There’s a fascinating — and heated — discussion going on over at Flowing Data about visualizations, and the quality thereof. It all started when Andrew Gelman commented on Flowing Data’s selection of the 5 best data visualizations of 2008, saying: “They all suck (for the purpose of data display).” Unsurprisingly, this kicked off a bit of a firestorm.
Nathan Yau’s response is well worth reading, particularly for the comments that follow. He suggests that statisticians are stuck in the mindset of Tukey and Tufte, and don’t “get” the “cool new stuff”. Point made, although statisticians are hardly a homogeneous bunch. I for one think many of the visualizations shown on Flowing Data and elsewhere are truly valuable, and have linked to some in the past. (I did however notice I’d omitted Flowing Data, a site well worth your time, from the blogroll — an omission since corrected.) Andrew did post a response in the comments with some justifications of his criticisms.
It occurs to me that this fooferaw is caused in some part by language. Charts / visualizations / graphics have different purposes in different venues: sometimes to present data, sometimes to explore and make discoveries, sometimes to offer an analysis or conclusion, sometimes to communicate a story. And sometimes they are by design not science but art, or intended “merely” to entertain. Yet we have no simple way to categorize these uses: as I just demonstrated, the words “chart”, “graphic”, “visualization” and others are all used interchangeably.
We don’t usually compare a comic book to a novel, or Rembrandt to a Clear Channel billboard. On the other hand, some novels are trash, and some comic books are sublime. When a creative work fails at its intended purpose, it’s worth calling it out and identifying the problems. But a heated argument on whether Battlestar Galactica is better or worse than Citizen Kane is only going to have the heat, but little light.
