visualizing uncertainty

Since there are so few authorities on this topic, I'm hoping my explorations can begin with sets of ongoing conversations with people from various fields. It only seems natural to try to better understand uncertainty not by referencing or becoming an authority on the matter, but by organically mapping out the questions we all have and hypothesis to test. 

There's currently an open call for speakers at Bocoup's OpenVis conference, set in Boston this April. A shared spreadsheet of topics the organizers would like to see covered has been going around, and it seems like one of the most agree-upon wants is for someone to talk about visualizing uncertainty. A number of others listed have to do with the role of art and aesthetics in data design. 

Uncertainty of various sorts:

  "Fundamentally inaccurate uses of the principle are also common in the academy, especially among social theorists, who often argue that it undermines science’s claims to objectivity and completeness. As Jim Holt has written, “No scientific idea from the last century is more fetishized, abused and misunderstood — by the vulgar and the learned alike — than Heisenberg’s uncertainty principle.”

In the swirling chaos of twenty-first century capitalism, everybody wants to know what’s next. “Data is the new oil” is a pithy little announcement. It reminds us how we got here, powered by the long energetic boom of fossil fuels, now entering its closing stages. it announces a successor, a new wealth (and just in time). But in drawing the analogy, it also constructs data in a certain way; as a sort of amorphous but precious stuff, a resource for exploitation, and a sort of promising abundance. Similarly The Economist trumpeted the “Data Deluge” on their February 2010 cover: a businessman catches falling data in an upside-down umbrella, funnelling it to water a growing flower whose leaves are hundred dollar bills.

We need not (and should not) accept this analogy; but it demonstrates how data is figured, or constructed, in our culture. Our everyday life and culture is traced, tangled and enabled by digital flows. We produce and consume data as never before. But what exactly is this data? What can it do, and what can we do with it? Who owns or controls it? How can we understand, appreciate, or even sense it? The construction of data as a cultural actor is vital because data itself is so abstract, so hard to pin down. We ought not leave it to the captains of industry, and their upside-down umbrellas. In Datascape we see artists working with data, applying and diverting it for their own ends, as well as offering their own figurations of its potentials and limits. In a culture increasingly built on data, these works provide moments of cultural introspection, reflections on this abstract stuff that is our new social medium.

Close your eyes and picture “data.” What do you see?

Jake Porway begins his lunch talk at Hyperakt, hosted by the studio on the first spring-like Friday in April, by posing this question to the audience. The responses paint a predictably dry portrait: spreadsheets, rooms of servers, and the anonymous, matrix-like flickerings of 0’s and 1’s. These commonly held assumptions, he laments, can obscure all the very intimate ways we produce and are produced by our data.

Realising the personal and meaningful potential of data is something he’s built quite the career out of. By now, hackathons are a familiar and celebrated staple of the urban landscape — a flexing of creative and technical muscle under hectic conditions. But at a session Porway once attended, he felt that the output was out of sync with big picture issues. The brightest minds were churning out apps to find parking and restaurant ratings — tools to make the lives of the comfortable just ever so slightly more convenient.

To help foster more meaningful collaborations, he co-founded DataKind, which funnels that surplus energy and talent towards nonprofits, NGO’s, and government agencies that need it the most. Their DataDives adopt the hackathon model, where data scientists spend a marathon weekend with groups that have well-defined problems to tackle. DataCorps is a longer venture, pairing volunteers with the organizations for 1-6 months for a more in-depth exploration.

Transforming intimidating fields of numbers into an appropriate visual expression is work that can easily be taken for granted. But the results vouch for themselves, providing the organizations the “ah-ha” moment of new insight into their own content. For most of DataKind’s partnerships, the effort goes into cleaning and visualizing the data to build exploratory tools, though this also sets the groundwork to make insight available to users at many literacy levels. Although a line is usually drawn between visualization for internal, academic purposes and those designed for public communication, the conversation between Porway and audience members turns to how even experts benefit from the use of intuitive, visual metaphors. As the saying goes, “if you can’t explain it simply, you don’t understand it well enough.”

With the growing popularity of data visualization have come a fair share of guidelines, tutorials, and manifestos. One of the often-repeated commandments tells us that "form follows data,” which seems to overshadow the nature of data as a cultural artifact. Porway shows a few examples of how deeper interrogation unravels these assumptions. Among them was an argument between a New York Times report unfavorably reviewing the new Tesla Model S, and the motor company disputing the review based on information collected by the car itself. Of course, this is only one of many examples of how the forensic process of finding the narrative in a spreadsheet can yield many, differing readings.

Porway finds his work indeed has more of the organic quirks of scientific processes, as initial expectations about a dataset fall apart when deeper investigation reveals missing information, noise, and biases. One of the benefits of DataKind collaborations to the data scientists is the exposure to policy experts, who have the qualitative expertise to help navigate the quantitative. It’s perhaps what influences his next open question to the audience: What responsibilities do we have as data scientists and visualizers? And to whom? No simple answers can be offered, though the audience — a mix from academia, statistics, and design — share concerns from their various perspectives.

  The last few slides give a glimpse at the recently released NYPD stop and frisk data. Instead of launching into the “raw” information first, Porway rounds out the talk with another question: What questions can we ask before we even look at the data? This time the room fires back with suggestions: whether or not there are quotas, how strictly stop and frisks are recorded, and how racial categories might be defined, to name a few. Even this short exercise seems to bring a more invigorated, critical perspective to discovering the problem at large, or, at the very least, bringing a heightened awareness to the problem at hand.

----

Catch Jake Porway on the latest Data Stories Podcast and follow @jakeporway and @DataKind

Spurred by Olympics and elections in 2012, the world of data and its visual manifestations celebrated exciting developments in new programming languages, journalistic insight, and design fluency that tended to be the rule rather than exception for once. However, the fervor over this growing domain comes with its fair share of disillusionment, and the calls for slow data, skepticism, and more intimate messages are growing into an audible din. As these things tend to go when preceded by a flood of hype, much of the nuance and interesting details surface after the dust settles.

One facet of this, the role of art in data visualization, has been subject to ongoing debate as the community has expanded and diversified. Among some high-visibility practitioners, there seems to be a tendency to put art and science on separate and opposing ends of the spectrum. In the most extreme cases, use of a personal aesthetic becomes associated with frivolous decoration and the pursuit of novelty for novelty’s sake. At the same time, efforts from designers working independently from data scientists and the original data sources can indeed produce the superficial results that critics point to in problematizing the use of style. But for the sake of progress, we may as well avoid the lowest common denominators.

 There are, of course, important technical considerations for exploratory visualizations contained within a field; and sometimes all it takes is a bar chart or scatter plot. But as the overall body of knowledge continues to grow, so does the dire need to capture the attention of others and accurately share insights to unfamiliar audiences. It’s worth keeping in mind that these pieces are ultimately communications to other humans—and not just an amorphous “public,” but individual interpreters having discrete experiences. If we take a step back, reexamine our intentions, and unpack some of these generalizations, we may be better able to understand the mechanisms by which art and science conspire to distill and communicate meaningful information.

  Process, not style 

“The biggest failure of data art, in my opinion, is in neglecting to address the individual character of a data set. We have all kinds of quite beautiful methods to visualize datasets of a specific sort – say network data sets – but they tend to be general use algorithms and techniques. As a result, visualizations of varied data sets tend to look the same.”

— Jer Thorp 

With powerful tools at hand to work with, it’s easy to see why many default to templated, established, formal solutions. It’s hard to disagree that visualizations should communicate clearly and refrain from too much novelty, but should we settle for visual sameness and stagnation? This isn’t to say that the answer is in arbitrarily applying veneers of style (hand-drawn sketchiness, nostalgic print treatments, etc). In any creative endeavor, it’s more important to focus on the process than the final “look” of a piece. No instagram filter could ever replace the thoughtful consideration of color and composition, of semiotics and visual perception that actually tighten a message and make it more deliberate and impactful. 

Of course, it’s easier to talk about these things in broad terms and metaphors than to understand and employ the right creative strategies. The following is merely a quick survey of designers and artists working around the periphery of data representation, hopefully in ways that can be inspiring and instructive across a broader range of practice.

Know the source.

When the topic of the human touch in information design come up, it’s impossible to overlook the data-crafted approach of Stephanie Posavec. Her work is often cited for the beauty of the final diagrams—uniquely organized through her own visual intuition—but it’s worth calling attention to the labor of love that is her data gathering process. As described in a short documentary by Protein, she explains: “by reading and rereading these texts, I’m able to understand more about a specific text or a specific subject matter than I would otherwise, than I would if I wrote a computer program to analyze that text for me.” Posavec chose to work with On The Road because of the novel’s personal significance, and in encoding the text by hand in her Writing Without Words series, she’s able to interpret the shades of meaning to be gained from great literature specifically by humans, just beyond any current algorithm’s reach.

Becoming more deeply involved in data collection and analysis may be one path towards crafting visualizations that resonate on a more poignant, intimate frequency. Also, in order to reach others at a personal scale, why not begin with personal interests and causes?

  Get Personal.

This is how Ai Weiwei began his investigation into the 2008 Sichuan earthquake. Following the devastating natural disaster and subsequent Chinese government cover-up, the artist took it upon himself to bring in a team to find the names and birthdates of child casualties that resulted from shoddy public school construction. Though the final visual outcome wasn’t a direct numerical representation of the original data, even the simple act of publishing those names to his blog formalized the message in a powerful way that rallied supporters and attracted harassment from local authorities. For a 2010 retrospective at Haus der Kunst in Munich, he used 9000 colored backpacks to write out a phrase from a mother of one of the victims—“she lived happily for seven years in this world.” In the documentary Never Sorry, he explains how the idea to use the backpacks came from being on the ground. Seeing those objects strewn about the rubble as the physical reminders of lost lives, and ultimately incorporated them into a message that is at once intimate and forceful.

Cultivating projects for a human scale doesn’t have to do with the size of the original datasets, but the point along the design process where you shift between the birds-eye-view and taking a reader through the experience on the ground. The map is not the territory, but there’s value to both take and give from getting closer to a first-person perspective.

  Maximize specificity.

Without venturing into the modern data visualization canon, I’ll end with an example from Berlin-based design firm ART+COM, which employs media specificity as a fully formed design philosophy and methodology. By embracing the fact that the tools, platforms, interfaces, and context all become part of the DNA of the message, they’re able to execute an impressive array of works that fit a unique niche for both their clients and audiences.

Mobility was a project built specifically to stand out in the 2010 Shanghai World Expo and speak to the concerns for handicapped people. Commissioned by Otto Bock HealthCare, they took the opportunity to utilize their client’s prosthetics in the installation, which combined neatly with an ancient form of information exchange—reflecting light in mirrors. Each hand was fitted with a mirror, then painstakingly positioned and choreographed to realign into the Chinese character for movement.

 A screen or wall of LED’s would also have done the job, but given the bright, chaotic surroundings at the 2010 and earlier World Expositions, their decidedly lo-fi route provided a moment of respite in an experience otherwise pervaded by grandiose architecture and a bombardment of flashing, colored lights. As with the previous examples, the combination of human intuition and sheer effort that went into researching and crafting these simple messages comes through.