Introducing Role Defeasibility in Description Logics

Accounts of preferential reasoning in Description Logics often take as point of departure the semantic notion of a preference order on objects in a domain of interpretation, which allows for the development of notions of defeasible subsumption and entailment. However, such an approach does not account for defeasible roles, interpreted as partially ordered sets of tuples. We state the case for role defeasibility and introduce a corresponding preferential semantics for a number of defeasible constructs on roles. We show that this does not negatively affect decidability or complexity of reasoning for an important class of DLs, and that existing notions of preferential reasoning can be expressed in terms of defeasible roles.