Each Tuesday, Eurry Kim, a student in our class, will pick one example of data visualization to share with us. Eurry writes:
This week’s theme: Data Art != Data Visualization
In no uncertain terms, I was annoyed to see this visualization tweeted by the Columbia Journalism Review. This was another classic mix-up of data art masquerading as data visualization — another demonstration of the domains within data science not talking to each other. First, notice that it was produced by Peter Orntoft of the Danish Design School. In and of itself, I love Design. But you’re playing a different game when you’re dealing with statistical visualizations. And that brings me to my second point, which is a question: why don’t the numbers add up to 100? They’re percentages, no? It appears that the width of the headscarves refer to the percentages. While you can compare them across categories (e.g., 66% of surveyed Danes think is it is unethical for judges to don the headscarf at work as opposed to 42.5% thinking the same for schoolteachers), it takes a second to mentally re-adjust to the fact that each headscarf is its own aggregate statistic. At first glance, it’s kind of ingenious and uber-creative. But at second glance, it’s counter-intuitive and data-light. Lastly, I had a tough time trying to find the underlying data. I wanted to verify the actual survey question with the words used to characterize the various populations (I was particularly interested in the use of the word “unethical”), but after ten minutes of search, I couldn’t find anything. Citing sources is another indicator of statistical integrity.
Daniel Kahneman, an economist/psychologist/Nobel Prize winner, recently wrote a book called “Thinking, Fast and Slow.” You can read Freeman Dyson’s review in the New York Review of Books. In my mind, a visualization targeted towards a broad audience is one that invokes both parts of the thinking processes about which Kahneman writes. First it catches the knee-jerk System One in any number of ways — “ohh pretty,” “hey, where am I in all this?,” etc. But the distinguishing factor of a good visualization is the right engagement of the critical thinking System Two. System Two makes us want to look deeper. It is a double-take of of our initial impressions. You can see how important this aspect would be for statistical visualizations. When there’s a lot of data embedded in a visualization, it requires more than a minute to appreciate its underlying message. There are a few visualization blogs out there that criticize pieces regularly. Two of my favorites are Kaiser Fung’s Junk Charts and Stephen Few’s blog.