Military commanders require means for detecting and anticipating long-term strategic instability. They have to get ahead and stay ahead of conflicts, whether those conflicts are within nation states, between nation states, and/or between non-nation states. In establishing or maintaining security in a region, cooperation and planning by the regional combatant commander is vital. It requires analysis of long-term strategic objectives in partnership with the regional nation states. Innovative tools provided by the quantitative and computational social sciences will enable military commanders to both prevent conflict and manage its aftermath when it does occur, write Sarah canna National security innovations (nsi), inc and Robert popp, Phd National security innovations (nsi), inc.
The Defense Sciences Office at the Defense Advanced Research Projects Agency (DARPA) under NGS2 is soliciting innovative research proposals to build a new capability (methods, models, tools, and a community of researchers) to perform rigorous, reproducible experimental research at scales necessary to understand emergent properties of human social systems.
A scientific team led by the University of Pennsylvania has received an award from the Defense Advanced Research Projects Agency to develop and validate reproducible methods for studying human social behavior. The award is part of DARPA’s new Next Generation Social Science program, or NGS2, which aims to revolutionize the speed, scale and rigor with which social science is performed.
“Many global trends, including conflicts among non-state groups and the growing influence of social media, point to the importance of social science for understanding the drivers of social and national stability,” Plotkin said. “We are excited about developing and applying cutting-edge science and technology to help social science become an even more predictive field and, in particular, to better understand the phenomenon of collective identity.”
Social science has its limitations, including technical and logistical limits to studying a large group that is representative of the population, says Adam Russell, DARPA program manager. “As a result, it’s been difficult for social scientists to determine what variables matter most in explaining their observations of human social systems and to move from documenting correlation to identifying causation. “This DARPA program will hopefully usher in a new research cycle of mechanistic modelling and hypothesis testing to make a predictive science of social phenomena,” Plotkin said.
DARPA NGS2 program
DARPA anticipates that the Next Generation Social Science (NGS2) program may require a fundamental reimagining of the social science research cycle and encourages participation from a wide and diverse combination of disciplines and skill sets – to include social sciences, but also physics, computer science, biology, game design, mathematics, and others. Specifically excluded is research that primarily results in incremental improvements to the existing state of practice.
Performers in the NGS2 program will work to determine fundamental measures and causal mechanisms that explain and predict the emergence of “collective identities.” A focus on “what matters most” for emergent social phenomena like collective identity presents an important and complex challenge to focus and validate new NGS2 research communities, tools, and methods. Note that while the NGS2 program will focus on collective identity formation as an exemplar research question,
DARPA anticipates that successful NGS2 capabilities will have benefits for tackling other complex problems and topics, including (but not limited to): resilience in social networks and structures, changes in cultural norms or beliefs, emergence of cooperation/competitions, social influences on preferences and cognition, etc.
Performance in the NGS2 program will occur in one of three categories: End-to-End (ETE), Enablers, and Test & Evaluation (T&E). Performers in the first two categories will focus on research and development and the third will perform independent testing and evaluation.
The NGS2 program is divided into two phases, a 24-month base period (Phase 1) with one 18- month option period (Phase 2). Each phase will consist of two research cycles. For the sake of clarity, this BAA will reference the conceptual elements of each research cycle in terms of the following Technical Areas (TAs), described in detail below:
TA1: Predictive Modeling and Hypothesis Generation;
TA2: Experimental Methods and Platforms;
TA3: Interpretation and Reproducibility
In Phase 1, performers will develop and demonstrate research tools and methods for rapidly testing and evaluating the accuracy of experimental hypotheses and predictions in multiple populations. DARPA-BAA-16-32/NGS2 BAA 5 Teams that demonstrate progress or promising technical approaches in support of NGS2 goals and TAs during Phase 1 may be encouraged to continue their research efforts in the next phase.
In Phase 2, performers will prove out their capabilities and demonstrate the replicability and rigor of their methodologies and models. Performers will evaluate their models’ external validity and robustness by scoring their predictive accuracy across the experiments of other Phase 2 performers. All proposers are expected to fully address how the proposed research will contribute to NGS2 program goals.
The proposal narrative should clearly explain the technical approach and fully describe the research and development required to build and test proposed methods, models and tools. Proposers must also describe how the work advances the state of the art as applied to social science in general and the exemplar problem in particular. Proposals should also identify the highest development, integration, and scaling risks involved and clearly describe how those risks will be mitigated early in the proposed effort.
Penn-led Team Receives DARPA Support to Develop ‘Next Generation’ Social Science
The grant provides the Penn-led, multi-disciplinary team with $2.95 million for two years, with a possible additional $2.3 million for a subsequent one-and-a-half years, dependent on progress, to further the goals of the NGS2 program, a key one being to develop a deeper understanding of the factors that drive the emergence or collapse of collective identity in human populations.
The team includes Joshua B. Plotkin, a professor in the Schools of Arts & Sciences’ Department of Biology, along with Erol Akçay, an assistant professor of biology at Penn; David Rand of Yale University; Simon Levin of Princeton University; Johan Bollen of Indiana University; and Alexander Stewart of University College London.
NGS2 also serves as a response to the so-called “reproducibility crisis” in the sciences, and the social sciences in particular, in which published findings have failed to be corroborated by follow-up studies. The program’s interest in applying rigorous methods to the social sciences aligns with a strategic strength of Penn Arts & Sciences, an emphasis on quantitative explorations of evolving systems.
The proposal by Plotkin and colleagues will encompass three scales of methods development and experimentation. On one level, the team will use game theory and evolutionary modeling to predict what factors govern group behaviors such as cooperation. The researchers will also put game theory into action, recruiting participants to play in-lab and online games in order to test model predictions for what conditions encourage a group to act as a cohesive whole. Finally, the research team will take advantage of massive datasets from such sources as Twitter to identify how social norms and collective identities arise and change over time in the real world.
“Our project is ambitious because it spans from mechanistic mathematical models to online experiments to observational studies of unfiltered social interactions,” Plotkin said. “We have assembled a group of researchers, drawn from a wide range of disciplines who all share a desire to help develop quantitative methods in the social sciences.”
Because the research involves studies on human subjects, it will be subject to IRB and human research protection offices’ review. Study subjects will be informed, consenting volunteers, and data will be de-identified to protect their privacy.
The DARPA award is structured with reproducibility built in: Each of the DARPA funded teams, after developing and testing its own models and hypotheses in the first phase of the project, will then cross-validate each other’s predictions in a second phase using their own study subjects. In addition, applying a relatively new practice in the social sciences, the researchers will pre-register all of their experimental plans in advance of performing them. This process, which requires laying out their hypotheses, protocols and planned analytical techniques, will help ensure proper, unbiased interpretation of results.
Social science modeling for Military and Security
In their book Poor Economics, Banerjee and Duflo propose that the development community has struggled to reduce global poverty because it does not adequately understand the subject and its root causes. The same may be true with population-centric conflict; the military has struggled due to an inadequate understanding of the effects of military interventions on the human dimension. Notably, despite billions of dollars spent to stabilize Iraq and Afghanistan, we know little about the causal effects regarding many programs, including the Commander’s Emergency Response Program (CERP).
Close collaboration between military and social scientists is one important factor in developing and delivering effective and efficient post-conflict reconstruction projects. As the author Jonathan Bate, an active duty US Army officer, an economics instructor in the Department of Social Sciences at West Point,rightly notes, both sides would benefit from this partnership. DoD can deepen its understanding of population-centric conflict and improve the effectiveness of its overseas stability operations through a stronger partnership with the social science research community.
The mission of the Department of Homeland Security Directorate for Science and Technology’s (DHS S&T) Human Factors Division is to apply the social and behavioral sciences to improve detection, analysis, and understanding of the threats posed by individuals, groups, and radical movements; to support the preparedness, response, and recovery of communities impacted by catastrophic events; and to advance national security by integrating human factors into homeland security technologies
Interdisciplinary quantitative and computational social science methods from mathematics, statistics, economics, political science, cultural anthropology, sociology, neuroscience, and modeling and simulation – coupled with advanced visualization techniques such as visual analytics – provide analysts and commanders with a needed means for understanding the cultures, motivations, intentions, opinions, perceptions, and people in troubled regions of the world.
Computational Social Science (CSS) is the investigation of social phenomena at any level of micromacro complexity enabled by computer models, especially (but not exclusively) through object oriented, agent-based models (ABMs) and other simulation tools. Social science modeling has evolved from statistic analysis, game, and decision theory to automated information extraction, social network modeling, social geography, complexity theory, systems dynamics modeling, agent based modeling, and many simulation methods.
These social science modeling and visualization techniques apply at multiple levels of analysis, from cognition to strategic decision-making. They allow forecasts about conflict and cooperation to be understood at all levels of data aggregation from the individual to groups, tribes, societies, nation states, and the globe. These analytic techniques use the equations and algorithms of dynamical systems and visual analytics, and are based on models: models of reactions to external influences, models of reactions to deliberate actions, and stochastic models that inject uncertainties. Continued research in the areas of social science modeling and visualization are vital. However, the product of these research efforts can only be as good as the models, theories, and tools that underlie the effort.
While the DoD and DHS are changing their orientation to quantitative/computational social science modeling, difficulties and challenges remain. First, opportunities and challenges in the theoretical domain include the need for better, scientifically grounded theories to explain socio-cultural phenomena related to national security. Better theories affect modeling on all levels. They yield clarified assumptions, better problem scoping and data collection, a common language to interpret analysis, and inform visualization options. Quantitative/computational social science presents the opportunity to integrate theories, explore their applicability, and test their validity.
References and Resources also include: