Incorporating end-users' attitudes towards uncertainty into an expert system

Uncertainty is a real and unavoidable part of the design process. Uncertainty in the area of manufacturing cost estimation can have a negative impact on the desirability of design alternatives, and some might be deemed ``too risky'' for further consideration in spite of potential benefits. Expert designers obtain a sense of what level of risk is acceptable for different types of applications. Expert systems for design thus might include a set of necessary and reasonable assumptions which reflect a particular expert's attitude towards risk and uncertainty. These assumptions may not be applicable for all users. In this paper we present an expert system for design which explicitly addresses the individual user's attitude towards uncertainty. We have developed a spreadsheet user-interface for assessing this attitude through a utility function, and for linking those results to a heuristic rule base. The heuristic rule base contains only technical expertise used to determine a set of feasible alternatives, while the process of evaluating and ranking these feasible alternatives is driven by the user's individual preferences and risk-taking profile.