Finding, Scoring and Explaining Arguments in Bayesian Networks
This work addresses the interpretability challenge in Bayesian Networks for users needing transparent AI decisions, though it appears incremental as it builds on existing probabilistic reasoning frameworks.
The researchers tackled the problem of explaining Bayesian Networks by proposing a new definition of probabilistic arguments and an algorithm to extract relevant, independent arguments given a network, target node, and observations, demonstrating its relevance by approximating message passing and providing natural language explanations.
We propose a new approach to explain Bayesian Networks. The approach revolves around a new definition of a probabilistic argument and the evidence it provides. We define a notion of independent arguments, and propose an algorithm to extract a list of relevant, independent arguments given a Bayesian Network, a target node and a set of observations. To demonstrate the relevance of the arguments, we show how we can use the extracted arguments to approximate message passing. Finally, we show a simple scheme to explain the arguments in natural language.