Normality and Faults in Logic-Based Diagnosis

Is there one logical de nition of diagnosis? In this paper I argue that the answer to this question is \no". This paper is about the pragmatics of using logic for diagnosis we show how two popular proposals for using logic for diagnosis, (namely abductive and consistency-based approaches) can be used to solve diagnostic tasks. The cases with only knowledge about how normal components work (any deviation being an error) and where there are fault models (we try to nd a covering of the observations) are considered as well as the continuum between. The result is that there are two fundamentally di erent, but equally powerful diagnostic paradigms. They require di erent knowledge about the world, and di erent ways to think about a domain. This result indicates that there may not be an axiomatisationof a domainthatis independent ofhowthe knowledge is to be used.

• Type: Conference
• Year: 1989
• Where: IJCAI
• Authors: David Poole
• Keywords: Di Erent Knowledge, Diagnostic Tasks, IJCAI 1989, IJCAI 2007, Powerful Diagnostic Paradigms