Facebook’s failure to remove hate speech and falsehoods may have contributed to real-world anti-Muslim violence in Sri Lanka in 2018, a human rights report has found.

The report, published by Article One on May 11, was the product of a two-year partnership with the social media giant, which involved conducting human rights assessments in Sri Lanka and Indonesia.

In a statement to Bloomberg, Facebook ‘deplored’ the misuse of its platform, adding, “We recognize, and apologize for, the very real human rights impacts that resulted.”

The Article One report gives much focus to the violence in Kandy during broader anti-Muslim and Islamophobic riots. Ethnonationalist and Buddhist nationalist groups had torched Muslim homes and businesses, mosques were also damaged, with at least one reported fatality, of a man trapped inside of a burning house.

According to Hidayath Saththar, a provincial council member, identified by the New York Times in its 2018 coverage, reported that four mosques, 37 houses, 46 businesses and multiple vehicles were damaged.

The Reuters news agency had reported how Buddhist nationalists had more broadly “protested against the presence in Sri Lanka of Muslim Rohingya asylum-seekers from mostly Buddhist Myanmar, where Buddhist nationalism has also been on the rise.”

Muslims in the village of Mullegama, speaking to the Associated Press, described how the police did nothing to stop the violence.

Coverage from multiple international sources reported how the violence was spurred on by the death of a Buddhist youth, with many defying a police curfew. There was no evidence of a bias motive in the attack, but anti-Muslim voices used the death as a pretext for further anti-Muslim violence.

The extent of the anti-Muslim and Islamophobic violence forced a temporary ban on Facebook.

Before and in the aftermath of the violence, Facebook extremists used dehumanising language in their calls for violence against Muslims, even posting detailed instructions about creating petrol bombs.

But for years, politicians and civil society groups had warned the social media giant that ethnonationalist and Buddhist nationalists were using the platform to target Muslims, women, and other minorities. One influential report, published by the think tank Centre for Policy Alternatives in 2014, identified twenty ethnonationlist hate groups targeting women and minorities, with a vast majority still active on Facebook four years later, according to a detailed report in BuzzFeed News.

A New York Times investigation, drawing from testimonies from victims, officials, and ordinary people, argued that Facebook’s newsfeed had a “central role in nearly every step from rumor to killing.” Facebook had declined to answer at length to specific questions about its role in the violence, as a spokeswoman quoted over email, referred to its steps to remove content when made aware.

The most infamous example of an anti-Muslim falsehood derived from Sinhalese-speaking Facebook, with memes and unfounded claims that Muslims were spiking food with sterilisation pills to destroy the country’s ethnic majority.

It resulted in a crowd attacking a Muslim restauranter destroying his shop before a mosque was set alight. The so-called ‘confession’ was filmed and would become another a viral anti-Muslim falsehood.

A similar conspiracy about Muslims and sterilisation appeared in a Reuters investigation one year later. It found no evidence to substantiate claims that a Muslim doctor had, in secret, sterilised 4,000 Sinhala Buddhist women.

The Reuters investigation also revealed that the falsehood appeared one week after Buddhist nationalists had used the Easter terror attacks which killed 269 people in coordinated suicide bombings, as a pretext to destroy Muslim homes, stores, and mosques in the northwestern province of the country.

Academics have documented examples of the exponential spread of anti-Muslim and Islamophobic content on Facebook from Buddhist nationalists in Sri Lanka. One page, with over 13,000 ‘likes’ had produced 71 anti-Muslim and Islamophobic image graphics between January and May 2013. The group’s name in English ‘Safe Buddhism’ had a more menacing name in Sinhala, which read as, “End Muslim barbarity against Sinhala Buddhist harmony”.

Facebook also lacked the staff to monitor hate speech in Sinhalese and Burmese, according to academics. BuzzFeed News identified through officials and civil society groups that, as of March 2018, it had just two Sinhalese-speaking moderators, who both live outside of Sri Lanka, in a country with over 4 million active users.

According to the Article One assessment, “Facebook’s lack of formal human rights due diligence in Sri Lanka prior to this HRIA and the limited cultural and language expertise among Facebook staff at the time of the May 2018 Kandy incident may have contributed to offline harm stemming from online engagement.” An issue which was exacerbated further by a now-removed algorithm designed to drive engagement on the platform.

Nor did Facebook implement its Community Standards in a conducive manner, allowing forms of harassment and hate speech to grow or remain on the platform, Article One found. Examples of anti-dehumanising Muslim language cited in the report included, “Kill all Muslims, don’t even save an infant; they are dogs.”

The Article One report also highlights a decade of frustrations for civil society groups who felt shut out by the platform following repeated attempts (including phone calls and published reports) to highlight the misuse of its platform and offline violence, until the government shutdown of social media in 2018.

Facebook published detailed statements in response to the Article One reports on May 12, adding how it is “using proactive detection technology to identify potentially violating hate speech, developing machine learning capabilities in Sinhala and Bahasa Indonesia.”

They have also employed more policy leads and program managers in Sri Lanka, Indonesia and Cambodia, and hiring more content reviewers who speak Sinhala, Tamil, Bahasa Indonesia, Javanese and Khmer.

Despite significant criticism, Article One found that a majority of the 74 stakeholders engaged, found value in Facebook in Sri Lanka when managed effectively, adding that marginalised groups including human rights activists and LGBTQI+ individuals found safe spaces on the platform. But the report did highlight how their fundamental rights and that of women and children, may have been upended.

Article One also called on Facebook to acknowledge the harms it may have caused and to make the findings of such reports public.