Speak Data, page 4
The Doomsday Clock was designed in the 1940s by the painter Martyl Langsdorf (who often went by just her first name, Martyl) and has since gained broad resonance in our culture, appearing in literature, television, and film. It’s a symbol, but also a surprisingly effective example of data visualization. In a few straightforward marks, the clock summarizes reams of quantitative knowledge about science, politics, economics, technology, demographics, and more, and then synthesizes that knowledge into a familiar graphic metaphor. But importantly, the clock is also emotional. It tells a story that everyone can understand: Time is limited, so we’d better get our act together. The reactions it elicits are visceral. Like the shot clock in the last moments of a championship basketball game, you just can’t look away.
The graphic designer (and our colleague at Pentagram) Michael Bierut has called the Doomsday Clock the “most powerful piece of information design of the twentieth century,” and we’re inclined to agree.2 But what makes a good data visualization? It’s a hard question to answer. Is it the most accurate explanation of a number’s meaning? The ease with which the reader understands? A sort of graphic elegance that makes the most complex numbers easy to understand and beautiful to look at? Data visualization expert Edward Tufte has written persuasively about what he calls the data–ink ratio: Expend the least amount of effort communicating the greatest amount of data.3 Is it about efficiency, speed, or power?
Cover of the Bulletin of the Atomic Scientists, 1947
Maybe. Considerations like these might well be appropriate depending on the situation. After all, a pilot needs a ruthlessly simple elevation gauge to steer their craft safely; a financial analyst wants a basic chart to plot investments strategically. We live in an image-driven world where visual literacy is low but the need to access information is high. But what such a zero-sum game often misses is fundamental to data visualization’s ultimate purpose: to convey a record of reality. Data encodes meaning, and visualization gives us access to that meaning in legible form. Instead of asking what makes a good data visualization, what if we asked a bigger question, free of past conventions or constraints: What is the most effective way to communicate meaning? And to answer that question, we need a fundamentally different approach.
A few years ago, the staff of the Bulletin approached our team with an intriguing request related to their upcoming clock reading. They didn’t want to change the iconic Doomsday Clock—who could improve on it?—but were seeking a way to better contextualize it. For outside observers, the Bulletin works under a shroud of secrecy. To maintain impartiality, its deliberations are confidential until the clock’s reading is revealed. Unsurprisingly, this leads to a barrage of questions each year about why the clock’s hands moved forward or back. For their latest announcement, the Bulletin wanted to offer an idea of the amount of research and discussion that goes into the decision to reposition the clock. Could we help?
We launched into a process that has become a hallmark of our information design practice, a tested methodology that, we believe, yields valuable and exciting results. Every project begins with research and a deep immersion into the content at hand—in this case, no easy task. From nuclear risk to climate change to disruptive technology, unless you’re a PhD scientist, the Bulletin’s work is not for the faint of heart. We pored over research papers; interviewed prizewinning experts; and were even allowed to listen in on meetings with the Bulletin’s Science and Security Board, the primary group of world-renowned experts who determine the clock’s reading each year.
Most importantly, we also were led by our own curiosity, asking “where,” “when,” and “why” at every turn. It soon became clear that to do justice to this information, we would need to leave behind the shopworn canon of conventional data visualization. Bar charts, pie charts, line graphs, histograms—in our tech-driven world, this standard lexicon is ubiquitous, showing up everywhere from news reports to phone bills to video games. Software like Microsoft Excel allows anyone to be a dataviz designer. But to us, this approach represented nothing more than blindly throwing technology at the problem. Instead, we were determined to let the data define the right design solution rather than the other way around. New meanings necessitated new models.
Data visualization pioneers like Charles Joseph Minard, Florence Nightingale, and W. E. B. Du Bois knew this well. Take Minard’s famous 1869 map of Napoleon’s 1812 Russian campaign: an undulating, layered diagram illustrating the movements of thousands of French soldiers over time. To track the advance and retreat of Napoleon’s troops, Minard devised an unconventional map that communicates six different variables—geography, path, direction, population, temperature, and time—in a single image, a feat that would never be possible with a traditional flowchart. The result is a striking portrait of these soldiers’ trials, analytically precise but just as emotionally evocative as the Doomsday Clock. It’s a complete story, and it rewards deep and lingering examination.
Charles Joseph Minard, Figurative Map of the Successive Losses of Men of the French Army During the Russian Campaign 1812–1813, 1869
Nuclear Risk: An Expanding Concern data visualization, 2020
Climate Change: A Warming Planet data visualization, 2020
Disruptive Technology: Trust and the Pandemic data visualization, 2020
For the Bulletin visualizations, we took a similar tack. To illuminate the expanding concern of nuclear proliferation, we imagined an unusual radial data visualization, mimicking the combustion of a nuclear warhead, and arrayed data about nuclear inventories, tests, and treaties around it. To represent the accelerating threat of climate change, we organized data about temperature, sea levels, and carbon emissions—often treated as isolated trends—along a single, curvilinear timescale. For a final visualization about trust in public institutions during the COVID-19 pandemic, we created bespoke graphs of public opinion sentiment data, which we then plotted on top of COVID-19 case counts.
Like Minard’s map from 1869, our three visualizations embrace new forms for communicating their meaning. They serve up information in layers, with strong visual hierarchies that lead the viewer’s eye through the content. For those with limited time (or attention spans), takeaways are structured and clear, but there are also intricate, detailed graphics with which the more intrepid reader can spend time if they wish. We believe that whenever the main purpose of data visualization is to open people’s eyes to fresh knowledge, it is not only impractical, but actually misleading, to avoid a certain level of visual complexity. Visualizing data is about providing access to our reality, not just simplifying it. We can write rich and dense stories with data and celebrate the true depth of complex realities. While some may find these visualizations unconventional to look at, we hope they encourage careful reading, and therefore a more invested engagement in the meaning behind the numbers.
Another pillar of our approach to data visualization is context. Data is never only what we see on a spreadsheet, and hiding below the surface of any single data point is a whole world of contextual information. This context is vital for telling richer, fuller, and ultimately truer stories. Augmenting hard data with layers of “softer” and more qualitative information is the only way to present this larger picture. For the Bulletin visualizations, we peppered each graphic with annotations, callouts, and tangents that provided context for what the statistics were telling us. This secondary gloss provided a first-person narration to what could otherwise have been a rather cold and impenetrable story. It helped the data make sense.
As a final flourish, behind each graphic we collaged in text from transcripts of the real deliberations of the Bulletin’s Science and Security Board. This subtle textual layer alluded to the amount of behind-the-scenes analysis that went into the clock’s reading, underscoring the essential human component often overlooked in data visualization. It’s a reminder that data is always human made: collected, analyzed, evaluated, and communicated by people.
Martyl’s original Doomsday Clock design is one way to use visual communication to convey meaning. Our three data visualizations discussed here are another. Both are valid responses to the same brief, albeit born of very different eras and design philosophies. But as the world becomes more and more complex, and its problems more and more urgent, it may be time to embrace new methods of information design that can fully speak to the issues we face. Ecological crises, global pandemics, artificial intelligence, economic inequality, political instability: such knotty topics demand data visualizations that prioritize complexity over simplification; contextualization over reduction; and customization over convention. And there’s no time to lose. The clock is ticking.
NOTES
1 John Mecklin, “A Moment of Historic Danger: It Is Still 90 Seconds to Midnight,” Bulletin of the Atomic Scientists, January 23, 2024, https://thebulletin.org/doomsday-clock/current-time/.
2 Michael Bierut, “Designing the Doomsday Clock,” The Atlantic, November 5, 2015, https://www.theatlantic.com/entertainment/archive/2015/11/doomsday-clock-michael-bierut-design/412936/.
3 Edward Tufte, The Visual Display of Quantitative Information, 2nd ed. (Graphics Press, 2001), 93.
Seth Godin
Seth Godin hardly needs introduction. A prolific writer on business and marketing topics, he is the author of more than twenty books. He is a virtuosic public speaker, and his TED talk on how ideas spread has received more than seven million views. For years, Seth has been obsessed with how humans think—and what might get them to change their behavior. His 2022 book The Carbon Almanac: It’s Not Too Late trains his formidable brain on how we can get the world to care more about the globe’s most urgent topic: climate change. In this conversation, Seth talks about the crucial differences between data, information, and truth; how absolute confidence in numbers is an ideal, but never a reality; and why USA Today might be to blame for society’s misunderstanding of quantitative information.
Over the years, you’ve written a fair bit about data and the communication of data from a marketer’s perspective. How do you wrap your head around this topic?
Let’s start with three words, because if we don’t understand them, it’s going to get confusing. The three words I’m going to pick are data, information, and truth. Data is a bunch of numbers that we choose to highlight. It’s a mass of verifiable numbers that have no information in them—yet. There is a sampling that goes on, as we cannot acquire all the data ever. Information is when a human being turns data into knowledge or understanding. And truth is impossible. When we present data in an attempt to create information, it always has a point of view.
Do you feel like the world has misunderstood what data is?
I think we don’t even understand the question. We should ask ourselves: Are we in the information business, or are we simply delivering data? If you just give me data, it’s boring. I won’t read it. But if you want to help me turn it into information, please own the fact that information always has a point of view, especially when it’s backed up by data. Giorgia and Phillip, you’re in the information business. You’re not in the data business. The magic of your work is that you’re very clear about that. Putting the word human in Data Humanism is really important. Because you’re saying: Here is one way to look at a bunch of data that can be verified.
I’ll give you another example. My wife and I recently listened to Robert Caro’s book The Power Broker on a long road trip. It’s sixty-six hours long. There’s a chapter in which the book’s subject, Robert Moses, lists all the traffic on the Long Island Expressway every year for ten years. The audiobook narrator just reads his chart. It’s as boring as it sounds. That’s data, not information.
All data is subjective, in that a human being decided what to collect and how to collect it. Yet when the general public sees a chart, too often they automatically trust it. Why have we all been trained to trust so blindly?
Since the Industrial Age, we have pushed people not to talk about their feelings. Instead, we have pushed people to talk about facts. The challenge is that as things get more complicated, people tend to feel stupid. And if you feel stupid, you don’t want to talk about it.
This happens for example with sunk costs and the simple truths of statistics. If the poll says there is a 60 percent chance that someone is going to get elected, most people think that means that 60 percent of people are going to vote for that person, and thus there’s a 100 percent chance that they’re going to get elected. When that isn’t what happens, they get angry. There isn’t a deep understanding of what the stuff on the screen even means. Then the media has an incentive to try to turn things that people don’t understand into things that will emotionally resonate because we are emotionally starved. My point is: If you don’t really understand, that’s what the system wants.
At the same time, people blindly trust data about say, an election, but when it comes to climate change, they are more suspicious. Why?
Forty years ago, Exxon’s chief engineer wrote a memo in which he described, in extraordinary detail, exactly what the climate was going to be like in 2022. And he was right. But Exxon decided, with a trillion dollars of resources in the ground, that they would sow disinformation instead. It’s much, much easier to sow disinformation than to sow understanding. It’s not that people don’t care about the climate. It’s that they don’t understand. There are a whole bunch of reasons for that: Human beings are bad at predicting the future. Human beings are bad at statistics. And human beings are bad at discerning between good charts and bad charts.
I can’t help but think about Jerry Siegel and Joe Shuster’s original Superman origin story. Superman’s dad figured out that Krypton was going to explode. He had a plan for all the people on Krypton to leave before it was too late. But none of the people in power wanted to look at his data. None of them wanted to understand what was happening. But when the earthquakes began, they understood. And then it was too late.
The challenge we have as communicators is not to create a stampede or a panic, because those don’t yield resilient results, but to create a body of work that isn’t political. There are people who want it to be political because political means, “Don’t talk about it!” There are other things in our world that it’s okay to talk about. It’s okay to talk about the rate of acceleration due to gravity, which is 32 feet per second squared. That’s not a political issue. With climate data, we need to just keep coming back to these first principles of explaining what it is, so people have a foundation for actual understanding.
What’s your personal definition of data?
Data is a mass of verifiable numbers that have no information in them—yet.
That puts the onus on us as designers—and, more broadly, anyone creating representations of data—to get it right.
My problem with the data visualizations published by outlets like The Washington Post and The New York Times is that they pretend they’re telling you something that is true—but they’re not. What they’re doing is capturing a bunch of data and trying to turn it into information, but it always carries with it some point of view. If we’re going to do a good job with data visualization, we’d better be clear about what the so-called alt text says for the chart or the graph we’re making. Editors at major newspapers don’t like having to use alt text honestly because they put up a picture of a firefighter and a four-year-old boy, and they think that a picture is worth a thousand words. But I think they should tell us why they chose that picture. Is it because the boy looks like this and is wearing these sorts of clothes? Is it because the firefighter is a woman? Editors should say those things, because they picked that photograph for a purpose.
I have never seen The New York Times run a correction that says, “The chart we ran didn’t honestly express what we wanted to say.” They have to run corrections all the time for mistakes made in text-based stories, but if you see a chart in The New York Times that’s banal or ineffective or poorly done, there’s no place to even comment on it. It’s viewed like the weather—you can’t do anything about it. But this is not the weather. It’s reporting. And if you are a reporter, you should own your work.
We need to be clear with one another before we make any graph or chart. What’s it for? What is the point that we are trying to make? Because it is not okay to say, “Well, it’s true.” There’s lots of ways that you can present information. If there’s more than one way, then you’re no longer dealing with what’s true, you’re dealing with: This is the point that I’m trying to make. This is the information that I think you should be willing to verify.
We saw people use and misuse data countless times during the pandemic. Has COVID-19 shifted your thinking on this?
Demagogues always misuse data to put forward invalid information. The very nature of being a demagogue is that you cannot sustain rational criticism and discourse. If you could, you wouldn’t be a demagogue. There aren’t any particularly vivid examples that come to me around COVID-19, but I can name plenty from the 1930s, 1940s, and 1950s—a peak moment of propaganda. Or, think about how North Korea is so good at manipulating imagery, using something that looks like data to make something that feels like information. Many people who have the power to do this in public have no training, understand nothing about statistics, and think what’s pretty is what’s important. I’ll blame a lot of that on USA Today, and some of that on Microsoft.
