Detecting Linguistic Markers for Radical Violence in Social Media

This article focuses on the use of linguistic “weak signals”—digital traces of intent—in social media as a tool of counterterrorism aimed at preventing lone-wolf attacks. The authors argue that, while a fully automatic tool for detecting lone-wolf terrorism is not possible, semi-automatic tools that work by tracing these “weak signals” can be of great value to intelligence agencies seeking to protect against attacks. The authors discuss the use of social media to detect three types of indicators that signal an accelerated risk of attack: leakage, fixation and identification. For the authors: leakage is the communication to a third party of intent to do harm to a target; fixation is any behaviour that indicates an increasingly pathological preoccupation with a person or cause; identification is behaviour indicating a desire to be a “pseudo-commando”. The authors differentiate between identification with radical action (such as “warrior mentality”), identification with a role model who incites them to violence, or group identification where the individual begins to identify with broader cause leading to violence.

This article will be particularly useful for practitioners and researchers looking to utilize social media as a source of predictive and preventative intelligence gathering. The authors develop a framework and associated process that includes the application of rights-based and ethical-traditions. The authors further identify a number of techniques for collecting and analyzing relevant data: 1). Translation Services: the advantage to which is the speed with which large amounts of data can be processed and accessed by the analyst; 2). Sentiment Analysis: teaching computer algorithms to learn the difference between radical and non-radical texts and to identify threats targeted to specific individuals or groups; 3). Mapping Websites: Text analysis can be used for automatic discovery of problematic websites and for creating networks of these sites, computer algorithms can also be used to identify users that express themselves in a radical manner; 4). Author Recognition: In cases where aliases are used, the further development of algorithms used to extract various features from text, to help identify the author of the text.