Reinstatement and Specificty in Argumentation Systems

Abstract

Reinstatement is a principle of argumentation systems that enables the justification of a defeated argument when all its defeaters are in turn ultimately defeated. Some counterexamples to reinstatement have been offered in the literature. Specifically, counterexamples suggest that reinstatement cannot be taken as a general principle of defeasible argumentation because the reinstated arguments may support incorrect conclusions. Some authors argued that the problems are not due to reinstatement but to the formalization of those examples. Then, the solution is to make the language expressive enough to obtain the correct results. They also warn that one should avoid tinkering with the formalization in concrete examples just to get a desired outcome. Therefore, this approach should be combined with the search of general principles for choosing the proper formalization. Taking into account that finding general principles of representation could be a hard enterprise, the goal of this thesis is to identify some criterion that allows i. neutralize the counterexamples, ii. preserve the original formal language as much as possible, and iii. maintain reinstatement as a general principle. To identify that criterion, counterexamples are analyzed and possible causes of the problem are detected. As a result it is found that the preference by specificity among arguments can be used to obtain that criterion. Three approaches based on specificity are proposed and evaluated. Two of them introduce alternative defeat relations among arguments. The third one is based on filtering the non maximally specific arguments.