Using the Moral-Situational Action Violence Risk Model for Assessing Women Involved in Extremist Violence: An Empirical Study

The intelligence community faces a complex and everchanging task in the monitoring of at-risk individuals who pose potential threats to national security. Recently, the National Academies of Sciences, Engineering, and Medicine (2019) published a decadal survey that emphasized the vital role the social and behavioral sciences play in assisting the IC in answering emergent intelligence questions and in developing human-machine interactions capable of enhancing that capability. In this empirical study, we explore a theoretically based violence risk model, the moral-situational action of extremist violence (MSA-EV), to determine its usefulness in identifying women who have become involved in direct action related to violent extremism. The MSA-EV risk model contains 3 domains: propensity, mobilization, and capacity-building. Each domain was coded quantitatively and qualitatively using a 41-p. protocol that captures open source online material from both the surface web (i.e., the part of the Internet that is accessible to everyone) and the dark web (i.e, the part of the Internet invisible to search engines and that requires an anonymizing browser [Tor] to be accessed). Using logistic regression analyses, the model classified 300 individuals into either a high-risk category or a combined medium-risk category/low-risk category. When relevant variables from all 3 domains were plotted in terms of specificity and sensitivity, the corresponding area under the receiver operating characteristic curve was .90, which is indicative of excellent classification accuracy. These findings should be further validated in future research using an updated version of the coding manual and replicated with male individuals. Public Significance Statement Given a recent executive order to maintain American leadership in artificial intelligence combined with the almost simultaneous publication by the National Academies of Sciences, Engineering, and Medicine (2019) of A Decadal Survey of the Social and Behavioral Sciences: A Research Agenda for Advancing Artificial Intelligence Analysis, there is no longer any ambiguity regarding the importance of pairing human skill with the potential of machine learning when assessing threats to national security. To realize this potential, social and behavioral scientists must assist in the development of theory-based risk models that can help federal law enforcement and intelligence analysts in monitoring at-risk individuals who manifest fluctuating l