Information about the eviction of Palestinians from an East Jerusalem neighbourhood was systematically removed from social media, say digital researchers who have collected evidence of the content takedowns.
Hundreds of posts and accounts documenting events in Sheikh Jarrah were deleted or restricted, the researchers said.
As violence escalated in Jerusalem at the weekend, Facebook, Instagram and Twitter were all accused of removing content or suspending accounts that shared information about the neighbourhood where Palestinian families are facing eviction from their homes.
Social media platforms have been dogged by issues with their moderation of content in non-English speaking conflict areas such as Palestine. As they tinker with their moderation systems, Palestinians are increasingly reporting that their digital rights are being violated by these platforms.
Instagram confirmed a system update resulted in a technical error that removed tens of millions of Stories, Highlights and Archives around the world, including in protest areas like East Jerusalem and Colombia.
"These errors result in infringing on people's rights, especially at critical moments where activists need social media tools to cover and document human rights abuses," Marwa Fatafta, the Middle East and North Africa policy manager at Access Now, told The National.
Instagram said the glitch was patched by Friday morning, but as violence in East Jerusalem was escalating, so were reports of content restrictions.
As Israeli forces stormed Islam's third-holiest site, Al Aqsa mosque, on Friday, the hashtags for the mosque were hidden on Instagram. The platform blocked content about Jerusalem just as Israeli forces stormed the mosque with stun grenades and rubber bullets, injuring 220 people, mostly Palestinians.
"We were made aware that the الاقصى# and الأقصى# hashtags were restricted in error," a spokesperson for Facebook, which own Instagram, told The National.
“This was unrelated to the technical issue, which affected Stories, Archives and Highlights. We sincerely apologise for both issues,” they said.
Beyond the admission that Stories and hashtags were affected, a litany of unusual social media behaviour surrounds content about Sheikh Jarrah.
Researchers have documented numerous restrictive behaviours experienced by users. They found that Twitter accounts were suspended, Facebook posts removed, graphic warning labels added to text-only posts on Instagram, and live-streams from Sheikh Jarrah blocked.
Nine digital rights organisations signed a statement saying the rapid escalation of content removal around Sheikh Jarrah was “egregious and pronounced”.
Having logged 200 examples of removed content, the Arab Centre for the Advancement of Social Media, known as 7amleh, contacted the various social media platforms and successfully pressured them to restore some of the content.
"Which means they did not really violate the community standards," Nadim Nashif, the executive director of 7amleh, told The National.
Palestine-based digital rights group Sada Social identified at least 50 Twitter accounts that shared information about Sheikh Jarrah and were suspended.
Twitter told The National it "took enforcement action on a number of accounts in error by an automated spam filter".
“We are expeditiously reversing this action to reinstate access to the affected accounts.”
Israel fuels social media takedowns
Meanwhile, Instagram confirmed to The National that no content was removed as a result of a government request.
The clarification is an important one, as Israel is known to run a cyber unit within its Ministry of Justice that systematically surveils Palestinian content and reports it to Facebook.
The number of content removal requests made by the Israeli cyber unit jumped from 2,241 in 2016, to 12,351 in 2017, to 14,283 in 2018 – an increase of 600 per cent over three years, reports from Israel’s state attorney’s office revealed.
Facebook complied with 90 per cent of the requests made by Israel. The targeted content was completely or partially removed, mostly for "identifying with a terrorist organisation" or "incitement offences", the state attorney’s report said.
“Israel already has enormous advantages politically, diplomatically, economically, and militarily. So the technological advantage only adds to the already massive asymmetry between the two sides,” said Khaled Elgindy, director of the Middle East Institute’s programme on Palestine.
It is increasingly common for Palestinians to see their social media posts disappear without explanation as Facebook has developed a number of policies that appear to disproportionately affect Palestinian content.
If the content is not removed at the behest of the Israeli government, it is likely the decision of artificial intelligence.
Moderating Arabic content without context
In 2017, Israeli police mistakenly arrested a Palestinian worker because an AI translation of his Facebook post mistook the words “good morning,” for “attack them” in Hebrew or “hurt them” in English. No Arabic speaker had reviewed the post before the arrest.
Social media companies rely heavily on moderation tools that use artificial intelligence and machine learning algorithms, but these systems struggle to digest the Arabic language and understand delicate contexts.
“For the global south countries, content moderation is problematic because it’s pretty much built on the experience of Americans,” said Mr Nashif.
“When it comes to occupation or conflict areas, it becomes much more complicated.”
In Palestine, anyone killed by Israeli forces is referred to as a shaheed, a martyr in Arabic. Facebook has determined the word falls under the “dangerous individuals and organisations policy”, meaning posts mentioning shaheed can be easily removed.
“Shaheed is a common word, it's part of the Palestinian lexicon,” said Ms Fatafta.
“So how did Facebook interpret shaheed as a vile word, as a word that is glorifying terrorism?”
Social media platforms are tasked with moderating content they do not understand and they are not willing to understand, added Ms Fatafta.
"When they use automated decision-making tools that are absolutely blind to context, the result is mass censorship and takedown," she told The National.
“It's a disaster in the making.”
Palestinian news organisations face similar barriers when it comes to utilising social media platforms to report from the front lines of violent battles. Content is routinely removed for being too graphic.
In 2016, the Facebook accounts of four editors at the Palestinian Shehab News Agency and three journalists from Al Quds News Network were disabled for violating community standards. Soon after, they were reinstated.
“This is an internationally recognised conflict zone. And it's important for media outlets to cover the reality for what it is, but that content sometimes gets taken down,” said Ms Fatafta.
Preventing Palestinians from sharing their stories has a tremendous impact politically, Mr Elgindy said.
"It distorts the debate and prevents policymakers and the general public from understanding the realities of occupation and dispossession," he told The National.
As Palestinian users find it increasingly challenging to navigate restricted online spaces, another challenge may soon await them.
Facebook is debating whether to designate the word "Zionism" as a protected term. If passed, it would mean critical conversations using the term would fall within the rubric of hate speech and could be removed.
“There are so many ways Palestinian voices have been marginalised, excluded and distorted. The oldest and most common has been to associate Palestinian resistance to the Zionist project and expressions of support for Palestinian rights with anti-Semitism. This has been going on for over a hundred years,” said Mr Elgindy.