Measuring affective responses to vehicle interior textures using Paired Comparisons

The design trend in the automotive industry is currently moving towards exploring more innovative ways of redefining the craftsmanship quality of the vehicle interior trim through perceived quality. Affective engineering has been widely used as a robust method for understanding people’ affective responses to vehicle quality attributes, also being aesthetically and physically pleasing. However, the research has identified the Semantic Differential scale and Likert scale are subject to biases and errors in assessing perceived quality attributes, resulting in non-linear measurements which ends up with poor reliability outcomes. In this study, the affective engineering approach has introduced pair comparison technique in order to measure valid and reliable participants’ affective responses using the multivariate statistical analyses of the Rasch model — with the objective to establish the linear correlation between participants’ affective responses to physical of multisensory cues of touch, vision and feeling of interior vehicle textures. In this research, the use of Rasch analysis of paired comparisons of products to derive a linear measurement of affective response is tested. Seven pieces of interior vehicle textures and nine unidimensionally fit statements to measure the dimension of perceived quality attributes. A computer-based self-report system presented one hundred and sixty-nine participants with pictures of pairs of stimuli and the evaluative statements in all combinations, and the participants were asked to indicate which stimuli satisfied the statement best. The analysis demonstrates the viability of using Rasch analysis to obtain measures of affective response from paired comparisons that participants find the choice faster and easier to make paired comparisons compared with evaluating products separately against the Likert scale. It has improved biases and error where the participants no longer make difficult judgements but that in this case, the fit of the data to the Rasch model is very poor.