Presenting Signiicant Information in Expert System Explanation Content Areas: Expert Systems, Explanation, User Modeling (using User Models) Presenting Signiicant Information in Expert System Explanation Content Areas: Expert Systems, Explanation, User Modeling (using User Models)

Machine-generated explanations of an expert system's conclusions can be made long and unwieldy by the inclusion of information that is insigniicant to the user requesting the explanation. A portion of an explanation can be insigniicant to the user for one of two reasons: (1) the portion is full of terminology and inferences that the user does not have the expertise to understand, or (2) the portion consists of a reasoning step that adds nothing to the expert's belief that the conclusion is correct. This paper presents a method for generating explanations of conclusions that attempts to present only information that the user will nd signiicant. The method uses a model of the user's understanding of the domain to select only those portions of the expert system's reasoning that add signiicantly to the user's belief that the conclusion is the right one. Several important features and limitations with this technique are identiied, and areas for future research are discussed.