Preference-Based Diagnosis Selection in Multi-Context Systems.

Nonmonotonic Multi-Context Systems MCS provide a rigorous framework and flexible approach to represent and reason over interlinked, heterogeneous knowledge sources. Not least due to nonmonotonicity, however, an MCS may be inconsistent and resolving inconsistency is a major issue. Notions of diagnosis and inconsistency explanations have been developed for this purpose, considering the information exchange as the primary culprit. To discriminate between different possible solutions, we consider preference-based diagnosis selection. We develop a general meta-reasoning technique, i.e., an MCS transformation capable of full introspection on possible diagnoses, and we present a natural encoding of preferred diagnosis selection on top. Moreover, for the more involved notions of diagnosis utilized, we establish that the complexity does not increase. However, this does not carry over to selecting most preferred diagnoses as the encoding is not polynomial.