Censorship of pro-Palestinian voices on social media soars amid war in Gaza

Sabah Khodir is an Egyptian-American writer and activist based in Cairo with more than 33,000 followers on Instagram, where she mixes posts about her work and life with posts about her political opinions. On average, she explains, her stories receive around 7,000 people views, a figure that sometimes rises to 12,000. But since Israel launched its latest military offensive in Gaza and Khodir began to share more about the situation in Palestine and express her solidarity with Palestinians, her stories have been viewed less than 100 times. “This is a recurring problem, it’s not new,” she explains. “And it only happens with Palestine.”

Like Khodir, many social media users and dozens of rights and civil society organizations claim that Facebook and Instagram — both owned by Meta — have been disproportionately censoring content and accounts in support of Palestine. The most common form of censorship is limiting the visibility of such posts and accounts. In other cases, the account is restricted, suspended or eliminated, or the user’s access to the platform’s functions is restricted.

The claims of censorship on social media — where many seek information on the Gaza war due to the widespread perception that mainstream Western media is following the pro-Israel line — has put moderation policies and practices once again in the spotlight. There is also growing criticism, as these practices limit access to information and user’s freedom of expression and political participation. “If it’s about Palestine in particular, [the visibility of my content] is always much lower. My profile has been attacked [previously] for reporting sexual predators, and yet I have never been restricted; it’s only when it comes to Palestine,” Khodir notes.

From the start of the Israeli offensive in Gaza until the end of November, Human Rights Watch documented over 1,050 takedowns and other suppression of Instagram and Facebook content that had been posted by Palestinians and their supporters in more than 60 countries. The organization, however, noted that the total number is likely to be much higher. The 7amleh-Arab Center for Social Media Advancement has also documented hundreds of cases.

One of the most striking cases is Facebook’s decision to suspend the Quds News Network, which has almost 10 million followers, according to 7amleh. Another example is the temporary suspension of the Eye on Palestine Instagram account, which has six million followers. In response to these takedowns, 90 human rights and civil society groups from around the world denounced the discrimination and disproportionate censorship of Palestinian content in a statement on October 13.

An Egyptian researcher in critical artificial intelligence studies, who prefers that her name not be published due to the sensitivity of the subject, explains that this large-scale practice is possible because, in addition to human reviewers, large social media platforms have a highly automated process of text recognition, machine learning techniques, computer vision programs and deep learning that allow them to identify all types of content that violate their moderation policies.

“What is problematic is determining what constitutes incitement to hate, antisemitism, terrorism and justifying terrorism, because it is an inherently subjective act,” she says. In the case of Palestine, she adds, “there have been occasions when pro-Palestinian and solidarity slogans have been confused with slogans that are considered to support or condone Hamas violence.” This, in turn, makes it difficult to achieve “a balance between freedom of expression and compliance with content moderation policies,” she says.

Fighting the algorithm

In response to the criticism, Meta declared on October 13 that it had formed an operations center, with Hebrew and Arabic speakers, to address the crisis, and admitted that, due to the increase in flagged content, some content was “removed by mistake.” The company said that they “removed or marked as disturbing more than 795,000 pieces of content,” between October 7 and 10, but did not detail how much of that content was in Hebrew and how much was in Arabic.

Meta said that in that three-day period, it removed seven times as many posts “on a daily basis” for violating its Dangerous Individuals and Organizations policy — a rule critics say disproportionately affects Palestinian voices. The companies also blocked hashtags associated with content that violated its community guidelines, and lowered its thresholds for using automated tools to detect and remove potentially violating content. In a second statement, following the HRW report, Meta denied it was “deliberately” suppressing a particular voice. EL PAÍS contacted Meta, but did not receive a response by the time of publication.

“They always claim that there are what they call technical errors or glitches, but they always end up resulting in discrimination, censorship and silencing of not only Palestinian voices, but also citizens around the world who advocate for the protection of Palestinian human rights and in solidarity with Palestine,” says Itxaso Domínguez de Olazábal, 7amleh’s EU advocacy officer.

An external report on Meta’s handling of content during the May 2021 Israeli military offensive in Gaza concluded that its actions appeared to have had “an adverse human rights impact [...] on the rights of Palestinian users” and their ability to share information. It also noted that Arabic content was more likely to be mistakenly removed, and that proactive detection rates of potentially violating content were significantly higher for Arabic content compared to content in Hebrew. It also identified unintentional bias where Meta policy and practice most affected the rights of Palestinian and Arab users.

Hate speech

A separate case is X, formerly Twitter, which is being criticized for giving free rein to misinformation after recent changes introduced by the company’s owner, Elon Musk. One of these changes is that paid-for accounts are more visible regardless of their credibility and who manages them. X has also been accused of failing to crack down on content that incites hatred. Since the start of the Israeli offensive in Gaza, 7amleh has recorded more than 2.5 million instances of violent speech in Hebrew against Palestinians on X — a figure it attributes to the fact that the platform does not have Hebrew-speaking content moderators.

On October 10, X said in a statement that it had taken action against tens of thousands of posts for their graphic, violent and hateful content. It added that it is proactively monitoring antisemitic speech, but made no mention of anti-Palestinian speech. The company’s CEO, Linda Yaccarino, published a letter the following day addressed to the European Union’s internal market commissioner, Thierry Breton, in which she explained that X had removed hundreds of Hamas-affiliated accounts, but did not mention any other cases. She also recalled that the platform’s rules prohibit content that threatens to “to damage civilian homes and shelters, or infrastructure that is essential to daily, civic, or business activities.” But it is not clear whether this rule is being applied to senior Israeli officials.

“On Twitter they are allowing — under the banner of freedom of expression — incredible amounts of hate speech and violent speech, which also come from [Israeli] authorities and are often accompanied by disinformation,” says Domínguez.

The Egyptian researcher and Domínguez note that this imbalance is taking place in a context of asymmetrical power: firstly between social platforms and their users, and secondly, between Palestinians and pro-Palestine supporter, and Israelis. For example, on October 19, the cyber department of the Israeli Public Prosecutor’s Office announced that it had reported more than 6,200 pieces of content on social media, and claimed that around 90% of its requests to Meta platforms had been accepted.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition