Mekelle፡Telaviv, Nairobi, Pretoria, London, (Tigray Herald)
Social Media Platforms Failed to Moderate Genocidal Content During Tigray War, Study Reveals
A new study by the Distributed AI Research Institute (DAIR) has found that social media platforms failed to adequately moderate genocidal content during the 2020–2022 war in Ethiopia’s Tigray region, allowing hate speech, incitement, and atrocity denial to flourish online.
The report, co-authored by journalists, activists, former content moderators, and DAIR founder Timnit Gebru, highlights major lapses in content moderation, particularly on platforms like Facebook. Internal documents leaked by whistleblower Frances Haugen show that despite Ethiopia’s population of 128 million people speaking over 100 languages, Facebook supported integrity systems in only two.
The study emphasized that superficial language moderation was not enough. “Platforms prioritized dominant languages but failed to engage with dialects, cultural nuances, and regional slang, which are crucial for identifying harmful content,” the authors wrote.
Researchers analyzed 340 posts in Amharic, Tigrinya, Arabic, and English over four months. They flagged posts like #FakeAxumMassacre under “Violent Event Denial,” pointing out that despite extensive survivor testimonies and corroborations by @amnesty International, @hrw Rights Watch, and Associated Press, such posts were allowed to circulate unchecked.
Another example included a post that translated to “Clean the cockroaches,” which was missed by many annotators. The study clarified that recognizing its genocidal implication required an understanding of how the Ethiopian government and its allies often referred to Tigrayans using dehumanizing terms.
The findings raise serious concerns about the role of tech platforms in conflicts and call for urgent reforms in multilingual and culturally aware content moderation practices.
