Plausibility Structures for Default Reasoning

Friedman and Halpern have introduced the inference by plausibility structures, which provides semantics for various default logics. This is a generalization of known inferences, such as those based on expectations, or on possibilistic logic. We argue that a slightly different inference by plausibility would be more appropriate for default reasoning. It would still be appropriate for all the default logics considered by Friedman and Halpern. Significant cases, such as subsets of normal defaults with multiple extensions, or a formalism extending circumscription to the cases where (AND) is falsified, can be translated into the new formalism only. In order to prove our results, we complete the list of the reasoning properties for the (two versions of the) inference by plausibility. Since these properties describe the behavior of a given inference in an intuitive and non technical way, this list is important for any potential user. Moreover, it happens that considering these properties only, without going into the technicalities of the plausibility approach, is enough to describe the main results given here.