Agent-based simulation of social violence situations

Michael Winsper & Michael Garlick
Computer Science, Aston University

Agent-based modelling is a method for studying complex systems like economies, societies, ecologies etc. Very often top-down mathematical analysis is limited in its ability to capture the complexity of such systems. In this case, agent-based modelling offers a practical, constructive method of analysis. It is particularly useful in developing an empirical understanding of a given complex system and investigating why certain regularities have evolved and persisted. Agents represent entities within the virtual world and are provided with a set of behaviours which define and control their interactions. Such micro level interactions cumulate to exhibit macro level observable phenomena, which may not be predictable from the individual agent behaviour. Replicating observable macro level features mirroring real world circumstances allows a modeller to identify the behaviours and parameters which cause such outcomes.

Drawing inspiration from Epstein?s paper "Modelling Civil Violence: An agent-based computational approach", we create an agent-based model to explore and observe two different civil unrest scenarios. In the first scenario we model episodes of decentralised rebellion involving citizens and cops, where cops represent the centralised authority attempting to suppress the outbreaks of violence. We investigate how the policies of (a) restricting social influence and (b) imposing curfews affect the level of unrest exhibited within the population upon different types of society. We find that restricting social influence pacifies rebellious societies, but has the opposite effect on peaceful ones. Curfews, however, exhibit a pacifying effect across all types of society.

In the second scenario we present a multi-agent based model of a population comprising two adversarial groups and a peacekeeping force. We compare the effect of random and ingroup migration strategies for civilian movement on the resulting violence in this bi-communal population. Ingroup migration leads the formation of clusters. Previous work in this area has shown that ingroup clustering instigates violent behaviour in otherwise passive segments of the population. Our findings confirm this. Furthermore, we show that while the adoption of ingroup migration by one group is a winning strategy in violently predisposed populations, in more peaceful settings it becomes a restricting factor.