Handling Borderline Cases Using Degrees: An Information Processing Perspective

Although linguistic vagueness is generally related to the existence of borderline cases, the main theories of vagueness are not primarily aimed at handling borderline cases explicitly. As a result, these theories are able to avoid the use of truth degrees, e.g. by focusing on truth value gaps or by taking the epistemic point of view that there is an unknown but sharp boundary between truth and falsity. We argue that, on the other hand, in information processing settings, borderline cases usually are a matter of degree, or at least of ranking, although such degrees may serve different purposes. To support our claim, this paper discusses a number of information processing scenarios, in relation with their use of degrees to handle borderline cases. First, we consider the use of fuzzy labels, contrasting the role of (graded) borderline cases in three different situations: (i) allowing for flexibility when specifying and evaluating fuzzy requests, (ii) allowing for abstraction when describing precise information, and (iii) allowing for uncertainty when stating imprecise information. Second, we discuss the idea of degrees of typicality in the setting of formal concept analysis, seeing e.g. penguins or kiwis as borderline cases of birds. Finally, we illustrate how degrees of similarity may be useful for maintaining consistency, taking advantage of a flexible understanding of linguistic terms, seeing e.g. civil unions as borderline cases of marriages.