Interpretable Optimization: Leveraging Large Language Models to Enhance Transparency in Operations Research Decision-Making

Abstract: Optimization‍‌‍‍‌‍‌‍‍‌ models are an essential part of operations research and are used to guide complex decision-making processes. However, a considerable number of practitioners find it difficult to understand how these models come to their solutions. The opacity of these models often leads to a decrease in the trust of the resulting recommendations and, as a consequence, the collaboration between human decision-makers and algorithmic tools becomes less effective (López & Comellas, 2025). The advancements that have been made lately in the area of large language models (LLMs) open up a variety of new possibilities for bridging the interpretability gap. Through the use of natural language that is easier to understand by non-specialists, LLMs can take the role of intermediaries that facilitate optimizations by non-experts (Biemans et al., 2025; Zhang et al., 2025). The document describes a model of interpretable optimization based on a structured approach that explanation modules using LLM are directly integrated into traditional OR workflows. We present the core methodological components i.e explanation extraction, prompt design, scenario-based reasoning, and validation and infer their potential by means of demonstrative cases in the supply chain and scheduling. Additionally, we gauge the effect of LLM-produced explanations on user understanding, trust, and the willingness to accept the decision. The main takeaway from our study is that the incorporation of LLMs into optimization workflows has the potential to significantly increase not only the transparency but also the practical application of OR models, if the risks, such as hallucination and oversimplification, are taken into account and managed ‍‌‍‍‌‍‌‍‍‌properly. Keywords: interpretable optimization; explainable OR; large language models; decision transparency; operations research.