Science Probes Why Tweets Go Viral

Whether it's a cat video or a news story, what is a Twitter post's likelihood to spread? The answer has more to do with the structure of the network and our limited attention spans than the content or author, says an Indiana University study.

David Carr

April 5, 2012

4 Min Read
Network Computing logo

Facebook Apps In Action

Facebook Apps In Action

Facebook Apps In Action (click image for larger view and for slideshow)

Whether a tweet dies quietly or gets retweeted relentlessly depends more on the structure of the social network and the limited attention span of recipients than it does on the content or author of that Twitter post, according to an academic study.

In other words, success or failure at spreading a message through social media looks kind of random. When a social media campaign goes viral, "you don't have to assume it's because some people are influential--you can still get some random thing that gets 50 million views on YouTube," said Filippo Menczer, director of the Center for Complex Networks and Systems Research at the Indiana University School of Informatics and Computing. "Maybe the answer is that there is no answer."

Or, to put it another way, when computer scientists (or social media marketing gurus) try to prove cause-and-effect, they need to do a better job of correcting for the element of chance. "You have to ask, is it just a spurious correlation, where someone was going to win," Menczer said in an interview.

The paper on Competition among memes in a world with limited attention was published on Nature Magazine's Scientific Reports website. Indiana University Ph.D. candidate Lilian Weng was the lead author, with co-authors Menczer, Alessandro Flammini, director of undergraduate studies for Indiana University Informatics, and Alessandro Vespignani, Sternberg distinguished professor of physics, computer science, and health sciences at Northeastern University.

The term meme, coined by the British evolutionary biologist Richard Dawkins in his book The Selfish Gene, describes ideas that reproduce and spread like genes in biological systems. The Indiana University study also alludes to a related theory that ideas and concepts spread like infections, with fads and viral videos breaking out like epidemics. Another theoretical underpinning of the study was the idea of an attention economy in which information is plentiful but human capacity to focus on it is scarce.

The researchers built a computer model of a system with a network structure similar to that of Twitter, and a limited attention span for the software agents simulating the role of users (meaning they would only "remember" or focus on any given meme for a short time). By pumping simulated memes into this virtual Twitterverse and picking winners and losers according to a few simple rules, the computer scientists were able to reproduce the same kind of behaviors seen in the real network. The real-world analysis was based on 120 million retweets connected to 12.5 million users and 1.3 million hashtags.

"Our question was, can we look at these things to explain why some become very popular--why some YouTube videos go viral and others don't. There might be one dancing cat that gets a million views, but another dancing cat where only two people watch," Menczer said. Other studies have tried to make connections to the time of day an item is posted or the number of connections the person issuing the post has, but those may be misleading, he said.

Like an ecosystem that can only support a limited number of species and individuals within that species, social networks pick winners that go viral and losers that die without a trace according to a brutal process of natural selection, where chance plays a big role, according to the study.

The rules of the game can be summarized as "ideas propagate on the network, and people forget after a while," Menczer said. The researchers found that if they altered the assumptions in the model--for example, by altering the degrees of separation between social media users or their ability to pay attention--their model no longer matched reality. On the other hand, when they stuck to their base model--in which no post was inherently more interesting or popular than the other--the data matched up closely.

This does not really mean that there are no ideas more interesting or people more influential than others, only that the role of those factors is easy to overestimate, Menczer said.

"Of course, there are things that are more objectively interesting than others and if you write about them, probably yes, that will receive a bunch of following. However, even if you don't you might get lucky. And even if you do, someone else might be posting the same thing and get the attention instead of you," he said.

Follow David F. Carr on Twitter @davidfcarr. The BrainYard is @thebyard and

The Enterprise 2.0 Conference brings together industry thought leaders to explore the latest innovations in enterprise social software, analytics, and big data tools and technologies. Learn how your business can harness these tools to improve internal business processes and create operational efficiencies. It happens in Boston, June 18-21. Register today!

About the Author(s)

David Carr

Editor, InformationWeek Healthcare and InformationWeek Government (columnist on social business)

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights