Sunday, November 1, 2020

Review of The Evolution of Cooperation by Robert Axelrod

This review was written by Eugene Kernes

Book can be found in:

Elaborate Description

Cooperation can occur in the most austere situation. Cooperation does not require both parties to be either friend or have a brain. Cooperation does require a chance that each party will meet in the future, but a chance future meeting is not sufficient. Axelrod uses situations where it is always better to defect in the short-term than cooperate, yet provides ways and reasons why parties still cooperate rather than defect. The generic representation for this situation is the Prisoners Dilemma.

The Prisoners Dilemma game comes from game theory and is the core for the book. Each player has a strategic incentive to defect, but the outcome is better for both players to cooperate than to defect. Prisoners Dilemma is not a zero-sum game which means a gain from one party does not directly takes from other in this game. Both players can benefit and lose depending on their individual choice and the other players simultaneous choice. The payoffs from the game are required to fit a rule specified for the short-term temptation to defect and the better payoffs for mutual cooperation than mutual defection, but the payoffs need not be equal or comparable.

A computer tournament was played where individuals sent in code representative of a strategy to cooperate or defect decisions in an iterated Prisoners Dilemma. The strategy that won is called Tit for Tat, meaning this code would reciprocate every cooperation and defection used by the opponent in the prior turn. The strategy may not have gotten the best scores compared with others, but it got the highest average. A simple reciprocity rule may be robust, but it certainly is not the best given different conditions. Under certain conditions, Tit for Tat actually made both players worse off due to an echo of defections. Tit for Tat was used as a base strategy for the book. Lessons drawn from why this strategy won was due to it being a nice strategy, easy to recognize, and non-exploitability.

A nice strategy means that the strategy cooperates on the first decision and is never the first to defect. Not nice strategies usually defect on the first decision and future ones. Defecting initially usually causes both players to do worse due to reciprocation of mutual defection for the rest of the game. In a non-zero-sum game like Prisoners Dilemma, having an easily recognized strategy means that the other player can adept to the strategy. If a strategy is exploitable, it provides an incentive for the opponent to defect without reciprocity of a defection.

There are many ways to promote cooperation such as more interactions between players, being provoked, forgiveness, and changing the playoffs. More interactions between players would mean that each player would believe they would meet again in the future. If there are future meetings, it increases the cost of defection thereby making cooperation more likely. Both cooperation and defection must be reciprocated. Being provoked means that a defection by one player will result in a defection from the other at some stage. If the player is not provoked then the player will most likely see much more defections from the other. Forgiveness to prior defections is needed in order to prevent mutual recrimination. The payoffs of the game can actually be changed by having an outsider such as government or group provide a cost to defection, thereby lessening the temptation to defect.

This book may seek to explain how cooperation works and is more beneficial for all, but Axelrod does express that cooperation may not be best for everyone else. In some circumstances such as in business collusion, cooperation may be good for the players but at the expense of everyone else. To reduce unwanted cooperation, all that is needed is to take the reverse of cooperation building factors. The best example provided how cooperation is built with enemies and how cooperation was reversed was provided by trench warfare during WW1. The soldiers on each side of the trench had developed a live and let live strategy whereby they would shoot to miss rather than to kill. Enemies they were, but due to a multitude of cooperative building factors, trying not to kill each other was reciprocated. Each side’s leaders did not want the soldiers cooperating with their enemy.

A slight problem with the book is that the even though Prisoners Dilemma is explained in detail, the framework for game theory is left out. Unexplained assumptions and reasons can cause a misunderstanding if the reader has no prior knowledge of basic game theory. A missed detail is that by making the Prisoners Dilemma iterated, it changed the very dilemma by changing the payoffs. A problem with game theory is that the decisions and strategies are deterministic. Either to cooperate or defect rather than how to cooperate or defect. Incorporating the real situation details for the generic game theory representation is needed to make them useful.

With a high enough chance of meetings reoccurring in the future, there is no single best strategy. The best strategy can be found via trial and error. Axelrod shows how cooperation can emerge and be sustained. Given a readiness to defect, as tempting as the payoff may be initially, can be impeded by an expectation of reciprocity and future defection by everyone.

Book Details

Edition ISBN:  9780465005642
Pages to read:  223
Publication:     2006
Format:            Paperback

Ratings out of 5:
Readability     5
Content           5
Overall           5