How do you make an “ethical” graphic? How do you avoid misleading a reader as to what is being said, while still editorializing enough to ensure legibility of meaning? Kieran Healy, Carl Bergstrom and Jevin West all attempt to answer these questions through their “Profiles in Badness,” often reaching similar conclusions on the ways in which aesthetic considerations affect the perception of a reader/viewer on what is being communicated.
Healy provides a much more thorough and wide-ranging discussion of these issues than Bergstrom and West, going through the reasons to look at data, what makes “bad” figures bad, and what visual elements are at play when examining readers’ perceptions.
Healy’s breakdown of problems into three varieties is a useful framework. Badness can come from: strictly aesthetic issues (tackiness/tastelessness), bad data (well-designed but misleading), or bad perception (not designed for legibility). His discussion of the aesthetic ideals of Edward Tufte— maximizing the “data-to-ink ratio” and minimizing “chartjunk”—is especially nuanced and compelling. While acknowledging Tufte’s conclusions as good rules of thumb, Healy provides examples based on perceptual science to acknowledge that sometimes increasing visual complexity enhances legibility (as is the case with box and whisker diagrams where Tufte’s preferred model was recorded as being less legible to users), and “junk” may lend visual distinctiveness and enhance retention of the graphic for the viewer.
Healy’s discussion of perception is equally nuanced and peppered with useful tips for the aspiring data visualization designer. Discussions of color (and the benefits of the HCL colorspace for pairing colors with appropriate variations across one element versus another) provide concrete examples and are easy to grasp. The discussion of “pre-attentive pop out” and the immediacy with which color differences scan versus shape differences is particularly illuminating.
Bergstrom and West’s analysis of “bad” graphics is more focused on editorial issues, but no less enjoyable or insightful. Their case that “proportional ink” should dictate the presentation of bar graphs and other shaded elements is persuasive and easily comprehended. Their inclusion of characteristic “bad” graphs — including one where shading and an inversion of axes leads a reader to potentially draw the exact opposite conclusion from the reality of the data – allows them to make their points in a relatively brief manner.