Robust Bayesian Decision Making under Adversarial Uncertainty

The article introduces a sequential adversarially robust decision-aware experimental design framework that accounts for information gain against worst-case unexpected effects modeled as adversarial variables.

• The approach formalizes an adversarially robust optimal decision within Bayesian decision theory.
• It derives a principled Bayesian experimental design criterion that explicitly targets decision stability rather than nominal optimality.
• Experiments on synthetic and real-world datasets show conventional methods converge to high confidence yet fragile decisions, while the proposed method yields significantly more stable and reliable outcomes under adversarial variation.