Policy and Complex Systems
Figure 6 . % population oscillation for threshold-level of . 6 .
These early results show a clear exponential increase in simulation run length as the global Threshold to Act level approaches . 5 . The first indication these results provide is that while the model is generally biased toward genocide as seen in Experiments 1 and 2 , it is flexible enough to be able to avoid genocide through making the Threshold to Act a globally uniform and static value in the middle of its range . The implication of this result is that even in a divided society with identitybased conflict , a genocidal outcome is far less likely if people are unwilling to act against the opposing identity group regardless of their level of animosity or grievance against the other . Thus , policies could have greater impact if geared towards providing incentives not to act , even if they do not remove 100 % of that willingness . Additionally , as noted above , endogenous factors such as a strong middle class may be a source of restraint that should not be underestimated if it indeed causes people to avoid taking action against opposing identity groups ( Straus , 2012b ).
This experiment yielded a second notable property . As the Threshold to Act level increased , the spread between average population counts of the two groups decreased exponentially . Running the model for 10,000 ticks with Threshold to Act levels ranging from . 5 to . 9 in steps of . 1 , and at . 95 produced the following results given below . Figure 6 shows group percentage population oscillations for a Threshold to Act level of . 6 . Here , it is clear that while neither population can attain dominance such that it completely exterminates the other , there is a great deal of volatility in population counts , indicating high levels of violence .
Figure 7 shows the trend for the average spread between population percentages at each Threshold to Act level :
43