EXALT: EXplainable ALgorithmic Tools for Optimization Problems

Explainability gaps exist beyond machine learning: While most Explainable AI research focuses on complex machine learning models, this work highlights that even traditional optimization algorithms lack human-interpretable explanations, suggesting the explainability challenge is broader than commonly recognized.

Four practical explanation methods can enhance algorithm transparency: The paper proposes concrete techniques—generating alternative solutions, creating robust solutions through input testing, building decision trees, and producing comprehensive reports—that can make algorithmic decisions more understandable to non-technical users.

Current tools lack flexibility and expert integration: Existing explainability tools are often designed for specific algorithms (like clustering methods) and fail to incorporate domain expert knowledge, limiting their real-world applicability across different decision-making contexts in healthcare, e-commerce, and other sectors.

Explainability is essential for algorithmic adoption: The paper identifies lack of interpretable explanations as a "critical challenge" hindering widespread adoption of algorithmic solutions, suggesting that transparency directly impacts whether organizations and individuals will trust and use these tools in practice.

Optimization problems deserve explainability focus: By starting with optimization problems rather than only complex "black box" models, this approach addresses a foundational layer of algorithmic decision-making that affects many real-world applications but has received less attention in XAI research.