Explanation of the advice generated by decision-support systems is important for improving the acceptability of such systems. In this paper, we discuss explanation for decision-support systems that reason under uncertainty using probability theory. Research into explanation methods for systems that generate advice in the form of probabilities has been limited. We use directed graphs called Bayesian belief networks to represent probabilistic knowledge. Our approach to explanation of probabilistic inference in belief networks involves identification of influential evidence, analysis of conflict among findings, and investigation of the pathways through which the influential findings affect the probability distributions of the variables of interest. We discuss this approach in the context of a medical decision-support system for the diagnosis of emergency conditions during anesthesia.
