Upgrading Ambiguous Signs in QPNs

Abstract

A qualitative probabilistic network models the probabilistic relationships between its variables by means of signs. Nonmonotonic influences have associated an ambiguous sign. These ambiguous signs typically give rise to uninformative results upon inference. We argue that a non-monotonic influence can be associated with a more informative sign that indicates its effect in the current state of the network. To capture this effect, we introduce the concept of situational sign. Furthermore, if the network converts to a state in which all variables that provoke the nonmonotonicity have been observed, a nonmonotonic influence reduces to a monotonic one. We study the persistence and propagation of situational signs upon inference and give a method for establishing the sign of a reduced influence.