The goal of this project is the development of a general approach to explanation generation for advice-giving systems that use mathematical or other quantitative problem-solving methods. Specific objectives are: (1) Development of a generalized methodology for this type of explanation; (2) Implementation of this methodology in a prototype system that can be used in different applications; and (3) Creation of systems for real-life environments. In this approach to explanation generation, the first task is determining the substance of the explanation, that is, discovering which quantitative values contribute to the decision process. The second task is translating the quantitative values into concepts that are understood by the non-technical user. The significance of this research lies in the development of rigorous methods for performing these tasks: how to represent the different conceptual ``worlds'' of the quantitative system and the lay user; how to determine significant factors for explanation; and how to translate between the two worlds. Practical implications include the possibility of d eveloping explanation generators for `off-the-shelf' decision support systems used by individuals who are not familiar with the formal models. As a result, these systems would be better accepted and more generally used.
