A taxonomy for argumentative frameworks based on labelled deduction

Abstract

Artificial Intelligence has long dealt with the issue of finding a suitable formalization for reasoning with incomplete and potentially inconsistent information. Defeasible argumentation [SL92,CML00,PraVre99] has proven to be a successful approach in many respects, since it naturally resembles many aspects of commonsense reasoning (see [CML00,PraVre99] for details). Besides, recent work [PraVre99,BDKT97] has shown that defeasible argumentation constitutes a confluence point for characterizing many different approaches to non-monotonic reasoning. Nevertheless, the evolution of different, alternative formalisms for modeling argumentation has resulted in a number of models that share some common features (the notion of argument, attack between arguments, defeat, dialectical analysis, etc.). This constitutes a motivation for the definition of a unified ontology, under which these different features can be analyzed and inter-related. As a byproduct from such an ontology, a classification (or taxonomy) of argumentation frameworks in terms of knowledge encoding capabilities, expressive power, etc. would be possible. In [Che01] a logical framework for defeasible argumentation called SDEAR was developed. The SDEAR framework is based on labelled deductive systems [Gab96]. Labelled Deductive Systems offer an attractive approach to formalizing complex logical systems, since they allow to characterize the different components involved by using different sorts of labels. One of the motivations for developing this framework was namely the definition of a single, unified ontology to capture the main issues involved in defeasible argumentation by specifying a suitable underlying logical language and its associated inference rules. In this presentation we focus on a particular research line which emerged from the above formalization, namely the classification of different defeasible argumentation frameworks based on features that can be ‘abstracted away’ in the SDEAR framework. The presentation is structured as follows: first, in section 2, we will briefly sketch how the SDEAR framework works. Then, in section 3 we will describe how different argumentation frameworks can be interrelated through SDEAR. Finally, section 4 concludes.