Designing Interactive Visualizations for First-time Novice Users

Information visualization tools provide visual representations of data (commonly known as visualizations), textual representations of data and interactive operations on both these representations. It is possible, in theory, to make use of only the textual representations in order to detect trends and patterns in data. However, it would be extremely laborious and ineffective and it defeats the purpose of a visualization tool. Novices have had the problem of relying on text and failing to successfully detect trends in data because it was too laborious. We want users to effectively use visual representations to detect trends in data.

Information visualization tools have been shown to be successful with experts. But can novice users using a visualization tool for the first-time, adopt visualization-based strategies to finding trends and patterns in data?

This thesis derives a framework of learnable elements in an interactive coordinated-view visualization tool. This framework provides an outline of prerequisites to be learned in order to effectively use visualizations. That is, the new aspects of visualization tools must be mastered so that novices can use the tool effectively.

Three interface design principles are derived to make these elements learnable to novice users:

• The data-first approach - Provide a prominent overview of all available data, as opposed to showing only visualizations. This helps understand the data structure - this may be essential knowledge in being able to navigate to required data attributes.

• The "less visualization, more explanation" approach - Show more explanations of visualizations at the expense of being able to pack in more visualizations. Explanations help identify how data is mapped onto visual marks, a crucial step in understanding visualizations.

• The "predetermined task-based coordinations" approach - Provide separate predetermined sets of coordinated visualizations that help achieve different tasks. This is different from techniques that simply present all visualization types and expect users to choose coordinations according to tasks.

Two versions of Datamaps visualization tool for Census data were tested. Both were equivalent in functionality and the kinds of visualizations offered. But the new version was specifically designed based on the three design principles. A usability study showed that the version that was implemented according to the three design principles successfully led novices to effectively use visualization-based strategies to detect trends and patterns in data.