Structural model refinement under uncertainty using decision-maker preferences

In this research, a framework is presented for making decisions with respect to model development. In particular, the proposed framework is applicable whenever a decision-maker (DM) must first choose between information sources, and then choose among structurally different models. Models serve as representations of reality that help DMs understand input-output relationships, answer what-if' questions, and find optimal design solutions. Model development seeks to improve the accuracy of models, with an ultimate goal of making better design decisions. Unfortunately, the desire to develop the most accurate model competes with the desire to reduce costs, and the DM is ultimately required to make trade-offs when making model development decisions. In the proposed framework, conjoint analysis is implemented in order to solicit DM preferences for competing attributes, and imprecision in attributes are systematically assigned and propagated through the framework. Using only preliminary experimentation results from a packaging design example, the framework is demonstrated to support model development decision-making under uncertainty. The primary contribution to model development is the capability of the framework to simultaneously evaluate multiple information sources and structurally different models.