There is evidence to suggest that Meta’s algorithms are continuing to amplify harmful content in Myanmar in violation of its own policies. Just one month after Meta banned the Myanmar military from the Facebook platform in 2021, an investigation by Global Witness found that Meta’s page recommendation algorithm was still amplifying pro-military content that violated many of its own policies on violence and misinformation.238 Michael, the aid worker who received death threats after becoming the subject of a viral Facebook post, recounted how the most inflammatory content seemed to gain the most traction in 2017: “The fact that the comments with the most reactions got priority in terms of what you saw first was big – if someone posted something hate-filled or inflammatory it would be promoted the most – people saw the vilest content the most. I remember the angry reactions seemed to get the highest engagement. Nobody who was promoting peace or calm was getting seen in the news feed at all.” 239 Facebook groups were also a major concern for local digital rights activists and others concerned about the risks that online vitriol against the Rohingya could spill over into a mass outbreak of violence. Victoire Rio, who worked at the Phandeeyar Foundation before co-founding the Myanmar Tech Accountability Network in early 2018, told Amnesty International: “Our biggest concern was hyper-local groups and pages – they became very popular as people naturally searched for their town… We saw them being used for spreading localized rumours, including unsubstantiated claims of imminent attacks by Muslims. These claims often referred to local landmarks, making them particularly believable and concerning. I also think that the group feature was being used for targeting certain groups.”240 Rio cited one example wherein two separate ‘chain’ messages were circulated among Muslims and Buddhists in Myanmar. Both urged preparedness for planned violence to be perpetrated by the other group ahead on the 2017 anniversary of the 9/11 terrorist attacks in the United States. Rio described how an interfaith activist had received both messages, leading her to believe that recipients of these chain messages were identified from religion-focused Facebook groups.241 According to reports, these chain messages triggered mobs to attack Muslim houses, shopfronts and a mosque in the towns of Magway and Taungdwingyi.242 The importance of groups in the Myanmar context was echoed by others. For digital rights activist Zaw (pseudonym), “groups were the most powerful things that happened on Facebook”. He explained how large numbers of people in Myanmar began joining groups from 2014 onwards, and how this fed hateful content: “you would get an alternative reality of the narratives happening in the political arena”.243 The independent human rights impact assessment for Myanmar commissioned by Meta has been sharply criticized by some human rights experts. Some of the most substantial criticism it has faced relates to the HRIA’s failure to account for the impact of Meta’s content-shaping algorithms in promoting advocacy of hatred against the Rohingya, in addition to a failure to adequately assess the specific risks presented by the Myanmar context prior to Meta’s deployment of certain technologies in the country.244 As has been demonstrated in this section, the Facebook platform’s signature features – including virality, groups, news feed, and recommendations – played a prominent role in Myanmar’s information landscape in the months and years preceding the 2017 atrocities in Northern Rakhine State. At the time, the platform’s content-shaping algorithms were largely shielded from scrutiny by a lack of transparency on the part of Meta, meaning that were not directly addressed by either the IIFFMM or the independent HRIA. However, recent revelations make it clear that these features were actively amplifying, promoting and recommending divisive Global Witness, ‘Algorithm of harm: Facebook amplified Myanmar military propaganda following coup’, 23 Jun 2021, globalwitness.org/en/campaigns/digital-threats/algorithm-harm-facebook-amplified-myanmar-military-propaganda-following-coup 238 239 Amnesty International interview by video call with Michael (pseudonym), 28 March 2022. 240 Amnesty International interview by video call with Victoire Rio, 12 April 2022. 241 Amnesty International interview by video call with Victoire Rio, 12 April 2022. John Reed, “Hate speech, atrocities and fake news: the crisis in Myanmar”, Financial Times, 22 February 2018, ft.com/content/2003d54e-169a-11e8-9376-4a6390addb44 242 243 Amnesty International interview with Zaw (pseudonym), 28 March 2022. See, for example: Mark Latonero, and Aaina Agarwal, ‘Human Rights Impact Assessments for AI: Learning from Facebook’s Failure in Myanmar’, Carr Center For Human Rights Policy Harvard Kennedy School, 2021. 244 THE SOCIAL ATROCITY META AND THE RIGHT TO REMEDY FOR THE ROHINGYA Amnesty International 47