A User Interface for Consistent AHP Pairwise Comparisons

Abstract

Decision Makers generally reason on several criteria, aiming to obtain a total consistency or partial order of several alternatives. MultiCriteria analysis is based on the assumption that such ordering exists. Decision Makers are supported by several kinds of approaches or tools. One approach consists in comparing the criteria two by two, i.e. pairwise comparison, to find the relative importance of each criterion. This relative importance, called weight of criteria, is used to find the final order of alternatives. One methodology, developed by Saaty, called Analytical Hierarchical Process (AHP), is based on this principle of pairwise comparison. Having the weights of criteria, the decision makers have then to compare the alternatives two by two for each criterion. Pairwise comparisons are simple to use; however, as the number of items to compare increases, so do the effort of conducting all comparisons and the probability of introducing inconsistencies. In this article we present an innovative approach to conduct pairwise comparisons based on a UI widget that resembles an interactive data plot. It uses the transitivity property of a consistent comparison matrix to infer comparisons. Our hypothesis is that this new approach is more efficient (as it reduces the number of actions the user must conduct to compare all items), more effective (as it limits the sources of inconsistencies), and yields better user satisfaction. We conducted a controlled experiment involving 50 participants. We observed that the proposed widget reduces the effort of making pairwise comparisons, improves the consistency of the comparisons, and leads to a better user experience.