To find out if a story is going viral on the internet, a crystal ball might seem handy. After all, any equation or formula for figuring it out should account for a magical blend of a story’s qualities, quirky human preferences, and online habits – at least that’s what you’d expect. But according to neuroscientists at the University of Pennsylvania, our brains may have a simple, generic calculation for it.
As we read news stories, our brains gauge whether a story is interesting to ourselves or to others, and whether sharing it could improve our standing or relationships, the researchers report. Then our crafty noggins seem to use default rating machines in the brain to essentially combine those ratings, score the story, and ultimately decide whether to share it or not. By monitoring those brain processes as people flipped through them New York Times health articles, the researchers could better predict which stories actually went viral on the web. Their results were published Monday in PNAS.
“It’s cool that the brain has developed this kind of specialized skill,” senior author Emily Falk, director of Penn’s Communication Neuroscience Lab, told Ars. Psychologists might expect that if you just provide health information to people in a logical way, that’s enough – people will have an objective understanding of its value. “But our data suggest that two of the really important inputs to ‘the value signal’ are these potentially holistic assessments of how self-relevant and socially relevant information is.”
And it seems that those inputs and calculations are generic. “There’s enough commonality between people that we can take brain activity and say something sensible about what’s going to happen around the world… and the whole world.” New York Times readership,” she added. (The study participants were not NYT‘s core audience, and many didn’t read the newspaper, she stressed.)
To stamp out the brain activity behind virality, Falk and colleagues, including lead authors Christin Scholz and Elisa Baek, first conducted large studies and meta-analyses that linked thought processes to specific brain regions. They singled out subregions and systems in the brain related to appreciation processing, self-relevant expectations and “mentalizing,” which is an understanding or prediction of the mental states of others.
They then used functional magnetic resonance imaging — which tracks activity in the brain based on blood flow — to follow 80 participants as they viewed more than 80 headlines and news story summaries. All stories had to do with health, but they had varying degrees of success on the Internet, which the participants were unaware of. Of the selected stories, the total number of shared posts across Facebook, Twitter, and email ranged from 34 to 12,743 shares.
The researchers analyzed the participants’ brain activity and recorded relative activity changes in the brain regions they had already identified. On data modeling, they found that activity in all regions was related to participants’ sharing decisions and real-life virality. But the researchers noticed a pattern: Activity suggesting self and social interest wasn’t enough to explain virality — the brain also had to involve its rating system. An increase in activity in regions of self-interest and community interest was linked to activity in valuation processing. And high value activity was key to sharing: “The more activity we saw in the brain’s value system, the more we saw in virality,” explains Falk.
That value processing center is generic, says Falk, and not focused on news. The activity is also involved in deciding what to eat or do.
The researchers also looked at whether aspects of the articles themselves could beat brain activity in predicting virality. The authors used computers and human raters to rate each story on whether it was perceived as positive, helpful, controversial, suggestive, or novel. Brain activity preceded these other measures.
But there’s still a secret sauce to virality on the web. While brain activity was the strongest predictor of virality tested, even when combined with survey participants’ responses, it could still account for only about 20 percent of the variation in article sharing. That’s double what we were able to do before, Falk points out, but still far from predictive.
For communicators trying to convey important health information, it comes in handy. But researchers have more work to do to fully understand information sharing in our digital worlds.
PNAS2017. DOI: 10.1073/pnas.1615259114 (About DOIs).