With a nudge from Data Elixir by my colleague Lon Riesberg, I read the Reddit Ask-Me-Anything marathon by Tamara Munzner, a respected professor, researcher and author in data visualization. She was asked whether virtual reality (VR) is a valuable tool for education and research in data visualization. Her short answer was ‘NO’ because 3D is only needed for spatial settings and the cost of VR, broadly defined, outweighs its benefits. Her long answer was thorough, deserving careful consideration. So, let’s unpack her response, surfacing take-aways for the IA community.
First, Munzner noted that, “so far the VR hardware technology forces you to pay a nontrivial price for immersion and a sense of presence.” In other words, current VR technology is awkward and unstable having drawbacks, such as pixel resolution limits, physical fatigue, latency distortion, and poor task switching. She admitted that the technology is getting better, citing her experience with SteamVR by Value Software. She was surprised at her scale-driven emotional response to virtual objects that initially appeared small but then appeared huge.
I agree! I remember the moments of extreme lag in Second Life and nausea from a roller-coaster using Oculus DK1. A helpful example comes to mind. SCUBA is a sport that is burdened with costly and awkward equipment, as I experienced during certification. Many have tried the SCUBA experience and have chose NOT to partake because of this and other reasons. On the other hand, once immersed in 30-40 feet of clear blue ocean next to a reef, the benefits, clearly and uniquely, outweighed the costs for me. The same is true for VR, regardless of VR improvements in the near future. The take-away is to tread lightly with VR technology, using the minimal to achieve the objective. If that dictates using old 2D vis, so be it.
Second, she noted that, “immersive VR is justified if and only if 3D spatial layouts are justified” such as naturally 3D systems like weather modeling, particle physics, or planetary science.
I agree, except for the if-and-only-if. There are stunning examples of 3D vis, which are both useful and beautiful. My son, Eric Hackathorn, leads a NOAA research team that released the SOS Explorer tool used for spherical earth projections by hundreds of museums and schools. The design of 3D visualization is delicate, requiring more skill and interactivity. The issue is what systems are naturally 3D, which may not be exclusive to science applications. Advanced analytics may transform boring 2D business data into naturally 3D systems, both in appearance and behavior …which is a hypothesis to be proven. The take-away is that we should place higher priority on examining the hypothesis that virtual data worlds driven by analytics (beyond static descriptive visualizations) changes the game.
Third, she has documented that human factors research shows little or no benefits from 3D visualizations [in her Visualization Analysis and Design text around page 129]. This opinion is shared by other leading researchers (as verified in personal conversations with Edward Tufte and Pat Hanrahan). The reasons are attributed to the distortion from parallax and occlusion effects that motion resolves but “imposes cognitive load or requires conscious attention”.
I agree that a static 3D visualizations are close to useless, by an epsilon ε. However, I have witnessed examples where interactive immersive spaces have been highly effective, such as 3D scenes by van Gogh and Near Star Plots. Like SCUBA, the benefits must outweigh the cost. And, Munzner’s mantra No Unjustified 3D deserves constant respect. Hence, it is not sufficient for virtual data worlds to be close to or even slightly better than the effectiveness of 2D infographics. The take-away is that virtual data worlds should provide unique capabilities (such as collaboration) and be synergistic to 2D visualizations (blending them in-world).
One final conclusion seems certain. If the IA community naively uses virtual data worlds to extend 2D visualizations by one more dimension, we will fail to find effective use cases.