Different goals, different looks: Infovis and the Chris Rock effect

Seth writes:

Here’s my candidate for bad graphic of the year:

times-map

I [Seth] studied it and learned nothing. I have no idea how they assigned colors to locations. I already knew that there were more within-city calls than calls to individual distant locations — for example that there are more SF-SF calls than SF-LA calls. The researchers took a huge rich database and boiled it down to nothing (in terms of information value) — and I have a funny feeling they don’t realize how awful this is and what a waste.

I send it to you because it isn’t obvious how to do better — at least not obvious to them.

My reply:

My first reaction is to agree–I don’t get anything out of this graph either! But let me step back.

I think it’s best to understand this using the framework of my paper with Antony Unwin, by thinking of the goals that are satisfied by different sorts of graphs.

What does this graph convey? It doesn’t tell us much about phone calls, but it does tell us that some people can make colored maps with lots of lines. It also tells us that someone has a bunch of telephone call data. Even though the lines on the graph are difficult to interpret, they (correctly, I assume) convey that they come from a big database.

The graph also has the pleasant feature of revealing things we already knew. For example, the country is divided into many localities by color. I don’t know how this was done but I assume it was some clustering algorithm applied to the telephone call data. In any case, we see Arizona and New Mexico together–hey, that makes sense!–also Virginia paired with Maryland and Western Pennsylvania connected with West Virginia. These make sense too. We also see Alabama connected with Georgia rather than Mississippi (which is what I’d expect), but, hey, no algorithm is perfect. The map with all the lines also shows a bunch of coast-to-coast calls–that makes sense too–and it confirms our intuition that Minneapolis, Chicago, and Detroit are in the upper midwest, whereas Boston, New York, and Philadelphia are tightly packed in the northeast.

I call this the Chris Rock effect. Chris Rock says things we all know are true. But he says it so well that we get a shock of recognition, the joy of relearning what we already know, but hearing it in a new way that makes us think more deeply about all sorts of related topics. Sure, you might have already known that Denver is not near any other large city–but seeing it on this map of phone calls brings this fact to life in a way that maybe never happened in your previous experiences looking at U.S. maps.

It’s just like that famous map of Napoleon’s march into Russia. It didn’t tell you anything you didn’t already know, but it presented familiar knowledge in an attractive, unfamiliar format, Sort of like if your spouse sent you a valentine written in pig Latin. Good old “I love you” sounds that much better if you have to work for it a bit.

OK, back to goals. The graphs that Seth hates so much do their job in that they look unusual and draw the viewer in to look more carefully and rediscover familiar truth. After that, though, there’s not much more there, and it would be great if they could link to something more informative.

P.S. See Chris Volinsky’s comments below.

17 thoughts on “Different goals, different looks: Infovis and the Chris Rock effect

  1. I'm amused by the "at&T Labs Research" in the lower right. I'm guessing some auto-correct feature in some software package converted "AT" into "at" and no one noticed.

  2. This is a great example of a graphic grabbing our attention and yet showing very little. I agree with Seth.

    @Andrew
    Does the graphic confirm our intuition that Alaska is South of Los Angeles and Hawaii more to the East?

    Does anyone know how the top map was constructed? I suppose it was some kind of clustering on the counties. It would be interesting to learn how they decided on that number of clusters.

  3. Antony:

    I agree–I hate the map too! But I also want to step back and try to understand what goals it is satisfying.

  4. It is a colourful representation with readily anticipatable less wrong versions.

    A few reactions:

    From my visits with my wife (fine arts major) to art galleries – the notable or key art pieces that give rise to new styles seldom look like anything to the untrained eye – the later stuff is usually more appreciated by others.

    Why the angst at something that’s more style than substance?

    I agree with Andrew that displaying in a novel way (maybe in a variety of unusual ways) something we all are supposed to know more directly, precisely or rigorously – has its value. (The nicest complement I ever got from one of my stats profs was “what a colourfully non-standard way to put that result”.)

    Anyway, think I need to be more aware of the angst to the colourful plots I work on to display Bayesian analyses. (Showed some of them to my teenage son a few days ago for the first time and his neat reaction was – should there not be an equation that does a better job of that!)

    K?

  5. I kinda like both graphs. I'm not American so the way states are connected and not connected is interesting. It would have been intersting if they could have linked it to ethnicity or age or whatever – ethnicity I assume for California.

    The second map is interesting but it would be even more interesting to see only the long distance calls not local. And it would be really interesting to try and remap it by amount of phone contact rather than distance – the coasts would move to the middle and the insides would move out.

  6. The actual website created different maps depending on whether they were using Texts, Data, or Phone Calls. They then talked about how this shows that texting is fundamentally more local than calling.

    Plausible, but if you're making model-dependent claims, you should show your model! Especially since it's non-obvious how the clustering model maintains contiguity when IRS migration data shows all sorts of ill-behaved exceptions (Miami should be much more closely related to New York than Jacksonville).

  7. Thanks Chris! Now at least one of us liked the bottom graph as well.

    I can think of two reasons:

    1. It has artistic promise.
    2. It is a raw data plot and no matter how ugly there should always be (access to) a raw data plot.

    Any further info?

    For 1, I am guessing HCL-Based Color Palettes in R might have been used.

    For 2 that is one of my pet complaints as in the almost complete absence of L'Abbe plots in published meta-analyses.

    K?

  8. Some may not like the top graph because it's "obvious", but I think it's not obvious and I like it. There are many plausible state groupings that don't pan out, so it does add information.

    For example, I would have guessed that Texas and Louisiana would be grouped, and Mississippi and Alabama, but would have been wrong. I would have guessed the Oregon-Washington grouping, but would have put Ohio with western Pennsylvania.

    (Most importantly, I think it also gives us reasonable priors for the location of Phineas' and Ferb's Tri-State Area. ;-))

  9. This is a pretty bad graphic, but it's not NEARLY as bad as the one I nominated for Worst Graphic of the Year a couple of months ago.

  10. The bottom chart reminds me of this facebook linkages map:
    http://www.dailymail.co.uk/sciencetech/article-13

    Using brightness instead of the "Tall, narrow arcs [to] show many calls [or, connections] within a small geographic area" really cleans out a lot of the messiness and garbage. And i don't think it's necessary to color each region to see which areas are connected…look at Hawaii and Guam being very clearly drawn to into US with the webbiness in between the two. The linked image is missing some sort of map overlay to compare borders, which isn't entirely necessary there, but may be useful for proving the point of the top graph listed here.

  11. Pingback: Why Aesthetics vs. Utility is beside the point, and Meaning matters | DATATELLING

Comments are closed.