Conformalized Decision Risk Assessment

2025-05-19

AI Summary

Prediction–optimization methods suffer from poor interpretability and insufficient robustness under multimodal or distributionally uncertain settings—critical limitations in high-stakes decision-making. Method: We propose the first risk certification framework integrating inverse optimization geometry, conformal prediction, and generative modeling. Crucially, it imposes no distributional assumptions and delivers statistically rigorous, distribution-free upper bounds on the suboptimality probability for any candidate decision. Contribution/Results: The resulting risk certificates are both semantically interpretable and computationally tractable, enabling human experts to perform verifiable robustness audits of decisions. Experiments demonstrate substantial improvements in reliability and human–AI trustworthiness of prediction–optimization pipelines under multimodality, establishing a novel decision-verification tool for mission-critical domains including healthcare, energy systems, and public policy.

Technology Category

• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Constraint Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

High-stakes decisions in domains such as healthcare, energy, and public policy are often made by human experts using domain knowledge and heuristics, yet are increasingly supported by predictive and optimization-based tools. A dominant approach in operations research is the predict-then-optimize paradigm, where a predictive model estimates uncertain inputs, and an optimization model recommends a decision. However, this approach often lacks interpretability and can fail under distributional uncertainty -- particularly when the outcome distribution is multi-modal or complex -- leading to brittle or misleading decisions. In this paper, we introduce CREDO, a novel framework that quantifies, for any candidate decision, a distribution-free upper bound on the probability that the decision is suboptimal. By combining inverse optimization geometry with conformal prediction and generative modeling, CREDO produces risk certificates that are both statistically rigorous and practically interpretable. This framework enables human decision-makers to audit and validate their own decisions under uncertainty, bridging the gap between algorithmic tools and real-world judgment.

Problem

• Assessing suboptimal decision risks in high-stakes domains
• Overcoming predict-then-optimize brittleness under distributional uncertainty
• Providing interpretable risk certificates for human-algorithm decision auditing

Innovation

• Combines inverse optimization with conformal prediction
• Generates distribution-free risk certificates for decisions
• Integrates generative modeling for interpretable results