VisWeek updates by Jérôme Cukier: Day 5

The IEEE VisWeek Conference 2011 is taking place in Providence, RI this week (23rd to 28th). VisWeek 2011 is the premier forum for visualization advances for academia, government, and industry, bringing together researchers and practitioners with a shared interest in tools, techniques, technology and theory.

The week is organized around three separate conferences IEEE Visualization 2011, the venue for all visualization research for data that has an intrinsic spatial component, IEEE Information Visualization 2011 focused on research relating to visual mappings of non-spatial data and interaction techniques and IEEE Visual Analytics Science and Technology 2011 which concerns the reasoning processes involved in visual analysis and the application of visual environments to generate useful insight about real-world problems.
I’m disappointed to not be able to attend the event this week but am delighted that Jérôme Cukier has very kindly agreed to provide updates of his discoveries, reactions and experiences. I’m particularly pleased to provide a platform for Jérôme’s updates because I consider him to be one of the most astute and thoughtful observers within the visualisation field.
 

Day Five – Thursday 26th October

Remarkable Presentations
Georgia Albuquerque and colleagues from TU Braunschweig presented an incredible tool in a paper called “Synthetic Generation of High-Dimensional Datasets”. The tool and the accompanying paper are available here. Simply put, what this does is let a user “paint” a dataset with a few mouse strokes. Let’s suppose you want a dataset with data following a certain pattern, or with certain correlations: with this interface you can easily generate data that approach the form you want. This is plotting data…. but in reverse.

In the applications section there has been two papers that I really liked. The first, ‘BallotMaps: Detecting Name Bias in Alphabetically Ordered Ballot Papers’, struck me as how visualization techniques can contribute actively to UK democracy. The authors, around Jo Wood, wanted to see whether the order in which names appear in a voting ballot have an influence on the voter choices. There findings confirmed this, and also that candidates with non-English/Celtic names garnered less votes than expected.

The image above shows one of the many visuals of the ballotmaps research. If there were no bias, green and purple squares would be equally spread on the screen. The fact that darker screens are on the top means that names which are towards the top of the ballot are favored. The presentation of the study was very aesthetically pleasing and convincing, and with these results, the researchers approached the British electoral commission with good hopes that their practices change.
Speaking of aesthetics, Danielle Albers presented Scalable Surveyor which is a very impressive tool to explore and compare genomes. I remembered seeing an earlier version as a poster last year that struck me as a balance of design and function. And for those of us who are not obsessed with genetics, the last slides suggested the tool could be used for literature analysis or to study datasets like the google N-grams.

I also really liked two of the papers in the Evaluations track and for similar reasons. By this I mean that I’m not necessarily talking about the most innovative techniques or the ones which will are likely to be most cited, which would be way beyond my capacities. Rather, what I’m interested in in VisWeek in general are techniques that relate to the real world, either because they try to tackle a practical problem, or because they could be easily adapted to a real world use.
This is why I was very impressed by Michelle Borkin’s presentation on “Evaluation of Artery Visualizations for Heart Disease Diagnosis”. The technique presented was a new way to detect heart disease.

In fact there were three remarkable things in the presentation. First, the subject and motivation. Heart disease is the first cause of mortality in all western countries, so anything that can help here has tremendous potential. The presentation style was also one of the best in the conference. While the subject was technical, Michelle made her presentation very accessible, I think everybody understood the problem, the contribution of the technique and the impact. And of course, the results of the technique were incredible. With state-of-the-arts tools, doctors are capable of detecting 39% of the zones impacted by disease. With the new technique, the rate bumps up to 91%! On top of this, this paper is iconic of the visweek discussions as one key change is the move from the rainbow colormap, universally decried by the Vis community.
The following presentation was from Lyn Bartram on Ambient and Artistic Visualization for Residential Energy Use Feedback, or in other words, in devices that show power consumption, with a catch. Those in the study were unobtrusive, designed not to stick out in the house, and artistic, so more focused on the aesthetics than on conveying exact information. The image below is one example of ambient power consumption displays, where the the more waves on the backsplash, the more power the house is using.

In any of the displays presented, in fact, users couldn’t read data directly. But they could guess if there was a power draw or not, or if their current activity was causing high consumption or not.
The Capstone Talk
We were very fortunate to welcome Amanda Cox to close this amazing week. Amanda is a very cool choice for a capstone. For one, everybody knew her work. Also, while what she does is very related to visweek-style visualization, it requires very different qualities. V isualization for storytelling is not the main focus of the infovis research. And conversely, this research isn’t always constrained by strict deadlines or design. Yet she and researchers have a lot of concepts in common.
With that in mind, she helped us understand how she work. The one point she really wanted to carry across is that as journalists, they don’t show a bunch of data, but first find an angle which they then illustrate with data. In her words: what if the front page of the Times read “Here are some words. We hope you’ll find some of them interesting”? and yet instead articles have carefully crafted, engaging titles.
To draw the user in, the NYT graphics team rely on two main techniques. First, they don’t try to show every data point they have (more on that shortly). Second, they rely on the annotation layer. Interestingly, I notice that virtually none of the visweek applications have any form of annotation layer. By this she refers to notes and text that appears close to the actionable parts of the visualizations. One common framework that is used in NYT visualizations is the “stepper”, their “interactive slides” system. You often see in one corner of their designs consecutive numbers one next to another with a “next” button, so the user can make their way through the explanation or go back to any previous point. Well, this stepper model allows the team to provide new explanations with each slide.

Since they are so comfortable with annotations, the NYT team is ok with using unique, non-recyclable visualizations. I really appreciated what she confirmed when she answered questions, that they don’t really care about using novel visualizations that most readers never saw before, versus very traditional ones. Amanda claims that if they use innovative (but commented) forms, they do lose some readers but “the net joy remains positive”. They are very mindful of finding clever, relevant ways to represent the data, versus either familiar templates or bleeding edge forms.
Finally, she insisted that design is hierarchy, that is, knowing is important and what is not as important. This goes beyond the choice to not show every data point that we alluded to earlier, but it’s also about what to highlight and what to downtone. Mastering this balance is critical to achieve the signature NYT style.