and data-driven services at scale — yet without requisite standards, tools and capacities for risk management, data ethics and digital security — such efforts may unintentionally expose affected populations to additional risk. Weaponization phenomena therefore raise very serious questions regarding the Tech4Good, Open Data, Humanitarian Innovation and Digital Transformation agendas, among others, within the sector. If we do not address our own failure to account for the range of emerging negative externalities associated with digital data-driven interventions, we risk turning ourselves into threat actors and eroding the very principles that provide the foundation for our work. Improving programmatic responses to weaponized social media This assessment identifies a range of needs and approaches to enhance the quality of programmatic responses to weaponized social media. 1) A working theory of harm must be developed in order to prevent, mitigate or counter weaponization phenomena. Since January 2019, Mercy Corps and The Do No Digital Harm Initiative have been working together to develop such a theory, based on systems approaches to information environments. We have learned that for matters regarding the weaponization of information — whether malicious and systematic disinformation operations, or the inadvertent propagation of viral rumors — the propensity of harmful or otherwise violent outcomes is a function of the dynamic interplay of a handful of environmental factors that characterize either robust and resilient information ecosystems on one end of the spectrum or weak, volatile, asymmetric or hostile information ecosystems on the other. The latter have a latent capacity for harm when misinformation, disinformation, propaganda and the like are introduced. These environmental factors include:    Foundations for digital harm: The environmental conditions that make a particular context more (or less) susceptible to digital harms.73 Pathways to digital harm: The ways in which such harms unfold in a particular context, whether intentionally (in a malicious way) or unintentionally (in an organic way), or a combination of the two.74 Signals of digital harm: Indicators (direct or indirect), symptoms, or early warning signs that alert us that weaponization of social media issues are actively unfolding in a particular context. To effectively address weaponization of social media at scale, any intervention must take into account the dynamic, interdependent and nonlinear nature of this system — and the confluence of these factors. 73 This includes, for example, lack of reliable information, high levels of ambient fear, a history of group grievance, distrust of local authorities to provide justice and/or security, etc. 74 For example, intentionally means that there are malicious actors actively deploying strategies for leveraging information to cause harm. The strategies and tactics used by Russia’s Internet Research Agency (IRA), cyber-operations divisions, and affiliated cyber-mercenary groups to deceive, distort, and disrupt information environments in targeted contexts is one such example. Unintentionally means that there are dynamics around the flow of information that exacerbate or amplify the propensity for harm, whether or not there are coordinated campaigns. For example, in Ethiopia, while social media fill a need for critical information in the face of limited press freedoms, periodic internet blackouts, and restriction of mobile data, in recent years, these platforms have inadvertently amplified rumors, misinformation, and dangerous speech that have played a significant role in promoting radicalization and extremism, fomenting ethnic tensions, and inciting violence during a tenuous period of transition and reform. MERCY CORPS The Weaponization of Social Media: How social media can spark violence and what can be done about it 26