The Oversight Board is urging Meta to change the way it moderates the word “shaheed,” an Arabic term that has led to more takedowns than any other word or phrase on the company’s platforms. Meta asked the group for help crafting new rules last year after attempts to revamp it internally stalled.
The Arabic word “shaheed” is often translated as “martyr,” though the board notes that this isn’t an exact definition and the word can have “multiple meanings.” But Meta’s current rules are based only on the “martyr” definition, which the company says implies praise. This has led to a “blanket ban” on the word when used in conjunction with people designated as “dangerous individuals” by the company.
However, this policy ignores the “linguistic complexity” of the word, which is “often used, even with reference to dangerous individuals, in reporting and neutral commentary, academic discussion, human rights debates and even more passive ways,” the Oversight Board says in its opinion. “There is strong reason to believe the multiple meanings of ‘shaheed’ result in the removal of a substantial amount of material not intended as praise of terrorists or their violent actions.”
In their recommendations to Meta, the Oversight Board says that the company should end its “blanket ban” on the word being used to reference “dangerous individuals,” and that posts should only be removed if there are other clear “signals of violence” or if the content breaks other policies. The board also wants Meta to better explain how it uses automated systems to enforce these rules.
If Meta adopts the Oversight Board’s recommendations, it could have a significant impact on the platform’s Arabic-speaking users. The board notes that the word, because it is so common, likely “accounts for more content removals under the Community Standards than any other single word or phrase,” across the company’s apps.
“Meta has been operating under the assumption that censorship can and will improve safety, but the evidence suggests that censorship can marginalize whole populations while not improving safety at all,” the board’s co-chair (and former Danish prime minister) Helle Thorning-Schmidt said in a statement. “The Board is especially concerned that Meta’s approach impacts journalism and civic discourse because media organizations and commentators might shy away from reporting on designated entities to avoid content removals.”
This is hardly the first time Meta has been criticized for moderation policies that disproportionately impact Arabic-speaking users. A 2022 report commissioned by the company found that Meta’s moderators were less accurate when assessing Palestinian Arabic, resulting in “false strikes” on users’ accounts. The company apologized last year after Instagram’s automated translations began inserting the word “terrorist” into the profiles of some Palestinian users.
The opinion is also yet another example of how long it can take for Meta’s Oversight Board to influence the social network’s policies. The company first asked the board to weigh in on the rules more than a year ago (the Oversight Board said it “paused” the publication of the policy after October 7 attacks in Israel to ensure its rules “held up” to the “extreme stress” of the conflict in Gaza). Meta will now have two months to respond to the recommendations, though actual changes to the company’s policies and practices could take several more weeks or months to implement.
“We want people to be able to use our platforms to share their views, and have a set of policies to help them do so safely,” a Meta spokesperson said in a statement. “We aim to apply these policies fairly but doing so at scale brings global challenges, which is why in February 2023 we sought the Oversight Board’s guidance on how we treat the word ‘shaheed’ when referring to designated individuals or organizations. We will review the Board’s feedback and respond within 60 days.”
This article originally appeared on Engadget at https://www.engadget.com/the-oversight-board-weighs-in-on-metas-most-moderated-word-100003625.html?src=rss