InfoViz Week 3
Graphical Practice — Excellence, Integrity and Origins
Graphical practice is all about establishing the good, bad and ugly in the history that led us to the fundamental methods of communicating quantitative information visually. There is a statement in the introduction, however, that doesn’t sound convincing by itself: “The design of statistical graphics is a universal matter […] not tied to the unique features of a particular language”. Of course, not all aspects may be subjective, but I tend to think that language (and the associated culture) will influence the interpretation and analysis of graphics. Color, for instance, has different connotations depending on culture, making different designs necessary to evoke similar responses across peoples.
This chapter traces the origins of graphical representations from those with familiar physical analogies, beginning with data maps, to more abstract ones that are commonplace today. It was impressive to read how innovations (such as the bar chart) were frankly critiqued by the inventors themselves; in present times, I believe we would be prone to address them as limitations, rather than flaws. It also tries to define the gold standard in terms of the intent behind visualization, quality of data and choice of variables, and finally, its presentation. One of the notions I’m curious about is the use of least ink — more specifically, how it would relate to current times, with computer-generated interactive graphics which can hide/reorder data with great ease.
A suggestion arises from the argument that only a picture can show immense volumes of data, which although apparently true, is somewhat exaggerated in my opinion. A better measure would be to count pieces of information of interest that a visualization represents. This leads us to a question: what are the most information-dense plots and is there an empirical limit beyond which they are too tedious to create or simply not usable? Next, what seems to have been swept under the rug is that there may be some ambiguities in graphical representations that inevitably need textual clarifications. How can we evaluate and better those?
The next chapter makes a case for the need for rigor in statistical charts by illustrating the lack thereof, which could be a symptom of a deeper malaise that is pointed towards by this phrase: lies, damned lies and statistics. Big data is what makes this affair convoluted: how does one make sense of data that is hardly obvious even after time-consuming exploration?
The importance of normalizing nominal data to real terms is emphasized repeatedly. This is something my teacher in high-school would stress to no end while teaching economics but I couldn’t appreciate well, since it is only now common practice to adjust data appropriately, while it clearly was not in earlier times. I wonder if the argument against it is due to the impreciseness of any such adjustment, although that does not anyhow condone comparisons of nominal values to come up with “ghost” trends.
An avenue for research could be adapting the Lie Factor to allow for scaling done to achieve accuracy after any distortion that our perception inflicts on say, figures such as circles. It would be even more useful to develop a formulaic approach to accommodating such perception anomalies to make statistical graphics with confidence that they have the desired (and not untoward/exaggerated) impact. Another constraint that a visual designer could have is to make use of pictures, which might not be uni-dimensional like the data they’re trying to depict: coming up with a fair, usable set of guidelines is a challenge.
Sources of Graphical Integrity and Sophistication
The lack of integrity in visual displays is examined, with the blame squarely at the increase in the perceived need for graphs and the simultaneous lack of training. This begs the question if it is prevalent (and possibly in a worse state) in today’s world where anyone can automatically plot data with computers? Tufte rightly disabuses us of the doctrines that support form over function and (over)simplification of data graphics, which are due to undue emphasis on beautifying data. It is a bold stance that opposes mere decoration of numbers and exposes the dichotomy between the depth of writing and graphics that is based on assumptions about the audience not being “smart” enough to understand school-grade relational graphics. I would nonetheless contend that there are limited opportunities for news publications to use relational graphics as frequently as we would like (and I would certainly like to know how it is so popular in the Japanese press, as compared to elsewhere in the world).
Something surprising was that this account did not consider studies (which quite possibly, may not have been done then) on the cognitive benefits of using visualizations, rather than just as a means to study complex data. It would be intriguing to know when attention was brought to the basis in cognition that supports use of such graphics.