabilities, but an analyst is needed to maintain clear systems of accountability and review for all issued forecasts…(More)”