On the Faithfulness of Graph Visualizations

Introduction

Graph drawing algorithms developed over the past 30 years aim to produce “readable” pictures of graphs. Here “readability” is measured by aesthetic criteria, such as few crossings or few edge bends or small grid drawing area. However, the readability criteria for visualizing graphs, though necessary, are not sufficient for effective graph visualization.

This poster introduces another kind of criterion, generically called “faithfulness”. Intuitively, a graph drawing algorithm is “faithful” if it maps different graphs to distinct drawings. Faithfulness criteria are especially relevant for modern methods that handle very large and complex graphs; data reduction or aggregation or generalisation are commonly exercised to enhance readability.