Modeling Genocide at the System and Agent Levels
In all runs using the above conditions , one agent group eventually dominates the other such that only its agents survive . There is a great deal of volatility in agent group populations , both numerically and spatially . Generally , there is a loose clustering of the minority group as
members that are more isolated die before they can reproduce . Because this is not a result of localized agent decision-making , it is not an emergent property , but is rather a result of the globally implemented rule structure . A typical environment is shown below :
Figure 3 . Base model environment .
Early fluctuations in each group ’ s population as a percentage of the total system population have the general pattern shown below in Figure 4 , with volatility and oscillations developing for simulations that run for extended periods :
500 runs of this model provide the following statistics :
1 . Genocidal outcome : 100 % of runs . 2 . Average run length : 1,117 ticks . 3 . Standard deviation : 748 ticks .
Figure 4 . Base model population percentages in time .
40