Visualization of Trustworthiness Graphs

Stephen D. Mayhew, Dan Roth

Abstract: Trustworthiness is a field of research that seeks to estimate the credibility of information by using knowledge of the source of the information. The most interesting form of this problem is when different pieces of information share sources, and when there is conflicting information from different sources. This model can be naturally represented as a bipartite graph. In order to understand this data well, it is important to have several methods of exploring it. A good visualization can help to understand the problem in a way that no simple statistics can. This paper defines several desiderata for a "good" visualization and presents three different visualization methods for trustworthiness graphs. The first visualization method is simply a naive bipartite layout, which is infeasible in nearly all cases. The second method is a physics-based graph layout that reveals some interesting and important structure of the graph. The third method is an orthogonal approach based on the adjacency matrix representation of a graph, but with many improvements that give valuable insights into the structure of the trustworthiness graph. We present interactive web-based software for the third form of visualization.