Does the Visual Information Seeking Mantra Hold in 3D?
Understanding physical navigation in 3D scatterplots.
Dr. Ben Shneiderman’s Mantra for Visual Information Seeking — “Overview first, zoom and filter, then details-on-demand” — is well-known and has given rise to several visual navigation techniques that follow it. Most of these techniques are used to interact with 2D visualizations on a flat screen, but their effectiveness at navigating and interpreting 3D data in a virtual space is still unproven.
There are some key differences between 2D and 3D visualizations. The first is that in 2D, the user can always see the entire screen whereas in an immersive environment, the user’s field of view is always changing and the user may need to physically move around to see the entire visualization. Research has also shown that larger displays consistently outperform smaller displays for 2D visualizations, but there is no conclusive evidence of the same for immersive visualizations.
In 2020, researchers from Harvard and Monash University conducted an experiment to check which method was more effective at interacting with a 3D scatterplot: an overview+detail interface or a zooming interface. The four main types of interfaces tested were the following:
1) A room sized interface in which the user could either move around physically or teleport to a desired point using handheld controllers.
2) A similar room sized interface, but with the addition of an ‘overview’ which was a miniature version of the room. Users could bring up the overview at any time and could shrink it, enlarge it, rotate it and use it to teleport themselves anywhere in the room. The overview also represented the user’s position and field of view.
3) A zooming interface that started by showing the entire scatter plot and let users zoom in and out as required. The handheld controls were latched to the interface to help orient the users while the zoomed in and out.
4) A similar zooming interface with an overview.
For details, see this video showcasing the four interfaces tested.
Participants were asked to perform certain common visualization tasks like comparing distances between two pairs of points and counting the number of points in a cluster. Their accuracy and response time was recorded for each type of interface along with the number of action performed, like teleportation and zooming. Participants were also asked about their individual opinions about each interface.
Results
It was discovered that while an overview was beneficial in the room sized interface, many participants felt that it was obstructive in the zooming interface. It can be concluded that an overview is not really required in a table sized zoomable interface. An overview should also be non-distracting and only come up on user demand.
Another surprising finding was that even though the room sized interface was efficient and user accuracy and response time was comparable to other interfaces, users did not prefer using it. One possible explanation for this is that the physical effort involved in navigating this interface is tiresome. Another reason is that since the users could not see the whole dataset at once, it placed a higher load on their memory.
Providing an overview in the room sized interface help significantly improve accuracy, at the cost of response time. This interface was also more usable than the one without an overview, according to users.
The zoomable interface without the overview was the most popular among users. It also performed better in the counting tasks and had much less physical movement and teleportation.
Therefore, there isn’t a single interface that is unquestionably superior while interacting with 3D visualizations in virtual reality. Different interfaces are suited for different tasks, overviews are helpful for room sized interfaces and zoomable interfaces are preferred for seated users. 2D visualization techniques have been developed and refined over decades, and with time, new techniques for 3D visualization will evolve too.
Citation
- Y. Yang, M. Cordeil, J. Beyer, T. Dwyer, K. Marriott and H. Pfister, “Embodied Navigation in Immersive Abstract Data Visualization: Is Overview+Detail or Zooming Better for 3D Scatterplots?,” in IEEE Transactions on Visualization and Computer Graphics, vol. 27, no. 2, pp. 1214–1224, Feb. 2021, doi: 10.1109/TVCG.2020.3030427.
