The prisoner's dilemma is such a story: There are two criminals in a gang. After being seized, they are locked in two cells and cannot collusion. Each prisoner can use two strategies: to confess or not. The police tells them that if they confess, they will be sentenced to 1 year's imprisonment. If they do not confess and evidences are found, they will be sentenced to 3 years. However, there is actually a possibility that there will be no evidence and immediate release. So the choice of two prisoners can constitute four combinations:

Chart 1  Prisoner’s game

The two digit for each cell in the table show the payment (utility) of the two prisoners in the strategy combination. For A, if he confesses, he will definitely be sentenced to 1 year; if he does not, he may be released immediately, or he may be sentenced to 3 years, depending on B. But will B confess ? A does not know. When A does not confesses, he has big benefit, and also risky. Then A will think, “If I don't confess, but you confess, then I shall be sentenced to 3 years, you are sentenced to 1 year, you're lucky making me unlucky, that will not work.” So after such a comparison, he chooses to confess. B is the same.

As onlookers, we see these four kinds of payments and think that it is best for them not to confess, but they will not choose such results. The prisoner does not choose the best result in the game, that is, it does not reach Pareto optimality, individual rationality leads to collective irrationality. Why is this happening? This is determined by the organizational relationship between the prisoners. There is no obligation between the two people. If you hurt me, I can't punish you; if I hurt you and you can't punish me. Therefore, the function of this system is very poor. The obligation relationship determines organizational efficiency.

When a group of refugees face a band of robbers, they will flee in disorder no matter how many