The Class of Preference Strategy Formulas

We have recently worked on two approaches towards a dynamic representation of the diagnosis process, namely using a preference relation over single diagnoses and using one over sets of diagnoses. The latter is formalized in the concept of strategies, while the former is captured by the notion of preference graphs. In this paper we show how to combine both lines by specifying the preference graphs in the strategy language we deened. The core of the paper is the deenition of the class of preference strategy formulas. We prove that this class and the concept introduced earlier are equivalent. Additionally, we show how interaction of preference strategy formulas with the class of simple strategy formulas (a subset of strategy formulas described in another paper) can be solved. 1. Introduction Recent work stresses the need for a dynamic view of diagnosis. After Struss coined the concept of diagnosis as a process Str92a], subsequent work focused on the reali-sation of speciic aspects of the process. Based on their notion of preferred diagnosis, BB ottcher and Dressler incorporated several strategies directly into a diagnostic system. Their approach is general, but it suuers from a lack of declarativeness and exibility. All strategies have to be implemented directly as part of the system BD93, BD94]. Damm asio, Nejdl and Pereira deened preferred revisions