"How Meta’s platforms normalize anti-Palestinian racism", 16 November 2023

The reluctance or inability of social media companies generally, and Meta in particular, to safeguard their users implicates them, significantly contributing to the dehumanization and normalization of calls for violence against Palestinians as well as the amplification of anti-Palestinian racism.

Since Oct. 7, 2023, social media platforms have been flooded with disinformation and hateful rhetoric that have contributed to worsening the situation. Meta’s platforms, including Facebook, Instagram, and WhatsApp, have played a particularly consequential role in this because of their especially wide use in the region and the repeated mysterious “technical glitches” that have intensified their algorithmic bias against Palestinians and Palestine.

Algorithmic bias amplifies stereotyping

The impact of Meta’s algorithmic bias on content generation and dissemination has become especially clear.

Meta’s persistent record of failures

In May 2021, efforts by Israeli forces to displace Palestinian residents in some Jerusalem neighborhoods sparked a nearly two-week-long attack on Gaza... Palestinian users immediately found their views suppressed or accounts removed on various social media sites. Advocacy groups pushed Meta to engage a third party to assess how its platforms, including Facebook and Instagram, moderated Palestinian-Arabic and Israeli-Hebrew content. The Business for Social Responsibility (BSR) human rights due diligence report,... affirmed claims of bias against Palestinians because of Meta’s policies, which effectively infringe upon their freedom of speech, assembly, and political engagement, as well as freedom from discrimination.

One of the significant recommendations from the BSR report was the need for Meta to continue to work on implementing functioning and efficient Hebrew hostile speech classifiers. ... By September 2023, the tech company informed its Oversight Board that it had successfully achieved this goal. However, soon after Oct. 7, Meta internally admitted it had refrained from utilizing its Hebrew hostile speech classifier for Instagram comments due to insufficient data.

Consequently, over the past month, these long-running challenges have contributed significantly to transforming violent rhetoric into real-world harm.

Meta’s insufficient investment in combating incitement and racist speech has also permitted the spread of online hate speech far beyond Palestine.

But even as the company permits this type of inflammatory speech without proper content moderation, it actively suppresses Palestinian content that seeks to document human rights violations in Palestine... Additionally, shadow banning... has disproportionately affected Palestinians and their allies, including journalists, human rights defenders, and media organizations, since Oct. 7.

After identifying a notable rise in hateful Instagram comments originating from Israel, Lebanon, and the occupied Palestinian territory over the past month, Meta tightened its comment moderation filters as a temporary risk response. Specifically, for Israel, Lebanon, Syria, Egypt, and the occupied Palestinian territory, Meta lowered the threshold of certainty its moderation systems require to remove an inflammatory comment on an Instagram post... Subsequently, purportedly to prevent hostile speech, it further reduced this threshold but only for Palestine... resulting in significant stifling of Palestinian expression online.

Additionally, Meta’s limited comprehension of local social and political contexts along with inadequate investment in moderation improvements have contributed to systemic defaming of Palestinians on its platforms. For example, ...an Instagram user documented a particularly egregious translation error. On his profile, he identified himself as Palestinian and featured a Palestinian flag and the Arabic phrase “alhamdulillah,” meaning “praise be to God” in English. However, clicking “see translation” generated an English translation that disturbingly read, “Praise be to God, Palestinian terrorists are fighting for their freedom.” Not only do such incidents slander and unfairly stereotype Palestinians, but the company’s apologies fall short of rectifying the damage, particularly given the recurrent nature of these failures.

Furthermore, some users’ comments on Instagram featuring the Palestinian flag emoji are reportedly being concealed, with users indicating that these comments are labeled as “potentially offensive.”

The company has yet to announce any specific investments or improvements to address its persistent technical and internal process issues. It continues to primarily rely on its automated moderation systems and biased algorithms for content moderation as well as on user reports regarding specific posts rather than proactively safeguarding its global user base.