Many portfolio selection problems require a decision maker``s preference structure. In this study, a procedure for elicitation of decision maker``s preference structure, especially mean-variance utility function, is developed. It is accomplished under the assumptions that, first, money cannot be borrowed and, second, the return may not be utility independent to the risk. The technique of experimental design is utilized to directly assess a quadratic utility function, and it reduces the decision maker(DM)``s burden and time consumptions. This porcedure provides a decision maker with practical interview process that is set up interactively for personal and micro-computer. After obtaining the decision maker``s preference structure, Nonlinear Goal Programming is implemented for solving portfolio selection problems. This study suggests a way of practical portfolio selection in real world.