People in groups behave in different ways, some are trusting and co-operative others are selfish and deceptive. This project creates a virtual society in which individuals behave in different ways. Each individual has a different strategy which may evolve as time goes on. The individuals may learn and improve their behaviour to obtain greater rewards.
This society is based upon the classic Prisoner's Dilemma 'game'. Under the rules of this game two prisoners are questioned separately about a crime they committed. Each may give evidence against the other or may say nothing. If both say nothing, they get a minor reprimand and go free because of lack of evidence. If one gives evidence and the other says nothing, the first goes free and the second is severely punished. If both give evidence, both are severely punished. The overall best strategy is for both to say nothing. However not knowing (or trusting) what the other will do, each prisoner's best strategy is to give evidence, which is the worst possible overall outcome.
This project expands on this basic game and implements an algorithm which allows strategies to evolve over several generations of prisoners. This program allows us to see more complex strategies evolve; possibly towards an optimal strategy.
The Prisoner's Dilemma 'game' seems simple, but it has generated huge quantities of research. It's a game whose principles may have wide-ranging explanatory power. For example, some biologists believe "that many wild animals and plants are engaged in ceaseless games of Prisoner's Dilemma" (Dawkins 1989). The game may also explain aspects of economic, social and political behaviour.