There’s a lot of information cycling around the Internet. Some things are only shared a few times, while others get shared millions of times. While going viral still seems like striking lightning, researchers at Cornell University may have cracked the code.
The researchers started with the basic premise of a cascade: When content is shared and reshared on social media, it cascades out to larger groups of people. When something goes viral, the cascade is wider, faster and reaches more people. And while the prevailing wisdom cautions that virality is unpredictable, according to the Cornell research [PDF], there are ways to reliably predict the cascade growth.
In order to prove this hypothesis, the researchers started with three basic questions:
- How accurately can cascade growth be predicted?
- Can large and small cascades be predicted?
- Can the shape and structure of the cascade be tracked?
The researchers observed the cascade patterns of images shared on Facebook during June 2013. While they acknowledge the limitations of the study, they were able to identify certain features that aid in the prediction of cascade growth: the content type, the original sharer, those who reshare, as well as structural and temporal features.
What they discovered was that it’s easier to accurately predict larger cascades than it is to predict smaller cascades: “It is easier to predict whether a cascade that has reached 25 reshares will get another 25, than to predict whether a cascade that has reached five reshares will obtain an additional five,” said the study.
Predicting the structure of the cascade proved to be more challenging, since the origin of the post impacted the structure. The research indicates some connectivity between different cascades, which means people are exposed to the same images via different cascades, and that earlier uploads usually generate the biggest cascades.