As I wait to board on the plane back from Salt Lake City, let me try to collect the last shreds of coherent thought I might have from this great week. Jerome Cukier mentioned on twitter he slept less than 30 hours throughout the whole week. I counted 35-40 hours over 7 days; not as bad, but still I can no longer think about anything very much.
One trend mentioned on the hallways, and at least unofficially confirmed by attendance data, is that SciVis seems to be having a harder time getting attention than InfoVis (I really don’t know about VAST). This seems to match, at least in part, with my experience throughout the week. For example, Sam Gerber and co-authors’ work on high-dimensional visualization paper was presented at SciVis, but it seems broadly applicable to InfoVis as well. As a whole, however, we are doing very well, and had record attendance. While VisWeek is nowhere as large as UIST or CHI, if the trend continues we will be soon flirting with 1000 attendees. That’s really great.
To me, the most intriguing papers this year had to do with foundational issues. I already talked about the work of Wickham et al., which tries to bridge the visualization and statistical ways of attacking problems. But I was particularly surprised to learn about many papers which, in a sense, are actively looking for principles of vis effectiveness. A comment by Pat Hanrahan on the VAST panel was particularly important; he said, roughly, that “the community has had enough failures that it is now worth studying them in detail”.
As visualization becomes an approachable tool for data analysts “out there”, we will have to fundamentally shift our way of doing research. This is an exciting future, but it’s a different one than what I saw in maybe a third of the papers this year. The way in which the vis community realigns itself along the axis separating basic science from engineering practice is going to have a large impact in the field. We should all pay attention.