Moderating Quechua Content on Social Media

This CDT report examines how social media platforms moderate content in Quechua, an indigenous language of the Andes, and finds that Quechua-speaking users face markedly different — and often worse — outcomes than users posting the same content in Spanish, describing this as a form of linguistic inequity built into platform systems.

Experts interviewed for the study said large language models are not yet reliable for Quechua-language moderation given the scarcity of training data, and moderation teams were found to generally lack Quechua-speaking staff, relying instead on content policies essentially transplanted from English and Spanish contexts. This produced contradictory effects: some users experienced unexplained removals or shadowbanning of Quechua content, while others found their Quechua posts went entirely unmoderated and, in some cases, deliberately switched to Quechua specifically to avoid enforcement.

The report recommends platforms consult the Global South's existing community of NLP researchers and language-technology experts, arguing their domain knowledge could substantially improve moderation quality for Quechua and other under-resourced languages.