The Meta Oversight Board recently released a Policy Advisory Opinion addressing the moderation of the word "shaheed" on Meta's platforms, emphasizing the need for cultural and linguistic sensitivity in content moderation policies. The term "shaheed" carries various meanings and is widely used across different languages, including Arabic, Urdu, Hindi, Farsi, and Turkish.
The Opinion highlighted that the term "shaheed" is the most heavily moderated word on Meta's platforms, particularly in connection to Meta's Dangerous Organization and Individuals (DOI) list. The Board expressed concerns about the lack of transparency and clarity in Meta's moderation processes, leading to a broad restriction on the term without proper disclosure to Arabic-speaking users.
According to the Opinion, the ban on "shaheed" has resulted in a disproportionate limitation on freedom of expression within the Arabic-speaking community, fostering distrust and reinforcing biased perceptions. The Board made recommendations for Meta to revise its DOI policy, ensure transparency in enforcement, and update moderation guidelines to allow for the use of "shaheed" in appropriate contexts.
The Opinion aligns with the Santa Clara Principles, urging companies to consider cultural and linguistic contexts in content moderation. It emphasized the importance of engaging with language speakers and experts to develop more inclusive moderation practices.
Furthermore, the Opinion called for a reassessment of automated content moderation tools, particularly in languages other than English. It highlighted the inadequacies in current training data and the potential for bias in moderation decisions, especially in low-resource languages like certain Arabic dialects.
The Advisory Opinion serves as a crucial milestone in promoting a more nuanced approach to content moderation and underscores the necessity for tech companies to invest in understanding diverse linguistic and cultural contexts. It urges Meta and other industry players to prioritize inclusivity and transparency in shaping content moderation policies for a more equitable online environment.
