Probabilistic Counterfactuals: Semantics, Computation, and Applications

摘要

We have reformulated Bayesian networks as carriers of causal information. Theresult is a more natural understanding of what the networks stand for, what judgments are required in constructing the network and, most importantly, how actions and plans are to be handled within the framework of standard probability theory. Starting with functional description of physical mechanisms, we were able to derive the standard probabilistic properties of Bayesian networks and to show: (1) how the effects of unanticipated actions can be predicted from the network topology, (2) how qualitative causal judgments can be integrated with statistical data, (3) how actions interact with observations, and (4) how counterfactuals sentences can be formulated and evaluated.