Judging from some of Facebook's most viral images in history—textbooks wrapped in paper bags, futuristic beach houses, Barack and Michelle mid-hug—it seems safe to say that, content-wise, a mix of nostalgia and aspiration makes 'book users click, like, comment, and share. On Flickr, it's cool nature shots. Instagram loves the Kardashians.
But favorite subjects come and go, while the viral cycle lives on. What if you could predict the kinds of photos most likely to strike a nerve? Given the incredible amount of data available on how online users engage with images, is there a way to measure the objective qualities of an image's shareability?
Saeideh Bakhshi is searching for the answer. The research scientist at Yahoo! Labs' Human-Computer Interaction group has spent her career trying to understand what makes certain images catch attention on social media. "Historically, there hasn't been a lot of understanding about multimedia from a social science perspective," she says. "It's been hard to extract information from images."
But now, with a wealth of social media data and image-analysis tools, social science is getting closer to linking certain types of human behavior to particular qualities of photographs.
In a study published earlier this month on PLOS One, and which Bakhshi completed while still a PhD student at Georgia Tech, Bakhshi established a link between a photograph's main colors and its shareability. Analyzing 1 million images on Pinterest, she found that pictures composed mainly of the colors red, purple, and pink were more likely to be repinned, or shared. Blue, green, black, and yellow, meanwhile, suppressed diffusion.
The study controlled for factors that most impact photo engagement, such as the number of followers seeing it and how active certain users were. It's important to note, however, that the study did not control for gender, or the kind of content most commonly posted to Pinterest. As of 2012, some 71 percent of Pinterest users were female. Food and drinks, DIY and crafts, and home decor dominate the images shared on the platform—a demographic and and set of interests that may skew towards a certain color scheme.
Bakhshi's color conclusions are likely what she calls "platform-dependent." So are her other, past findings: On Instagram, photos with faces in them are much more popular than those of natural settings. On Flickr, filters that saturate color make certain types of images more shareable. But the most engaging images on any one of these platforms wouldn't necessarily be on another.
So while there's probably not a unified formula for crafting the perfect viral image, Bakhshi's research adds depth to old theories on the social science of media—and in the case of this most recent paper, on how the psychology of color affects emotion and behavior.
There are a few landmark studies on color psychology: Sports teams that wear red, for example, are more likely to win. Waitresses wearing red are more likely to receive tips from male customers. Prisoners who sit in cells painted with a peptic shade of pink have been shown to be calmer and better behaved.
Bakhshi's work is among of the first to consider how color affects online behavior and information dissemination—a question of enough significance to interest Yahoo!, of course, as well as the Defense Advanced Research Projects Agency, which "commissions advanced research" for the U.S. Department of Defense. DARPA funded the study.
"Colors tend to have a very implicit effect on mood and behavior, and these effects are not observed [by science] that strongly, so it's not easy to generalize about what colors people prefer and why," says Bakhshi. "But when you look at it on a larger scale, an online setting, then you have a collective behavior that can't otherwise be observed in things like politics, or culture, or food."