What is visualization? When asked of data and information visualization professionals, answers will generally swirl around one of two punch lines…a visual tool that aids in (1) analysis or (2) communication of information. Last week in Chicago at IEEE VIS—an annual conference focused on the tools and technology of visualizing information—Purdue Polytechnic professor Vetria Byrd reminded me that visualization is also a process, one that requires us to think critically and sequentially though the information at-hand.
Themes and Benchmarks
As a first-time attendee of IEEE VIS, I was hoping to identify themes across presentations, in an effort to weave together a story and make sense of the barrage of information ahead of me. It quickly became clear that with 169 paper presentations, 122 posters, and seven panels across three major tracts—Visual Analytics Science and Technology, Information Visualization, and Scientific Visualization—and several co-located sub-conferences, identifying research themes would be almost arbitrary. I could pretty much choose a central idea, and find enough presentations to support my preconceptions. I'm a long time practitioner of data visualization in science communication, but just now getting familiar with the primary literature. So, rather than attempt to identify major themes from my relatively naive viewpoint, I decided to defer to the research experts and highlight a handful of benchmark projects; specifically, papers that were deemed award-worthy by the program committee. For the full list, see the conference portal. A select group of papers will appear in a January 2016 issue of IEEE Transactions on Visualization and Computer Graphics (TVGG).
Best paper: Visual Analytics Science & Technology
Reducing Snapshots to Points: A Visual Analytics Approach to Dynamic Network Exploration
By Stef van den Elzen, Danny Holten, Jorik Blaas, and Jarke J. van Wijk
Best paper: Scientific Visualization
Visualization-by-Sketching: An Artist’s Interface for Creating Multivariate Time-Varying Data Visualizations
By David Schroeder and Daniel F. Keefe
A few researchers revisited celebrated publications from the past. Matthew Kay and Jeffrey Heer, from the University of Washington Interactive Data Lab, looked at the data behind a study by Lane Harrison et al. in 2014 that ranked the effectiveness of visualization types, and emerged with a refined punch line. When it comes to choosing a plot type that allows the reader to make the most accurate determination of the relationship between two variables, Kay and Heer suggest that scatterplots are your best bet, not parallel coordinate charts. And in the process, they highlighted the importance—and their appreciation—of the public release of data and open communication across research groups, as a way to strengthen the field as a whole.
I was also happy to see that Northeastern University’s Michelle Borkin and colleagues are pushing forward with research that builds on their 2013 study that looked at the elements that make a visualization memorable. Their latest questions revolve around recognition and recall. A standout—although perhaps unsurprising—finding; "…during the recall phase of the experiment, titles were the element most likely to be described if present" (my italics).
Bridging the Researcher/Practitioner Divide
Whereas the primary sessions focused on research projects, skewing largely towards perspectives and projects from academia, several panels throughout the week provided a forum for discussion across academia, government, and industry.
Sadly, I missed the color discussion, and thus meeting Cynthia Brewer, editor of the revered ColorBrewer tool. Here's one comment that emerged from the panel:
For further discussion on alternatives to the overused rainbow palette, check out a classic post on the topic, by Rob Simmon.
A panel on teaching across the researcher-practitioner gap focused on college, graduate school, and continued education. Insights and resources from the panel were rich and varied, from Tamara Munzer's literature-focused syllabus at the University of British Columbia, to Marti Hearst's active-learning and tool-centric approach at UC Berkeley. Jon Schwabish of PolicyViz declared that even as the community embraces cutting-edge tools for developing visualizations, we shouldn’t ignore that Microsoft Excel is the default tool across many industries. He added that we wouldn’t be realistic if in addition to the latest tools, we don't also teach folks how make effective visualizations with the ones they already have.
Byrd articulated her above comments in a panel that nodded to visualization as a pathway to STEM for under-represented groups. Visualization is not discipline-dependent. She suggested that we could engage students that don't necessarily self-identify as computer scientists, and encourage them to think critically about—and visualize—data and information in their existing area of interest, from science, to the humanities, sports, and beyond. Joseph Cottam, a researcher at Indiana University, echoed this sentiment in the same panel when speaking of his experiences teaching high school workshops: When a visualization project allows individuals to see themselves—or their interests—in data, it can act as a powerful engagement tool.
Perhaps that's why I gravitated towards panel discussions during the course of the week. I wanted to hear from folks that share my concerns and challenges. My questions were best addressed by multidisciplinary conversations between researchers and practitioners. What conclusions from the research side could, and should, be applied in industry? And what can researchers learn from those of us creating visualizations outside of academia?
I got some answers, I suppose. Scatterplots and titles are good. Rainbow color palettes are bad (duh). But perhaps more importantly, I'm better acquainted with the larger visualization community, and ongoing efforts to bridge the gap between research and practice.