Charting without prejudice: is it even possible?

As if it ain’t hard enough to slog through a wet, cold & gritty Canadian early Spring, Maclean’s Magazine published Cary Wu and Rima Wilkes’ article, “The stats bear it out: In Quebec, trust is low” on March 30th of 2017 just to bum us all out even further. Well, that, and to present statistically defensible quantification of Andrew Potter’s incendiary assertion, in an earlier Maclean’s article, that “Quebec is an almost pathologically alienated and low-trust society, deficient in many of the most basic forms of social capital that other Canadians take for granted.”

That’s pretty strong stuff. It remains to be seen what will become of Potter’s career at McGill, but I’m less concerned with the message of the article or its veracity than I am with the graphics that accompanied it initially and, rather conspicuously, the graphics which hastily replaced them. Below, the dreary picture of pathological Francophone hyper-vigilance with which the article was initially published. When asked, “Who do you trust?” Canadians responded:


OriginalBarChartMulti

Note the vertical scale. All of the data is presented in percentiles, of course, but the actual range of each graph varies. Some start at the absolute bottom of 0% and go to 25%, while others pick up halfway and terminate at 75%. The reason, or rather, the reason I’m assuming the graphic artist would cite for laying out the data this way, is to feature the relative difference between the Rest Of Canada (ROC) and Quebec rather than featuring the absolute difference between all or none. While the former is pertinent to the question of comparing Canadians, the latter means defining a scale between…

  1. a hypothetical, idiotically gullible rube who trusts everybody all the time, and
  2. a hypothetical, cripplingly paranoid hermit.

I mean, that’s what “100%” and “0%” would represent on an absolute scale, and such naifs / Tolkienian Gollums are not pertinent to the point the authors want to make. At least, that’s how I’d have made this mistake.

And I would be mistaken because there are in this story characters other (and more important) than the those hypothetical extremes described above. In the graph presented as above, the ROC positively towers over Quebec’s trust-score in all but the “members of family” category. Why, just look at that great, amicable red stripe of mostly English-speaking friendliness dwarfing those tiny, paranoid blue stubs of Quebec. As you might expect, people complained.

Maclean'sGraphTweet(Les Perreaux)
And the magazine reacted. They replaced the relative-scale graph with one which showed the same data in an absolute scale:

QuebecTrustChartAFTER01

Well, that puts things in a different light. Even the printer can see this is a fairer representation just by looking at the ink reservoirs when the job finishes.

What can we learn from this? Was the first chart an attempt to deceive or obfuscate the facts? Edward Tufte sagely observed that “data graphics are no different from words [in so far as] any means of communication can be used to deceive.” Indeed, he cites as an example of such deception a graph using non-zero Y-axis values in Visual Display of Quantitative Information (2001), but he does so because the graph at issue compares quanta from two dissimilar Y-axis starting points (dirty pool!). More to the point, he laments later in the chapter that he finds it “particularly disheartening […] that the reported perception of something as clear and simple as line length depends on the context and what other people have already said about the lines.” Or, as Donald Rumsfeld famously put it, even the most elemental data visualization is subject to distortion from “that which we don’t know we don’t know.”

So: know. That’s the lesson. Know, or know that if you don’t know, somebody smarter than you will let you know. I don’t believe the graphic designer who produced the initial graph for Maclean’s was “deceptive,” nor intentionally trying to dwarf Quebec with cloying Anglophonic bonhomie. But the fact that the graph had to be replaced shows they did that anyway, which must have been mortifying, and for which they have my deepest sympathies.

But with the article already hot like a hot potato in the world of Canadian letters (having just got Andrew Potter dealt the Spanish Archer* at McGill), the editor should have known that their position just might be read graphically as a bit defensive concerning their accuracy in assessing how trusting are Quebeckers relative to the rest of Canada vs. Quebeckers relative to Tolkein’s Gollum.

If they didn’t, it’s safe to say they do now.

*El Bow

Advertisements

How to evaluate materiality: crafty analysis!

You know the drill: first post of the week discusses an interesting design issue currently causing chins to be scratched in high places; the second examines one of the images presented earlier beneath the coppery analytical bristles of Theo van Leeuwen and Gunther Kress’ quite excellent Reading Images: The Grammar of Visual Design (2006). There is no third post, ’cause by the time you’ve finished the second you’re already reeling back on your heels and whistling through clenched teeth. I mean, what more do you want?

In our last chapter we considered the current trend toward handcrafted, “genuine,” non-computer-generated design in today’s consumer culture and put it into some historical perspective alongside the late Victorian Arts and Crafts movement. To that end, I decorated my discussion with this pithily ironic needlepoint:

Created by Briar Mark
Created by Briar Mark

Here it is in close up so’s you can see the delicate embroidery:

Detail from Briar Mark's "I could have done this on my mac."
Detail from Briar Mark’s “I could have done this on my mac.”

What are we to make of this saucy locution in finger-blistering needlepoint? Let us consider the thing first as a text, as the artist has clearly gone out of his/her way not just to convey a language-based message, but to present that message in the recognizable forms of a regularized, sans-serif typeface.
The thing about texts as such is that they can be infinitely varied in their appearance without affecting their meaning; you can write the phrase, “I did this on my PC” in any typeface you please without significantly changing the meaning of the phrase. Of course, if you render it in, say, Impact (the typeface of warning labels) the message may take on some new connotation further subject to contextualization, but that won’t change what we like to think of as its “literal meaning.” So, at a textual level, this piece has the following characteristics: it’s all caps, has no punctuation and a subjunctive direction: the reflexive subject (“I”) tells us that which is not the case. At the most ephemeral, immaterial level of analysis, this is a message, presumably from the artist to you the viewer, about that which is not the case.

But no message exists in a purely ephemeral form: in order to cook the proteins in the rods you keep buried in your retinas or knock loose the little bones in your ears or whatever, a message has to be manifested physically. This is an important thing that people who work with words all the time often forget: in order to work, words have to have volume and mass. Oh, you may have lovely ideas about words that float magically on gossamer banners in your fecund imagination, but if you’re ever going to make anyone else aware of them you’ll have to articulate them in some quantifiably volumetric manifestation. All visual texts are, therefor, first and before they’re anything else, pictures of text.

Kress and van Leeuwen make this important distinction to help us navigate the next step (which is good because it’s a doozy). There are a vast number of ways in which we might turn when considering the word manifested in writing or speech or the careful arrangement of dozens of Franklin Mint figurines or whatever. But if we approach the problem like good semioticians, we shall look for motivated signs only. It’s not enough that the viewer should say, of the image above, for instance, that “it’s white background reminds me of snow and of winter; therefor, white is a symbol of winter in this image,” because the whiteness may or may not be a signifier intended by the creator or, even if it was, it may not be recognized as such within the culture of its production or reception. To signify, for our analytical purposes, the proto-sign must meet both conditions. Another way to say this is “The white=winter value would have to meet the criteria for motivation.”

If it’s only motivated signs we’re looking for, that narrows the field considerably. Kress et. al. make the next step easy, too. There are three categories into which we might like to group our motivated signs:

  1. the surfaces involved, such as  “paper, rock, plastic, textile, wood, etc” (216);
  2. the substances with which the message is materialized, like “ink, gold, paint, light, etc” (216);
  3. the tools of production with which (1) was realized upon (2), such as “chisel, pen, brush, pencil, stylus, etc” (216).

What makes the image at issue interesting and worthy of my fine analysis here is that it delivers a duality of surfaces and structures simultaneously in order to make a point about the tools of production. The artist was clearly motivated to pair dissimilar signs from our shared culture in order to make a point about craft: the delicate, time-consuming art of embroidery meets the effortless, instantaneous convenience of word processing. The white surface on which the embroidery is featured is a suggested (especially by the former) as a symbol for a blank computer monitor, especially because it is quite at odds with the way in which delicate embroidery is commonly backgrounded in our culture.

The colours chosen back this up neatly and cast the whole thing in an editorial slant. The Cyan-Magenta-Yellow gamut is of course the arena in which computer printers create the millions of colours we perceive, but the artist has chosen a) to not blend them at all despite their overlapping, b) to cast them ghostishly apart from each other, as though the result of some Tyndall effect-inflected trapping error, which c) puts the very technical concept of computer-controlled print alignment in the foreground of the image’s meaning.

The result is a message about what the artist refused to do in a manner that suggests an intentional malfunction: rather than go the easy way, the artist went to great lengths to represent a printing error with tools that require such meticulous attention as to obviate any interpretation of the error as an error. It’s a pretty neat trick.