Memes, like Rickrolling or LOLcats, are the invasive species of social network ecosystems such as Facebook and Twitter. “Viral hashtags are so interesting that even at first sight, you just start to use them,” says Yong-Yeol Ahn, assistant professor at Indiana University’s School of Informatics and Computing. Ahn and his coauthors have isolated the network properties of memes and turned them into a forecasting tool, enabling the prediction of which Twitter hashtags will go viral nearly two out of three times based on how the hashtag is shared in its early stages. Ahn says later this spring they’ll be publishing follow-up research that looks at predicting just how big a splash a viral meme will make.
Circles represent communities—densely connected groups of users. Circle size represents the number of tweets with a given hashtag made per community. The color indicates when in the hashtag’s life span the most tweeting occurred in a community (darker is newer).
Ahn’s group considered more than 10 million different Twitter hashtags generated during March and April 2012. #ProperBand was a nonviral example, never spreading beyond one degree of separation from the original community of users that created it.
By contrast, #ThoughtsDuringSchool started in one community and, even in its early stages, rapidly spread to other communities, some distantly connected to the hashtag’s original source. This is a hallmark of virality, Ahn says. By the end of the rapid diffusion of #ThoughtsDuringSchool, it’s nearly impossible to discern which community originated the hashtag. But the cross-community behavior it displayed at its outset is still in evidence. Forecasting virality, then, is about identifying cross-community connectivity at its earliest stages and then betting that a given meme will keep behaving the same way, just on a larger scale.
This article originally appeared in print as “Forecasting Virality.”