Systemic analysis reveals how AI and human bias intersect to normalize antisemitism in digital spaces
Original framing: “Uncovering coded antisemitism online takes both human expertise and AI automation” — The Conversation - Global
Coded antisemitism has been a tool of statecraft for centuries, from medieval blood libels to 20th-century 'anti-Zionist' rhetoric in the USSR, where 'cosmopolitanism' served as a euphemism for Jewishness. The current AI-human detection model mirrors historical collaborations between state and private actors (e.g., IBM’s role in Nazi census data) to surveil and exclude Jewish populations. This continuity suggests hate speech detection is less about technology and more about institutionalizing exclusion under the guise of neutrality.
The mainstream narrative’s focus on AI-human collaboration to detect coded antisemitism obscures how these tools are embedded in a colonial, capitalist, and Ashkenazi-centric framework that reproduces historical patterns of exclusion.
Systems like those deployed by Facebook and Twitter are not neutral arbiters but extensions of surveillance capitalism, where 'hate speech' is defined by Western legal norms and enforced by algorithms trained on biased data. The erasure of Mizrahi, Sephardic, and Palestinian Jewish voices—who experience antisemitism differently—further entrenches these biases, while the framing ignores how platform governance itself fuels polarization. A decolonial approach would center marginalized epistemologies, mandate historical accountability, and redistribute power to those most affected, transforming detection from a tool of control into one of repair. The trickster’s insight—exposing the absurdity of techno-solutionism—reveals that the real work lies not in parsing language but in dismantling the systems that weaponize it.