Monday, November 28, 2022

Idealism and pragmatism in visualization

Some reactions to the Washington Post chart on the right, designed by my former student Luís Melgar, reminded me of a passage from The Art of Insight that I shared the other day. This morning, a few readers of this chart asked in social media: “Why isn't time on the X-axis?” implying that there's something wrong with that, as it breaks some convention or rule.

A better question might be: “Can you understand the chart anyway?” I bet you can. It may take a few more seconds than usual, but does that make a significant difference when you might gain something else? What makes a graphic good isn't whether it tries to approximate an ideal of truth, beauty, goodness, or excellence; rather, it's whether actual people can make sense of it, learn from it, or enjoy it.

A follow-up question could be: “Do you think that sometimes there are good reasons to flip the axes of a time-series chart?” There might be. In this case, I believe, the reason is rhetorical: to emphasize the left-right partisan divergence. We may agree or not with the choice of flipping the axes, but it's a choice that can be justified, discussed and eventually tested.

The following is a section of The Art of Insight where I talk about the difference between idealist and pragmatist discourses in visualization. I've shared this very early draft on Mastodon, so let me know what you think there, if you wish:

My conversation with Ed Hawkins made me reflect on the ways that visualization professionals and scholars talk about what we do and why we do it. I also remembered a question that I’m asked often: What are the rules of data visualization?

I was born in Galicia, Spain. We Galicians are known for replying to questions with even more questions, so I often ask in return: What are the rules of writing?

I find such questions puzzling. I think of visualization as a technology that humans have devised to engage with the world. It’s a form of speech whose syntax consists of mapping data onto symbols. In this sense it isn’t that different to writing. In writing, beyond a loose observance of the vocabulary and grammar of the language we employ, there are no universal rules that are applicable to all its varieties regardless of purpose or audience.

This misconception about the existence of universal rules in visualization is in part due to the inheritance —and also to a slight misreading, I’d say— of what I call “the Tuftean consensus,” after Edward Tufte, author of several influential books, the best among them being The Visual Display of Quantitative Information (1983).

In The Infographic: A History of Data Graphics in News and Communications (2020), Murray Dick, a journalism professor at Newcastle University, describes various discourses in visualization, which are ways to think and talk about the craft. The first is the functionalist-idealist discourse. Tufte is the most popular exponent of it, although Murray cites several other statisticians and cartographers, such as Jacques Bertin, author of the foundational Semiology of Graphics (Sémiologie Graphique. Les diagrammes, les réseaux, les cartes, 1967.)

The functionalist-idealist discourse, Dick explains, has been dominant for decades. It conceives of visualizations “first and foremost as a scientific methodology” and as form of “visual logic [based on] the rigorous application of a monosemic system that depends on a priori rules (present in standards, and in conventions, such as the use of grid lines, legends, labels, etc.) These provide a means by which signs may be used to connect propositions in a logical sequence.”

According to Dick, to someone who favors the functionalist-idealist discourse “graphics necessarily deal in complex, multivariate ideas and they must explain clearly and efficiently, telling the truth about the data.” Then he adds a key sentence: “The notion that designer and audience may not share a common and irreducible understanding of what “the truth” means is not countenanced.” To the functionalist-idealist discourse, the quality of a graphic depends exclusively on the nature of the information presented and on the decisions that its designer makes about it within existing norms.

That’s why Tufte has written that “if your statistics are boring, you’ve got the wrong numbers” or “the only worse design than a pie chart is several pie charts.” However, being boring —or clear, or efficient— isn’t a property of things, but of the relationship between those things and the people who experience them. The best statistics can be boring, and the clearest and most efficient graphic to you can be confusing and inefficient to others. Pie charts are often misused, but this is also true of any other chart type. No chart type is good or bad in essence.

Designers who favor a functionalist-idealist discourse often come up with comprehensive and coherent systems of thought and action —sometimes involving transcendentals such as Truth, Good, or Beauty,— derive rules from them, and use them to judge whatever they see. Tufte doesn’t employ the term “rules” in his books, but he has defined a series of uppercase “Principles of Graphical Excellence.”

I’ve seen people try to apply these principles to their work too strictly—and fail. The reason is that they should be taken instead as sensible guidance whose pertinence depends on many considerations. Tufte himself is subtler than his most devoted fans, as he usually adds the caveat “within reason” to soften his otherwise strongly worded recommendations. For instance, he suggests maximizing “the share of a graphic’s ink devoted to data [...] within reason.” These last two words shall remain undefined.

Murray Dick’s discourses aren’t mutually exclusive; their boundaries are porous, and the same person can use different ones. I’ve employed a functionalist-idealist discourse myself, but The infographic says that my predominant discourse is pragmatist. 

Such a label encapsulates my ideas well. I respect the functionalist-idealist discourse, as I’ve learned a great deal from those who favor it. However, I’ve come to think, teach, and talk about visualization not as a series of allegedly universal principles, but in terms of ad hoc reasoning. This reasoning is informed by aims, constraints, conventions, trade-offs, likely outcomes, personal experience and taste, and by an ever-imperfect but ever-evolving body of scientific knowledge.

I don’t deny the possibility of norms, but I don’t understand them as rules or principles that can be given, top-down, by “leaders” in the field. Instead, norms are tacit patterns of behavior that emerge, bottom-up, from the ongoing historical dialogue between those who practice the craft. Because of this, norms can and should always be subject to examination.

It’s for this reason that, instead of a hierarchical professional landscape where authoritative figures assemble lists of principles, I prefer to foster a level field where conversations among kind peers flourish. Kind not in the sense of being nice, but of being welcoming, helpful, and constructive. My most cherished critics, mentors, and colleagues were and are kind, but not all mince their words.

Finally, I also believe that a central goal of visualization is to benefit an audience somehow —and that this audience could be the designer alone. “Benefitting,” by the way, doesn’t always mean “getting the most information in as little time as possible.” That’s just one of the many purposes that a visualization may have.

To be concise, I’d argue that my discourse isn’t just pragmatic. It’s also pluralistic —if visualization is a language, many dialects are possible— and even a tad hedonistic*.

* I’ll break my self-imposed goal of avoiding footnotes to clarify what I mean by “hedonistic”. Hedonism has nothing to do with the contemporary vulgarization of the term, meaning sensory excess. Hedonism is a set of philosophical schools with ancient roots that, in broad terms, agree that pleasure —roughly equivalent to welfare and the diminishing of suffering— is a moral good (for some schools it’s the key moral good.) Whatever contributes to pleasant lives is generally considered more desirable than what doesn’t.