Online Library

There is an emerging consensus that ideologically-based narratives play a central role in encouraging and sustaining radicalization to violence, and that preventing, arresting, or reversing radicalization requires some means by which to address the effects of these narratives. Countering violent extremism (CVE) is a broad umbrella phrase that covers a wide array of approaches that have been advanced to reduce the radicalizing effects of extremist narratives. There is considerably less agreement, however, regarding the most appropriate means by which the mitigation of extremist narratives might best be accomplished. An important emerging area of interest is the role of the Internet, both as a forum through which narratives are transmitted and as an avenue for delivering CVE programs. At present, very little is known about which principles and practices should inform online CVE initiatives. This study attempts to establish a foundation and framework for these programs: first, by identifying the concepts and constructs which may be most relevant to countering violent extremism online, and second, by examining the available material from six online CVE programs in relation to these concepts. This examination suggests that these programs are lacking strong theoretical foundations and do not address important elements of radicalization, such as contextual factors or identity issues. It is important that future iterations of CVE programs consider not just the specific content of the narratives, but also take into account why these narratives have resonance for particular individuals.

VOX-Pol Blog post.

There is an emerging consensus that ideologically-based narratives play a central role in encouraging and sustaining radicalization to violence, and that preventing, arresting, or reversing radicalization requires some means by which to address the effects of these narratives. Countering violent extremism (CVE) is a broad umbrella phrase that covers a wide array of approaches that have been advanced to reduce the radicalizing effects of extremist narratives. There is considerably less agreement, however, regarding the most appropriate means by which the mitigation of extremist narratives might best be accomplished. An important emerging area of interest is the role of the Internet, both as a forum through which narratives are transmitted and as an avenue for delivering CVE programs. At present, very little is known about which principles and practices should inform online CVE initiatives. This study attempts to establish a foundation and framework for these programs: first, by identifying the concepts and constructs which may be most relevant to countering violent extremism online, and second, by examining the available material from six online CVE programs in relation to these concepts. This examination suggests that these programs are lacking strong theoretical foundations and do not address important elements of radicalization, such as contextual factors or identity issues. It is important that future iterations of CVE programs consider not just the specific content of the narratives, but also take into account why these narratives have resonance for particular individuals

A key challenge for automatic hate-speech detection on social media is the separation of hate speech from other instances of offensive language. Lexical detection methods tend to have low precision because they classify all messages containing particular terms as hate speech and previous work using supervised learning has failed to distinguish between the two categories. We used a crowd-sourced hate speech lexicon to collect tweets containing hate speech keywords. We use crowd-sourcing to label a sample of these tweets into three categories: those containing hate speech, only offensive language, and those with neither. We train a multi-class classifier to distinguish between these different categories. Close analysis of the predictions and the errors shows when we can
reliably separate hate speech from other offensive language and when this differentiation is more difficult. We find that racist and homophobic tweets are more likely to be classified as hate speech but that sexist tweets are generally classified as offensive. Tweets without explicit hate keywords are also more difficult to classify.

This report represents the interim findings of a two-year study designed to increase understanding of the social media footprint of right-wing extremism (RWE) in Canada. This work is part of a larger project designed to understand RWE in Canada led by Ontario Tech University (OTU), in partnership with Michigan State University and the University of New Brunswick. This team are currently working on a similar project designed to map offline RWE in Canada. The project follows a similar study delivered in 2015, enabling researchers and policymakers to understand how RWE has changed in the past five years.

This report outlines the results of a programme trialling
a methodology for identifying individuals who are
demonstrating signs of radicalisation on social media,
and engaging these individuals in direct, personalised
and private ‘counter-conversations’ for the purpose of
de-radicalisation from extremist ideology and
disengagement from extremist movements. This is
the first programme globally which has trialled the
delivery of online interventions in a systematised
and scaled fashion.