A formalization of defeasible argumentation using labelled deductive systems : Preliminary report

• Autores: Chesñevar, Carlos Iván; Simari, Guillermo Ricardo
• Año de publicación: 1998
• Idioma: inglés
• Tipo de recurso: documento de conferencia
• Materia: Ciencias Informáticas; Informática; Inteligencia Artificial; defeasible reasoning; argumentative systems; labelled deductive systems
• Institución: Universidad Nacional de La Plata

Argumentative systems [SL92, Vre93, Che96] are formalizations of defeasible reasoning [Pol87, Nut88]. An argument is a tentative piece of reasoning an intelligent agent can use to reach a given conclusion. In case there is information available supporting counterarguments which defeat the argument, its conclusion will no longer be valid. In order to determine whether a conclusion h is justified belief, it is necessary to consider a tree-like structure (having an argument A for h as its root), in which defeaters for A, defeaters for these defeaters, and so on, must be taken into account. If the argument A prevails over all its associated defeaters, then A is called a justification for h. Currently there exist several alternative formalizations of defeasible argumentation. recent work [PS96, KT96, BDKT97] has shown that defeasible argumentation constitutes a point of confluence for the characterization of different approaches to monotonic reasoning (NMR). from the early '90 ther have been several attempts to find a unified logical framework for NMR. In this aspect, the labelled deductive system [Gab96a] (or LDS) constitute an attractive approach, allowing to characterize different logics by introducing labels as part of the logic 's object language and keeping a single inference mechanism for all logics. This paper presents a formal approach for characterizing defeasible argumentation in terms of LDS. Inference rules are present in the style of natural deduction, and they capture the process of defeasible argumentation as defined in the MTDR framework [SL92, SCG94]. We contend that this approach make easier to state and prove properties and characteristics of defeasible argumentation within logical-deductive setting.
V Workshop sobre Aspectos Teóricos de la Inteligencia Artificial (ATIA)
Red de Universidades con Carreras en Informática (RedUNCI)