I knew him for the baby name visualizer which he originally designed to help publicize his wife's baby names book, though its audience grew far beyond what he originally imagined. He tracked its uses by Googling for it and reading all of the blog entries and other references on it, and found that bloggers were treating it like a game, setting data-mining challenges for themselves. They roughly fell into the same categories that MUD users did: achievers, who were actually looking for baby names; explorers, who looked for quirks in the data such as I, O, ETH, LAT; socializers, who related the visualization to their own lives and used it as a conversation piece; and killers, who used it to make fun of names they thought were stupid. Because the system was interactive and playful and discoveries could be replicated easily, people could easily be drawn deeply into the data. Also, everyone had a fairly distinctive starting point - often their own name - which meant that the data set got a lot of coverage.
But his visualizations are far more numerous than just that one. He made a squarified treemap of 500 of the Smithsonian's "favorite" objects for an online exhibit, and another squarified treemap of the average color of pictures returned by Yahoo photos for the 33,000 nouns in WordNet. When he was working for a financial company, he made a treemap view of the Dow Jones market performance which saw a lot of use, and he also made a visualization that let you draw a stock trend and find similar ones.
Moving away from treemaps, he made a hypnotic lensing timeline of internet art for the Whitney museum. He talked a bit about the difficulty of getting fisheye views to work. Typical fisheye views expand the motorspace, so that moving the mouse one pixel corresponds to moving the view, say, ten pixels. In his interfaces, he prefers to do a two-level fisheye, like in his Whitney timeline. Once a stripe opens, there's no more zooming, and there's a fair amount of space to move the mouse around so navigation is easy.
He mentioned that Whitney had also commissioned another project from him as part of their CodeDoc, where artists released both the art and the source). He also made a revision visualization for Wikipedia and found some fascinating patterns such as edit wars and wiki vandalism, and he made a chess game visualization called "thinking machine" that traced the computer's "thoughts" about what move it should make next.
He made an artistic visualization that placed words that were typed in in certain rooms of a house and then organized people's houses into a "city," echoing the ancient technique of "memory palaces." He found that people used this in very clever ways for their own self-expression, and sometimes even as a confessional of sorts. For a museum piece, he made a couple more interactive versions, one that projected images returned by the input words on the walls and floors of a room, and another that had multiple input sources. He had more artistic projects on his personal page.
More recently, he made a "forum reader," which he tried out on Slashdot. He's hoping to release the source for it soon. For this visualization, he started out with fancy branching trees to represent the conversation, but found that though the visualization was aesthetically pleasing, even he didn't want to look at a conversation that way.
He had some design advice, too. One of the first lessons of human-computer interaction is to separate the model and the view when coding. For visualizations, he advocates sandwiching geometry between those and maintaining the modularity of all three layers. Having to reimplement visualizations in other languages has taught him this the hard way. He generally makes all of his visualizations Java applets, and codes them in Java 1.0, which he actually really likes though he's heard all of the usual complaints about how constraining Java is. He also admitted that he didn't do any fancy pre-processing for his visualizations, even the stock market grapher - they're responsive enough by just implementing with "brute force" in Java, so why optimize further? Any on-screen point just needed to know it's current and target x,y coordinates, and possibly velocity and mass and other parameters specific to a visualization.
When he's designing a visualization, the first thing he does is look through the data, preferably in notepad. (He also finds this is a great "bozo filter" for people who want him ot make visualizations: he'll ask for the data and more often than not, they'll say, "well, actually, I don't have it ...") He generally doesn't spend much time identifying his audience or what they expect to see - after all, in a good visualization one sees the unexpected, so he just tries to display the data in a useful way instead. Once he designed a visualization for Nasa, which could only give him the database schema until he got appropriate clearance. He designed a visualization based on the schema data, which included a "date" field - only to find later that the date was actually just the date the database was created.
After he looks through the data, he sees what has been done before in this area. Then, as he's designing, he makes sure to actually use the visualization himself. He finds that if he actually likes to play with it, he's on the right track. Finally, he does some sort of crit or usability test. He admitted that one doesn't often do formal user studies on visualizations, especially the artistic ones. He would more often have an artist's critique or an informal evaluation from some of his super-critical friends.
Of course, many of us wanted to know how he had gotten into doing visualizations like these (and by extension, how we could too). He got a math degree at Berkeley, though he had been doing design in high school. He worked for a financial company and did artistic designs on the side, sandwiched between his work and his home life. Even now, he has to make time to do the artistic installations outside of work, which makes for a stressful existence at times. He told us to take advantage of the time we had now in school and experiment, though he admits that when he was in school, he probably would have scoffed at such at suggestion. (At least when you have a job, you generally have your evenings and weekends free!) Between his talk and Mirrormask, I feel freshly inspired to continue my dabblings in art and design - it makes me feel so alive.
Utah Quicksilver, Park City and Salt Lake City's best transportation service