Title page 1

Contents 1

Introduction 4

Challenges to Modeling Social Risk 6

Unleashing the Power of Big Data and Machine Learning: Seeing through the Fog of Social Risk 7

Modeling to Inform World Bank Group Operations Affected by Complex Social Risk 9

Conclusion 18

References 20

Abstract 23

Tables 18

Table 1. Considerations for the viability of machine learning model-produced predictions 18

Figures 7

Figure 1. A conceptual framework of social risk modeling 7

Figure 2. Correlation between variables and violence events 10

Figure 3. Correlation between all variables in Democratic Republic of Congo model 11

Figure 4. Model-informed factors' influence over change in violence 12

Figure 5. Expert-informed theory of migration, violence, and livestock 13

Figure 6. Built structure identification in a Wajir, Kenya, neighborhood 13

Figure 7. Model-identified factors' association with population change 14

Figure 8. Potential live visualization of violence model forecasts 15

Figure 9/Figure 10. Change in crime preceding social unrest events 16

Figure 10/Figure 11. Predicted versus actual volume of reported crime events 17

Figure 11/Figure 12. Model applications in development policy and operations 19

Boxes 10

Box 1. Conflict-affected eastern provinces of Democratic Republic of Congo 10

Box 2. The Horn of Africa 13

Box 3. Small Island Developing State 16