Extracting Pairwise Comparisons Data from Best-Worst Scaling Surveys by Logical Inference

Best-Worst Scaling (BWS) based user interfaces allow efficient evaluation of many items through a low number of user actions. In addition to the explicit selection of the best and the worst item, more paired comparisons can be extracted from a BWS set using the unselected items. In this report, we explore how to extract further paired comparisons by logical inference between two BWS sets. Logically inferred pairs can be interpreted as implicit choices and used to validate the consistency of explicitly extracted pairs or to enrich datasets or resp. reduce data collection costs.