WikiFindings

Clever game players that extort their partner in the Prisoner’s dilemma do well head-to-head but do poorly in a mixed population of different players

Background

Researchers have understand how cooperation can evolve by using game theory, especially the Prisoner’s dilemma game. In this game players can do well by cooperating a lot, but there is always the temptation to defect on the deal for a better payoff. A player that appreciates that their partner will respond to their behaviour can extort them to ensure they win by choosing probabilities from certain ranges (‘zero-determinant’ strategies). The more the partner cooperates, the more both their scores increase, but the extortioner always wins.

Findings

We pitted lots of different strategies against one another in computer-based tournaments. Some strategies always cooperated or always defected no matter what the opponent did. Tit-for-tat does whatever the partner did last time, which punishes defection. A ‘generous’ version is a bit more forgiving. Win-stay-lose-shift changes its choice if it would have done better last time. Extorting strategies get overall poor averages scores in the tournament. Although they always beat partners, they do so by causing both scores to be lower. Generous zero-determinant do the opposite: the partner always wins, but both do really well on average

Implications

Extorting strategies are fascinating in that they give players a way to win head-to-head games. However, they are unlikely to evolve in the brains of real animals (including humans). This is because what matters to natural selection is not whether individual interactions are won, but the overall scores from lots of interactions. It is likely to be better to have generous strategies, which helps to explain why so many social animals are so cooperative.