Scientists have created an algorithm that can identify extremists in social networks before posting posts, according to the journal Operations Research . To do this, they analyzed five thousand accounts on Twitter.
Extremists associated with the “Islamic state” (the organization’s activities are banned in Russia), as well as other terrorist organizations, usually use social networks to propagate, intimidate the population and recruit new members. Companies such as Facebook and Twitter are already blockingsuspicious users using special algorithms, but this happens after posting posts. Now researchers have created an algorithm that will in the future allow deactivating extremist accounts immediately after they are created.
Jytte Klausen of the University of Brandeis in Massachusetts, along with colleagues from the Massachusetts Institute of Technology, collected the data of five thousand accounts on Twitter, which were either members of the IGSF or related people. Their names were obtained through the media, blogs, law enforcement agencies and think tanks. Researchers collected all the information about the users that was listed in the profile, and also received similar data for their friends and subscribers. Because of this, the number of profiles in the database has increased to 1.3 million. Also, scientists analyzed 4.8 million tweets related to extremist accounts: some of them belonged to users themselves, and some simply mentioned them or contained the corresponding hashtags.
Using statistical modeling of extremist behavior, results of post analysis and user data, the Clausen group has created a predictive model that can define accounts as extremist, even if they have not made a single entry in the microblog. This is due to the fact that the algorithm relies not on the description of posts, but on how the typical profile of the extremist looks as a whole. So, scientists found out that 70 percent of the accounts from the new data set belong to extremists, and the error was only 2 percent.
“Users who engage in some form of online extremism will have very similar behavioral characteristics in social networks. They will be associated with a specific set of users who form an extremist grouping. They will create new similar accounts after locking the old ones, and then return to the social network and with a high probability to restore links with the same persons, ” Clausen explains .
Sometimes scientists use a non-standard approach to finding extremist accounts and groups on the Internet. For example, a model was recently built that relies on the theory of gelling and describes the growth of extremist groups and showed with its help that extremist groups usually unite people with similar interests, and it is impossible to get rid of network extremism by destroying several “instigators”.