Abstract: The emergence of interactive online spaces and the evolution of internet-based communication have dramatically changed the way the individual relates to the world and interacts with other web users. The specificities of online communication such as anonymity and mutual reinforcement of web users have led to an increase and normalisation of hate speech (Troschke and Becker 2019. “Antisemitismus im Internet. Erscheinungsformen, Spezifika, Bekämpfung.” In Das neue Unbehagen. Antisemitismus in Deutschland und Europa heute, edited by Günther Jikeli and Olaf Glöckner, 151–72. Glöckner Hildesheim: Olms; Becker and Troschke 2023. “Decoding Implicit Hate Speech: The example of antisemitism.” In Challenges and perspectives of hate speech analysis: An interdisciplinary anthology, edited by Christian Strippel, Sünje Paasch-Colberg, Martin Emmer and Joachim Trebbe. Berlin: Digital Communication Research). This paper presents the results of our qualitative analysis of antisemitic content on Facebook profiles of British, French and German mainstream media, generated in the framework of the Decoding Antisemitism research project. The online debates of interest were identified in the context of discourse events – real-world events that have the potential to trigger antisemitic reactions – such as the Russian invasion of Ukraine, escalation phases in the Middle East conflict, including the events of October 2023, or scandals and instances of hate crime in Europe and beyond. The results of our analyses point to several commonalities in the three language communities in how Israel is conceptualised and evaluated through stereotypes in these comment sections. On the other hand, there are also consistent differences between the three corpora in the choice of stereotypes. Another significant difference concerns the verbal immediacy and frequency with which these mental concepts are communicated in online debates. This article will attempt to map the qualitative and quantitative patterns, compare and contrast the analyses for the three language communities and at the same time put forward for discussion possible socio-historical and -political reasons for this discursive behaviour (cf. Ascone et al. 2022. Decoding Antisemitism: An AI-driven Study on Hate Speech and Imagery Online. Discourse Report 4. Berlin: Technische Universität Berlin. Centre for Research on Antisemitism).
Abstract: The proliferation of hateful and violent speech in online media underscores the need for technological support to combat such discourse, create safer and more inclusive online environments, support content moderation and study political-discourse dynamics online. Automated detection of antisemitic content has been little explored compared to other forms of hate-speech. This chapter examines the automated detection of antisemitic speech in online and social media using a corpus of online comments sourced from various online and social media platforms. The corpus spans a three-year period and encompasses diverse discourse events that were deemed likely to provoke antisemitic reactions. We adopt two approaches. First, we explore the efficacy of Perspective API, a popular content- moderation tool that rates texts in terms of, e.g., toxicity or identity-related attacks, in scoring antisemitic content as toxic. We find that the tool rates a high proportion of antisemitic texts with very low toxicity scores, indicating a potential blind spot for such content. Additionally, Perspective API demonstrates a keyword bias towards words related to Jewish identities, which could result in texts being falsely flagged and removed from platforms. Second, we fine-tune deep learning models to detect antisemitic texts. We show that OpenAI’s GPT-3.5 can be fine-tuned to effectively detect antisemitic speech in our corpus and beyond, with F1 scores above 0.7. We discuss current achievements in this area and point out directions for future work, such as the utilisation of prompt-based models.
Abstract: Developments in Artificial Intelligence (AI) are prompting governments across the globe, and experts from across multiple sectors, to future proof society. In the UK, Ministers have published a discussion paper on the capabilities, opportunities and risks presented by frontier artificial intelligence. The document outlines that whilst AI has many benefits, it can act as a simple, accessible and cheap tool for the dissemination of disinformation, and could be misused by terrorists to enhance their capabilities. The document warns that AI technology will become so advanced and realistic, that it will be nearly impossible to distinguish deep fakes and other fake content from real content. AI could also be used to incite violence and reduce people’s trust in true information.
It is clear that mitigating risks from AI will become the next great challenge for governments, and for society.
Of all the possible risks, the Antisemitism Policy Trust is focused on the development of systems that facilitate
the promotion, amplification and sophistication of discriminatory and racist content, that is material
that can incite hatred of and harm to Jewish people.
This briefing explores how AI can be used to spread antisemitism. It also shows that AI can offer benefits
in combating antisemitism online and discusses ways to mitigate the risks of AI in relation to anti-Jewish
racism. We set out our recommendations for action, including the development of system risk assessments,
transparency and penalties for any failure to act.
Topics: October 7 2023 attacks + aftermath, Antisemitism, Antisemitism: Far right, Holocaust Denial, AI, Discourse and Discourse Analysis, Antisemitism: Discourse, Social Media, Internet, Main Topic: Antisemitism, Holocaust: Distortion
Abstract: Reflecting on the months since the recent October 7 attack, rarely has the theme of Holocaust Memorial Day 2024, ‘The Fragility of Freedom’, felt so poignant. Communities globally experienced the shattering of presumed security, and antisemitic incidents responsively spiked.
Antisemitism rose across both mainstream and fringe social media platforms, and communities resultantly reported a rise in insecurity and fear. CCOA constituent countries have recorded significant rises in antisemitic incidents, including an immediate 240% increase in Germany, a three-fold rise in France, and a marked increase in Italy.
The antisemitism landscape, including Holocaust denial and distortion, had shifted so drastically since October 7 that previous assumptions and understands now demand re-examination. In the run up to Holocaust Memorial Day 2024, this research compilation by members of the Coalition to Counter Online Antisemitism offers a vital contemporary examination of the current and emergent issues facing Holocaust denial and distortion online. As unique forms of antisemitism, denial and distortion are a tool of historical revisionism which specifically targets Jews, eroding Jewish experience and threatening democracy.
Across different geographies and knowledge fields, this compilation unites experts around the central and sustained proliferation of Holocaust denial and distortion on social media.
Topics: Antisemitism, Antisemitism: Discourse, Antisemitism: Monitoring, Internet, Social Media, Main Topic: Antisemitism, Attitudes to Israel, Israeli-Palestinian Conflict, Methodology, October 7 2023 attacks + aftermath, Hate and Hate Speech, AI
Abstract: Over the past 3.5 years, the Decoding Antisemitism research project has been analysing antisemitism on the internet in terms of content, structure, and frequency. Over this time, there has been no shortage of flashpoints which have generated antisemitic responses. Yet the online response to the Hamas attacks of 7 October and the subsequent Israeli operations in Gaza has surpassed anything the project has witnessed before. In no preceding escalation phase of the Arab-Israeli conflict has the predominant antisemitic reaction been one of open jubilation and joy over the deaths of Israeli Jews. As demonstrated in the sixth and final Discourse Report, this explicit approval of the Hamas attacks was the primary response from web users. The response to 7 October therefore represents a turning point in antisemitic online discourse, and its repercussions will be felt long into the future.
The report contains analysis of the various stages of online reactions to events in the Middle East, from the immediate aftermath to the Israeli retaliations and subsequent accusations of genocide against Israel. As well as examining online reactions in the project’s core focus—the United Kingdom, France, and Germany—this report also, for the first time, extends its view to analyse Israel-related web discourses in six further countries, including those in Southern and Eastern Europe as well as in North Africa. Alongside reactions to the escalation phase, the report also examines online responses to billionaire Elon Musk’s explosive comments about Jewish individuals and institutions.
Additionally, the report provides a retrospective overview of the project’s development over the past 3.5 years, tracking its successes and challenges, particularly regarding the conditions for successful interdisciplinary work and the ability of machine learning to capture the versatility and complexity of authentic web communication.
To mark the publication of the report, we are also sharing our new, interactive data visualisations tool, which lets you examine any two discourse events analysed by our research team between 2021 and 2023. You can compare the frequencies and co-occurrences of antisemitic concepts and speech acts by type and by country, look at frequencies of keywords in antisemitic comments, and plot keyword networks.
Topics: Internet, Social Media, Main Topic: Antisemitism, Attitudes to Israel, Israeli-Palestinian Conflict, Football, Hate and Hate Speech, AI, Antisemitism: Online, Conspiracy Theories, Antisemitism: Discourse and Tropes
Topics: Antisemitism, Antisemitism: Discourse, Antisemitism: Monitoring, Internet, Social Media, Main Topic: Antisemitism, War, Terrorism, Attitudes to Israel, Israeli-Palestinian Conflict, Boycott Divestment and Sanctions (BDS), Hate and Hate Speech, AI
Abstract: This article introduces the pilot project “Decoding Antisemitism: An AI-driven Study on Hate Speech and Imagery Online.” The aim of the project is to analyse the frequency, content and linguistic structure of online antisemitism, with the eventual aim of developing AI machine learning that is capable of recognizing explicit and implicit forms of antisemitic hate speech. The initial focus is on comments found on the websites and social media platforms of major media outlets in the United Kingdom, Germany, and France. The article outlines the project’s multi-step methodological design, which seeks to capture the complexity, diversity and continual development of antisemitism online. The first step is qualitative content analysis. Rather than relying on surveys, here a pre-existing “real-world” data set-namely, threads of online comments responding to media stories judged to be potential triggers for antisemitic speech-is collected and analysed for antisemitic content and linguistic structure by expert coders. The second step is supervised machine learning. Here, models are trained to mimic the decisions of human coders and learn how antisemitic stereotypes are currently reproduced in different web milieus-including implicit forms. The third step is large-scale quantitative analyses in which frequencies and combinations of words and phrases are measured, allowing the exploration of trends from millions of pieces of data.
Abstract: The 3-year pilot project presented here aims at analyzing antisemitic hate speech and imagery on mainstream news websites and social media platforms in different European contexts. Current forms of antisemitism will be examined in various ways by three international research teams from Germany, France, and the UK.
First, the datasets will be studied in detail (qualitative analysis based on pragmalinguistic, image analytical and historical approaches), taking into account explicit as well as implicit forms of communication (TU Berlin).
The resulting annotated datasets will provide training, validation, and test data for supervised machine learning techniques (King’s College London).
Eventually, all studied phenomena will be measured over time through statistical/quantitative analysis (TU Berlin and King’s College London).
The project stands in contrast to previous quantitative research on antisemitism online due to a) its awareness of verbal and visual complexity in the respective cultural and situational contexts, and b) its detailed, multimodal approach. Thus, it will provide the most accurate picture yet of the full extent of Jew-hatred on the interactive web.
The focus of the pilot project will be on German, English and French websites and their respective social media platforms. After the initial three year period, the focus will broaden out to investigate other European language communities.
The project will make a major contribution to the study of viral hate in different cultural contexts. Moreover, the researchers will engage in an ongoing dialogue not only with academia, but also with political, media and pedagogical institutions. An additional output will be an open source tool that will help to identify the full extent of antisemitism in various web milieus.
The half-yearly discourse reports share central insights of the ongoing research outcomes of the project "Decoding Antisemitism" and review unfolding trends.
The second discourse report presents the definitional basis of our analyses and for the first time provides comprehensive insights into our corpus analyses relating to Great Britain, France and Germany.
Abstract: The 3-year pilot project presented here aims at analyzing antisemitic hate speech and imagery on mainstream news websites and social media platforms in different European contexts. Current forms of antisemitism will be examined in various ways by three international research teams from Germany, France, and the UK.
First, the datasets will be studied in detail (qualitative analysis based on pragmalinguistic, image analytical and historical approaches), taking into account explicit as well as implicit forms of communication (TU Berlin).
The resulting annotated datasets will provide training, validation, and test data for supervised machine learning techniques (King’s College London).
Eventually, all studied phenomena will be measured over time through statistical/quantitative analysis (TU Berlin and King’s College London).
The project stands in contrast to previous quantitative research on antisemitism online due to a) its awareness of verbal and visual complexity in the respective cultural and situational contexts, and b) its detailed, multimodal approach. Thus, it will provide the most accurate picture yet of the full extent of Jew-hatred on the interactive web.
The focus of the pilot project will be on German, English and French websites and their respective social media platforms. After the initial three year period, the focus will broaden out to investigate other European language communities.
The project will make a major contribution to the study of viral hate in different cultural contexts. Moreover, the researchers will engage in an ongoing dialogue not only with academia, but also with political, media and pedagogical institutions. An additional output will be an open source tool that will help to identify the full extent of antisemitism in various web milieus.
The half-yearly discourse reports share central insights of the ongoing research outcomes of the project "Decoding Antisemitism" and review unfolding trends.
The first discourse report provides insight into the methodological approaches and the nature of antisemitic hate speech in selected discourse spaces.
