Elamur Research⌕
☰
home / papers / 2026 / The Surge of Anti-Semitism in German Social Media following the October 7 Attacks

The Surge of Anti-Semitism in German Social Media following the October 7 Attacks

Summary

The paper measures how the Hamas attacks of October 7, 2023, changed anti-Semitic talk in German social media. The authors gathered Judaism- and Israel-related posts from Facebook and Telegram (N = 125,718) using keywords, covering three months on either side of the attacks. They built a codebook grounded in the IHRA definition, with 26 sub-categories, diagnostic indicators, and positive and negative examples. A domain expert labelled about 800 posts, and this sample was used to compare open-weight LLMs and tune prompts. Each model was tested in two variants: post text alone, and post text plus information about the posting user. The best configuration reaches up to 83% F1 for the yes/no anti-Semitism decision. Adding user information helps most models and sharply reduces false alarms on posts that report or criticize anti-Semitism. Applied to the full corpus, the classifier shows a significant rise in anti-Semitism on both platforms after October 7. Telegram carries roughly ten times the prevalence of Facebook, and the two platforms' category profiles become more alike after the attacks. The study shows LLMs can scale expert-defined content analysis, though they still need human oversight.

Key points

  • Corpus: 58,411 Facebook posts from 622 users and 67,307 Telegram posts from 1,000 channels. Only 21.2% (Facebook) and 9.6% (Telegram) of posts predate October 7.
  • The taxonomy has 26 categories. They split into "classic" tropes, such as Jewish power (K07), evil (K02) and blood libel (K03), and Israel-related forms, such as boycott calls (K20), Apartheid analogies (K13) and genocide accusations.
  • Per-category results for gpt-oss 120b vary widely. Boycott calls (K20) reach F1 0.81 and K15 reaches 0.79. Rare categories such as K21 (0.07) and K06 (0.00) essentially fail.
  • Before the attacks, Facebook's anti-Semitism clustered around the alleged Gaza genocide narrative, while Telegram was dominated by power and conspiracy stereotypes.
  • After the attacks, Facebook saw classic tropes rise (e.g., K02, K04), and Telegram saw Israel-related categories surge, with genocide accusations growing most.

Interesting ideas

  • I1. Giving the classifier metadata about who wrote a post (name, handle) lets it separate voicing a view from reporting or condemning it, which cuts false positives. — For Mova (frames): Speaker and source attributes could feed speech-act and attribution analysis, helping decide whether a clause is asserted by the author or merely reported.
  • I2. The codebook pairs each category with diagnostic cues and with negative examples that mark where a neighbouring, non-matching case begins. — For Mova (learn): LLM teacher prompts could include explicit near-miss negatives for each label, so silver annotations respect boundaries between similar frames or roles.
  • I3. A compact expert-labelled set (~800 items) serves only to pick the model and tune the prompt, after which one setup is run over the whole corpus. — For Mova (learn): A small gold set could select which LLM teacher tier or witness parser to trust, before scaling silver data generation.
  • I4. Reporting precision, recall and F1 per category alongside the count of positive instances exposes rare labels that are unreliable despite a good aggregate score. — For Mova (learn): Evaluation reports could list per-relation or per-frame scores with support, so weak rare classes are flagged rather than hidden by averages.
  • I5. Comparing category shares rather than raw counts across two sources and two time periods reveals how a discourse's composition shifts and converges. — For Mova (world): The same proportion-based view could track how the mix of event types or states changes across text segments or corpora.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2609.36290 · arXiv 2609.36290

@misc{wiedemann2026surge,
  title = {{The Surge of Anti-Semitism in German Social Media following the October 7 Attacks}},
  author = {Wiedemann, Gregor and Wehrend, Daniel},
  year = {2026},
  doi = {10.48550/arxiv.2609.36290},
  eprint = {2609.36290},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2609.36290}
}

References

28 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. ADF. 2018. Quantifying Hate. A Year of Anti-Semitism on Twitter. Technical report, Anti-Defamation League, New York, NY, USA. find
  2. Vered Andre’ev and Lara Portnoy. 2024. Arabic and English Antisemitism on Social Media Platforms Post-October 7: Analysis on Discourse and Prevention. Technical report, The Institute for National Security Studies (INSS). find
  3. Matthias J. Becker, Laura Ascone, Matthew Bolton, Veronika Bundzíková, Alexis Chapelan, Pia Haupeltshofer, and Alexa Krugel. 2024. Decoding Antisemitism: An AI-driven Study on Hate Speech and Imagery Online. Discourse Report 6. find
  4. Matthias J. Becker, Laura Ascone, Matthew Bolton, Alexis Chapelan, Pia Haupeltshofer, Alexa Krugel, Karolina Placzynta, Marcus Scheiber, and Victor Tschiskale. 2023. Celebrating Terror: Antisemitism online after the Hamas attacks on Israel. Technical report, Technische Universität Berlin. find
  5. Matthias J. Becker, Laura Ascone, Matthew Bolton, Alexis Chapelan, Jan Krasni, Karolina Placzynta, Marcus Scheiber, Hagen Troschke, and Chloé Vincent. 2022. Decoding Antisemitism: An AI-driven Study on Hate Speech and Imagery Online. Discourse Report 4. Technical report, Technische Universität Berlin. find
  6. Matthias J. Becker and Hagen Troschke. 2023. Decoding Implicit Hate Speech: The example of antisemitism. In Christian Strippel, Sünje Paasch-Colberg, Martin Emmer, and Joachim Trebbe, editors, Challenges and Perspectives of Hate Speech Research. Freie Universität Berlin. ISSN: 2198-7610. find
  7. Maura Conway, Ryan Scrivens, and Logan Macnair. 2019. Right-wing extremists’ persistent online presence: History and contemporary trends. ICCT Policy Brief. find
  8. DeepSeek-AI, Daya Guo, Dejian Yang, and Haowei Zhang. 2025. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. arXiv preprint. ArXiv:2501.12948 [cs]. ↗ our card
  9. Team Gemma, Aishwarya Kamath, Johan Ferret, and Shreya Pathak. 2025. Gemma 3 Technical Report. arXiv preprint. ArXiv:2503.19786 [cs]. ↗
  10. Gemma 4 Team. 2026. Gemma 4 Model Card. find
  11. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 others. 2024. The Llama 3 Herd of Models. Preprint, arXiv:2407.21783. ↗
  12. Gunther Jikeli, Sameer Karali, Daniel Miehling, and Katharina Soemer. 2023. Antisemitic Messages? A Guide to High-Quality Annotation and a Labeled Dataset of Tweets. arXiv preprint. ArXiv:2304.14599 [cs]. ↗
  13. Günther Jikeli, Deepika Awasthi, David Axelrod, Daniel Miehling, Pauravi Wagh, and Weejoeng Joeng. 2021. Detecting Anti-Jewish Messages on Social Media. Building an Annotated Corpus That Can Serve as A Preliminary Gold Standard. In ICWSM Workshops. find
  14. Wolfgang Koch. 2022. Reichweiten von Social-Media-Plattformen und Messengern. Media Perspektiven, pages 471–478. find
  15. Jitendra Singh Malik, Hezhe Qiao, Guansong Pang, and Anton Van Den Hengel. 2025. Deep learning for hate speech detection: a comparative study. International Journal of Data Science and Analytics, 20(4):3053– 3068. find
  16. OpenAI, Sandhini Agarwal, Lama Ahmad, and Jason Ai. 2025. gpt-oss-120b & gpt-oss-20b Model Card. arXiv preprint. ArXiv:2508.10925 [cs]. ↗ our card
  17. Jay Patel, Hrudayangam Mehta, and Jeremy Blackburn. 2025. Evaluating Large Language Models for Detecting Antisemitism. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 35356–35385, Suzhou, China. Association for Computational Linguistics. find
  18. Qwen Team. 2026. Qwen3.5-Omni Technical Report. arXiv preprint. ArXiv:2604.15804 [cs.CL]. ↗
  19. Hanna Rose and Paula Matlach. 2024. Narratives of Hate: Post-7 October Antisemitism and Anti-Muslim Hate on Social Media. Technical report, Institute for Strategic Dialogue. find
  20. Rebecca Ruth Gould. 2020. The IHRA Definition of Antisemitism: Defining Antisemitism by Erasing Palestinians. The Political Quarterly, 91(4):825–831. find
  21. Joni O. Salminen, Sercan Şengün, Juan Corporan, Soon gyo Jung, and B. Jansen. 2020. Topic-driven toxicity: Exploring the relationship between online toxicity and news topics. PLoS ONE, 15. find
  22. Anna Schmidt and Michael Wiegand. 2017. A Survey on Hate Speech Detection using Natural Language Processing. In Proceedings of the Fifth International Workshop on Natural Language Processing for Social Media, pages 1–10, Valencia, Spain. Association for Computational Linguistics. find our card
  23. Elisabeth Steffen, Helena Mihaljevic, Milena Pustet, Nyco Bischoff, Maria Do Mar Castro Varela, Yener Bayramoglu, and Bahar Oghalai. 2023. Codes, Patterns and Shapes of Contemporary Online Antisemitism and Conspiracy Narratives – an Annotation Guide and Labeled German-Language Dataset in the Context of COVID-19. Proceedings of the International AAAI Conference on Web and Social Media, 17:1082–1092. find
  24. Florian Tippelt. 2025. Social Media zwischen Wachstum und Sättigung: Nutzungsmuster und Plattfor-mdynamiken in Deutschland. Media Perspektiven, (31/2025). find
  25. Gregor Wiedemann, Felix Victor Münch, Jan Philipp Rau, Phillip Kessling, and Jan-Hinrik Schmidt. 2023a. Concept and challenges of a social media observatory as a DIY research infrastructure. Publizistik, 68(2-3):201–223. find
  26. Gregor Wiedemann, Jan-Hinrik Schmidt, Jan Rau, Felix Victor Münch, and Philipp Kessling. 2023b. Telegram in der politischen Öffentlichkeit. Zur Einführung in das Themenheft. Medien & Kommunikationswissenschaft, 71(3-4):207–211. find
  27. Gregor Wiedemann, Seid Muhie Yimam, and Chris Biemann. 2020. UHH-LT at SemEval-2020 task 12: Fine-tuning of pre-trained transformer networks for offensive language detection. In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1638–1644, Barcelona (online). International Committee for Computational Linguistics. find
  28. An Yang, Anfeng Li, and Baosong Yang. 2025. Qwen3 Technical Report. arXiv preprint. ArXiv:2505.09388 [cs]. ↗ our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wiedemann-2026-surge-anti-semitism-german-social/