Susan F. Martin Sidney Berkowitz Jeffrey R. Collman Lisa Singh Georgetown University
This project will assemble a multidisciplinary community of scholars and practitioners to create community and capacity for a large-scale, data intensive early warning system for detecting forced migration or population displacement. The system will be based on Raptor, a vast unstructured archive at Georgetown University of over 600 million publicly available open-source media articles. Mobilizing vast amounts of open source data will enable discovery of patterns of acute events (triggers) and/or slow-onset processes (trends) in the context of pre-existing stressors. Developing an effective early warning system of population displacement requires collaboration and shared learning between subject matter experts who understand the factors that contribute to forced migration at the macro, meso and micro levels and technical experts who understand how to collect, store, mine and analyze masses of data derived from international, national and local sources. Bringing together social scientists and computer scientists will expose social scientists to new modeling approaches for analyzing their subject matter. At the same time, computer scientists will exploit domain expertise in the social sciences. This expertise will provide insight for the development of beyond state of the art data mining of very large open source data bases for event detection, sequential mining and change detection. Participants will include scholars from many universities and practitioners from relief and migration-oriented organizations. The results of this endeavor will be: 1) methods and algorithms that can serve as a blueprint for integrating computational models into new avenues of social science research, and 2) a community and a plan for improving early warning of forced population displacement through human-computer analysis that address two key societal concerns, population changes and social disparities.
Broader Impact Effective early warning of forced population displacement will help in state- and organization-level planning and preparation for such movements, as well as directly aid potential refugees and displaced persons before, during and after their exodus. Planning can lead to action to try to avert mass displacement, preferably by tackling the triggering events and stressors and providing options to those who would otherwise be forced to relocate (e.g., getting food to villages at risk of famine). Earlier warning may also help divert forced migrants from risky modes of movement (e.g., via non-seaworthy boats or across landmine infested borders). Early warning of displacement would enable the pre-positioning of shelter, food, medicines and other supplies in areas that are likely to receive large numbers of refugees and displaced persons.
With funding from the National Science Foundation, Georgetown University has assembled a multidisciplinary community of scholars and practitioners to create a pilot of a large-scale, data intensive early warning system for detecting forced population displacement. The system is based on the Expandable Open Source (EOS) database (formerly known as Raptor), a vast unstructured archive at Georgetown University of over 600 million publicly available open-source media articles, and supplemented by social media such as Twitter. In addition to the open-source and social media data, our system has incorporated data from fieldwork (supported by previous funding mechanisms) with Iraqi refugees, whose recounts of personal experiences regarding forced displacement enriched our dataset and research methods. Mobilizing vast amounts of open source data will enable discovery of patterns of acute events (triggers) and/or slow-onset processes (trends) in the context of pre-existing stressors. It also supports simulation models that will help planners test the efficacy of different responses based on scenarios consistent with the early warning information produced. Developing an effective early warning system of population displacement requires collaboration and shared learning between subject matter experts who understand the factors that contribute to forced migration and technical experts who understand how to collect, store, mine and analyze masses of data derived from international, national and local sources. Our Project team members have included scholars renowned in their respective fields, from Georgetown (US), Fairfield (US), Fordham (US), York (Canada), University of Toronto (Canada), Sussex (UK) and Kultur (Turkey) Universities, Lawrence Livermore National Laboratory, and practitioners from the Jesuit Relief Services, Refugees International, Women’s Refugee Commission and the Brookings-LSE Project on Internal Displacement. Bringing together social scientists and computer scientists during this planning period has exposed social scientists to new modeling approaches for analyzing their subject matter. At the same time, computer scientists have benefited from domain expertise in the social sciences, enhancing the intellectual merit of our project. This expertise has provided insight for the development of beyond state of the art data mining and machine learning of very large, incomplete and potentially biased open source databases for topic modeling, event detection, sequential mining, change detection, sentiment analysis and dynamic graph mining. The outcomes of this endeavor have been: 1) a community of international and regional partners of practitioners and researchers collaborating to improve early warning of forced population movements through human-computer analysis; 2) methods and algorithms for using massive data that can serve as a blueprint for integrating computational models into new avenues of social science research; 3) a plan for implementing our approach to reach new audiences and broaden the social impact of our project; 4) knowledge transfer to a broader community through presentations of papers and model simulations at international conferences; 5) pilot studies on forced displacement in and out of Syria, Iraq and Somalia; and 6) presentations and publications aimed at social and computer scientists as well as practitioners and policy makers in the forced migration field.
