This chapter presents an approach that takes advantage of the different roles that computers and humans play in an interactive engineering design environment. It draws on the positive features of learning-oriented methods and searching-oriented methods, thus adapting design trade-off strategy when more precise preference information is learned during the evolutionary search process. The rationale and advantages of evaluating design fitness based on a fuzzy-set based preference aggregation are provided, which not only relies on specifying parameters about importance weights of different design attributes, but also the degree of compensation among them. The designers’ preferences are elicited, and the parameter learning of the preference aggregation function is implemented in an artificial neural network. Guided by online adaptive fitness evaluation, the current favorable solution set is generated by means of evolutionary computation through a component-based design synthesis approach. An example problem of panel meter design configuration is provided to demonstrate the approach.