BaSh

TWELVE EASY GAMES

Introduction: Prisoner's Dilemma

Now we turn to a very basic problem in all human conduct: cooperation. Not all interactions between people can be captured in price for a transaction, but must also include expectations of future action conforming to promises. If promises can be depended on, then huge possibilities for cooperation over time become possible.

For this purpose we use a table of outcomes for cooperation long used by economists, called "The Prisoner's Dilema". If two persons have been caught by the police and put in separate cells, they might have a motive to cooperate by refusing to testify against each other, or one or the other might testify against the other. By setting the penalties for a prisoner who does not confess when his partner does, the police can create enough risk for each to testify agains the other in most cases. I believe this kind of thing does occur all the time.

When you start the game you should see a table of outcomes. Initially, we set these so that if you cooperate with the other player, you both make a modest return of 3 units. If one of you defects and the other still tries to cooperate, the defector does much better at 5 units and the one whom he betrays gets zero. These payoffs provide a huge tempatation to defect. But if both of you defect, you get a very low payoff of 1 unit each.

If you play the game only once, it will be hard to avoid defecting, because your payoffs, 5 and 1, are so much better than the payoffs for cooperating of 3 and 0. But things are very different if you are going to replay again and again. Then the disadvantage of getting stuck in a repeated no-trust 1-1 payoff situation are far worse over many turns than attaining a degree of confidence and cooperation that allows a run of 3-3 payoffs. Grabbing an opportunistic 5 unit payoff may not be worth the damage of a falling into a string of low no-trust payoffs after that.

You will be assigned random partners, and the game will begin. When one player takes his turn, nothing seems to happen because the game is paused until both players have decided what to do. Then you should see the outcome of the turn, and whether your partner has cooperated. Because of the tendency to double-cross in the last move of this game, we stop the game at a fixed number of moves, and that number is unknown to you while playing. But all player-pairs play for the same number of moves so we can see the advantages of cooperation.

If we have time, we will play the game with an adjusted payoff matrix. Suppose we, as a community decide to enforce cooperation for the public good. Defection payoffs can then be reduced, to say 4, 3, or 2. As far as I know, rough and ready efforts to enforce cooperation among criminals take the form of threatening to kill defectors in real prison situations. In the real world, penalties for any failure to perform as expected in contracts can be modelled the same way.

On to the Prisoner's Dilemma

What Did We Learn?