Dynspace was a system that supported multiple workspaces for Visual Analytics, which enabled users to create multiple workspaces to express and reflect upon the workflow needed for their analytical tasks, such as testing multiple hypotheses. Yet, the authors of DynSpace never evaluated their system with such activities. In this paper, we present two variants of DynSpace (SINGLE and MULTIPLE) to investigate the impact of ease of chart creation and the use of multiple tabs for testing multiple hypotheses.
The first improvement eases chart creation. While DynSpace required dropping dimensions into specific axes, DynSpace MULTIPLE removes this restriction by following VizInteract and automatically creating an appropriate chart depending on the properties of the dimension(s). For a single dropped dimension DynSpace MULTIPLE thus shows a histogram, making it faster for users to see the data. We also added support for transposing (swapping vertical and horizontal axes), as the user may want to have the same dimension occupy the same axis in all charts in a workspace. Relative to the original version of DynSpace, this reduces the need to recreate charts. To verify the effectiveness of this approach, we conducted a user study. According to our overall observations, simple and quick chart creation made it very easy for participants to rapidly explore the data. Sometimes a participant wanted to change the dimension displayed along one of the axes and unsuccessfully dropped a dimension onto a chart area, with the intent to. Still, participants also quickly figured out that creating a new chart enabled them to find answers, often by creating a new chart beside an existing related one. The most frequently performed actions included chart creation, creating a new workspace, switching between workspaces, and using bookmarks. From the set of interactions observed and clustering them across common threads, we were able to identify a list of emerging patterns. Many users used the Overview tab as a focal point for all tasks. Further, many participants created a new tab for each new task they received, but a few participants used a single tab for all tasks, while some used tabs in an unstructured manner.
The second contribution of our work investigates how users leverage multiple tabs in a VA system to pursue multiple hypotheses and analysis paths. In DynSpace SINGLE, the user can only access an “overview” tab, and cannot create or switch to other tabs/workspaces. However, a new “clear” button enables the user to clear all the charts in the current tab (i.e., the overview). In a second user study, we investigated the differences between analysts working with single or multiple workspaces (DynSpace SINGLE and MULTIPLE, respectively). Overall, all users were able to successfully use both versions of the system to analyze the data. Yet, in contrast to DynSpace MULTIPLE, we observed with DynSpace SINGLE that many participants were frustrated by the implicit need to clear all charts after every task. This problem became more acute with the increasing complexity of the tasks. With DynSpace MULTIPLE, we did not observe this issue, as participants naturally used multiple tabs to organize their analysis, similar to the findings from our first study. The idea behind offering the ‘’clear charts’’ button was to see whether clearing the workspace could compensate for the need for multiple tabs. However, this was not the case. The fact that several users explicitly expressed a preference towards multiple tabs for analysis, over clearing existing charts and doing further analysis on the same tab, supports this further. We also observed that participants quickly demonstrated that they understood how to generate new tabs, which contrasts with the slight reluctance to clear charts. None of the participants using multiple tabs mentioned (or displayed) any reluctance to use new tabs for analysis. The slight hesitancy to clear the charts may stem from the concern of loosing important information or previous analyses. This is relevant for complex tasks, where the participant may need even more information for analysis. Overall, participants found it thus easier to perform their analysis through multiple tabs. Several users explored the world dataset in more depth, suggesting that they were curious about the data, even beyond the task questions. Here, it seemed that users intuitively considered several parallel hypotheses and -- once they saw the data -- used the dashboard to explore and verify which hypothesis held. This illustrates that users were more focused on the data than the interface in our study. Our investigation showed that multiple workspaces have benefits for more complex analysis tasks; particularly as users preferred opening a new tab (workspace) rather than clearing an existing tab (a single workspace) to conduct further analysis. We also identified some usage patterns for multiple workspaces.