Reference: Klein, D. A.; Weber, M.; & Shortliffe, E. H. A Framework for Refining Multiattribute Value Functions in Computer-Based Decision Aids. 1991.
Abstract: Computer-based decision aids encourage the iterative, interactive refinement of preference functions by providing convenient facilities for inspecting, modifying, and executing such functions. The operational contexts in which such automated tools are employed, however, introduce at least two new challenges in refinement: (1) We require a methodology for focusing the decision maker's attention on both sensitive and insensitive parameter values that may be inconsistent with his preferences and (2) this methodology must be accessible to decision makers who lack substantial decision-analytic training. In this paper, we present a computer-based framework for refining multiattribute value functions that addresses these challenges. The framework clarifies the qualitative relationships among value-function parameters and decisions to promote refinement, imposing organization on the sensitivity- analysis tools and parameter-management facilities commonly provided by automated aids for multiattribute decision making. Our methodology illustrates the feasibility of interpreting and guiding the refinement of preference functions automatically in computer-based decision aids.