A System to Detect Inconsistencies between a Domain Expert's Different Perspectives on (Classification) Tasks

This paper discusses the range of knowledge acquisition, including machine learning, approaches used to develop knowledge bases for Intelligent Systems. Specifically, this paper focuses on developing techniques which enable an expert to detect inconsistencies in 2 (or more) perspectives that the expert might have on the same (classification) task. Further, the INSIGHT system has been developed to provide a tool which supports domain experts exploring, and removing, the inconsistencies in their conceptualization of a task. We report here a study of Intensive Care physicians reconciling 2 perspectives on their patients. The high level task which the physicians had set themselves was to classify, on a 5 point scale (A-E), the hourly reports produced by the Unit’s patient management system. The 2 perspectives provided to INSIGHT were an annotated set of patient records where the expert had selected the appropriate category to describe that snapshot of the patient, and a set of rules which are able to classify the various time points on the same 5-point scale. Inconsistencies between these 2 perspectives are displayed as a confusion matrix; moreover INSIGHT then allows the expert to revise both the annotated datasets (correcting data errors, and/or changing the assigned categories) and the actual rule-set. Each expert achieved a very high degree of consensus between his refined knowledge sources (i.e., annotated hourly patient records and the rule-set). Further, the consensus between the 2 experts was ~95%. The paper concludes by outlining some of the follow-up studies planned with both INSIGHT and this general approach.