The tiny fraction of headlines that news editors push out on Twitter draw a large share of eyeballs, but it's the stories recommended by friends that trigger more clicks.
In what may be the first independent study of news consumption on social media, researchers at Columbia University and the French National Institute (Inria) found that reader referrals drove 61 percent of the nearly 10 million clicks in a random sample of news stories posted on Twitter. The researchers present their results on June 16 at the Association for Computing Machinery's Sigmetrics conference in Nice.
"Readers know best what their followers want," said the study's senior author, Augustin Chaintreau, a computer science professor at the Data Science Institute and Columbia Engineering. "In the future, they will have more and more say about what's newsworthy."
Social media in 2014 overtook organic search as the top way people accessed content on the web, driving 30 percent of all traffic. But despite the social web's growing influence, relatively little is known about how people consume news on these proprietary platforms. Facebook, and to a lesser extent, Twitter, filter and personalize news for users and closely track the results, but because this data is fundamental to their advertising business very little is made public.
The researchers attempted to peer under the hood by collecting all the open data they could find-- the number of Twitter's 280 million followers who potentially viewed and shared a news link shortened by the web app, Bit.ly, and how many clicks those links received. From the one percent of tweets made public by Twitter, the researchers picked all URLs linked to five news outlets--BBC, Huffington Post, CNN, New York Times and Fox--during a one-month period last summer.
The goal was to find out which stories in their sample of tweets would be shared and clicked on more: the less than 2 percent of headlines news editors picked to promote from their official Twitter feed, or the headlines readers found on Twitter and shared themselves.
Though far more readers viewed the links news outlets promoted directly on Twitter, the study found that most of what readers shared and read was crowd-curated. Eighty-two percent of shares, and 61 percent of clicks, of the tweets in the study sample referred to content readers found on their own. But the crowd's relative influence varied by outlet; 85 percent of clicks on tweets tied to a BBC story came from reader recommendations while only 10 percent of tweets tied to a Fox story did, the researchers found.
Their results also suggest that people are quicker to share, than read, news discovered on Twitter. Though social networks commonly measure a story's popularity in shares, researchers found that 59 percent of all links shared in their sample went unclicked, and presumably unread.
"People are more willing to share an article than read it," said study coauthor Arnaud Legout, a research scientist at Inria. "This is typical of modern information consumption. People form an opinion based on a summary, or summary of summaries, without making the effort to go deeper."
For those willing to read, the study finds that stories on Twitter have a relatively long shelf life. While more than 90 percent of links in the study were shared within a few hours, most links were clicked on, and presumably read, much later; 70 percent of clicks happened after the first hour, and a full 18 percent happened in the second week, the study found.
"Our results show that sharing content and actually reading it are poorly correlated," said Legout. "Likes and shares are not a meaningful measure of content popularity. This means that the industry standard for popularity needs to be rethought."
Social media has upended many professions, including journalism, by allowing everyone to become a creator and publisher with a built-in audience. The study optimistically affirms the social web's democratizing power.
But it's also a reminder that popular news platforms like Facebook and Twitter are now the effective gatekeepers, controlling what we see and read. The researchers worked since 2013 to develop their method for inferring behavior on Twitter with limited public data.
"This is an unregulated field," Columbia journalism professor Emily Bell writes in a recent piece for Columbia Journalism Review. "There is no transparency into the internal working of these systems."
In pursuit of transparency, the researchers said they hope to build on their method and continue to explore, as Chaintreau put it, "which voices really get heard and shape public opinion."
The study's other authors are: Maksym Gabielkov, Inria; and Arthi Ramachandran, Columbia.