Reasoning over Attack-incomplete AAFs in the Presence of Correlations
Attack-Incomplete Abstract Argumentation Frameworks (att- iAAFs) are a popular extension of AAFs where attacks are marked as uncertain when they are not unanimously per- ceived by different agents reasoning on the same arguments. We here extend att-iAAFs with the possibility of specifying correlations involving the uncertain attacks. This feature sup- ports a unified and more precise representation of the differ- ent scenarios for the argumentation, where, for instance, it can be stated that an attack α has to be considered only if an attack β is considered, or that α and β are alternative, and so on. In order to provide a user-friendly language for spec- ifying the correlations, we allow the argumentation analyst to express them in terms of a set of elementary dependen- cies, using common logical operators (namely, OR , NAND , CHOICE , ⇒). In this context, we focus on the problem of verifying extensions under the possible perspective, and study the sensitivity of its computational complexity to the forms of correlations expressed and the semantics of the extensions.