Online Library

In this article, we conduct a comprehensive study of online antagonistic content related to Jewish identity posted on Twitter between October 2015 and October 2016 by UK-based users. We trained a scalable supervised machine learning classifier to identify antisemitic content to reveal patterns of online antisemitism perpetration at the source. We built statistical models to analyze the inhibiting and enabling factors of the size (number of retweets) and survival (duration of retweets) of information flows in addition to the production of online antagonistic content. Despite observing high temporal variability, we found that only a small proportion (0.7%) of the content was antagonistic. We also found that antagonistic content was less likely to disseminate in size or survive for a longer period. Information flows from antisemitic agents on Twitter gained less traction, while information flows emanating from capable and willing counter-speech actors—that is, Jewish organizations—had a significantly higher size and survival rates. This study is the first to demonstrate that Sampson’s classic sociological concept of collective efficacy can be observed on social media (SM). Our findings suggest that when organizations aiming to counter harmful narratives become active on SM platforms, their messages propagate further and achieve greater longevity than antagonistic messages. On SM, counter-speech posted by credible, capable and willing actors can be an effective measure to prevent harmful narratives. Based on our findings, we underline the value of the work by community organizations in reducing the propagation of cyberhate and increasing trust in SM platforms.

Judenfeindschaft als kulturelle Konstante und kollektiver Gefühlswert im digitalen Zeitalter (Kurzfassung) In welchen Manifestationen tritt Antisemitismus im 21. Jahrhundert in Erscheinung? Wie, wo und von wem werden judenfeindliche Inhalte artikuliert und verbreitet? Welche Stereotype werden kodiert, welche Argumente benutzt? Welche Rolle spielen Emotionen und irrationale Affektlogik beim aktuellen Einstellungs- und Verbalantisemitismus? Inwiefern hat das Internet die Verbreitung und Intensivierung von Antisemitismen akzeleriert und forciert? Wie lassen sich die modernen Ausprägungen des Judenhasses wissenschaftlich beschreiben, einordnen und erklären? Die vorliegende Langzeitstudie im Rahmen der empirischen Antisemitismusforschung zur Artikulation, Tradierung, Verbreitung und Manifestation von Judenhass im Internet hat sich mit diesen Fragen beschäftigt und quantitativ umfangreiche sowie inhaltlich detaillierte Untersuchungen vorgenommen.

This article compares two universes of official Islamic State media that were compiled 18 months apart. It explores the nuances of the group’s worldview and illustrates the extent to which external and internal situational exigencies impacted the Islamic State’s brand during its formative years as caliphate. It finds that the organization’s media infrastructure was about half as productive in early 2017 as it had been in mid-2015. The data also show that, even though the group had internationalized its theater of terrorist operations during the time period in question, the brand itself actually contracted to become markedly less globalized in 2016. Finally, the data indicate a substantial thematic rearrangement in the organization’s propaganda, one that saw its story shifting away from the millenarian “utopia” towards military denialism. In sum, the data indicate that the Islamic State propagandists were far less productive by January 2017, and that their aggregate product was less international and less utopian but more militant and more defiant, a shift that suggested a new phase in their political marketing operations, one focused on framing the caliphate as an embattled but still defiant pseudo-state struggling to maintain past momentum.

Moderators of online discussion forums often struggle with controlling extremist comments on their platforms. To help provide an efficient and accurate tool to detect online toxicity, we apply word2vec’s Skip-Gram embedding vectors, Recurrent Neural Network models like Bidirectional Long Short-term Memory to tackle a toxic comment classification problem with a labeled dataset from Wikipedia Talk Page. We explore different pre-trained embedding vectors from larger corpora. We also assess the class imbalance issues associated with the dataset by employing sampling techniques and penalizing loss. Models we applied yield high overall accuracy with relatively low cost.

This paper was presented at the 2nd European Counter-Terrorism Centre (ECTC) Advisory Group conference, 17-18 April 2018, at Europol Headquarters, The Hague.

The growing presence of modern terrorism on the Internet is at the nexus of two key trends: the democratization of communications driven by user-generated content on the Internet; and the growing awareness of modern terrorists of the potential of the Internet for their purposes. How best can the terrorists’ use and abuse of the Internet be countered? As this article argues, the answer to violent radicalization on the Internet lies not in censorship of the Internet, but in a more sophisticated and complicated strategy, relying on the theoretical notion of “noise” in communication process theory.