Skip to Main content Skip to Navigation

Preference-based argumentation frameworks: application in decision making and negotiation

Abstract : Argumentation is a promising approach for reasoning with uncertain or incoherent knowledge or more generally with common sense knowledge. It consists of constructing arguments and counter-arguments, comparing the different arguments and selecting the most acceptable among them. This thesis contains three main parts. The first one concerns the notion of equivalence between two argumentation frameworks. We studied two families of equivalence: basic equivalence and strong equivalence. We proposed different equivalence criteria, investigated their links and showed under which conditions two frameworks are equivalent w.r.t. each of the proposed criteria. The notion of equivalence is then used in order to compute the core(s) of an argumentation framework. A core of a framework is its compact version, i.e. an equivalent sub-framework. Hence, instead of using an argumentation framework which may be infinite, it is sufficient to consider one of its cores, which is usually finite. This core determines the stability of the status of each argument. The second part of the thesis concerns the use of preferences in argumentation. We investigated the roles that preferences may play in an argumentation frame- work. Two particular roles were identified: i) to privilege strong arguments over weaker attacking arguments when computing the standard solutions of a frame- work, and ii) to refine those standard solutions. We showed that the two roles are completely independent and require different procedures for modeling them. Besides, we showed that almost all the existing works have tackled only the first role. Moreover, the proposed approaches suffer from a drawback which consists of returning conflicting extensions. We proposed a general approach which solves this problem and which presents two novelties: First, it takes into account preferences at a semantic level, i.e. it defines new acceptability semantics which are grounded on attacks and preferences between arguments. Second, a semantics is defined as a dominance relation that compares any pair of subsets of arguments. The third part illustrates our preference-based argumentation frameworks (PAF) in case of decision making and negotiation. We proposed an instantiation of our PAF which rank-orders options in a decision making problem. Then, we studied the dynamics of this model. More precisely, we showed how the ordering on options changes in light of a new argument. We also used our PAF in order to show the benefits of arguing in negotiation dialogues. For that purpose, we proposed an abstract framework for argument-based negotiation, investigated the different types of solutions that may be reached in such dialogues, and showed for the first time under which conditions arguing is beneficial during a negotiation.
Document type :
Complete list of metadatas
Contributor : Srdjan Vesic <>
Submitted on : Tuesday, July 19, 2011 - 12:47:46 PM
Last modification on : Tuesday, December 8, 2020 - 10:06:04 AM
Long-term archiving on: : Thursday, October 20, 2011 - 2:22:16 AM


  • HAL Id : tel-00609508, version 1



Srdjan Vesic. Preference-based argumentation frameworks: application in decision making and negotiation. Computer Science [cs]. Université Paul Sabatier - Toulouse III, 2011. English. ⟨tel-00609508⟩



Record views


Files downloads