Improving Human Scaling Reliability

Abstract

In order to measure the subjective experience of system users, the field of human factors makes extensive use of classical scaling methods. In this paper, the applicability of a new scaling method for human factors research is demonstrated. Constrained scaling, a technique for training individuals to translate mental magnitudes to numeric scales, is introduced. Constrained scaling has been found to reduce significantly the variability in scale use between individuals. Prior research has focused extensively on psychophysical constrained scaling. As an example of how constrained scaling can also improve the quality of psychometric measures available to human factors researchers, new research is presented that compares classical scaling with constrained scaling for rating the visual appeal of Web pages. Constrained scaling is found to increase scaling reliability in this subjective domain.