Why Cultural Context Matters for Detecting Disinformation with AI

Dresden Research Insights · Generative AI and the New Disinformation Lifecycle · 22 September 2026

A disinformation narrative can be perfectly intelligible to its intended audience and almost invisible to an automated system trained on another language or cultural context. That is the problem the kuKI project set out to address.

The Dresden presentation described a multilingual research infrastructure covering German, Turkish and Russian information environments. The goal is not simply to classify texts in three languages. It is to understand how narratives depend on cultural references, community-specific communication and the way people encounter information.

The project combines interviews with relevant communities, cross-platform monitoring, AI-assisted classification and a knowledge graph. It also connects the technical pipeline to participatory fact-checking through CORRECTIV’s Faktenforum. This creates an opportunity to bring human interpretation back into automated analysis.

The researchers discussed extending monitoring across platforms such as YouTube, Telegram and TikTok, while acknowledging that platform data access is a practical constraint. The project also aims to develop educational material, policy recommendations and opportunities for exchange across related research initiatives.

This approach exposes a broader weakness in many AI benchmarks: performance measured on convenient datasets may tell us little about culturally embedded manipulation.

Research question: How should we evaluate an AI disinformation detector when the crucial context is cultural rather than merely linguistic?

Watch the talk: YouTube recording. Explore the complete workshop archive and Dresden programme.

This editorial research insight summarizes a workshop presentation. It is not a verbatim transcript or a peer-reviewed assessment of the presented claims.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *