consume potentially fabricated or misleading information on digital platforms with the power to increase communication speed, volume (of output and input), variety (of content), reach and coverage. Social media amplifies hate at scale. Finally, the inherent design of social media begets selective exposure, information bubbles, homogeneous echo-chambers, confirmation bias, and hyper-personalized, hyper-sensory and hyper-insular information environments that reduce our cognitive capacity to objectively evaluate information.25 Digital hate speech and virulent rumors warrant a unique aggregation of environmental factors, malicious strategies, and inadvertent actions in a logical narrative to know when thresholds for violent conflict are reached. In considering the patterns, conditions, features and drivers above, Mercy Corps, Do No Digital Harm and Peacebuilding have developed a theory of harm: When certain underlying conditions are present, we would expect social media to have an amplifying effect on conventional conflict dynamics. In situations of security-related anxiety, rumors – especially if they conform to pre-existing worldviews and emotionally-relevant narratives, and especially if audiences are repeatedly exposed to them – can perpetuate unfounded threat claims, amplify in-group/out-group tensions, and motivate rational actors to engage and justify collective violence in the name of self-defense.26 Techniques and tactics Digital communication amplifies conflict dynamics through the following extra-factual sources of information27: Rumor: Unverified information that is transmitted from one person to others. Rumors can be true, false or a mixture. At their core, mis- and disinformation are rumors.28 Hate speech: Any form of expression (speech, text, images) that “demeans or attacks a person or people as members of a group with shared characteristics such as race, gender, religion, sexual orientation, or disability.”29 Dangerous speech: Speech that has a special capacity to catalyze or amplify violence by one group against another.30 25 Padalino; 7-8; Fraenkel 10-11; Freelon, 38. 26 Greenhill and Oppenheim, 2-3; Benesch, 2014 3, 21; Padalino, 9. 27 According to Greenhill and Oppenheim, extra-factual sources of information are (a) unverified at the time of transmission, (b) serve as a source of actionable knowledge, (c) intended to influence recipients’ attitudes or behavior, (d) are emotionally resonant, and (e) are framed in ways that fit pre-existing societal narratives. See Kelly Greenhill and Ben Oppenheim, “Rumor Has It: The Adoption of Unverified Information in Conflict Zones,” in International Studies Quarterly (2017), 2 28 See John Bugge, “Rumour Has It: A Practice Guide to Working with Rumours:” (Communicating with Disaster Affected Communities (CDAC): 2017), 8; and Greenhill and Oppenheim, 2. Importantly, a rumor can also take on multiple forms over time: “For example, a human trafficker can spread a rumor amongst refugees … with the intent to deceive (disinformation), and a refugee can then pass this rumor to his friends and family not intending to deceive them (misinformation).”(Bugge, 8). 29 See Robert Faris et al., “Understanding Harmful Speech Online” Berkman Klein Center for Internet and Society (2016), 5-6. See also Kagonya Awori and Susan Benesch, “Umati: Kenyan Online Discourse to Catalyze and Counter Violence” (Conference Paper: IFIP 2013), 470. 30 Awori, 470; Theo Dolan et al, “Youth and Radicalization in Mombasa, Kenya: A Lexicon of Violent Extremist Language on Social Media” (PeaceTech: 2018), 6. This kind of speech is predicated upon the risk of violence (e.g. instilling fear by warning of impending threats, or by making an incitement to violence). MERCY CORPS The Weaponization of Social Media: How social media can spark violence and what can be done about it 12