Rank Aggregation and Belief Revision Dynamics

Abstract

In this paper, we compare several popular rank aggregation methods in terms of their accuracy of finding the true (correct) ranked list. Our research reveals that under most common circumstances simple methods such as the average or majority actually tend to outperform computationally-intensive distance-based methods. We then conduct a study to compare how actual people aggregate ranks in a group setting. Our finding is that individuals tend to adopt the group mean in a third of all revisions, making it the most popular strategy for belief revision.


Back to Table of Contents