Conformal Risk-Averse Decision Making with Action Conditional Guarantee

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.

Machine learning systems are increasingly used to support important decisions, such as medical treatment choices, financial actions, or automated recommendations. In these settings, it is not enough for a system to be accurate on average: each type of decision it makes should also be reliably safe. For example, a medical decision tool may look safe overall while still making risky recommendations for a particular treatment option. This paper develops a method for making cautious decisions with clearer safety guarantees for each possible action. Instead of only measuring whether the system is safe on average, our approach checks safety separately for every action the system might choose. This helps prevent risks for one decision type from being hidden by better performance on others. We also provide a practical algorithm that works with existing prediction models and gives mathematical guarantees from limited data. Experiments on medical diagnosis and recommendation tasks show that the method reduces harmful decisions while maintaining useful performance.