In order to maximize the potential of decision based design, optimization should be an integral part of the process. In this paper, several guidelines are proposed while assessing and fitting utility functions, defining constraint sets, defining backtracking rules and choosing optimization algorithms. We discuss the central role convexity plays in optimization and how it is relevant to maximizing utility functions of risk-averse decision makers, which are the most common type of decision makers. Strong convexity and Lipschitz continuity are also relevant to the discussion since they allow finding optimal designs and facilitate proofs of convergence. The commonly used exponential form is shown to have these properties. We further discuss the motivation to backtrack within an optimization process in order to account for varying preferences of the decision maker across the design space. Design implications of these guidelines are also covered. We conclude with a summary of guidelines and observations. A discussion on future work regarding optimization under uncertain designer preferences is also presented.

This content is only available via PDF.
You do not currently have access to this content.