The Dynamics of Twitter Cascades

Matt Yglesias writing about how the now famous protest sign (below) went viral, argues that it’s not the quantity but the quality of your followers that matters.

herein lies one of the secrets of the faux-meritocracy of the internet. [Ben] Furnas may appear to be a mild-mannered law student with only 325 Twitter followers. But in an earlier life he was a key player in a ton of CAPAF’s policy products and he’s extremely well socially and professionally connected to the younger cohort of political media people. So that 325 includes reporters and editors from The Washington Post, Politico, Slate, Good, ThinkProgress, Mother Jones, and the Nation and think tank folks from CAP, Third Way, New America, and the Manhattan Institute. Given that particular nexus of people, it’s hardly a long and winding path to wide exposure for something interesting. This, I think, is an illustration of something important. People sometimes talk about the Internet as if it somehow supplants or replaces personal relationships. But in practice, it often acts as a force multiplier for them.

We actually have social science on Twitter cascades . Eytan Bakshy, Jake Hofman, Winter Mason and Duncan Watts test a model based on (a) the Twitter followers that a particular Twitter user has, and (b) the previous ‘success’ that the relevant user has had in influencing and others and finds that:

In fact, the model fit without averaging predicted and actual values at the leaf nodes is relatively poor (R2 = 0.34), reflecting that although large cascades tend to be driven by previously successful individuals with many followers, the extreme scarcity of such cascades means that most individuals with these attributes are not successful either. Thus, while large follower count and past success are likely necessary features for future success, they are far from sufficient. These results place the usual intuition about influencers in perspective: individuals who have been influential in the past and who have many followers are indeed more likely to be influential in the future; however, this intuition is correct only on average.

What this suggests is that better connected people (insofar as one can infer this through their past success) and people with more Twitter followers, are on average those who will start cascades. Nonetheless, very few posts actually do start cascades, and it is going to be extremely hard to predict in advance which posts will start cascades and which will not. Bakshy et al. furthermore argue that if you are a direct marketer or other person seeking to influence Twitter debate, it may be more cost effective to focus on the ‘non-influentials’ than the influentials. This also has implications for debates about the role of networks in international relations theory – which I hope to return to one of these days …

[Cross-posted at The Monkey Cage]

Henry Farrell

Henry Farrell is an associate professor of political science and international affairs at George Washington University.