ABSTRACT
The need for explanation capabilities in intelligent systems, and especially expert systems, has been widely recognized. Most commercially available expert system building tools provide at least rudimentary explanation capabilities. Yet, few rigorous and thorough evaluations have been attempted. This is in part due to the fact that an explanation system is not stand-alone, but occurs as a component of a larger, intelligent system and the quality of a system's explanations depends on factors outside the control of the explanation component. Criteria and methods for evaluating explanation systems are only now beginning to emerge.
