I spent last week at Harvard doing research in the archives of William James, the psychologist, philosopher, psychical researcher, brother of Henry James, and all around interesting person. He was a brilliant, charming, self-defeating, deeply strange man (exhibit A: he believed taking a high dose of nitrous oxide helped him finally understand Hegel). That mixture of qualities comes across vividly in his papers.
What doesn’t, at least at first, is that he was a talented visual artist. In fact, before he became a psychologist, William James dreamed of being a professional painter. He studied for several years in his late teens and early twenties under the painter William Morris Hunt.
Although none of William’s paintings appear to survive, a careful reader of his archive will find evidence that he continued to draw throughout his life.
Here he is, for instance, doodling on an envelope addressed to him from Geneva:
Readers of The Metaphysical Club, Louis Menand’s wonderful book about James and his circle, will recognize the image below as William’s sensitive drawing of his brother Wilkie while he was recovering from being shot during the Civil War.
The visual creativity of James is not just a clue about how his own mind worked. It’s also part of a larger shift in the culture of science during his generation: in the nineteenth century, design and the nascent world of big data came together, for the first time, to create the modern concept of data visualization.
Although James and his collaborators are rarely mentioned in discussions of the origins of data visualization, they actually played a very important role in shaping it. They were the consolidators and extenders of a new paradigm — the generation after the famous names in the field like William Playfair and Florence Nightingale.
The generation of William James came into adulthood in the Machine Age of the 1870s, ‘80s and ‘90s. Information buffeted the human brain like never before. And the new technique of data visualization pushed into new domains: the mapping of the mind, the sociology of race, and the pursuit of explicit political ends.
James is famous among historians of science for what Francesca Bordogna calls “boundary work” — he moves across disciplines and fields with a manic, restless energy that makes it hard to figure out exactly what he was.
As we’ll see, this included an unusual approach to visualizing mental activity that yielded significant firsts, including the first schematic of a neural network. Along the way, James also had important relationships with two pioneers of modern data visualization who are rarely put in the same sentence: Francis Galton and W.E.B. du Bois.
Francis Galton (center) was a sort of intellectual frenemy for James, initially a mentor and influence, then later — in his guise as a founder of eugenics and ardent imperialist — as an exemplar of the dangers of scientific hubris. Du Bois (pictured at right) studied with James at Harvard and was deeply influenced by his philosophy of pluralism.
What they shared is a conviction that drawing, diagramming, and composing images was not a decorative step added after the thinking was done. It was how the thinking got done. This seems to me a crucial point in light of new AI tools like Claude Design, which automates the design process (on which more below).
Quite a few people have pointed out that writing is a form of thought (my favorite entry in this genre is this 2025 essay by Derek Thompson). But it’s worth thinking more about what else counts as thinking too — specifically, the sorts of important, creative thinking we don’t want to accidentally mislabel as “drudgery that we are happy to let AI take over for us.”
Design, I would argue, is not drudgery.
The images below also remind us how handmade data visualization once was. W.E.B. du Bois’s visualizations from the 1900 World’s Fair are rightfully famous online, and they look great as compressed jpgs shared on social media. But it’s important to remember that these are large, hand-drawn, hand-lettered posters. These are not just the product of mental work but manual work, too.
That link between the hand and mind is harder to come by in a world where all research is digital, but it can be fun and important to access when doing serious research. When I was deepest in researching the history of 20th century psychedelic science for Tripping on Utopia, I filled up a yellow notebook with collaged images and primary source snippets from my archival research. I started it as a sideline, almost a hobby to distract me from what I saw as the “real” work of actually writing the book.
It turned out to be one of the most significant forms of research I did, precisely because it was so freeform and undirected. I started noticing links between different documents, and thinking more deeply about the motives (public, private, and even subconscious) of the people I was writing about.
The rest of this post is a gallery of some of the ways that James, Galton, and Du Bois visualized data along with some desultory commentary. I end by experimenting a bit with Claude Design to see what gets lost when we automate this type of exploratory visual thinking with data.
If you find it interesting, please consider subscribing.
You’d be hard-pressed to find any reference to James in books about the history of data visualizations or design. But I think he deserves a page or two. For one thing, his Principles of Psychology (1890) contains what I believe to be the earliest ever visual representation of a neural network:
The most interesting data visualization in James’s work is this however, IMO:
James intends the image above as a representation of how consciousness “moves through” the process of uttering of a simple sentence over time. The numbers on one axis are showing moments in time, and the other axis shows the words being said or thought. The Joy Division-esque crest running through the middle is the changing attention we pay to each word over time.
It is really striking to me how much this looks like a computer rendering — but it’s from 1890! For this attempt at a sort of faux-four-dimensional modeling of thought, plus the neural network chart alone, I think James deserves a lot of credit for his data visualization.
There are other interesting images in Principles of Psychology, especially in chapter two which you can read here, although they are a bit more familiar.
For instance, here is a schematic of how a child perceives a candle flame:
And an early attempt to map which parts of the brain correspond to specific body regions and sensations:
James read Galton closely, and cited him throughout Principles of Psychology — especially on questions of mental imagery and visual perception. Galton’s famous “breakfast-table questionnaire,” which asked hundreds of correspondents to describe how vividly they could picture the objects on their breakfast table that morning, was one of the first systematic attempts to gather data on subjective visual experience at scale. James was fascinated by it. He replicated versions of Galton’s mental-imagery surveys in his own classes and used the results in his writing.
Galton was also a pioneer of meteorology and produced many, many beautiful charts relating to weather phenomena — things like this:
But the visualization that appears to have most interested James (and which I find most striking too) is this jam-packed color plate about mental imagery and synesthesia which comes at the end of Galton’s book Inquiries into Human Faculty and Its Development (1883):
The big idea that linked Galton and James was not just that inner life could be represented as data, but that this data could be rendered as pictures.
Galton, obsessed as he was by measurement and averages, took this in a novel and discomfiting direction. Elsewhere in the same book that the beautifully strange image above comes from, Galton introduced composite portraits — photographs of criminals, tubercular patients, “types” of any kind — layered individual faces on top of each other to produce a kind of statistical average in visual form.
The project was inseparable from his invention of eugenics — a term which Galton coined in the same 1883 book. Galton’s visualizations encoded his conviction that human variation could be sorted, ranked, and ultimately improved through selective breeding.
In this work, to a large extent, the visual was the argument. Later, Galton even used his own photographs and biometric measurements as an example to encourage others to submit to biometric data collection. He wanted to visually model not just a new approach to data but a new mode of life in which all aspects of the human were reducible to data.
Reading his work from the perspective of the 2020s — with our turn toward “post-literate,” image and video based apps powered by mass data collection — it’s hard not to conclude that in this, he was hugely successful.
William James, who prized pluralism and unconventionality, could not follow Galton on his journey into the world of averages and biometric data. But one of James’s most talented students, a young W.E.B. Du Bois, did. Sort of.
As the philosopher Colin Koopman writes, Du Bois took Galton’s obsession with measurement and flipped it on its head.
W.E.B. Du Bois — the first African-American to earn a Harvard PhD, and later one of the most important sociologists and civil rights thinkers of all time — studied philosophy with James in the late 1880s. Du Bois had a remarkably visual mind, with an approach to imagery that verged on the synesthetic.
Du Bois absorbed James’s insistence that human experience could not be flattened into a single scale. But he also absorbed Galton’s conviction that the world could be made legible (and changeable) through creative use of data visualization. What he did with this combination was something neither of his predecessors had quite imagined.
Below are some of the charts that Du Bois made for the 1900 Paris Exposition Universelle — the same world’s fair that gave us the first public moving walkway, Rudolf Diesel’s engine, and the Art Nouveau métro entrances that still dot Paris today. Du Bois was there as the lead curator of the “American Negro Exhibit,” a small pavilion aiming to show a European audience what Black Americans had made of themselves in the thirty-five years since emancipation.
The roughly sixty charts he produced, hand-drawn and hand-lettered on large poster boards by Du Bois and his students at Atlanta University, are some of the most arresting data visualizations ever made. Like Galton’s charts, they visually plot the result of careful statistical investigation using idiosyncratic and arrestingly odd formats; also like Galton, they are using data visualization to pursue political goals. But where Galton leveraged his visualizations to argue for the lasting, ahistorical superiority of people who resembled him (Protestant, British, scientific, and from “good families”), Du Bois was charting the rapidly changing role of Black people in American society.
Change over time runs through virtually all of his charts from 1900, a few of which are reproduced below:
Writing this post got me thinking about a question which we are figuring out the answer to in the 2020s: by automating away the design process, do we also cede away our ability to use design itself (and not just writing) as an act of thinking?
Using Anthropic’s new Claude Design feature, I wanted to see what happened when I asked Claude to make something in the style of these nineteenth-century visuals. Here’s what I got for a Du Bois pastiche (my prompt is available at this footnote):
Asking for Galton-style charts yielded a similar stew of serif fonts and faux-aged paper effects, but in more of a Victorian bric-a-brac style which, on closer inspection, adds up to nothing much at all:
These failed attempts to capture a distinctly Victorian data visualization style should not be surprising. Claude Design is optimized to make slides and other data visualizations that suit modern sensibilities, and we can guess that there wasn’t much Du Bois or Galton in its training data. But aside from the by-now-familiar fact that AI models tend toward the median and the expected, there is the larger absence here: these charts don’t have a perspective.
James, Du Bois, and Galton, different as they were, had one big thing in common, and this thing is part of why we are still thinking and debating them today: they were deeply idiosyncratic, deeply personal. You see in their visualizations not just raw data but an odd, distinctive intellect at work. And more importantly, an intellect working through an argument that is both mental and physical. They are instantiating the act of thought in the mechanical work of their hands.
Claude Design is appealing for those of us (like myself!) who find themselves having to make fiddly tweaks to Powerpoint slides. It is, in this sense, just one more step in a process of design automation and abstraction that has been happening for centuries: from hand-lettered to printed book, from drawing to photograph, from scissors and glue to “desktop publishing,” and now from humans designing things to humans telling AI designers what to make.
As the act of thought gets more and more removed from the union of human mind with human hands, the new ideas, the palpable jolt, that you get from designing something by hand gets rarer, and thus more valuable.
This is why I am going to be making more scrapbooks and weird collages on yellow legal pads as I embark on my new book project about the Machine Age.
After all, what could be more human?
