Network Visualization Literacy: Task, Context, and Layout

Abstract

Information visualization as a practice is becoming increasingly global, being conducted by and distributed to increasingly diverse stakeholder groups. Visualizations are being viewed in casual contexts and for a variety of purposes. The use of network visualizations has likewise increased in recent years, in part because network visualizations have properties that are applicable to datasets ranging from academic journal and patent citations to molecular interactions to the movement of refugees across national borders. Unlike charts based on numerical or categorical axes, common network visualizations operate under a set of rules that are largely unexplained to the users of the diagrams. For example, unlike axis-based charts, there is no stable reference system across node-link diagrams. The same dataset can produce many visualizations that look very different from each other, depending on the choice of layout algorithm, rotation, data thresholding, etc. Research on the skills required to interpret network visualizations and the prevalence of those skills have typically been small in scale–limited to a small group of users or a limited set of visualization design choices. With the broadening of the audiences for visualizations and the dissemination of more sophisticated visualization types, a detailed examination of the typical skills of a novice viewer of network visualizations is crucial to the development of appropriate and successful visualizations.This dissertation advances our understanding of network visualization literacy by studying performance of both novices and experts in network science on a variety of network analysis tasks and datasets using a variety of visualization designs. The empirical results will provide a baseline for understanding network visualization usage and will offer advice to visualization designers on the design features that best support particular tasks.