Data Visualization and Journalism

The word “visualization” once more readily described the act of creating a mental image in one’s mind, whereas today it’s far more likely to mean the graphical representation of information. We are living in an increasingly visual world, peering into screens of different sizes with incrementally superior resolutions at every device upgrade. We are also living in a world with more data available to us than ever before. IBM says that 90% of the world’s data was produced only in the last two years and that we produce 2.5 quintillion bytes of data daily, contributing to what some are calling a new planetary nervous system. Marry the large amounts of data with how our brains understand images faster than they do text and you get the bourgeoning world of data visualizations; it’s the most efficient way for us to understand and grasp buried patterns, relationships, and to find stories in the data. Just as text, photos, and videos represent journalistic formats that inform and empower the public to make decisions, data visualization is fast joining the ranks as an equally important format.

Julie Steele of O’Reilly categorizes visualizations into three buckets, which can be a useful framework to keep in mind:

• Infographics use a small data set and significant manual design, such as this National Geographic graphic.
• Data visualizations use a large data set with less manual design; it’s algorithmic. For example, this New York Times interactive.
• Visual art is unidirectional coding; it’s beautiful, but the brain can’t decode the information, for example, computational art by by Kunal Anand.

What Is the Problem?

Data visualizations in journalism are often influenced by fields such as computer sciences and mathematics. As a result, exotic forms, shapes, and interactions are being used to illuminate data in journalistic works. This is amplified by the proliferation of tools that make all sorts of visualizations easier to create, such as Many Eyes, Tableau, and open source libraries including D3. These are some reasons why many works end up appealing to super-users and hinder non-specialists from delving in, which is harmful when the goal is to inform the public. This is why some understanding and awareness of the growing pains around visual literacy with data visualizations is so important.

A New “Visual Grammar” in Journalism

The journalism community celebrates the following examples of recent interactive visualizations, which are finalists in the 2013 data journalism awards. All three experiment with ways to present interactive journalism and provide rich views of important data. They look impressive but can also be intimidating to interpret.

Gay Rights in the US, State by State

The Guardian’s visualization about gay rights in the United States created a lot of buzz in the data visualization and journalism worlds because of its experimental format. The chatter and debate keenly highlight the tension between trying new things and sticking to more accepted formats. The graphic represents the country in a circle, which allows more information to be displayed at once than in a map format. The shapes and colors are eye-catching but they are also an additional layer for the reader to have to commit to trying to understand before getting to the matter.

The Guardian’s interactive on Gay Rights.

Connected China

Reuters presents a beautiful, ongoing package that visualizes thousands of complex relationships among China’s power structure. The design is clean and minimal, though the information feels overwhelming and disembodied from the stories, which are all packaged in a separate section. A deeper internet search leads to a series of related YouTube videos that don’t seem to be linked from the package itself. These video explainers of the navigation are a good idea though would be much more useful if they also helped users read the visualizations and understand the content itself.

Connected China interactive from Reuters.

Decisive Demographics

AP’s look at demographic trends during U.S. presidential elections includes smooth interactions and opens with a heat map view that is clear to understand but could have benefited from a layer of annotation that provides some explanation. Once you dig deeper, there is a confusing percentage bar view with color values that is broken apart, with no anchor for the eye.

The AP’s Decisive Demographics interactive.

Without sufficient thought to the visual literacy of audiences, graphics risk alienating news consumers. Many data visualization forms may look novel and interesting but if average visual literacy is not taken into account, the message and the story will get lost.

What is Visual Literacy?

Visual literacy deals with the ability to understand and interpret images. In Colin Ware’s book Information Visualization: Perception for Design, he talks about the debate around the “sensory” and “arbitrary” distinctions of images. The sensory defines images that can be processed by the brain without explanation; aspects our brains have evolved to understand quickly without learning, such as cave drawings and textures. Arbitrary defines representations that must be learned, such as the alphabet, conventions in diagramming, or cultural associations with color. Computers allow us to create new graphical codes, which means arbitrary forms need more explaining to users since they aren’t intuitive and must be learned.

In order to understand which visual encodings (for example length, direction, and color) are most accurately understood by people, William S. Cleveland and Robert McGill conducted a series of experiments in their seminal 1984 paper “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods.”

Encoding quantitative data into visuals is complex—but the most efficient decoding of the data, according to the paper, is ranked as follows:

1. Position along a common scales (e.g. Bar graphs, scatter plots)
2. Position along nonaligned scales (e.g. Multiple scatter plots)
3. Length, direction, angle (e.g. Bar charts, pie charts)
4. Area (e.g. Bubble charts, tree maps)
5. Volume, curvature (e.g. Volume charts)
6. Shading, color saturation (e.g. Heat maps)

The paper concludes that the ranking should be referred to as a guideline, not a complete prescription, since there are often other complexities at play.

Ask Questions about Visual Literacy and More

For maximum impact, a visualization needs to fulfill its purpose: what is it trying to demonstrate or clarify? It’s equally important to focus on how to make it accessible to the audience. Once you have processed the data and are experimenting with visualizations for the audience, you might ask yourself:

• Who am I creating this for?
• What journalistic impact should the visualization have?
• If I opt for novel graphical/interaction styles, what guidance will I provide the audience?
• Should I blend exploratory aspects with explanatory aspects?
• How will I expose the story?
• Can I add a narrative, causation information, or a news peg?
• Although I’ve edited the data already, is there superfluous data that I can still edit out?

Accolades to Accessibility

The statistician Hans Rosling does an excellent job of illuminating visualizations and making them accessible to audiences. In this TED presentation he inserts himself into the charts, blending context and causation to create a narrative around global trends in health.

In one example Texas Tribune not only explained what a tree map is but actually created a video tutorial on how to read their visualization of ads that appeared in the 2010 race for governor in Texas. Rather than overly simplify information graphics and interactivity at a time when data is complex and presentation possibilities are vast, they attempted to bring their audiences along with them.

Using narrative to guide the audience is an effective way to provide context for visualizations, making them more accessible, as seen here in Quoi Info and the BBC.

Deep Background: More on How-To

• In Geoff McGhee’s 2010 documentary, Journalism in the Age of Data, he explores traditional narratives merging with sophisticated displays. It’s an excellent primer about the visual journalism landscape.
• Juan Velasco of National Geographic explains overcoming the challenges of data-rich graphics when thinking about mobile users.
• Jeffrey Heer and Michael Bostock seek to validate future cheaper, crowdsourcing techniques in their perception experiments using Amazon’s Mechanical Turk. Check out their 2010 paper, “Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design,” where they replicate previous studies and compare the results.
• Edward Tufte’s classic The Visual Display of Quantitative Information is startlingly clear as he guides the reader through the basics of excellence in statistical graphics. Chock-full of examples, his focus includes graphical integrity, data density, and chartjunk.

Start Here

There isn’t a magic formula that will give a data visualization balanced appeal to both initiated and uninitiated audiences. But there are frameworks and guidance around visual literacy and best practices available, some of which I’ve outlined here.

There is a new grammar that is being created in the world of data visualization, which needs to be balanced in the context of journalism’s responsibility to inform the public and help them make decisions. That doesn’t mean we have to resort to only simple forms and large chunks of explainer text, but it does mean we have to be sensitive to folding teaching into our work, and provide clear guidance to the reader; we have to be creative and look for new formulas to do that.

The need to explain a design may seem at first counterintuitive. After all, truly good design shouldn’t require a separate explanation. But if we want to create visualizations with more depth and complexity, we also need to heighten public awareness of the techniques we use because the payoff to the deeper understanding of the world is worth it.

It’s very easy to get distracted by the amazing tools we have at hand today and as they evolve and we get better at using them, we should remember to stop and check our mission. We should be sensitive to the journalistic goal, respect our audiences and their time, and remember that we are not the audience.

Every visual story involves negotiation between intuitive and learned aspects of visuals. But also, more importantly, they all require a big dose of human judgment.