Handling Revision Inconsistencies: Creating Useful Explanations

With ever growing knowledge bases the presence of inconsistencies has become a major challenge. In the past mainly the problems of finding and removing inconsistencies from knowledge bases have been addressed. However, knowledge is not static and needs to be adapted to changes on a regular basis. Consequently, in recent years methods to revise knowledge have been introduced. The belief change operation called revision has been developed to adapt probability distributions, a common representation for knowledge, to new sets of beliefs. However, in complex real world applications inconsistencies are almost unavoidable when trying to assign new sets of low-dimensional (conditional) probability distributions in order to revise a high-dimensional initial distribution. It is therefore important to handle inconsistencies properly in order to avoid the introduction of new inconsistencies into an initially consistent knowledge base. One relevant aspect is the resolution of occurring inconsistencies. In this article we examine the topic of creating and presenting explanations by discussing different components of inconsistencies and resulting consequences for explanations. Furthermore in the end we present a prototype UI featuring a view of an inconsistency from different angles.