Spatial Prisoner's Dilemma with Stochastic Reaction Strategies

摘要

We extend the spatial game proposed by Nowak in 1992. In Nowak's original model, agents can only choose two strategies: AllC (unconditional cooperation) and AllD (unconditional defection). In our model, agents are allowed to adopt stochastic reaction strategies, and can learn from neighbors' according to payoff differences. Only the lattice network was considered in Nowak's model, while both homogenous (lattice, small-world) and non-homogenous(scale-free) networks are investigated in our paper. On the lattice networks, we found that the evolving process of the population's strategies is quite similar to that in the replicator dynamics model [13]. On the small-world networks, we investigated the effect of randomness of the topology to the evolving result, and found that a more clustering topology tends to sustain a more cooperative world. On the scale-free networks, we investigate how the hub nodes' initial states affect the evolving result.