Conformalized Decision Risk Assessment

Wenbin Zhou, Agni Orfanoudaki, Shixiang Zhu

Keywords: Conformal prediction, inverse optimization, risk assessment, decision making under uncertainty

TL;DR: A framework for quantifying the probability that a candidate decision is suboptimal in an optimization setting

Abstract: We introduce \texttt{CREDO}, a framework that provides distribution-free upper bounds on the probability that any candidate decision is suboptimal. By combining inverse optimization geometry with conformal prediction and generative modeling, \texttt{CREDO} generates statistically rigorous and interpretable risk certificates. This enables decision-makers to audit and validate choices under uncertainty, bridging the gap between decision-making algorithms and human judgment.

Submission Number: 165