Periodic Reporting for period 1 - ViEWS (ViEWS – a political Violence Early Warning System)
Período documentado: 2022-09-01 hasta 2024-02-29
Wars not only exact an immediate toll on the battlefield, but also leave scars that endure for decades – war is development in reverse. The imperative to prevent and mitigate armed conflict is clear. However, effective action demands timely and precise foresight.
The political Violence Early-Warning System (ViEWS, an ERC-AdG) offers a cutting-edge solution to this challenge, harnessing the power of advanced machine learning techniques and sophistical statistical methods to deliver monthly updates of forecasts for armed conflict up to three years in advance.
Seeking to equip policymakers with the tools they need to prevent blind spots in risk analyses and help allocate scarce resources where most direly needed, the idea taken to Proof-of-Concept sought to change the ViEWS pilot model into a financially sustainable early-warning system at global scale.
This has required fusing the various standalone explorative model developments and bespoke solutions in the AdG pilot into a more integrated technical solution. Furthermore, it sought to address a usability challenge: repackaging the pilot model and the information in technical academic papers into an accessible digital tool designed to meet the needs of the practitioners' community.
Finally, the financial sustainability of the project had to be made less uncertain in order to ensure long-term reliability of our services.
The political Violence Early-Warning System (ViEWS, an ERC-AdG) offers a cutting-edge solution to this challenge, harnessing the power of advanced machine learning techniques and sophistical statistical methods to deliver monthly updates of forecasts for armed conflict up to three years in advance.
Seeking to equip policymakers with the tools they need to prevent blind spots in risk analyses and help allocate scarce resources where most direly needed, the idea taken to Proof-of-Concept sought to change the ViEWS pilot model into a financially sustainable early-warning system at global scale.
This has required fusing the various standalone explorative model developments and bespoke solutions in the AdG pilot into a more integrated technical solution. Furthermore, it sought to address a usability challenge: repackaging the pilot model and the information in technical academic papers into an accessible digital tool designed to meet the needs of the practitioners' community.
Finally, the financial sustainability of the project had to be made less uncertain in order to ensure long-term reliability of our services.
The PoC project has awarded us time and resources to move the pilot of the ViEWS model towards a comprehensive, financially viable, industry-grade conflict early-warning system (EWS).
We have successfully enhanced the public version of our forecasting system with a number of model expansions and new data services derived from standalone developments funded by external collaborations alongside the AdG.
Highlights include migration to a new data backend, release and continuous upgrades to a web-based client allowing users to interact with this, development of our API, a new model predicting fatalities in conflict, expansion of the geographic scope of the model (to global coverage at the country level and Africa and the Middle East at the sub-national level), and – to assist interpretation of our forecasts – exploration of the use of "surrogate models" to identify the isolated contribution to our forecasts from the information contained in selected indicators.
Prior to the PoC project, the results from these collaborations were running separately from the live early-warning system. By means of the PoC, they have gradually been integrated into the main model pipeline, allowing both our partners and our early-warning system to benefit from the joint results.
Furthermore, we have released a new dataset containing time-series data on some of the key indicators informing our model, aggregated to the VIEWS units and levels of analysis. Most importantly, this includes a derived version of conflict history data from the Uppsala Conflict Data Program (UCDP), the most important driver of our forecasts as well as the outcome against which we compare and evaluate our predictions. The dataset is updated on a monthly basis, available via the VIEWS API.
The prediction model results, the surrogate model results, and the adapted conflict history data from the UCDP, are all visualised and available for download in the new version of the interactive VIEWS data dashboard that will be released in 2024.
Beyond the technical and scientific results of the PoC, we have worked extensively with representatives from our key user group to to ensure that the products we offer are maximally useful for anticipatory action, preparedness, and strategic planning. The new version of our data dashboard is the main product of this effort, complemented by a concerted effort at documenting and describing our model, data, and methodology. This has entailed a major overhaul of the content on our website, launch of a model documentation series, and development of a semi-automatic tool kit for quick production of commonly requested material and analyses of VIEWS data to help justify use of our services (validate the model and evaluate results in a format suitable for the end users outside of the academic community) and further support their daily operations.
Last, but certainly not least, the PoC has allowed us to explore options for the long-term organisational structure of ViEWS, resulting in the ongoing formation of a research consortium between Uppsala University and Peace Research Institute Oslo (PRIO). The consortium unites a suite of interrelated research projects and external collaborations that study and predict armed conflict and its impact of society and human development, led by the PI of the AdG project, Prof. Håvard Hegre. The AdG project called "the political Violence Early-Warning System (ViEWS)" has consequently transformed into the research consortium "Violence & Impacts Early-Warning System (VIEWS)".
We have successfully enhanced the public version of our forecasting system with a number of model expansions and new data services derived from standalone developments funded by external collaborations alongside the AdG.
Highlights include migration to a new data backend, release and continuous upgrades to a web-based client allowing users to interact with this, development of our API, a new model predicting fatalities in conflict, expansion of the geographic scope of the model (to global coverage at the country level and Africa and the Middle East at the sub-national level), and – to assist interpretation of our forecasts – exploration of the use of "surrogate models" to identify the isolated contribution to our forecasts from the information contained in selected indicators.
Prior to the PoC project, the results from these collaborations were running separately from the live early-warning system. By means of the PoC, they have gradually been integrated into the main model pipeline, allowing both our partners and our early-warning system to benefit from the joint results.
Furthermore, we have released a new dataset containing time-series data on some of the key indicators informing our model, aggregated to the VIEWS units and levels of analysis. Most importantly, this includes a derived version of conflict history data from the Uppsala Conflict Data Program (UCDP), the most important driver of our forecasts as well as the outcome against which we compare and evaluate our predictions. The dataset is updated on a monthly basis, available via the VIEWS API.
The prediction model results, the surrogate model results, and the adapted conflict history data from the UCDP, are all visualised and available for download in the new version of the interactive VIEWS data dashboard that will be released in 2024.
Beyond the technical and scientific results of the PoC, we have worked extensively with representatives from our key user group to to ensure that the products we offer are maximally useful for anticipatory action, preparedness, and strategic planning. The new version of our data dashboard is the main product of this effort, complemented by a concerted effort at documenting and describing our model, data, and methodology. This has entailed a major overhaul of the content on our website, launch of a model documentation series, and development of a semi-automatic tool kit for quick production of commonly requested material and analyses of VIEWS data to help justify use of our services (validate the model and evaluate results in a format suitable for the end users outside of the academic community) and further support their daily operations.
Last, but certainly not least, the PoC has allowed us to explore options for the long-term organisational structure of ViEWS, resulting in the ongoing formation of a research consortium between Uppsala University and Peace Research Institute Oslo (PRIO). The consortium unites a suite of interrelated research projects and external collaborations that study and predict armed conflict and its impact of society and human development, led by the PI of the AdG project, Prof. Håvard Hegre. The AdG project called "the political Violence Early-Warning System (ViEWS)" has consequently transformed into the research consortium "Violence & Impacts Early-Warning System (VIEWS)".
The PoC has allowed us to start transforming the AdG pilot of the ViEWS model into a comprehensive, financially viable, industry-grade conflict early-warning system, and widen its use by collaboration with organisations such as UNESCWA, UNHCR, UN Kenya, and UK FCDO. The system is supported by iterative research and development (R&D) conducted by the broader VIEWS consortium and associated research funding, ensuring our standing as a state-of-the-art early-warning system and a valuable public good.
The cutting-edge research conducted in the ERC-AdG project and the enhancements implemented over the course of the PoC has generated significant interest from the academic and policy/practitioners communities alike. It has generated over 7 million EUR in research funding since the PoC proposal was submitted, including a new ERC AdG (ANTICIPATE, Grant agreement No. 101055176) and some funding for the EWS itself.
To be fully viable as a technically and financially robust early-warning system, continuous investment in both the infrastructure and the underlying research is necessary given the rapid development in AI technology. The PoC has strengthened VIEWS' outreach and communication capacity, widened the usage of the system, and helped the team in understanding the needs and requirements of the users.
The cutting-edge research conducted in the ERC-AdG project and the enhancements implemented over the course of the PoC has generated significant interest from the academic and policy/practitioners communities alike. It has generated over 7 million EUR in research funding since the PoC proposal was submitted, including a new ERC AdG (ANTICIPATE, Grant agreement No. 101055176) and some funding for the EWS itself.
To be fully viable as a technically and financially robust early-warning system, continuous investment in both the infrastructure and the underlying research is necessary given the rapid development in AI technology. The PoC has strengthened VIEWS' outreach and communication capacity, widened the usage of the system, and helped the team in understanding the needs and requirements of the users.