Kjellin et al. examine how several test subjects interpret and predict the eventual outcomes of three visualization experiments. They created 2D and 3D maps of a fictitious landscape, using tracks to show the progress of vehicles. The object was to have the test subjects predict where the vehicles would meet, given their recent histories.
The paper provides a generous discussion of 2D and 3D representations. The discussion is based mainly in the military domain. With traditional 2D maps, icons on the map represent the progress of military units. (Remember the War Rooms during WWII, where clerks received intelligence and moved elements of units across the map board.) Here, the time marks on the trail left by the unit can represent the speed of an item of interest. The closer together the marks are, the slower the units move. With the advent of computers and extensive databases, we can automate these maps. The question arises: “How can we program the computer to create a better representation of reality?”
The authors thus created their three modes:
- A flat 2D model with vehicle tracks. Each track has hash marks at constant time intervals to indicate the speed of the vehicle. Terrain is indicated by shades of gray.
- A 3D “time cube,” where the added third dimension is time. The steeper the slope of the vehicle in the time domain, the faster the vehicle is going.
- A 3D animation showing the motion of the vehicles as they go along, and move up and down hills.
The authors created a map environment for use by the subjects. In the second and third mode, the test subjects wore polarized glasses so that they could make use of the illusion of 3D representation.
Kjellin et al. describe the great care taken to make certain that the test subjects only relied on the map visuals. They employed a series of three experiments. The first asked the question, “Given the past several hours of tracks, where would the vehicles meet 4.5 hours in the future?” The second used an improved rendering technique. The third asked, “What is the arrival order of the vehicles at the rendezvous?”
Each subject looked at 72 trials: six pretests, 60 learning tests, and six post-tests.
Although the test results indicate that the 2D visualization provided better results, the authors observe that the 2D representation is the most familiar. Furthermore, with the advances in the computer gaming world and the advent of glorious 3D (see Avatar, and even Shrek), there is room for more research on 3D representation.
The “time cube” is probably a nonstarter. We can perhaps await the next iteration of the experiment with better graphics.
