Adaptive Design Augmentation Strategies for Visual Product Evaluation

In order to assess design of new products, the visual evaluation (especially in the concept phase) gained an extraordinary usefulness thank to the availability of more and more realistic digital moke-ups. Then several alternative features could be affordably compared in a very short time, so giving significant glues to the designer toward an optimal design. Obviously, a careful design of the experiment is the essential prerequisite to avoid confounding as idiosyncrasies and halo effects. Although the cost constrains are at a large extent relaxed (since in expensive VR experiments the candidate designs could be simultaneously assessed by a jury of many members who secretly express their evaluations) the duration of the experimental session (due to the tiring of each respondent) keeps as a limiting factor in the design of such an experiment. Conversely the need to compare many candidate designs in the form of different combinations of alternative features, ask for factorial designs which remain cumbersome even after high reduction. We reviewed some strategies to (also dynamically) augment an initial cheap and very coarse experimental design in order to obtain a quick convergence to the design evaluated as optimal by a single respondent. Therefore we can demonstrate how these strategies perform on the base of some testcases developed at Department of Aerospace Engineering. Finally the strategy which revealed the most efficient will be discussed in the statistical issues of data pooling too.