IEEE Visualization Conference - some of the best visualizations in the world
Last week I was in Sacramento, California at the IEEE Info Visualization conference. It was great learning more about the latest research in human/computer interaction as it relates to information. Plus, there was a lot of brutally cool graphics to look at. They had a poster contest, not to mention an art display, and a number of product demos and seminars. There was also a contest to create a visualization for interacting with an entertainment database.
The keynote speaker was one of the graphics editors from the NY Times. He talked about some of the behind the scenes work that's done to find the "best" way to present information - in some cases it takes them several iterations to find the right representation. Other times, they work on an incredibly short deadline, starting only hours or minutes before their press deadline. They also make online visualizations. Some of the more interesting online interactive charts can be found here and here.
One of my favorite charts was this human head. Yes, it really is a human head chart. The hair represents several dimensions at one time by using different characteristics. The different items were: length of the hair, the hair color, the branchy-ness (whether the hair has split ends), and the density and location of the hair (from front-to-back). Kudos to the students who came up with this one!
Theoretically, you could bind real data to each of these characteristics if you had a data cube containing at least 5 dimensions - let's imagine the head represents the greater Seattle area. You could picture the front of the head is North Seattle, and the back of the head is South Seattle. Hair growing in each area may represent the population density in that area. The hair length might represent the average income, and the hair color might represent the level of education, and the split ends might represent crime rate. This is based on the theory that split ends are bad, so I think that works. Given that these dimensions are all independant characteristics it is possible to represent any combination of data. For example, you could have a light colored hair that is either long or short, just like a dark hair can also be long or short. So looking at the overall picture, you could detect trends - for example, is most of the short hair dark colored? This implies that less education results in less income. What if you find an area where the hair was light and short, this means the people are highly educated but still aren't making much money. They must have selected the wrong major, or perhaps it represents an area where people are more highly educated but still in school, near the University District for instance.
It's interesting to compare this human head display to the charts most people really use - imagine if the next version of Excel had a human head chart! That sounds a bit nuts, but why not?! Why should Office only produce conservative professional businessey looking graphics? Well, I have a theory that there are hundreds of visualizations that people could dream up besides the usual lines, bars, and pies. Some of them would certainly be off-the-wall like the human head, but others would be both innovative and useful to others. And to be fair, some of them would be completely incomprehensible. I call for a revolution of sorts - let there be so many visualizations that we need to invent some way to search through all of them and rank them. I'm ready to go solve that problem; I've been thinking about it for the past few weeks.
If you have an interest in data visualizations - I want to hear from you. Send me email if you've got something unique to share. I'll publish the best links I get in a future blog post. Also if you are a coder, and you want to make thinking about visualizations your day job - send me your resume. I am currently hiring.
Now, I'll leave you with another link which will probably keep you busy for at least 30 minutes if you haven't seen it before: Data Visualization Modern Approaches