The Genocide Facebook Helped Broadcast
Years before Myanmar's military drove more than 700,000 Rohingya into Bangladesh, anti-Rohingya hate speech circulated widely on Facebook. United Nations investigators and human-rights groups later concluded that the platform's design helped amplify it.
In August 2017, Myanmar's armed forces launched what they described as clearance operations in Rakhine State. Within weeks, more than 700,000 Rohingya Muslims had fled to Bangladesh, according to United Nations agencies, amid accounts of killings, sexual violence and burned villages. A UN fact-finding mission later said the campaign had been carried out with “genocidal intent” and recommended that senior generals be investigated and prosecuted.
Much of the groundwork, investigators found, had been laid online. In a country where Facebook is for many users synonymous with the internet, pages and accounts — some linked to the military — had for years portrayed Rohingya as outsiders and threats.
On 12 March 2018, the chairman of the UN mission, Marzuki Darusman, told reporters that social media had played a “determining role” in Myanmar. Facebook, he said, had “substantively contributed to the level of acrimony and dissension and conflict” among the public, according to Reuters. A separate UN investigator, Yanghee Lee, said the platform had “turned into a beast” in the country.
Four years later, Amnesty International went further. In a September 2022 report, “The Social Atrocity,” the organization concluded that the recommendation systems used by Facebook's parent company, Meta, had proactively amplified content that incited violence against the Rohingya. “In the months and years leading up to the atrocities, Facebook's algorithms were intensifying a storm of hatred against the Rohingya which contributed to real-world violence,” Amnesty's secretary general, Agnès Callamard, said in a statement accompanying the report. Amnesty called on Meta to pay reparations.
Meta has acknowledged shortcomings. The company commissioned an independent human-rights assessment, released in 2018, which found it had not done enough to prevent its platform being used to incite violence, and it has said it was too slow to act in Myanmar. Civil-society groups in the country had warned Facebook about anti-Rohingya content for years before the 2017 exodus.
Researchers who study platform design point to a common thread. Feeds ranked by engagement tend to reward the content that provokes the strongest reactions, and hostility toward an identifiable group is among the most reliably engaging material. Content moderation, particularly in languages and regions where platforms invest little, has repeatedly lagged behind the speed at which such content spreads. Many academics and regulators argue that binding rules on platform accountability are the durable remedy, though such measures remain contested and unevenly applied.
The Rohingya case remains among the most heavily documented examples of online hate speech preceding mass violence offline — a sequence that researchers and human-rights bodies continue to cite as they press governments and companies to act earlier.