Tech platforms’ treatment of pro-Palestinian content raises bias allegations
A visitor to Arabic-speaking corners of Facebook or Instagram in recent days might have encountered a strange anachronism: thousand-year-old Arabic script salvaged from the past in order to trick modern algorithms.
Over the last week, as Israeli jets have rained missiles and bombs on the Gaza Strip and Hamas fighters based there have launched rockets at Israeli cities, social media users have taken up the classical letterforms in an effort to circumvent what they describe as a wave of censorship and over-moderation affecting Palestinian and pro-Palestinian voices across the internet.
Marking it as archaic is the lack of dots and marks — called diacritics — that were originally added to make the Quran easier for Muslims from North Africa, Spain and India to read, said Mohamed Gaber, a Cairo-born student pursuing a master’s degree in Arabic typography. The diacritical marks help readers distinguish between words that would otherwise look identical.
Its users believe the ambiguity of diacritic-free Arabic is one way to foil the humans and software charged with enforcing the networks’ content policies, which they see as muzzling legitimate political speech.
“The machine cannot really break this,” Gaber said. “You need to just have complete familiarity with the shapes of the letters, and even the lingo of who’s writing.”
The revival of centuries-old writing styles is one of several tactics Palestinians and their allies are employing to keep their posts from getting blocked or suppressed.
Amazon faces questions from senators over a reported contract with Dahua, a Chinese security camera company that touted on its website the ability to alert police when its facial recognition software identifies members of the Uighur ethnic group.
Some users have been intentionally misspelling keywords such as “Palestine” or “Israel”; others have begun interspersing posts about the conflict with more mundane images from their everyday lives. On Instagram, some users are advising their followers that adding a COVID-19-themed “Get Vaccinated” or “Stay Home” sticker to a pro-Palestine post will make it more visible.
For Palestinians, the struggle with Israel has always felt imbalanced — a poor displaced and refugee population pitted against one of the world’s most advanced militaries, backed by the American superpower.
But the latest flare-up of this long-running conflict is marked by a widespread feeling that the U.S.-based communications platforms people use to read and talk about what’s happening are putting their thumbs on the scale in a way that echoes long-standing American policy.
“It’s a double-edged sword,” said one social media user, a recent college graduate, who asked that her name be withheld to avoid online harassment. “On one hand our social media helps spread the word, but we are also restricted.”
The companies deny bias, pointing to technical glitches and good-faith efforts to uphold their own policies on matters such as hate speech and violent incitement, as well as U.S. and international law, at a scale and a pace that make some rate of error unavoidable.
But those who’ve been affected are reluctant to take such assurances at face value or trust the platforms to get it right. On Venmo, the social payment service, activists have begun advising one another to mask the true purpose of donations they make to Palestinian relief efforts amid a crackdown by the platform on related keywords. “Label it as lunch, coffee, nails, or anything mundane,” one Twitter user advised.
A cease-fire between Israel and Hamas militants kicked in early Friday after 11 days of war and appeared to be holding in its opening hours.
The most recent spate of accusations that the platforms have been enacting wholesale bans and so-called shadow bans against pro-Palestinian users began in early May, shortly after an Israeli court ruled in favor of evicting Palestinian families from their homes in the Sheik Jarrah neighborhood of East Jerusalem. After video was captured of stun grenades erupting inside the Al Aqsa Mosque — the third-holiest site in Islam — on the final Friday of Ramadan, a wave of posts, images and videos with hashtags such as #saveSheikJarrah and #HandsoffAlaqsa swept Facebook, Instagram, TikTok and Twitter.
On May 5, the Institute for Middle East Understanding tried to post a version of an informational slide show the organization had shared in February on Instagram. The new cover image read, “The Israeli Politicians Emboldening Anti-Palestinian Violence.” It was removed for violating Instagram’s community standards on hate speech and symbols.
The Instagram account LandPalestine, which has 167,000 followers, had a comic criticizing Israeli settlements removed for hate speech. The owner of the account, Nour Elbash, then received a notice that her account was at risk of deletion for the post.
A Save Sheikh Jarrah Facebook group started by outspoken Palestinian activist Mohammed El-Kurd with 132,000 followers was temporarily taken down. A pro-Palestinian account with 43,000 followers, Majid96e, has had several posts taken down for violating community standards and has posted several times that he believed his account was being shadow-banned. (The Times confirmed that when searching for the account, even when following it, it didn’t show up unless the full account name was spelled out, and then appeared much lower than other accounts with fewer followers.)
Twitter also suspended several pro-Palestinian accounts, which it attributed to an algorithmic error.
And Palestinian American hijabi influencer Maria Alia posted on Instagram that several pro-Palestinian Instagram stories she shared and added to her highlights were deleted even after she re-added them.
Nadim Nashif, executive director and co-founder of 7amleh — the Arab Center for Social Media Advancement, said the organization has been collecting reports of censorship on social media and has received 500. Fifty percent of those complaints have been about Facebook, and 35% have been about Instagram. Other organizations such as the Jewish Voice for Peace and MPower Change — both of which have been working closely together with 7amleh — say they’ve also received hundreds of reports of censorship on these platforms.
The group said Facebook has removed tens of thousands of posts from its platform at the behest of Unit 8200, the Israeli military’s cyber unit. In 2016, Unit 8200 flagged some 2,241 posts; that number swelled to 14,238 in 2018. Facebook complied with 90% of those requests, 7amleh says.
In a statement posted to Facebook, Israel’s Office of the State Attorney said Facebook and Instagram have complied with takedown requests less frequently than TikTok and Twitter.
In a statement, Facebook spokesperson Dani Lever acknowledged that “there have been several issues that have impacted people’s ability to share on” the company’s apps, “including a technical bug that affected Stories around the world, and an error that temporarily restricted content from being viewed on the Al Aqsa mosque hashtag page.”
“While these have been fixed, they should never have happened in the first place,” Lever said via email. “We’re so sorry to everyone who felt they couldn’t bring attention to important events, or who felt this was a deliberate suppression of their voice. This was never our intention — nor do we ever want to silence a particular community or point of view.”
While consumers find ways to work around what many perceive to be algorithmic bias, tech workers have launched efforts to put pressure on companies from the inside. On Tuesday, a Google employee group called the Jewish Diaspora in Tech published an open letter asking the company to condemn the actions of the Israeli government as well as “ensure that any support for Israeli humanitarian efforts be matched by support to Palestinian-led human rights and relief efforts.”
Google leadership has yet to respond to the letter or individual organizers. Last month, the company and Amazon were granted a $1.2-billion cloud contract by the Israeli government. Google did not respond to a request for comment. The company also would not say whether it would update its maps to include high-resolution images of the region now that a law banning the use of the images has been lifted. (Apple said it is working on an update to make its maps high-resolution.)
Meanwhile, Venmo told The Times that the company “takes its regulatory and compliance obligations seriously, including adherence to U.S. economic and trade sanctions administered by the U.S. Department of the Treasury’s Office of Foreign Assets Control” — the government organ that enforces American sanctions. Multiple Palestinian rights organizations are listed under the office’s directory of sanctions.
Facebook too has seen rising employee dissent over the issue. Discussions on Facebook’s internal chat system have been especially tense, with groups of Arab and Muslim employees calling out senior executives for a perceived internal bias against Palestinian voices and content, said Ashraf Zeitoon, who ran Facebook’s Middle East policy unit for three years until mid-2017.
“I’m told these have been some of the hardest and most vocal discussions ever at Facebook, on the same level of BLM,” he said, referring to the Black Lives Matter movement.
Mai El-Mahdy, who left her native Egypt to join Facebook in Ireland as a member of the Escalations Team and left the company in 2017, said she had heard from her former colleagues that the discourse was turning “ugly,” with employees reporting one another to Facebook’s human resources division.
“Arabs and Muslims, they see that Facebook is biased. Israelis say that what is happening is correct, and that to the contrary, Facebook should be harsher,” El-Mahdy said in a phone interview Thursday.
El-Mahdy noted that previous conflagrations between Israelis and Palestinians had led to similar internal discussions but to no effect.
Lever, the Facebook spokesperson, said: “Our policies are designed to give everyone a voice while keeping them safe on our apps, and we apply these policies equally, regardless of who is posting or their personal beliefs.”
“We have a dedicated team, which includes Arabic and Hebrew speakers, closely monitoring the situation on the ground, who are focused on making sure we’re removing harmful content, while addressing any enforcement errors as quickly as possible.”