Conformal Risk-Averse Decision Making with Action Conditional Guarantee

2026-06-03

AI Summary

Existing machine learning decision systems offer only marginal safety guarantees, which are insufficient for risk-averse applications. This work proposes an action-conditional conformal prediction framework that explicitly links uncertainty quantification with specific decision actions, thereby constructing feasible decision sets conditioned on actions and establishing their theoretical connection to action-conditional Value-at-Risk (Action-Conditional VaR) optimization. Leveraging the pinball loss, we design an efficient optimization algorithm with finite-sample guarantees. Experimental results demonstrate that the proposed method significantly outperforms existing conformal baselines on two real-world datasets, achieving substantial improvements in action-conditional performance metrics.

Abstract

Reliable decision making pipelines powered by machine learning models require uncertainty quantification (UQ) methods that come with explicit safety guarantees. Conformal prediction provides such UQ by wrapping ML predictions into prediction sets, and recent work by Kiyani et al. (2025b) established that these sets can be translated into optimal risk-averse decision policies -- yet only inheriting marginal safety guarantees. We generalize and strengthen their results by (i) introducing action-conditional conformal prediction, which yields safety guarantees conditioned explicitly on each action taken by the decision maker, (ii) showing that action-conditional prediction sets serve as a proxy for the feasible decision space for risk-averse decision makers aiming to optimize action-conditional value-at-risk, and (iii) proposing a principled finite-sample algorithm based on pinball-loss minimization, connecting the framework of Gibbs et al. (2025) to action-conditional guarantees. Experiments on two real-world datasets confirm that our approach significantly improves action-conditional performance over conformal baselines.

Problem

• conformal prediction
• risk-averse decision making
• action-conditional guarantee
• uncertainty quantification
• value-at-risk

Innovation

• action-conditional conformal prediction
• risk-averse decision making
• value-at-risk
• uncertainty quantification
• pinball-loss minimization