Title: From Explainable AI to Explainable Optimization

Abstract: While explainable AI (XAI) has made significant progress in making complex machine learning models transparent, mathematical optimization problems that drive critical decisions often remain opaque black boxes. This talk introduces the emerging field of explainable optimization, demonstrating how techniques developed for XAI can be extended and adapted to optimization models. I will present two complementary approaches: Coherent Local Explanations for Mathematical Optimization (CLEMO) and Counterfactual Explanations for Linear Optimization (CELOPT). CLEMO provides explanations for optimization models that respect their underlying mathematical structure, enabling stakeholders to understand how model parameters influence both objective values and decision variables. CELOPT offers three types of counterfactual explanations—weak, strong, and relative—that answer stakeholder questions like "What minimal parameter changes would lead to my desired outcome?" Through examples, I will demonstrate how these methods can enhance transparency, facilitate negotiations among stakeholders, and provide valuable insights for decision-makers.