Combining Belief Revision and Abstract Dialectical Frameworks

Master's thesis by Umer Mushtaq

Supervisor Sebastian Rudolph, Sarah Alice Gaggl
Computational Logic
11 Februar 2016 – 13 Oktober 2016

An important area of study in Knowledge Representation is to investigate how an agent deals with contradictory truth assertions by different entities and emerges with a coherent epistemic outcome in light of these contradictions. Belief Revision and Abstract Argumen- tation are two Knowledge Representation formalisms, among others, that deal with this problem. In Belief Revision, an agent is concerned with what truth assertion(s) to exclude from it’s original belief state in light of a newly encountered truth assertion(s) which con- flicts with it’s existing belief set. Researchers have suggested many belief revision operators, categorized as formula-based and model-based operators, in literature each with a unique approach to solving the inconsistency in the knowledge of an agent. Abstract Argumenta- tion is a mechanism to arrive at a coherent result in a scenario with multiple conflicting truth assertions. Dung-style frameworks (AFs) are a widely studied abstract argumentation formalism. Abstract Dialectical Frameworks (ADFs) are a special class of argumentation frameworks where the acceptance of a truth assertion can be made dependent upon the satisfaction of an associated condition. Semantics of ADFs specify what coherent outcomes can be reached given conflicting truth assertions. In this work, we seek to find an equiva- lence between the process of belief revision, as defined by various belief revision operators, and syntactic and semantics dynamics of ADFs. We shall also venture a brief look into the complexity of belief revision operators encoded as equivalent ADF updates.