Decision-Making under Uncertainty

Decision-making under uncertainty studies how agents should choose actions when outcomes, states, preferences, or even the governing model are imperfectly known. Common approaches include Bayesian beliefs and partially observable Markov decision processes for sequential learning, robust optimization for misspecified distributions, risk-sensitive criteria such as CVaR, and conformal methods that calibrate when to act, gather information, or abstain. These methods are tested in settings such as autonomous driving and robotics, clinical diagnosis, energy and supply-chain scheduling, finance, and human–AI interaction. A central tension is that better-calibrated or more conservative uncertainty estimates do not automatically improve decisions: methods must balance safety and tail risk against utility, computational cost, and the value of additional information.