Chapter Two: On Rational, Scientific, Objective Viewpoints from Mythical, Imaginary, Impossible Standpoints
In 2012, twenty kindergarten children and six adults were shot and killed at an elementary school in Sandy Hook, CT. In the wake of this tragedy, and the weight of others like it, the design firm Periscopic started a new project – to visualize gun deaths in the United States. While there is no shortage of prior work in the form of bar charts or line graphs of deaths per year, Periscopic, a company whose tagline is "do good with data", took a different approach. When you load the webpage, you see a single, arcing line that reaches out over time. Then, the color abruptly shifts from orange to white. A small dot drops down, and you see the phrase, "Alexander Lipkins, killed at 29". The arc continues to stretch across the screen, coming to rest on the x-axis, where you see a second phrase, "could have lived to be 93." Then, a second arc appears, displaying another arcing life. The animation speeds up over time, and the arcing lines increase, along with a counter that displays how many years of life have been "stolen" from these gun victims. After a couple of (long) minutes, the visualization moves through all of the year 2013, arriving at 11,419 people killed and 502,025 stolen years.
What is different about Periscopic's visualization than a more conventional bar chart of similar information such as "The era of 'active shooters'" from the Washington Post?
The Post's graphic has a proposition - that active shooter incidents are on the rise - and demonstrates visual evidence to that effect. But Periscopic's work is framed around a singular emotion: loss. People are dying, their remaining time on earth has been stolen from them. These people have names and ages. We presume they have parents and partners and children who also suffer from that loss. The data scientists who worked on the project used rigorous statistical methods and demographic information in order to infer how long that person would have lived, which are documented in their notes. But in spite of the statistical rigor, and its undeniable emotional impact, “U.S. Gun Deaths” drew mixed responses from the visualization community. They couldn’t decide: should a visualization evoke emotion?
The received wisdom in technical communication circles is, emphatically, "NO." In the recent book, "A Unified Theory of Information Design," the authors state: "The plain style normally recommended for technical visuals is directed toward a deliberately neutral emotional field, a blank page in effect, upon which viewers are more free to choose their own response to the information." Here, plainness is equated with the absence of design, and thus greater freedom on the part of the viewer to interpret the results for themselves. Things like colors and icons, it is implied, work only to stir up emotions and cloud the viewer's rational mind. In fact, in the field of data visualization, any kind of ornament has historically been viewed as suspect. Why? Well, as historian of science Theodore Porter puts it, "quantification is a technology of distance" and distance was historically imagined to serve objectivity by producing knowledge independently of the people that make it. This echoes nineteenth-century statistician Karl Pearson's exhortation for people to set aside their own feelings and emotions when it came to statistics. The more seemingly neutral, the more rational, the more true, the better. At a data visualization master class in 2013, workshop leaders from the Guardian newspaper called spreadsheet data– those endless columns and rows– "clarity without persuasion."
Back in the olden days of visualization, before the rise of the web elevated the visual display of data into a prized (and increasingly pervasive) artform, Edward Tufte, statistician and statistical graphics expert, invented a metric for measuring the superfluous information included in a chart– what he called the "data-ink" ratio. In his view, a visualization designer should strive to use ink only to display the data. Any ink devoted to something other than the data itself – such as background color, iconography, or embellishment – should be immediately erased and, he all but says, the designer spat upon. Visual minimalism, according to this logic, appeals to reason first. ("Just the facts, ma'am" says Joe Friday to every female character on Dragnet). Decorative elements, on the other hand, are associated with messy feelings– or, worse, represent stealthy (and, of course, unscientific) attempts at emotional persuasion. Data visualization has even be classified as the "unempathetic" art, in the words of designer Mushon Zer-Aviv, because of its emphatic rejection of emotion.
The gendered dimension of this thinking should be as clear as when President Richard Nixon, in a now-infamous statement, declared that no “woman should be in any government job... because they are erratic. And emotional.” (He subsequently admitted that “men are erratic and emotional, too,” but insisted that “a woman is more likely to be.”) We saw this same sentiment play out in the presidential debates of 2016, between Hillary Clinton and Donald Trump, but it’s not limited to politics alone. The truth is that most Anglo-Western cultures have long prized reason over emotion for its supposedly greater neutrality and universality. And the belief that women are more emotional than men (and, by contrast, that men are more reasoned than women) is one of the most persistent stereotypes in the world today. Indeed, psychologists have called it the "master stereotype," and puzzled over how it endures even when certain emotions– even extreme ones, like anger and pride– are simultaneously coded as male. One need only compare any number of “Hulk smash!” GIFs to the equal number of crazy-lady PMS memes floating around the Internet to prove, first of all, that everyone is crazy; and second, that the stereotype of the unstable, irrational woman persists.
But what happens if we let go of the binary logic for a minute and posit two questions to challenge this master stereotype. First, is visual minimalism really more neutral? And second, how might activating emotion – leveraging it, rather than resisting emotion in data visualization – help us learn, remember, and communicate with data?
Until recently, data visualization was a rather specialized form of communication, more common in scientific papers and earnings reports than on the front page of The New York Times. Because of this history, the field’s theorists and practitioners have come from technical disciplines aligned with engineering and computer science, and have not been trained in the most fundamental of all Western communication theories: rhetoric. In the ancient Greek treatise of the same name, Aristotle defines rhetoric as "the faculty of observing in any given case the available means of persuasion." Rhetoric does not (only) consist of political speeches made by men in robes on ancient stages. Any communicating object that makes choices about the selection and representation of reality is a rhetorical object. Whether or not it is rhetorical (it is) has nothing to do with whether or not it is "true" (it may or may not be).
Why does the question of rhetoric matter? Well, because "a rhetorical dimension is present in every design," as Jessica Hullman, a researcher at the University of Washington, says of data visualization. This includes visualizations that do not deliberately intend to persuade people of a certain message. We would say that it especially and definitively includes those so-called “neutral” visualizations that do not appear to have an editorial hand. In fact, those might even be the most perniciously persuasive visualizations of all!
Hullman and co-author Nicholas Diakopoulous wrote an influential paper in 2011 introducing concepts of rhetoric to the information visualization community. Their main argument is that visualizing data involves editorial choices – some things are necessarily highlighted, while others are necessarily obscured. When designers make these choices, they carry along with them "framing effects," which is to say they have an impact on how people interpret the graphics and what they take away from them. For example, it is standard practice to cite the source of one's data. This functions on a practical level – so that a reader may go out and download the data themselves. But this choice also functions as what Hullman and Diakopoulous call provenance rhetoric designed to signal the transparency and trustworthiness of the presentation source to end-users. This trust between the designers and their audience, in turn, increases the likelihood that viewers will believe what they see.
So if plain, "unemotional" visualizations are not neutral, but are actually extremely persuasive, then what does this mean for the concept of neutrality in general? Scientists and journalists are just some of the people that get nervous and defensive when questions about neutrality and objectivity come up. Auditors and accountants get nervous, too. They often assume that the only alternative to objectivity is a retreat into complete relativism, and a world in which everyone gets a medal for having an opinion. But feminists would beg to differ. (Feminists, generally speaking, do not like alternative facts any more than scientists do). Rather than valorizing the "neutrality ideal," feminist thinkers like Sandra Harding would posit a different kind of objectivity that strives for truth at the same time that it considers– and discloses– the standpoint of the designer. This has come to be called “standpoint theory.” It is defined by what Harding calls "strong objectivity" which acknowledges that regular-grade, vanilla objectivity is mainly made by mostly rich white guys in power and does not include the experiences of women and other marginalized groups.
This myopia inherent in traditional “objectivity” is what provoked renowned cardiologist Dr. Nieca Goldberg to title her book "Women Are Not Small Men," because she found that heart disease in women unfolds in a fundamentally different way than in men. The vast majority of scientific studies– not just of heart disease, but of most medical conditions– are conducted on male subjects, with women viewed as varying from this "norm" only by their smaller size. Harding and her followers would say that the key to fixing this issue is to acknowledge that all science, and indeed all work in the world, is undertaken by individuals, each with a particular standpoint - gender, race, culture, heritage, life experience, and so on. Rather than viewing these standpoints as threats that might bias our work– for, after all, even the standpoint of a rich white guy in power is a standpoint– we should embrace each of our standpoints as valuable perspectives that can frame our work. Our diverse standpoints can generate creative and wholly new research questions. We discuss standpoint theory further in Unicorns, Janitors, Ninjas, Wizards and Rock Stars, where we assert that a participatory process that centers the standpoints of those most marginalized is what makes for strong objectivity.
Along with this embrace of our various standpoints goes the rebalancing of the false binary between reason and emotion.
Once we embrace the idea of leveraging emotion in data visualization, we can truly appreciate what sets Periscopic's Gun Deaths apart from the Washington Post graphic, or any number of other gun death charts that have appeared in newspapers and policy documents. The Washington Post graphic represents death counts as blue ticks on a generic bar chart. If we didn't read the caption, we wouldn’t know whether we were counting gun deaths in the U.S., or haystacks in Kansas, or exports from Malaysia, or any other semi-remote statistics of passing interest. But the Periscopic visualization leads with loss, grief, and mourning. It provides a visual language for representing the years that could have been– numbers that are accurate, but not technically facts. It uses pacing and animation to help us appreciate the scale of one life, and then compounds that scale 11,419-fold. The magnitude of the loss, especially when viewed in aggregate, is a staggering and profound truth– and the visualization helps recognize it as such through our own emotions. The generic WaPo bar chart cannot do the work of communicating this truth.
Skilled data artists and designers know these things already, and are pushing the boundaries for what affective and embodied data visualization could look like. In 2010, Kelly Dobson founded the Data Visceralization research group at the Rhode Island School of Design (RISD) Digital + Media Graduate program. The goal for this group was not to visualize data but to visceralize it. Visual things are for the eyes, but visceralizations are data that the whole body can experience– emotionally, as well as physically.
The reasons for doing visceralizing data have to do with more than simply creative experimentation. How do visually impaired people access charts and dashboards? According to the World Health Organization, 253 million people globally live with some form of visual impairment. This might include cataracts, glaucoma and complete blindness. Creators in the visceralization mode have crafted haptic data visualizations,
For example, in the project "A Sort of Joy (Thousands of Exhausted Things)", a theater troupe joined with a data visualization firm to craft a live performance based on metadata about the artworks held by New York’s Museum of Modern Art. With 123,951 works in its collection, MoMA's metadata consists of the names of artists, the titles of artworks, their media formats, and their time periods. But how does an artwork make it into the museum collection to begin with? Major art museums and their collection policies have long been the focus of feminist critique because the question of whose work gets collected translates into the question of whose work is counted in the annals of history– and, as you might guess, this history has mostly consisted of a parade of white male “masters.”
In 1989, for example, the Guerrilla Girls, an anonymous collective of female artists, published what we would today call an infographic: Do women have to be naked to get into the Met. Museum? The graphic was designed to be displayed on a billboard. However, it was rejected by the sign company because it "wasn't clear enough." (If you ask us, it's pretty clear). The Guerrilla Girls then paid for it to be printed on posters which were displayed throughout the New York City bus system, until the bus company cancelled their contract, stating that the figure "seemed to have more than a fan in her hand." (It is definitely more than a fan). The figure is certainly provocative, but the poster also makes a data-driven argument by tabulating gender statistics for artists included in the Met collection, and comparing them to the gender stats for the subjects of art included in the collection. As per the poster, the Met readily collects paintings in which women are the subjects, but not those in which women are the artists themselves.
A Sort of Joy tackles a similar subject, but with wholly different tactics. The performance starts with a group of white men standing in a circle in the center of the room. They face out towards the audience, which stands around them. The men are dressed like stereotypical museum visitors: collared shirts, slacks, etc. Each wears headphones and holds an iPad on which the names of artists in the collection scroll by. "John," the men say together. We see the iPads scrolling through all of the names of artists in the MOMA collection whose first name is John: John Baldessari, John Cage, John Lennon, John Waters, and so on. Three female performers, also wearing headphones and carrying iPads with scrolling names, pace around the circle of men,. "Robert," the men say together, and the names scroll through the Roberts alphabetically. The women are silent and keep walking. "David," the men say together. It soon becomes apparent that the artists are sorted by first name, and then ordered by which first name has the most works in the collection. Thus, the Johns and Roberts and Davids come first, because they have the most works in the collection. But Marys have fewer works, and Mohameds and Camilas are barely in the register. Several minutes later, after the men say "Michael", "James", "George", "Jean", "Hans", "Thomas", "Walter", "Edward", "Yan", "Joseph", "Martin", "Mark", "José", "Louis", "Frank", "Otto", "Max", "Steven", "Jack", "Henry", "Henri", "Alfred", "Alexander", "Carl", "Andre", "Harry", "Roger" and "Pierre", "Mary" finally gets her due. It’s spoken by the female performers; the first sound they’ve made.
For audience members, the experience starts as one of slight confusion. Why are there men in a circle? Why do they randomly speak someone’s name? And what are those women walking around so intently? But "Mary" becomes a kind of a-ha moment– the same that researcher Robert Kosara says that data visualization is so good at producing– when the highly gendered nature of the collection is revealed. From that point on, audience members start to listen differently, eagerly awaiting the next female name. It takes more than three minutes for "Mary" to be spoken, and the next female name, "Joan," doesn't come for a full minute longer. "Barbara" follows immediately after that, and then the men return to reading, "Werner", "Tony", "Marcel", "Jonathan".
From a data analysis perspective, “A Sort of Joy” consists of simple operations: only counting and grouping. The results could easily have been represented by a bar chart or a tree map of first names. But rendering the dataset as a time-based experience makes the audience wait and listen. It also contradicts long-held wisdom in visualization design, as expressed by Ben Shneiderman in the mid-1990s: "Overview first, zoom and filter, then details-on-demand." Instead, in this data performance, we do not see "the whole picture". We hear and see and experience each datapoint one at a time. The different gender expressions, body movements, and verbal tones of the performers draw our collective attention to the issue of gender in the MoMA collection. We start to anticipate when the next female name will arise, and begin to speculate on whether it will be a Rosa, a Rhonda, or another name entirely. We feel the gender differential, rather than see it. This feeling is affect. It comprises the emotions that arise when experiencing the performance and the physiological reactions to to the sounds and movements made by the performers, as well as the desires and drives that result– even if that drive is to walk into another room because the performance is disconcerting or just plain long.
Designing data visceralizations requires a much more holistic conception of the viewer. The viewer isn’t just a pair of eyes attached to a brain. They are a whole body– a complex, feeling one. Theirs is a body located in space, with a history and a future. This notion of the visceral viewer, and of the additional knowledge that can be conveyed when designing with visceralization in mind, can be found in many current projects, even if they don’t always describe their work in those terms. For example, Catherine (one of the authors of this book) and artist Andi Sutton led walking tours of the future coastline of Boston based on sea level rise. And Lauren (the other author) and her team of Georgia Tech undergrads recreated Elizabeth Peabody's living-room-rug-sized charts from the 19th century using touch sensors and individually addressable LEDs. Mikhail Mansion made a leaning, bobbing chair that animatronically shifts based on real-time shifts in river currents. Teri Rueb staged “sound encounters” between the geologic layers of a landscape and the human body that is affected by them. Simon Elvins drew a giant paper map of silence in London that you can actually listen to. While these projects may seem to be speaking to another part of brain than your standard Sankey diagrams or network maps, there is something to be learned from the opportunities opened up by visceralizing data. In fact, scientists are now proving by experiment what designers and artists have long known through practice: activating emotion, leveraging embodiment, and creating novel presentation forms help people learn more from data-driven arguments, and remember them more fully.
It turns out that visceralizing data may also help designers solve a pernicious problem in the visualization community: how to represent uncertainty in a medium that’s become rhetorically synonymous with the truth. Its “truthiness” is both a feature and a bug. One of the best things about data visualization is that it does look so certain, so factual, and so authoritative. But why? After doing a sociological analysis, Helen Kennedy determined that four conventions of data visualization reinforce people's perceptions of its factual basis: 1) two-dimensional viewpoints, 2) clean layouts, 3) geometric shapes and lines, and 4) the inclusion of data sources at the bottom. These conventions contribute to the perception of data visualization as objective, scientific and neutral.
But even if you use these conventions with the best and most pure intentions, it's something of a problem for feminist design, because feminist theory maintains that there is no such thing as a purely objective view of the world. Knowledge is always partial, as Sandra Harding has shown us, and these conventions would seem to contradict that basic philosophical tenet. Haraway's "God trick," which we discuss in Bring Back the Bodies, is exactly that: a trick to make you believe that you can see everything, all at once, from an imaginary and impossible standpoint.
This is the argument from philosophy. But representing uncertainty is also a known problem in data journalism and visualization research. In these realms, people may care a tiny bit less about feminist epistemology but do care deeply about end users' ability to accurately interpret graphic depictions of data and use them to make decisions. To this end, designers have created a huge array of charts and techniques for quantifying and representing uncertainty. These include box-plots, violin plots, gradient plots, and confidence intervals. Unfortunately, however, people are terrible at recognizing uncertainty in data visualizations, even when they’re explicitly told that something is uncertain. According to work by Geoff Cumming and colleagues, researchers themselves have a hard time understanding confidence intervals. And even everyday weather forecasts such as "There's a 30% chance of rain tomorrow" are generally interpreted by the public to mean "It will rain 30% of the time" or "It will rain in 30% of my area.”
For example, let's consider the Total Electoral Votes graphic, which was displayed as a followup to the New York Times live online coverage of the 2016 election. The blue and red lines represent the Times’s "best guess" of the outcomes over the course of election night and into the following day. The gradient areas show the degree of uncertainty that surrounded those guesses, with the darker inner area showing electoral vote outcomes that came up 25% to 75% of the time, and the lighter outer areas showing outcomes that came up 75% to 95% and 5% - 25% of the time, respectively. Note that, at 6pm on election night, the outcome of Trump winning and Clinton losing easily falls within the 5 - 25% likelihood range.
Thus while many election postmortems pronounced the 2016 election the Great Failure of Data and Statistics, most reputable forecasts always included the possibility of a Trump victory. The underlying problem was that people are not trained to see the uncertainty in graphics like this. Rather than interpreting the gradient bands as probabilities (e.g. Trump had a 20% chance of winning at 6pm), people interpret it as votes (e.g. Trump had 20% of the vote at 6pm). This is called "heuristics" in psychology literature - using mental shortcuts to make judgements - and it happens all the time when people are asked to assess probabilities. Part of the problem is that visualization conventions reinforce those misjudgements. The graphics look so certain, even when they are trying their very hardest to visually illustrate uncertainty!
So, when people make decisions under conditions of uncertainty (which is most of the time), they use heuristics that help them arrive at a conclusion. Sometimes, these heuristics are helpful, like detecting snakes lurking in the grass. But in other cases, they can lead to patterns of systemic cognitive bias. For example, it is hard for people to remember all information they have learned, so they tend to recall vivid examples more than mundane examples. This is called the "availability heuristic" and leads us to overestimate the probability of things like being attacked by a shark.
Visualization researcher Jessica Hullman offers one solution to this problem. Instead of creating fixed plots like the New York Times example that represents uncertainty in aggregate or static form, Hullman advocates for rendering experiences of uncertainty. In other words, leverage emotion and affect so that people experience uncertainty perceptually. Or, to invoke a common refrain from design schools: "show, don't tell." Rather than telling people that they are looking at uncertainty while employing a certain-looking graphic style– which creates conditions ripe for those pesky heuristics to intervene– make them feel the uncertainty.
We can see a good example of showing uncertainty in action on the same New York Times live election coverage webpage. At the top of the page was a gauge that showed the Times' realtime prediction for who was likely to win the race, with a gradient of categories that ranged from medium blue ("Very Likely" that Clinton would win) to medium red ("Very Likely" that Trump would win). But like a speedometer in a car, the needle did not stay in one place. It jittered between the 25th and 75th percentile, showing the range in the outcomes that the Times was then predicting, based on simulations using the most recent data. At the beginning of the day, the range of motion was fairly wide but still only showed the needle on the Hillary Clinton side. As the night went on, its range narrowed, and the center moved closer and closer to the red side of the gauge. By 9PM, the needle jittered just a little, and on the Trump side only.
A number of Times' readers were aggressive in their dislike of the jitter, calling it "irresponsible" and "unethical" and "the most stressful thing I’ve ever looked at online and I’ve seen a lot of stressful shit." In response, one of the designers of the gauge defended the decision, explaining that "we thought (and still think!) this movement actually helped demonstrate the uncertainty around our forecast, conveying the relative precision of our estimates." So was this “unethical” design, or the sophisticated communication of uncertainty?
Hullman would argue for the latter. The jittering election gauge was actually exhibiting current best practices for communicating uncertainty. It gave people the perceptual, intuitive experience of uncertainty in order to reinforce the quantitative depiction of uncertainty. The fact that it unsettled so much of the Times readership probably had less to do with the ethics of the visualization and more to do with the outcome of the election– about which the left-leaning readership of the paper continues to despair. So score one for emotion in the task of representing uncertainty.
Emotion is not only useful for communicating uncertainty. Let's return to the question around visual minimalism: Are minimal charts really "better," as Edward Tufte has claimed? Charts like "Monstrous Costs," depicted above, have been historically dismissed as "chart junk." In this case, the data has to do with presidential campaign spending from 1972 - 1982. You will notice that there is a very low data to ink ratio– recall Tufte’s other major pet peeve. For example, the monster's eyes and body would be deemed as "not data" and thus embellishment. Likewise, there is gratuitous spittle emanating from the monster's mouth, and the tail forms an S-curve that is then turned into a dollar sign with two utterly superfluous grey lines. The monster is also wearing a VOTE badge, which, again, does not represent any aspect of the data at hand. Plus, monsters do not vote– so how irrational is that?
There are many popular blogs devoted to ridiculing this so-called chart junk. But recently, researchers have challenged the notion that chart junk is junk at all. A 2010 study by Scott Bateman and colleagues in Computer Science at the University of Saskatchewan that found that "embellished" charts do not hinder people's ability to accurately read them, and in fact, they are actually better for memorability. When polled 2 to 3 weeks later, people were much more likely to recall the message of the embellished chart over a minimalist chart that displayed the same data. People also thought the junk charts were more attractive, and enjoyed them more (Duh. Who doesn't like monsters better than bar charts?!) Likewise, in 2016, Michelle Borkin and colleagues showed that visualizations that make use of novel presentation styles are more memorable, and therefore more effective. Relating the visual form to the topical content of a chart works. So, as data journalist Mona Chalabi says, "If it’s about farts, draw a butt for god’s sakes."
After all this, you might be wondering, "So why is it that people like junk so much?" But that is not really the right question. The question we should all be asking is, "How has the view from an imaginary and impossible standpoint– the 'whole picture,' the 'overview,' – come to be seen as rational and objective at all?" The rational, scientific, objective viewpoint actually comes from a mythical, imaginary, impossible standpoint. The view from no body. "The god trick of seeing everything from nowhere," as Donna Haraway says it.
But let's say it more simply, "How did we arrive at conventions in data visualization that prioritize rationality, devalue emotion, and completely ignore the human body except for two eye stalks attached to a brain?" Any knowledge community inevitably places certain things at the center and casts others out, in the same way that male bodies have been taken as the norm in scientific study and female bodies imagined as deviation from the norm, or that rationality has been valued as an authoritative mode of communication and emotion cast out. But, following feminist theorist Elizabeth Grosz, what is regarded as "excess" in any given system might possibly be the most interesting thing to explore because it tells us the most about what and who the system is trying to exclude.
In the case of data visualization, this excess is emotion and affect, embodiment and expression, embellishment and decoration. These are the aspects of human experience coded "female," and thus devalued by the logic of our master stereotype. But how might we intentionally flip this? All design fields, including visualization, are fields of possibility. We must actively strive to question what (and who) is at the center of the discourse of our field, what (and who) is at the periphery. And then we must work to center the things and people that have been cast off. The first step in this re-centering process is to legitimize affect and embodiment in data visualization. This means to moving emotion to the center of visualization design, and to start to imagine data experiences for whole human bodies in all of their glorious, situated, uncontainable excesses.