Assessing the Real-to-Artificial Generalizability of Multiattribute Attitude Models in Tests of New Product Designs

Abstract

The authors investigate the neglected problem of generalizing from multiattribute attitude models based on data for actual brands of real products to reach conclusions on affective responses toward artificial objects or hypothetical concepts, such as those that might be used to develop new product designs. They propose an approach involving four stages of analysis to examine the intervening role of multiattribute attitude models for real products in accounting for the effects of artificial design features on affective responses. The method decomposes the overall relationship of affect to artificial design features into (1) the mediating effects of a multiattribute attitude model based on real products, (2) other indirect effects, and (3) a residual direct effect. This general approach is illustrated and supported by an application to data on responses to male vocal recordings (real products) and alternative recording mixes (artificial objects). The results suggest the need to proceed with caution in using multiattribute attitude models based on real products to guide the design of artificial objects in laboratory tests oriented toward new product development.