The prisoner's dilemma with image scoring on networks: How does a player's strategy depend on its place in the social network?
In, 3rd Australian Conference on Artificial Life, Gold Coast, AUSTRALIA,
04 - 06 Dec 2007.
Full text not available from this repository.
We investigate the evolution of cooperation in the prisoner's dilemma on different types of interaction networks. Agents interact with their network neighbours. An agent is classified by a value S is an element of [-1, 1] denoting its strategy and by its image score. It will cooperate if the opponents relative image score is above S and defect otherwise. Agents spread their strategies to their network neighbours proportionally to payoff differences. We find that network topology strongly influences the average cooperation rate; networks with low degree variance allowing for the largest amount of cooperation. In heterogeneous networks an agents place in the network strongly influences its strategy. Thus, agents on hub nodes are found to 'police' the population, while being on low degree nodes tends to favour over-generous less discriminating strategies.
Actions (login required)