On Countering Online Violent Extremism Using Social Network Analysis
In an era where the relative anonymousness of online activity encourages bad behavior, weeding out truly bad actors willing to manifest violent behavior from those that are just spewing rhetoric proves a difficult task. To this end, analyzing social network structure to assess the generation and propagation of violent ideologies is increasingly becoming a necessary tool in the government’s arsenal for countering violent extremism (CVE). In a study published on March 2013 for the International Center for the Study of Radicalization and Political Violence, J.M Berger and Bill Strathearn discuss methods of linking players in a radical network and assigning those links scores based on the probability that the people (nodes) the link connects will act out. As the study points out, violent extremism can cover a wide range of ideologies from jihadism and race related nationalism to eco-terrorism.
The study set out by first examining the network structure of 12 twitter accounts held by known individuals and organizations that openly identified as white nationalists. Public information including tweets from the followers of the 12 “seed” accounts (neighboring nodes to the “seed” accounts) were collected for the time period spanning three months. The study defined three main keywords – Influence, Exposure and Interactivity – and gave them scores respectively. Influence, according to the study, was given a score depending on how often one initiated and elicited a response from others while Exposure was graded on how often one was engaged in an interaction by the actions of others. The Interactivity score was a measure of the combined effects of both Influence and Exposure. So, a highly active account that draws out other accounts with ideological tweets or opinions would have a high Influence score while those that are drawn out into interaction with several twitter accounts will have a high Exposure score. An account can have high and low scores in one or both of these variables. In such a way one is able to use computers to quickly and easily map out the structure of potential radical networks and assign a strength/weakness measure to each edge.
Various things factored into assigning scores to these edges. In this case, keywords within the tweet relating to their ideology – white supremacy – was not factored in so that the study can have validity predicting behavior in cases of other forms of extremism. References to shifting communication from tweeter to other methods like calling, IM or text messaging between accounts was weighted highly due to the hidden nature of such a communication. From the top 100 accounts with high Influence scores, 86 were found to be overtly and intensely engaged in tweets of white nationalism. From the top 100 accounts with high Exposure scores, 93 were overtly engaged with that ideological sentiment. And through manual examination, it was seen that from the top 100 accounts with high interactivity scores, 95 were overly engaged. A similar study on extreme anarchism showed encouraging results on using strong and weak ties in social networks to assess threats.
Despite the incipient and highly classified nature of investigating network structures for countering violent extremism, this study gives us a glimpse into the inner workings of some such methods employed by agencies across the country. Network analysis cannot be the only tool we use for violence prevention, but it’s proving to be a vital one.
Works Cited
Berger, J., & Strathearn, B. (2013). Who Matters Online: Measuring influence, evaluating content and countering violent extremism in online social networks. Developments in Radicalisation and Political Violence.
Retrieved from http://icsr.info/wp-content/uploads/2013/03/ICSR_Berger-and-Strathearn.pdf