The Dynamics of Information Diffusion in the Turkish Protests

by Henry Farrell on June 9, 2013 · 7 comments

in IT and politics,Protest

Below is a guest-post from Sandra González-Bailón at the Oxford Internet Institute and Pablo Barberá at NYU’s Social Media and Political Participation (SMaPP) laboratory.

Social networking sites, such as Twitter, Facebook or Tumblr, appear to be playing a prominent role in the coordination of the still ongoing protests in Turkey. There is abundant evidence suggesting that social media have been pivotal in the spread of information, especially in the absence of coverage by traditional media [1]; to recruit and mobilize protesters [2]; to coordinate the movement without the infrastructure of formal organizations [3]; and to draw the attention and support of the international community [4]. That social media is at the heart of these protests was defiantly acknowledged by the Turkish Prime Minister himself when he described them as “the worst menace to society” [5]. There are also reports that 25 people were arrested because of their use of Twitter to spread information about the protest [6].

The protests in Turkey add up to a long list of popular uprisings and massive demonstrations around the globe that took shape and gained momentum with the help of social media. However, there are still many open questions about how social networks facilitate the diffusion of information and whether some users play special roles in increasing the visibility of the protests. Results from a preliminary analysis of data collected at the SMaPP lab tracking protest activity in Twitter reveal patterns that are consistent with previous findings about protests in Spain and the US [7,8,9,10].

First, the distribution of centrality is very hierarchical: 1% of users concentrate about 80% of all retweets received; three quarters of users active in the protests do not receive any retweets at all. The implication of this asymmetry is that a minority of users act as the main sources of information: they are the ‘authorities’, the authors of the messages that resonate through a higher number of users. It also suggests that communication in online networks relies on a division of labor: those who generate valuable content (authorities), and those who facilitate its dissemination (the rest).

Second, these authorities tend to have large audiences – or a larger number of followers – but many of them are not particularly central in the Twitter network. Figure 1, which plots activity data for the first five days of the protests, illustrates this. Users that act as the ‘authorities’ in this stream of protest information are located above the dashed red line. Most of them have a large number of followers, as celebrities or public figures tend to have; but many of these ‘authorities’ (37%) have networks that are quite symmetrical and average in size, or who follow more users than follow them back. We call these users “hidden influentials” because they are not globally visible in the Twitter network, but they are very visible in this stream of protest-related information. They are at the heart of protest communication even though they are not at the heart of the followers network.

Figure 1: Distribution of users according to network position and message activity
(Replication of Figure 4 in González-Bailón et al, 2012)


At the same time, there are also a few differences that make the Taksim Square protests unique:

First, compared to the ‘indignados’ protests in Spain, this network is more hierarchical, with more extreme outliers falling in the category of influentials (they have significantly larger audiences, that is, the potential to directly reach many more people); in particular, the network has many more broadcasters (31% versus the 7% in the Spanish case). Although the majority of people sending messages are still common users, that is, not particularly visible or central (59% fall in this category vs 66% in the Spanish protests), their relative weight is lower, which suggests that the influence of prominent people and public figures, like journalists or celebrities, is greater in the Turkish protests.

[Note: this figure only takes into account activity generated during the first five days of the protests, whereas the analysis of the Spanish case tracked activity for the period of one month; this might affect the number of hidden influentials, the visibility of which accumulates over time – some users currently classified as common might upgrade as protests unfold].

Second, the size and composition of the network of retweets shows certain volatility. In this network, a user (A) is connected to another user (B) if A reposts a message previously published by B. There is a lot of variation on who is more central in this authority network over the first few days of the protest. While the protests started as a local, grass-roots mobilization opposing plans to remove Gezi Park, they soon escalated into anti-government demonstrations, and were quickly internationalized, with the “occupy” movement being particularly active after the third day of the protest: by the fourth day, more than 30% of unique users employing protest hashtags were English speakers; Twitter accounts like “YourAnonNews”, “AnonOpsLegion” or “AnonOpsMob”, part of Anonymous, a network of hacktivists, also started to appear among the most retweeted.

Figure 2: Network Size, Daily and Cumulative, and Distribution by Language


To conclude, the study of social media can shed interesting light into the dynamics of information diffusion in the organization of collective action. This is particularly the case when social media, as in the Turkish protests, supplies information that is suppressed by traditional media. Evidence suggests that 15K users sent at least one tweet from Gezi Park [10] which points at the spillover effects of online activity into offline action. More research, however, is needed to identify the mechanisms that drive the self-organization of tens of thousands of people in the form of massive protests, and the features that are generalizable and unique to each particular case.




Comments on this entry are closed.

Previous post:

Next post: