ERIHAD.OS - OS Ethical Research Infrastructure for Hateful, Antidemocratic, and Discriminatory Data
Science cluster
Summary
The Open Science Ethical Research Infrastructure for Hateful, Antidemocratic, and Discriminatory Data (ERIHAD.OS) addresses a central dilemma within Open Science: researchers require access to data containing hateful, antidemocratic, and discriminatory content in order to understand and counter these phenomena — yet openly sharing such material is unethical and, in many cases, legally prohibited.
The project develops a secure, federated research infrastructure (RI), integrated within the EOSC ecosystem, that enables ethical discovery, controlled access, and analysis of sensitive datasets across repositories. ERIHAD.OS brings together academic archives, NGO-held datasets, monitoring initiatives, and Trusted Flaggers to support interdisciplinary research on racism, xenophobia, antisemitism, anti-Muslim hatred, antigypsyism, science skepticism, and other radicalising and socially divisive narratives.
In doing so, ERIHAD.OS inverts the usual Open Science logic: rather than making existing data more open, it makes currently inaccessible data researchable for the first time.
TERO PC, Thessaloniki — TERO, Instituto de Creatividad e Innovaciones Educativas, Universidad de Valencia — IUCIE
Challenge
Hate and discriminatory content across social media, online forums, and citizen-reporting channels poses a growing threat to social cohesion. Yet this material cannot be openly shared: GDPR, national anti-hate regulations, and ethical constraints confine it to institutional silos. Rich datasets held by universities, observatories, NGOs, and monitoring agencies therefore remain inaccessible for cross-sector research, even to researchers with legitimate scholarly need.
The emerging Common European Data Spaces framework — including the Research and Innovation Data Space — provides the policy foundation ERIHAD.OS is designed to complement, demonstrating at proof-of-concept scale what governed access to sensitive research data requires.
Solution
ERIHAD.OS builds a federated platform that integrates academic repositories, social media hate research datasets, and non-public datasets from Trusted Flaggers and monitoring bodies into a governed network within EOSC. Access is managed through a three-tier system: full anonymised data for vetted researchers, snippet-level access for exploratory queries, and tool-based access to aggregated analytical results. This ensures GDPR and anti-hate law compliance across all user levels.
Automated anonymisation, leveraging NLP tools such as spaCy via CLARIN and metadata hashing, preserves research utility while protecting privacy. The platform will enable secure analysis of otherwise inaccessible datasets despite their strict data-sharing restrictions. Registered researchers will be able to draw on tools from SSHOC (topic modelling, sentiment analysis), CESSDA (data discovery, metadata management), and EOSC (computational infrastructure) to trace hate patterns and develop evidence-based counter-strategies.
Scientific Impact
ERIHAD.OS enables a kind of research that is currently not possible at scale: the systematic, cross-border, cross-linguistic tracking of hateful narratives as they mutate and travel across languages, platforms, and media formats. Consider a concrete example: tracing how a specific hateful (antisemitic, racist, or otherwise discriminatory) trope spreads from German-language text forums to Spanish-language platforms, and mutates from written argument into image-based memes or short video content along the way, currently requires manual, resource-intensive work across disconnected archives. ERIHAD.OS will support automated cross-dataset querying across text, image, and multimodal sources that makes this comparison routine. Historians can trace a narrative's origins using DARIAH's text annotation tools, linguists can map its diffusion via CLARIN's NLP pipelines, political scientists can correlate it with state agencies’ monitoring statistics, and visual culture researchers can track its transformation into antisemitic imagery - all within a single virtual research environment.
By connecting fragmented datasets from academic repositories, NGOs, observatories, and media monitoring agencies, the project enables unprecedented interdisciplinary research on the dynamics of hate, discrimination, and antidemocratic narratives. The platform will support evidence-based policymaking, educational initiatives, de-radicalisation strategies, and public communication campaigns. It also extends to fields where similar data constraints apply (including public health research on vaccine hesitancy and environmental science on climate denialism) demonstrating that the governance model ERIHAD.OS develops is transferable well beyond hate research.
Beyond its immediate research outputs, ERIHAD.OS will produce a concrete and reusable governance contribution of this concept: the technical, legal, and ethical documentation of what it takes to make legally constrained sensitive data interoperable within EOSC. This is infrastructure knowledge applicable to any research domain where data exists but cannot be openly shared, and it will be deposited openly so that other communities can build on it.
Principal investigator
Prof. Dr. Uffa Jensen is Deputy Director of the Center for Research on Antisemitism (ZfA), TU Berlin. A historian of modern antisemitism and European Jewish history, he previously held a DFG Heisenberg Professorship (2017–2022). He leads the Arthur Langerman Archive for the Study of Visual Antisemitism (ALAVA) and promotes the integration of digital and computational methods into the study of hate and discrimination.
Dr. Jan Krasni is a media scholar specializing in discourse analysis, digital humanities, and AI-based research on online discrimination. He previously led the German team in the international project Decoding Antisemitism. As creator and coordinator of ERIHAD.OS, he develops its scientific framework, Open Science architecture, and cross-sector stakeholder network.
