6 lessons academic research tells us about making data visualizations
As more media outlets embrace data, and data visualizations become integral to storytelling, it's increasingly important to understand what makes an effective graphical presentation. While much has been written about data visualization tools and ways to teach data visualization, much less has been written about the academic research on effective visualization. Journalists may not be aware of the body of scholarly work on data visualization other than a passing awareness of Edward Tufte's popular books. But examining the academic literature on data visualization can have real implications for the practical, working journalist.
Here are 6 lessons that we can learn from academia:
1. Incorporate charts to increases persuasion.
Researchers from Cornell University found that merely including a graph in an article significantly increases reader persuasion. While only 68% of participants believed a scientific claim without a graph, a whopping 97% of participants believed that same claim when a graph was included. While their study examined persuasion in science in particular, it's not a stretch to extend their conclusions to news in general. The clear lesson for visual journalists is to incorporate graphs to have greater sway with your readers.
2. When accuracy counts, use bar charts over bubble or pie charts.
William Cleveland and Robert McGill were one of the first researchers to provide scientific results to back up what had been mostly common sense arguments and critiques of existing charts. In their landmark 1984 study of graphical perception, Cleveland and McGill ordered graph types according to how accurately people were able to "decode" the data from visual properties (for example, the area of a circle, the angle of lines, or the length of bars). A simplified version of their resulting hierarchy of graphical encodings is as follows, from most accurate to least:
1. position (dot plots, scatter plots)
2. length (bar and column charts)
3. angle (pie charts)
4. area (bubble charts)
5. color (choropleth maps)
[caption id="attachment_381170" align="alignleft" width="740"] Bubbles force readers use area to compare data, while bars rely on length. According to research, length helps readers make more accurate comparisons. In this New York Times graphic, bubbles provide the big picture immigration patterns and bars provide the more accurate details.[/caption]
3. Bigger is not always better.
Researchers studying the effect of size found that, as charts become shorter, reader accuracy decreases. However, while making charts taller increases accuracy, the effect quickly plateaus. The studies suggest that there is an optimal chart size that maximizes accuracy, which designers must consider along with the layout of any given page. Big charts may have more visual impact, but there is no added value in terms of reader comprehension.
4. Eyes beat memory: use spatial comparisons rather than temporal ones.
Tamara Munzer, a professor of computer science at the University of British Columbia, advises designers to help readers make comparisons as easily as possible. This means putting the things to be compared side-by-side (spatially) rather than in sequence (temporally). In her words, "eyes beat memory" since our brain is much better at looking at two items to compare than it is at remembering two items that were shown one after the other, as in an animation or video. She's not the first to advocate spatial comparisons, but it's good advice for any data visualization. What does it mean in practical terms? To show change, don't assume an animation is the answer. Reveal change or difference by presenting each stage right next to each other. Your end result is a grid of charts that Edward Tufte calls "small multiples."
5. Innovative graphs are more memorable.
Researchers at Harvard and MIT pursued a different question about data visualization. Rather than effectiveness or accuracy, they explored what makes a visualization memorable. After collecting, classifying, and conducting perception studies on over 5,000 visualizations, they conclude that the most memorable visualizations are those that include embellishments, more color, higher visual densities, faces and other recognizably human elements. They also find that unconventional graph forms are more memorable. However, they warn that their research only considers the visualizations as if they were images—their study ignores the comprehension of the actual data. Despite that limitation, their study imparts concrete, but perhaps obvious, advice on how to design a graph with lasting impact—go with unique, colorful, and decorated presentations.
Looking at award-winning visualizations suggests that the much of their research on memorability bears out. Many of the award recipients celebrated by Malofiej, the premiere venue for information designers, feature innovative and colorful forms that demand our attention.
[caption id="attachment_381172" align="alignleft" width="740"] The award-winning graphic from the Guardian features a large circular design as a color matrix of gay rights. According to research, unusual graphical forms are more memorable than conventional ones. Would this be as compelling if it were laid out in a more conventional rectangular grid?[/caption]
6. Embellishing your graphs with illustrations? The jury is still out.
[caption id="attachment_381181" align="alignleft" width="350"] Researchers compared embellished charts by graphic designer Nigel Holmes (top) with plain charts (bottom) and found that embellished charts were more memorable and afforded no less accuracy.[/caption]Some of the most contentious research revolves around the assertion that visual decoration distracts from the readability of charts, first championed by data visualization guru Tufte. Researchers from Canada questioned that idea and compared minimalist charts to lavishly embellished ones. They discovered that reader accuracy for embellished charts was actually "no worse than for plain charts", contradicting Tufte. Moreover, participants of the study were able to recall data from embellished charts significantly better after more than two weeks. So, should charts be decorated with illustrations? Advocates of minimalist chart presentations have balked at the thought and criticized the study's methodology. The debate between embellished versus unembellished graph effectiveness continues with no clear consensus, leaving the decision more in the hands of the artist than the scientist.
The practice of data visualization straddles the line between science and art with the science—empirical research on effectiveness—informing journalists how best to implement the art. It isn't always easy to move from academic conclusions to practical guidance, however. Mike Bostock, who until recently was a graphics editor at the New York Times, warned in a recent Reddit forum, "the danger of academia is that it can easily become too abstract." Translating that abstraction is worth the effort. Academia can provide practical guidance to creating charts and graphs, and newsrooms should keep one eye on the efforts in the ivory towers.
Russell Chun is Assistant Professor of Journalism at Hofstra University, teaching multimedia storytelling and data visualization. Twitter @russellchun