Analyzing the defeat relation in observation-based defeasible logic programming

Abstract

In the last decade several ways to formalize defeasible reasoning have been studied. A particular approach, defeasible argumentation, has been particularly successful to achieve this goal. The inference process of argument-based systems is based on the interaction of ar- guments for and against certain conclusions. The relations of attack and defeat among arguments are key elements in these inference process. Usually a preference criterion is used to calculate the defeat relation to decide, in case of con ict, which argument is preferred over its contender. Speci city is a domain independent principle that has been used in several formalisms. In this work we analyze the problem of incorporating speci city to characterize defeat in a particular argumentative framework, called Observation Based Defeasible Logic Program- ming. Since e ciency is an important issue in ODeLP, we have devised a new version of this criterion, that optimizes the computation of the defeat relation. We also present a formal proof to show that this new version is equivalent to the old one.