Linguistic Blind Spots in Platform Content Moderation in Ethiopia: The Case of Facebook

Language blind spots have hindered Facebook’s efforts to curb dangerous content, such as hate speech and calls for violence, during content moderation in Ethiopia. As a result, questions have emerged over whether – and to what extent – platforms such as Facebook support or undermine linguistic human rights (LHRs) in the country. This issue is especially pressing amid political violence and instability, as Facebook has applied mitigation measures unevenly, prioritizing English within a North American context while overlooking major Ethiopian languages. Approaches to remedying this gap must be delicate: while proactive takedown efforts are commendable in helping to stamp out dangerous content, inconsistent AI-based moderation of content without sufficient linguistic depth will trammel freedom of speech and LHRs in Ethiopia. This article seeks to examine the ways in which Facebook’s content moderation has floundered in the Ethiopian context due in part to language blind spots (and the lack of investment in methods to overcome these barriers), using three specific case studies from 2019–22. The article demonstrates that Facebook has shirked its due diligence duty in Ethiopia across all three case studies. As such, this article suggests that victims, civil societies, and human rights lawyers should consider suing Facebook (or Meta, Facebook’s parent company) before domestic and international tribunals as a result of the company’s complicity in fomenting atrocities in these three Ethiopian case studies.