Let's not pretend there's more controversy here than there really is. If anything, the point he's making is too banal. Almost the whole of _The Visual Display of Quantitative Information_ argues against the style of visualization this post argues against.
Virtually all infographics are indeed terrible visualizations of information.
> Virtually all infographics are indeed terrible visualizations of information.
I do not agree. I've seen quite a bunch that – although obviously not a definitive source of information – give a good sense of perspective. Here's one [0] that came back to mind.
Really? Most of that chart is just big text. Big text is not an infographic. I would say the most interesting one I've come across recently is http://xkcd.com/radiation/.
Notice how that graphic uses graphics primarily to illustrate, and often uses illustration in lieu of words; to fully communicate the same ideas in prose you might need many paragraphs. Notice how it uses color exclusively as a subtle callout. Notice how the color choice highlights effectively but doesn't draw a blinding neon sign.
By contrast, the infographic you like uses illustrations incorrectly; for instance, isn't it confusing the area and radius of circles? Notice how it uses multiple pictures each drawing a single comparison on a single axis. Think about how many words the same comparison takes in prose (not hard, because all those words are included in the infographic in gigantic type above the illustration, often in radioactive neon green).
--- Also: note that this [unfavorable] comparison is against a relatively weak example of explanatory illustration --- the best examples in _VDQA_ are way, way better than these NYT examples, which are a bit chartjunky themselves. If there's a good critique to be leveled at E.T., I think it may be that he tends to be overly favorable to friends, students, and, most importantly, people who share the same aesthetic sense that he does: simple line drawing style, low-saturation colors, arts & letters themes. Look at that "Game of Life" illustration, for instance; ouch.
Look at the "airspeed velocity of an unladen swallow" by, I believe, one of his students --- a truly excellent, effective example of an information graphic done in E.T.'s favored style [literally, with E.T.'s favorite palette]:
Is that a particularly effective graphic? Look at how the green doses transition to the orange ones (with a legend "all green doses combined" corresponding to an irregular number of orange boxes), and then look at the transition from orange to yellow; are all the orange doses equivalent to 50sv, or are all the orange doses equivalent to one box?
I agree that the XKCD "radiation" chart is an instance where an XKCD-style comic does a better job of explaining something than an equivalent well-written paragraph. But I don't agree that it's a particularly great visualization.
I would definitely agree that it isn't the best graphic, but I think it does an admirable job of portraying the primary thing it was created to portray (I assume) -- the context of radiation dose levels. I can very quickly eyeball the order of magnitude difference between a plane flight and a chest X-ray, while I could also dial down to the more precise differences if necessary.
The "radiation doses combined" transition seemed fairly clear to me. In every case the collective doses are measured in the new SI-prefixed unit delineated by the scale on the left. If I were to criticize anything I would be a little more wary of the color choices and the box labeling. The = sign and parenthetic dose notation seems a little confused (the blue box dose size uses a different convention from the others).
While the chart could layout some things a little more clearly, I think that Minard's chart shows that one shouldn't reduce information content solely for the sake of simplicity.
I think the chart would do better if removed from the arbitrary constraint of XKCD's page framing; if, for instance, a suitably large canvas was used, so that the eyes could immediately grasp the comparatively minimal exposures from common radiation sources to (say) the gigantic exposures from Chernobyl.
It's clearly within Munroe's capability to make such a graphic; he did this one on a deadline. I'm not criticizing Munroe (though: not an XKCD fan), just making an objective assessment of the graphic.
And, like I said: here's a case where a graphic, even an imperfect one, probably communicates rich information more effectively than prose. Unlike the "relative sizes of data" infographic upthread.
One of the most recurrent metrics in Tufte's books is "density". Another is "comparisons on multiple axes" (I'm going from memory, I think he had a more technical term). How do you think this particular graphic does?
Graphics like this seem more like posters promoting an idea than tools for understanding complex data.
Complex data relative to what? You and I can (quite) easily grasp how huge is a petabyte, an exabyte, a zettabyte even; yet someone not "in the field" will have a hard time grasping what represents such an amount of information.
Of course this one being ordered by Mozy should trigger some warnings about it being a covert ad, but it does not detract from the potential veracity of the data.
This one is admittedly quite simple, but should we discard all infographics on the merit they're not of Tufte level?
>You and I can (quite) easily grasp how huge is a petabyte, an exabyte, a zettabyte even
Can you really? I work with video every day. Terabytes are nothing when we're talking HD content and lots of it. It's still pretty hard for me to visualize, even sitting at a system with the details up in front of me, what exactly a petabyte is. Let alone hundreds of thousands of them.
It's like asking an average person to visualize a trillion dollars. There exists a point once a quantity becomes sufficiently large that you tend to stop consciously processing it.
-edit, realize this sounds really snarky, it's not. I am legitimately curious how you quantify that much stuff.
I've been working on problems at various scales from subatomic to galactic scale and about everything in between. At some point, when you're genuinely interested about what you do, you build mental bridges between domains and scales and values and units, you... connect some dots. You start understanding what a log/exp is, what that "shooting through the roof" profoundly means. By understanding I mean you start feeling it with your gut. So when someone talks about one Zm I could take it as a pure value, but I could try and get the feel of it. It's overwhelming, so much so that I can physically feel the cognitive weight of it on my mind. So when I make some conversion and find that 1Zm is about 100'000 light years, and I know (whatever the light-year value means, I take it as pure, transitional value to an item of knowledge) that the Milky Way is about that size in diameter, only then do I feel the absolutely immense size of our galaxy. I must insist that it's no more the vision of the galaxy that gives weight to a Zm to me, but the Zm that gives the incredible sense of dimension of that pack of matter.
That graphic isn't even internally consistent. I measure the 2000 "computer usage" circle at 56 pixels in diameter and the outer 2008 circle at 195 pixels. The text states usage in that time frame has increase 342%. Visually the change is 1,212%.
In text it states that Hitatchi came out with the first terabyte drive which holds 1000 GB. Later, in graphic form, it shows 1024 GB = 1 TB.
In the Internet Users by millions chart the author uses a bar graph (time is continuous, why use a bar graph) with no x-axis labels and then to express hard drive price over time we get a line chart (still no x-axis label). The missing labels are okay because there are callouts specifying the dates, but the change in graph type is weird.
I disagree. It fairly consistent. By your numbers, the computer usage in 2000 is 56 pixels in diameter while the diameter for the circle representing 2008 is 195 pixels. By area of the circle, The bigger cicle (half circle really, would that matter?) is 1212% larger. By diameter, the larger circle is 348% larger. That is a couple pixels (about 3 pixels) or percentage points off (about 5 percent). That is accurate enough to show scale on an infographic in my mind.
As for Hitatchi creating a 1TB drive that holds 1000GB, but later stating that 1TB is 1024 GB, isn't 1 terabyte technically 1000GB which the hard drive companies use, but also frequently used—incorrectly—to reference 1024 GiB or 1024^4 bytes? This other number, 1024GiB is actually a tebibyte, not a terrabyte. They are technically incorrect, but from the standpoint of the majority of the population, the difference between a terabyte and a teribyte is just small print on the back of box. The infographic could be clearer, but honestly, do you think this detail is worthwhile for trying to show the scale of what a petabyte is? That is what, a 5% difference in size? I think that is, when looking a petabyte. If I did my math correctly, there is 50GB of loss (or is that GiB?)
In the end, they are trying to show that a petabyte is a fuck-ton of data. More data than many people can wrap there mind around. They are trying to show scale. How you define the numbers are complicated because we have SI units for both binary and decimal which don't match up and IEC units for the binary side as well. The average consumer isn't going to know or care. They could care that a petabtye is HUGE. The infographic portrays this without trying to bog the consumer down in differences in technical jargon that most of use technical people know but don't bother using correctly ourselves most of the time.
Again, this probably isn't controversial: it's simply a mistake to make one-dimensional comparisons with a two-dimensional scale. Tufte called this specific example (abusing area as a one-dimensional scale) an examplar of the "lie factor" in statistical graphics.
That may be true, but Portent is an SEO firm and Ian is addressing a certain type of braindead SEO who jumped on the infographics bandwagon as a link building strategy.
They tend to know not, nor care, a thing about visual design of quantitative information. They think they're executing a strategy because their fingers are moving on the keyboard.
Then they don't get results and wonder why. The point is to tell them they aren't even in the right ballpark.
Virtually all infographics are indeed terrible visualizations of information.
reply