My final project on information visualizations was well received by my professor and class, so I want to share it here for thoughts and comments. The goal of the project was to synthesize the key guidelines about creating visualizations and invite exploration of the subject, even from people not well versed in it.
Chernoff faces of vizualization (PDF, 32Kb)
I used information glyphs called Chernoff faces, line drawing cartoons where each face stands for a record and each part of the facial expression, one data dimension. The literature on Chernoff faces is mixed. Some sources report they can be effective at quick categorization of large, detailed datasets. Other research seems to show they can lead to confusion and distraction, as people read emotions that the glyphs didn't intend to display.
Even if the use of Chernoff faces is not as well-regarded as scatter plots and Tufte railroad maps, I think the project was successful in inviting exploration of the subject in multiple ways. One reason I chose Chernoff faces was they're a little humorous. I had a long, intense semester's worth of detail to summarize, and I knew I had some burnout to overcome. It's okay to be a little goofy sometimes, if it gives you fresh energy for an important project.
As a result, I felt really good working on this project, from start to finish. Not only was I in flow the whole time, at the end all the work I put in came together efficiently. My preparation was enough, but not too much, and the final layout almost assembled itself. I hope you enjoy the visualization too.
I love that! I can understand the criticism, because there's clearly an emotional element to all the faces... it's impossible to avoid. But that's what makes it interesting - the faces are easy to view holistically, but it's also easy to drill down into the parts. For example, it's easy to glance at the faces and see that heat maps and hyperbolic trees support the largest data sets.
Interesting exercise. I'm not sure these are appropriate for wide use, because as you say they are a little humorous, but if nothing else it is a good way to teach the concept of trying to convey multiple types of data in a single visualization.
Posted by: Terry Bleizeffer | May 08, 2008 at 11:15 AM
Thanks for the comment Terry. The ability to focus on details without losing the overall context is a critical piece of a successful information visualization. Glad these seem to be supporting that!
Posted by: Joshua Ledwell | May 09, 2008 at 10:25 AM
i have one complaint. you make this look so easy to produce!
great stuff...
Posted by: Sean | May 09, 2008 at 03:01 PM