Introduction

Main Menu Previous Topic Next Topic

**Prisoner's Dilemma**

In general, one can imagine any kind of dependence between the two characteristics of social interactions proposed in the previous slide. One extreme is when each of the two individuals derives maximal benefit from treating the other in the best possible way. In such a case, they will automatically cooperate. For example, two players on the same soccer team have every reason to help each other out, since the ultimate goal for both of them is the score attained by the entire team. Another extreme is what is called a "zero-sum game" - when each individual's benefit is proportional to the other's loss. A chess player, for instance, will always go for the sequence of moves that will bring down her opponent.

"Evolution of Cooperation", however, focuses on a scenario that lies between these two
extremes, one where each individual's benefit is in conflict with that of her partner, but not
hopelessly so. The scenario is known to economists, psychologists, and philosophers
as the *Prisoner's Dilemma*. In it, each of the two players independently makes a decision
either to cooperate or to defect - whether to be nice to the other player. From the point of view of
one of the players, defection is always the more beneficial choice. On the other hand, when both players
defect, each of their payoffs is lower than when both cooperate. Furthermore, mutual cooperation brings the
two players the highest possible total payoff. The specific numeric payoffs
we choose for simulating agent interactions satisfy each of the above conditions.

Here, then, is the dilemma: two rational selfish players will necessarily defect given the payoffs described above, and get very low scores as a result. However, if only there were some way to make the two players cooperate instead, not only each one of them would get a higher score, their total payoff would be greater than in any other situation.