Visual analytics is an emerging discipline in which human analysts are tasked with using interactive visualization tools to make judgments and derive insight from large amounts of dynamically changing information. In a complex and longrunning analysis process, we hypothesize that there exist meaningful structures of user interaction behavior. Moreover, we believe that these structures may be used to better understand a user’s analytic goals and reasoning, and to guide the design of a visual analytic system.
To validate our hypothesis, we have conducted an empirical study which examines structures of user behavior during visual analysis and their implications. Unlike previous findings, our study focuses on examining fine-grained user visual interaction behavior over the course of a realistic analysis task. We report both our observations and analysis which uncover two key structures of user visual analytic behavior. Both structures are found to influence user task performance and often reflect limitations or user-desired features of visual analytic systems. Based on our findings, we present a series of design recommendations outlining how these structures can be used to improve visual analytic systems.