In this project we seek to critically examine how content moderation systems operate in non-English contexts, particularly in indigenous and other languages of the Majority World (i.e., the Global South). This includes how social media platforms conduct analysis of user generated content, as well as processes for evaluation, enforcement, and appeals of content moderation decisions made by the platforms (Kamara et al., 2021). In addition, many platforms are using large language models (LLMs) and other automated tools to support their content moderation systems. While there is important research on content moderation systems (or trust and safety systems) as it applies to Western contexts and languages, there is much less research on languages in the Majority World, particularly those that are “low-resource” (or those with very little digitized textual data available for training and developing machine learning models) and indigenous languages (Shahid & Vashistha, 2023). Similarly, the application of LLMs for content analysis for low-resource languages introduces several potential challenges (Nicholas & Bhatia, 2023a), but there is little research on the implications of doing so in specific low-resource languages.
[…]
Source: Moderating Quechua Content on Social Media – Center for Democracy and Technology