2019-01-07

Book Review: "The Evolution of Cooperation" by Robert Axelrod

I've recently read the book The Evolution of Cooperation by Robert Axelrod. (Note: this is somewhat of a technical book, so I will dive right into the review with jargon, with more on this point at the end of the review.) It's a primer on results from that time showing how in an iterated prisoners' dilemma, tit-for-tat strategies are remarkably robust for their combination of simplicity, clarity, tendency toward cooperation and forgiveness, and prompt & effective retaliation when needed, and that such strategies can effectively propagate environments even where other strategies are in place, provided that those who play the tit-for-tat strategies can find & cluster around each other to interact often enough, and provided that the value each player places on the next round compared to a given round in an iterated game is not too small. The author also uses examples from trench warfare in World War I, biological evolution, and international trade policy to illustrate the seeming universality of the principles of the prisoners' dilemma and its iterated variant.

Although this book was written in the 1980s, making it a little dated in terms of the complexity of models that could be tested on computers and the formalism of game theory itself, it was great to see the author anticipate a lot of more recent developments by discussing the importance of clustering, stereotypes, reputation, regulations, et cetera. Additionally, while the author stresses that cooperation can take place even among egotistical (non-altruistic) or antagonistic individuals in the absence of central authority, the author does take care to convey the nuance that this is not always a good thing per se, rather than taking the utopic view of libertarian philosophy; such cooperation is detrimental to the public at large in situations like economic collusion in an oligopoly, while the incentives to cooperate or not change such that government (or other societal) intervention is needed to do things like collect taxes & deter evaders to fund public goods, correct historical (and present) racist marginalization of minority groups, mediating conflicts among heterogeneous populations in large cities, et cetera. It was also cool for me to understand that any iterated game where the players are unsure of when the game will end but others controlling the game know it will end after a finite number of rounds can be rewritten as a similar game where the players believe the repetition will be infinite but with a different discount factor. My only minor complaints are that while the author does acknowledge that changing parameters of a social interaction can change the prisoners' dilemma into a different sort of game altogether, it would have been nice to see a more nuanced discussion of the degree to which the prisoners' dilemma is really a universal feature of human interactions as opposed to being culture-specific, given its seeming universality in other domains, and that the author rather glibly claims that sequential versus simultaneous play by players in each round of an iterated prisoners' dilemma doesn't make much of a difference, which I find suspicious in the absence of further explanation/context in the book itself. Overall, I enjoyed reading this and could read it quickly because of my minor in economics in college & interest in the subjects of economics, game theory, and network science, so it may be appropriate to others with similar interests & backgrounds as myself; it is a fairly technical book, so it may not be appropriate to general readers without this background, while specialists in the fields of evolution, game theory, or complexity science may find this book to be too dated.