Large-scale political violence kills and maims thousands of people every month across the globe. The challenges of preventing, mitigating, and adapting to large-scale political violence are particularly daunting when it escalates in locations and at times where it is not expected. Policy-makers and first responders to conflict-induced humanitarian disasters would benefit greatly from a system that systematically monitors all locations at risk of conflict and assesses the risks of conflict escalation.

The objective of this project is to carry out the basic research to build ViEWS – an ambitious political Violence Early-Warning System that would meet these needs. The project will build a pilot for a world-wide system with uniform coverage and frequent updates to avoid blind spots; provide alerts that are location- and actor-specific, and, most importantly, be transparent, replicable, and publicly available, including public assessments of predictive performance. No system satisfying these criteria currently exists. A successful execution of the research required to build one would constitute a major gain in conflict research, in terms of the impact for society but also in terms of methodological and theoretical development. The objective involves some risk in the sense that it is ambitious and challenging methodologically and theoretically, but is feasible. The conflict research community has laid the ground for such a system through careful isolation of theoretically manageable sub-components of complex phenomena, and concomitant systematic, disaggregated data collection efforts. By providing a framework for integrating them theo­retically and methodologically, ViEWS will fully harness the insights of these isolated efforts. This integration will not only allow an early-warning system of unprecedented scope and performance, but build theoretically informative bridges between these numerous compart­mentalized conflict research programs.

To lay the groundwork for ViEWS, the project will pursue five sub-objectives.

Sub-objective 1: Develop data-collection and programming routines for a pilot ViEWS. We will build a pilot ViEWS that covers a selection of component models of conflict process. It will cover Africa, a continent particularly affected by political violence, and have an update schedule that is sufficiently frequent to allow demonstrating its potential usefulness as a real-time system. It will provide warnings for armed conflicts, violence against civilians and between non-state actors, and forced population displacements, based on the data suite of the Uppsala Conflict Data Program (UCDP), currently the leading source of armed conflict data for researchers. The pilot will function as a laboratory that facilitates extensive experimentation with different methods, data, and models.

Sub-objective 2: Formulate theoretically informed models of violent processes. Modelling complex armed conflicts where many of the drivers of violent behaviour are inherently non-observable must depend on theoretically informed component models that connect useful observable indicators to specific conflict dynamics. A rich, systematic quantitative conflict literature has developed over the last three decades. Initially, the focus was on identifying structural conditions that increased the risk of war at the country level. Increasingly, however, this literature has moved towards disaggregation into smaller geographical and an actor-oriented focus. This allowed attending to more dynamic factors: triggers of violence, and escalation processes. It also highlighted differences between distinct types of political violence, such as violence against civilians and conflict between non-state actors, suggesting a need for refined theories to understand the particulars of each type. We will adapt and reformulate a number of selected disaggregated component models of distinct conflict processes for use in the forecasting ensemble.

Sub-objective 3: Solve methodological challenges to incorporate spatial and temporal dynamics across multiple levels of analysis to integrate and weight forecasts. Several important innovations in forecasting conflict violence over the last several years make a near-real time early-warning system possible. We will in particular expand on two methodological approaches to combine the disaggregated component models into a unified structure. First, ensembles of theoretically informed models combined through Bayesian Model Averaging (BMA) clearly improve political predictions (Montgomery et al., 2012; Raftery et al., 2005). Second, Hegre et al. (2013) suggest using dynamic simulations to model escalatory processes, integrating results obtained for detailed units of analysis into implications for larger temporal and spatial units. ViEWS will work to adapt and combine these approaches into a consistent system, while continuously separating data used to train models from the information used to validate them.

Sub-objective 4: Highlight the theoretical implications of the integrated and evaluated modelling approach. There is an increasing awareness that statistical significance testing is insufficient to evaluate theoretical propositions (Gill, 1999; Ward et al., 2010). ViEWS will inform the theoretical debates through systematic use of out-of-sample evaluation of component models’ predictive performance. The integrative work in ViEWS promises a considerable contribution to theory development. Studies tend to focus on one issue at the time, and pay insufficient attention to how various forms of violence or causal mechanisms are interlinked. Integrating these systematically will be a major methodological task in the project. This systematization, however, has to be theoretically informed. Hence, also this effort will speak to the theoretical debates in the field. The integrative work at the scale proposed here is completely novel. In this way, ViEWS will be important beyond its direct applicability as a high-quality early-warning system.

Sub-objective 5: Compare ViEWS with other forecasts in the field. We contend that the proposed system will provide more accurate and more comprehensive forecasts than any other sources of conflict risk assessments. We will substantiate this bold claim by publishing the results of systematic evaluation and comparison between the risk assessments produced by ViEWS and forecasts retrieved from other systems.