Authors: Kathrin Ballweg, Margit Pohl, Guenter Wallner, Tatiana von Landesberger
Abstract: A precise and valid measurement of perceived similarity is essential for evaluating comparative visualizations. Our experiment on the perceived similarity of directed acyclic graphs showed that the common Likert scale measurement lacks precision and challenges the measurements' validity due to a mismatch of the Likert scale with the preferences and the mental model of humans. We therefore searched for an alternative measuring instrument. By broadening our search scope we found Tversky's model of perceived pairwise similarity. It provides a valid and well-known model of human similarity perception. So, we transferred it into a measuring instrument. While our proposal still needs to be validated. We think it is capable of solving the Likert scale's issues and, therefore, it will provide substantial benefit for comparative visualization research.