At the Association of Threat Assessment Professionals (ATAP) DC chapter, I had the opportunity to share my perspectives on using a computational social science / complexity science approach for the prevention/mitigation of mass violence.
The ATAP DC group convened in person in Northern Virginia and by videoconference in several other locations along the East Coast. (I’m very grateful to the technical staff at Northern Virginia Community College for all their prework to make sure the technology worked and everything ran smoothly!). The ATAP DC members were a fantastic audience and humored me for what I understand was a slightly different take on mass violence than their usual.
Following a brief overview of computational social science and complexity science, I discussed some of the challenges of researching mass violence: mass violence is rare, complex, difficult to study, lacks agreed-upon theoretical models of causation, and is unfortunately often politicized.
We discussed different types of modeling, from verbal and mental models to mathematical and computational models. I believe that computational models are particularly suited to studying mass violence, and previously constructed one such model – Active Shooter: An Agent-Based Model of Unarmed Resistance. Computational models offer many benefits for mass violence research, education, and training, including the fact that they pose no risk to human subjects, they are infinitely repeatable, they are superlative for studying processes in systems, and can incorporate network features to study the influence of network ties in a particular process or outcome.
I demonstrated several computational models, including my active shooter ABM, as well as Epstein’s civil violence model and an epidemic model showing the spread of a virus between populations. If mass violence is, at least in part, germinated through the spread of the idea of perpetrating mass violence, whether by mass media or the internet, such models are useful in exploring how quickly and broadly such ideas could spread.
Finally, I discussed my view that the cumulative strain model proposed by Levin & Madfis is a verbal model that is ripe for a computational implementation.
I concluded by sharing my view that the threat assessment/threat prevention community could make use of computational modeling for training, for research, and perhaps ultimately for pre-warning. Computational modelers have demonstrated the value of working collaboratively with process stakeholders – for example, key officials in threat planning and response in schools and organizations – to perform “participatory” or “companion” modeling, in which stakeholder input is used to iteratively refine a model such that it is useful for the stakeholders in policy development or in response planning.
In the discussion following my presentation, I received several excellent and thoughtful questions, including whether psychopathy could be represented in agents (yes, through an additional modeling effort), whether my prior model could be extended to include armed responders or law enforcement officers (yes), whether these models can be validated (yes, but validation is challenging for many of the same reasons that mass violence research is difficult), and whether such models can be used for pre-scenario planning (yes).
I thoroughly enjoyed my time with ATAP DC and appreciated the opportunity to contribute.
Mass violence is a rare event. Attempts to empirically study episodes of mass violence can present a variety of challenges. The complex nature of episodes of mass violence, which may have germinated in years prior to actual attacks, make attempts to use conventional statistical techniques problematic.
Complexity science and the relatively new field of computational social science offer new paradigms and computational tools suited to the study of this dynamic problem. This talk reviews some of the challenges of mass violence research, provides an overview of complexity and computational social science, offers a live demonstration of a computational model of an active shooter scenario, and discusses a potential use case to computationally implement the cumulative strain model proposed by Levin and Madfis in 2009.
Why is this research important? Computational approaches enable new and innovative ways of studying, thinking about, and communicating with stakeholders about mass violence, and should become part of the threat assessment community toolkit.