Dissecting Visual Analytics: Comparing Frameworks for Interpreting and Modelling Observed Visual Analytics Behavior

Abstract

This paper provides an empirical, comparative exploration of the role of analytic frameworks in interpreting and modelling visual analytics behavior through data gathered in observational studies. The crucial research on understanding the complex and multi-faceted interplay between visual analytics tools and their users is often done through controlled or naturalistic observations of analysts engaging in the visual analytic process, followed by the interpretation of the observation data. The researchers in Human Computer Interaction and Cognitive Sciences have long used structured analytic frameworks for such analyses, where a guiding set of principles and questions direct attention to relevant aspects of the studied behavior, eventually leading to more complete and consistent analyses. Such frameworks are rarely applied in the visualization domain however, and information about how to apply them and their benefits is scarce. With this paper, we contribute a comparative account, grounded in empirical data collected in a user study with 10 participants using Tableau to analyze domain-specific data, of the types of insights we can glean from interpreting observational data using three different frameworks: Joint Action Theory, Distributed Cognition, and Situated Cognition.