Epstein (2002) conducts a thought experiment to understand what conditions create civic violence. Through computational modelling, he focuses on two cases: first exploring how `a central authority seeks to suppress decentralised rebellion', and second, how `a central authority seeks to suppress communal violence between two warring ethnic groups'. To achieve this, he models the situation as an agent-based simulation. There are two different kinds of agents in the simulation: general population and cops, or forces of central authority that seek to supress rebellion. The general population has variables that guide their behaviour, experienced hardship, preceded legitimacy, the central authority and risk aversion. Based on these factors, agents determine if they are quiet or aggravated. Cops detain active members of the general population and put them in `jail' for a given time. (For the second sub-goal, the model is defined differently; we hereafter only examine the first sub-goal.)
The model is used to run five different models. For four of them, parameter values change, as Epstein (2002) examines the impact of model behaviour. One model is run twice, thus examining in more detail the robustness of the model. For example, the model leads to conclusions such as `expected emergence of individually deceptive behaviour, in which privately aggrieved agents hide their feelings when cops are near but engage in openly rebellious activity when the cops move away.' Furthermore, it led Epstein (2002) to discuss tipping points that transformed the society from a non-rebel station into a rebel station. He also used the model to discuss different ways that societies might manage these, such as the `Salami Tactics of Corruption', highlighting that societies may be able to handle small but continuous incremental drops of legitimacy better than one larger drop - thus answering his research question.