For decades, social scientists have been studying empirical tendencies of human subjects to cooperate in laboratory Prisoner's Dilemma settings. Here is what we are learning.
In a classic Prisoner's Dilemma, players choose independently whether to contribute toward a mutually desirable outcome. Cooperation comes at a cost, so that (for example) a dollar's contribution yields a return of more than a dollar to the group but less than a dollar to the individual contributor. Thus contributing is optimal for the group and suboptimal for the individual. Not contributing is a dominant strategy, which is to say, standard game theory predicts defection. If players prefer higher to lower monetary playoffs, and if other things are equal so far as a player is concerned, a player will not contribute.
And yet, among the more robust findings to come out of experimental economics is that human players do not behave as game theory predicts. In laboratory Prisoner's Dilemmas, individual contributions toward the group good can be substantial.
In iterated Prisoner's Dilemmas – multi-period games where each period viewed in isolation has the structure of a Prisoner's Dilemma – game theory does not so straightforwardly predict defection. Iterations allow players to implement strategies of reciprocity, including the simple but effective strategy of Tit-for-Tat (Axelrod 1984), which means responding to cooperation by cooperating and responding to defection by defecting. In iterated laboratory games, the phenomenon of human subjects making substantial contributions is likewise robust. However, it also is a robust result that cooperative relations tend to crumble in an iterated game's final period (when players know in advance that a given period is the final one). Typically, levels of cooperation remain positive, but decay in the direction predicted by standard game theory.
Laboratory experiments can test theories about how people behave in a Prisoner's Dilemma. Experiments also can test policy proposals.