Simulating Tolerance in Dynamic Social Networks depends on the strategy of an agent's neighbor. However, when this game is played with multiple neighbors, the optimal strategy is the strategy adopted by the majority of an agent's neighbors. decreases. This relationship makes sense when considering that the condition for breaking a connection follows formula (4): r ij < EXPERIMENTS To demonstrate the theorems, we have run several preliminary experiments and watch for the causes and effects of tolerance. Agents are connected by a random network and play the two-action pure coordination game defined in Table 1. The time discount w=0.95. The memory limit L=10. The experiments are run in a relative small scale with only 300 agents. Each result is the average of 30 trials. The system is initialized with 50% "cooper- ate" agents and 50% "defect" agents. Throughout the trial, we keep track of four attributes: r total Effect of Weight W on Tolerance As the weight on past rewards increases in this experiment, agents begin to value their neighbors' past actions more strongly (shown in Figure 2). This means that n-tolerance increases as the weight increases. The effect of weight on toler- ance, however, is much weaker than the effect of the threshold on tolerance. This makes sense, as the weight has a smaller effect on the condition for breaking a connection.