Reasoning by hypothesizing causal models

The authors present some aspects of a reasoner that can handle uncertain knowledge and that hypothesizes causal models to explain the observed evidence. Such reasoning is useful where the objective of the reasoner is either to pursue an investigation or to construct a desired type of argument. The authors present the intuitive properties that may be displayed by a causal model and formalise them in the context of the hypergraph structure that is used for representing the causal knowledge. They use probability inferences made in the context of each causal model as a basis for preferring one causal model over the other. They use an algorithm based on A* search to construct efficiently those models which derive preferred probabilistic inferences