Online Library

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.

VOX-Pol Blog post.

This article analyses the online communication of the Estonian extreme right that appears to be characterized by an echo-chamber effect as well as enclosed and hermetic meaning-making. The discussion mainly relies on the theoretical frameworks offered by semiotics of culture.
One of the aims of the article is to widen the scope of understanding of autocommunicative processes that are usually related to learning, insight and innovation. The article shows the conditions in which autocommunicative processes result in closed interactions, based on reproducing stereotypes and redundant content. We detect antithetical meaning-making, an orientation towards normative (“correct”) texts and the prevalence of phatic communication as the main dominants that guide closed autocommunication. Such communication leads to polarization of dissimilar views and hinders dialogue. Our case study focuses on the discussion that arose in the context of the European Refugee Crisis that started in spring 2015.

This chapter illustrates that online and offline measures to counter violent extremism in Australia have targeted jihadism over other forms of extremism. With attention to the open and intensifying state of far-right extremism in Australia, it advocates for increased attention to this situation. It contends that, given the recursive nature of online extremism and its political and international dimensions, material disseminated by far-right and jihadist groups should be addressed from a comparative perspective.

Non-profits, as well as the media, have hypothesized the existence of a radicalization pipeline on YouTube, claiming that users systematically progress towards more extreme content on the platform. Yet, there is to date no substantial quantitative evidence of this alleged pipeline. To close this gap, we conduct a large-scale audit of user radicalization on YouTube. We analyze 330,925 videos posted on 349 channels, which we broadly classified into four types: Media, the Alt-lite, the Intellectual Dark Web (I.D.W.), and the Alt-right. According to the aforementioned radicalization hypothesis, channels in the I.D.W. and the Alt-lite serve as gateways to fringe far-right ideology, here represented by Alt-right channels. Processing 72M+ comments, we show that the three channel types indeed increasingly share the same user base; that users consistently migrate from milder to more extreme content; and that a large percentage of users who consume Alt-right content now consumed Alt-lite and I.D.W. content in the past. We also probe YouTube’s recommendation algorithm, looking at more than 2M video and channel recommendations between May/July 2019. We find that Alt-lite content is easily reachable from I.D.W. channels, while Alt-right videos are reachable only through channel recommendations. Overall, we paint a comprehensive picture of user radicalization on YouTube.

On October 27th, 2022, Elon Musk purchased Twitter, becoming its new CEO and firing many top executives in the process. Musk listed fewer restrictions on content moderation and removal of spam bots among his goals for the platform. Given findings of prior research on moderation and hate speech in online communities, the promise of less strict content moderation poses the concern that hate will rise on Twitter. We examine the levels of hate speech and prevalence of bots before and after Musk’s acquisition of the platform. We find that hate speech rose dramatically upon Musk purchasing Twitter and the prevalence of most types of bots increased, while the prevalence of astroturf bots decreased.