Avoiding the demise of intention

A few thoughts about generative A.I., visualization—and Michel Foucault

Photo: Surrounded by communication students and faculty from Universidad de Puerto Rico, which I visited in late August. More details about this event at the end of this post.

The most frequent question I’ve received in my most recent workshops and talks is how I think that generative A.I. will impact visualization and information design. As my current thoughts about this are rather trivial, I refer people to articles by Andy Kirk, Enrico Bertini, and others.

I’ve used generative A.I. to write elementary code—that I always double-check for errors and botched syntax,—and I’ve tried a few song generators that are amusing, but ultimately dull and forgettable; here’s Data Blitz, a goofy metal song about data visualization. I also use ChatGPT to translate this newsletter to Spanish; I edit the results manually.

At least for now, my interest in generative A.I. is tepid. It might be because the work that I do is intentional, deliberate, and handcrafted, even if it’s done with a computer and the software installed in it. My writings and designs require concentration and deep engagement with information and the symbols that encode it. As a consequence, the more mediated, abstract, and detached my connection to that information and those symbols is, the poorer the work becomes.

Moreover, as a reader I love to feel that behind the graphics and stories I encounter there are authors, actual humans who care about my experience of delving into their work. I dislike to sense that whatever I’m reading was generated by a machine with little or no human intervention.

This is closely related to advice that I give students every semester: Never trust software defaults, and always supervise, tweak, and edit the results of the automated processes we all rely on. Make readers feel that your chart isn’t an R or a Tableau or an Excel chart; your charts should be yours, personal, unique, idiosyncratic.

The same prudence we ought apply when using R, Illustrator, Tableau, Data Wrapper, or any other graphics tool, we should apply to generative A.I. It should be an aid to human agency, not a substitute for it.

This long preamble is an excuse to recommend Ted Chiang.

Chiang is a computer scientist and a deservedly famous Science Fiction author. In my opinion, he’s one of the the greatest writers in the history of the genre, on par with, say, Clarke, Le Guin, Butler, or Silverberg, (a favorite of mine; don’t miss Nightwings, Dying Inside, or Downward to the Earth.)

Since 2023 at least Chiang has been writing about A.I. for the New Yorker magazine. In his three critical essays (1, 2, 3) I saw a few of my own intuitions expressed better than I ever could; others were challenged and proved to be misconceptions. Here’s a central quote from the third essay, Why A.I. Isn’t Going to Make Art:

The task that generative A.I. has been most successful at is lowering our expectations, both of the things we read and of ourselves when we write anything for others to read. It is a fundamentally dehumanizing technology because it treats us as less than what we are: creators and apprehenders of meaning. It reduces the amount of intention in the world. 

We should all ponder whether we want to contribute to that reduction or to resist it.

What I’ve been reading

A lot of Michel Foucault. Why Foucault, you might wonder? Different reasons.

Even if I studied some philosophy in High School and college, I never engaged seriously with Foucault’s oeuvre beyond a few chapters from his writings on sexuality, epistemology, and power. However, in the past few years I’ve seen several unsavory right wing “intellectuals” in the U.S. and abroad attack Foucault for his alleged contribution to the fall of something they call “Western Civilization”, a construct that anyone with even just a basic grasp of history should find profoundly idiotic.

Being allergic to con artists as despicable as, say, Christopher Rufo, I decided to see for myself what the fuss around Foucault was about.

What I (re)discovered is a prodigiously erudite teacher who had a vast knowledge of the history of thought and ideas.

Contrary to the myth that portraits Foucault as an labyrinthine writer, he was a clear communicator—at least late in his career, as I’m reading him backwards, beginning with the transcripts of the last courses that he delivered at the Collège de France. So far I’ve read The Courage of Truth (his last lectures, 1983-1984) and The Government of Self and Others (1982-1983); I’ve also read an early course, Abnormals (1974-1975).

Part of what I’ve found in those books, particularly the detailed analysis of the notion of parrhesia—frank discourse that puts the speaker at risk of severe consequences—is applicable to an analysis of the craft of visualization, so there might be some Foucault in my next book.

(Funnily, and because I’ve been traveling too much lately, I read the first of the books above in a Portuguese edition, the second in English, and the third in Spanish. A friend of mine who knows Foucault well told me that the English translations are likely the worst of those three, but it’s unlikely that any of them fully capture the nuances of Foucault’s discourse. Might need to learn some French!)

Besides the venerable philosopher, I’ve also found plenty of interesting resources and readings mainly through Bluesky, the social network that I’m using the most these days (do join; the visualization community seems to be slowly regrouping in it):

• Susan Kruglinski and Andrew Gellman’s article chronicling the parallel histories of statistical charts and comics.

• Data Navigator. Frank Elavsky’s latest accessibility-in-visualization project, intended to make web graphics easier to interact with.

• Andrew Heiss’s great introductory class Data Visualization with R.

• Narges Mahyar’s overview of novel ways of imagining and talking about visualization, ‘Reimagining Data Visualization to Address Sustainability Goals'.’

What I’ve been doing

I’m working on several projects that I’ll share here as soon as I can. I’ve also kept thinking about and taking notes for a fifth book. As for conferences, if you are planning to attend IEEE Vis 2024, you may find me there. The tentative title of my talk is ‘The Golden Age of Visualization Dissensus’.

Recently I visited the Universidad de Puerto Rico, where I gave three talks about how to learn (and teach) visualization; here’s a summary in Spanish. I loved it.

—

That’s all for today. I leave you with Chris Stapleton and his hit song Either Way. I’m not a fan of country music, but I make an exception with Stapleton because of the elegant simplicity of his compositions and his unique and powerful voice: