Explainable AI for Reducing Fresh Produce Food Waste: A Systematic Review

Abstract

The fresh produce sector is affected by major economic, environmental and food-security problems as a result of perishability and supply-chain complexities. Explainable Artificial Intelligence (XAI) promotes the implementation and use of AI in forecasting, spoilage detection, and optimization by increasing transparency. This is a systematic review of the empirical applications of XAI in mitigating food waste in fresh produce supply chains that have gaps that prevent the implementation process. It reviewed peer-reviewed articles published between January 2012 and January 2025 according to PRISMA 2020, which were obtained through Scopus, Web of Sciences, IEEE Xplore, SpringerLink, and ScienceDirect. Inclusion concerned empirical methods of XAI such as SHAP, LIME, counterfactual explanations, Grad-CAM, saliency maps, integrated gradients, rule-based models, and decision-tree surrogates applied to waste reduction. Forty-five (45) articles passed the eligibility test. SHAP (32%), LIME (21%), and counterfactuals (18%) prevailed, 29% were visual/rule-based. Its main uses were in production (38%), logistics (27%), retail (23%), and consumer systems (12%), such as yield prediction, cold-chain monitoring, quality grading, inventory forecasting, and shelf-life support. Critical weaknesses are the absence of standardized measures of quality of explanation, little use in low income and middle-income nations, inapplicability to non-experts, and disjointed data to limit model accuracy. The review provides a typology of XAI applications on the stages of the supply chain and suggests: (1) standard benchmarking; (2) stakeholder-oriented explanation design with domain integration; (3) scaled, context-relevant XAI systems that are bound to sustainability objectives. These considerations encourage open, fair AI in waste minimization. Keywords: Explainable artificial intelligence, food waste, fresh produce industry, supply chain, transparency.