HistoryViewer: Instrumenting a visual analytics application to support revisiting a session of interactive data analysis

Abstract

Visual analytics applications (VAApps) rely heavily on visual representations and notations to communicate information and support user interaction. The visual representations themselves are the main communication form used by designers, reflecting their interpretation of what data aspects should be highlighted and how users can explore the data. Ultimately, the combination of representation and interactivity GUIdes the user’s data analysis in VAApp. To keep the VAApp development centered in their users, the design team must pay special attention to the visual representations and their underlying interaction mechanisms. Our team is developing HistoryViewer, a VAApp itself for users to explore log data obtained from interactions with other (“source") VAApps (VAAppsrc). During the development of HistoryViewer and the instrumentation of a VAAppsrc, we noticed unanticipated benefits for the development of the VAAppsrc itself. This paper presents our HistoryViewer system, proposes an architecture for VAApps integrated with our log model, discussing some ancillary benefits of such instrumentation. Finally, it reports two empirical studies conducted to gather potential users’ opinions about HistoryViewer, which revealed positive attitudes towards the system and pointed to opportunities for improvement.