Tala Schlossberg made the nice animations. It was a pleasure to work on this project with her and Stuart Thompson for the past few days. I've discussed the piece in these interviews at NPR and BBC Mundo (Spanish).
Background
I started writing about hurricane maps and uncertainty in visualization exactly two years ago. If you're interested in research about these maps, follow Lace Padilla and Le Liu (his dissertation is available online). A difference between this research and the one we're doing at UM is that we're mainly focusing on populations who are very vulnerable to hurricane risks, such as minorities, people without higher education, etc.
In the past two years I've become even more alarmed about how we misread charts like the one above and, more importantly, about how journalists—particularly TV anchors—explain these maps wrong to their audience. Just yesterday Noah Pransky complained about the sloppiness of TV journalists, and for good reason. They still misinterpret hurricane forecast maps in the way I explained. That's irresponsible.
Reactions
The reactions to the op-ed have been interesting. Some people asked why we don't design better visualizations. That's easier said than done. The cone is sometimes misinterpreted as an either-or display—“if I'm inside, I'm in danger; if I'm outside, I may be fine”—but spaghetti maps and other alternatives have their own problems, making viewers focus too much on the center line of forecast models and neglecting the uncertainty that surrounds them.
Also, knowing where the storm center may go tells us little about the risks we may be exposed to: wind is just one of the threats during a hurricane; storm surge, flooding, and heavy rain can also be deadly, and they may affect you if you live far from the cone center or even outside of it. That's why the NHC and other entities design additional visualizations.
The responsibility of journalists and readers
This is just a personal conjecture, but I don't think we'll be able to design maps that everyone will interpret correctly, at least in the short term. This is related to a complaint some people expressed about the NYT article: it seems that I blame readers, and not the scientists or designers who create the maps. That isn't so, but readers do have a responsibility. And also journalists, who are the translators, or mediators, between scientists and the public.
Scientists and designers certainly ought to create visualizations that are as transparent as possible, of course, but readers often have the unrealistic expectation that any graphic should be understandable without effort. We designers have fed that expectation, and we must stop. I go over a lot of detail about this problem in How Charts Lie. I think that it's related to several common myths, such as the popular “a picture is worth a thousand words”.
We need to accept—and help others accept—that a visualization isn't a picture or an illustration. It's an argument, or a text in the semiotic sense. You need to make an effort to grasp what it means. No graphic, no matter how well designed it is, can resist the test of an inattentive or careless audience.
Another mantra designers love to repeat is “show, don't tell”. Ideally that'd be the case, but when a visual is abstract and complex because it represents complex information, we need to show and tell.
Audiences often lack knowledge of graphic symbols, grammar, and conventions, and therefore may misread certain visualizations. My conjecture is that this is what leads to some misinterpretations of the cone of uncertainty. The cone seems to be an area under threat because it resembles the visual conventions often used to represent areas of relevance or interest on maps, such as distinct colors or sharp boundaries. To create a different mental schema of how to read the cone in readers's brains we need to calmly explain not only what the chart is saying, but also how to read this type of chart in general. Words matter.
Remember Hans Rosling's classic talks and documentaries. The first time he showed one of his now famous bubble scatter plots he didn't jump to its content. He first explained the scaffolding of the chart and its encodings: “position on the X-axis means such and such, position on the Y-axis...” Rosling did what TV news anchors should do, and what I tried to do myself in The New York Times article and in How Charts Lie: help increase readers's graphicacy.
