Explainable and Robust Multi-Criteria Decision Making: An Integrated Pipeline of Predictive Analytics, Preference Learning, and Distribution

Quantitative decision-making challenges in management, technology, and finance increasingly rely on data-driven forecasts and heterogeneous preferences, which are often only partially observable. This chapter presents an integrated decision-making framework that explicitly incorporates forecast and preference uncertainty to facilitate robust and explainable decisions. The proposed explainable and robust multi-criteria decision-making (xMCDM) approach integrates a predictive layer with uncertainty quantification, employs a Bayesian preference-learning method to model uncertain weights, and uses distribution-robust decision aggregation. Empirical case studies in supplier selection, renewable energy site planning, and credit portfolio allocation demonstrate that xMCDM enhances rank stability, reduces regret, and enables transparent evaluation of robustness trade-offs. Consequently, xMCDM advances the focus from score-based evaluation to robust decision-making under uncertainty.