I used to be convinced that data visualization is not so much the visualization as the data. I long thought that with interesting data, slapping on any visualization can make it look cool. I think this is related to my affinity towards particle effects in video games: you don’t need to really have a reason, just the more, and the smaller, the better.
More recently, I think I’ve become a bit inspired. Call it a zen moment, but I’m now willing to believe that all non-trivial data is in fact interesting and visualization is needed to reveal it. Actually, I don’t really think it’s the visualization that makes it interesting. Visualization is just some cheap gimmick. It’s finding the right way to interpret data.
Some may argue that it is actually the other way around. That smart visualization informs our interpretation. From what I know about data-viz, our approach does just the opposite. (Now this is entirely speculative because I am definitely no expert in data visualization.) I think when people want to find an interesting way to visualize a particular set of data, they immediately try to determine the right axis to dissect it. At least for me, when I have a potentially interesting data set, I don’t try all the possible graph/chart formats and select the most informative one. Instead, I analyze what is the best way to present the data. The visualization aspect is all eye candy.
I suppose the importance of data visualization is not to have the computer automatically interpret data for you, but to instead crunch out a certain interpretation on a large scale. For some reason, I was always hoping that data visualization could be one of those fields that took away most of the thinking we need. If that were the case, I’d imagine data-viz software to be hugely successful because then you could write your solid data-viz app once, and use it on general data. But if not, it means that a programmer or analyst of some sort will always need to be sitting by its side, teaching the program new tricks and tailoring it to new data sets. No one package could handle all possible avenues at once. I’m thinking this is where Palantir is hitting the nail on the head.
Anyhow, the only reason I started thinking about data visualization was because of last.fm. They have so much cool data and it would be very interesting to be able to analyze. (Part of me also thinks that data visualization just needs a lot of data to look cool, but I think that condition may be necessary but not sufficient). As far as I can tell, the information they give you access to isn’t so finely granulated. I’m tempted to roll my own service similar to twitter/last.fm. Who knows.
Leave a comment