Radicalization under deep uncertainty: a multi-model exploration of activism, extremism, and terrorism

Many security-related phenomena are both dynamically complex and deeply uncertain. The consequences of failing to address these two characteristics may be severe in the security domain. Radicalization, possibly culminating in terrorism, is a phenomenon with these characteristics. In this article, we use an exploratory multi-model approach to generate and explore plausible dynamics of radicalization under deep uncertainty. Three system dynamics simulation models related to phenomenon-based radicalization are introduced and used to generate ensembles of plausible dynamics. These ensembles are analyzed with machine-learning techniques to design adaptive policies that are robust under deep uncertainty. Finally, counter-intuitive findings–related to radicalization as well as to model-based analysis of radicalization-like phenomena–are presented. These findings would not have been discovered without a multi-models approach, broad exploration of the uncertainty space, and the use of advanced machine learning techniques.