On Cooperation — Takeaways from Axelrod’s Iterated Prisoner’s Dilemma
Source: Axelrod, Robert (1984), The Evolution of Cooperation, Basic Books, ISBN 0-465-02122-0
http://www.eleutera.org/wp-content/uploads/2015/07/The-Evolution-of-Cooperation.pdf
In the past, students on this same blog have alluded to Robert Axelrod’s famous Iterated Prisoner’s Dilemma tournament, choosing to highlight the success of the TIT-FOR-TAT strategy (and rightfully so), and yet they fail to make full sense of the paper and some of the other key points that should be taken away.
To summarize, professor Robert Axelrod ran a study on the iterated prisoner’s dilemma (in which the two players participate in a multiround version of the famous prisoner’s dilemma game, neither player knowing when the game will end) by hosting a tournament where participants would submit a strategy to be used against all other competitors, and a default competitor RANDOM (who would choose to defect/cooperate on a random level). Many famous game theorists, economists, biologists, computer scientists, etc submitted strategies varying from exceedingly complex to incredibly simple, but the winner in the end was the simple program TIT-FOR-TAT, that simply began by cooperating and then proceeded by copying its competitors last action (whether that be defection or cooperation). Axelrod published this result and some following analysis, and then hosted another tournament with the same rules some time later, giving everyone time to read the results and adjust their strategies accordingly. Surprisingly, in the second competition, TIT-FOR-TAT won AGAIN, besting even strategies developed to outthink TIT-FOR-TAT (such as programs that followed TFT’s strategy except for defecting every once in a while randomly to score some “free” points). The rest of Axelrod’s 1984 book discusses why TIT-FOR-TAT’s simplistic strategy was so effective while also providing real world examples of its effectiveness in other fields such as biology, polisci, and economics.
To abridge Axelrod’s interesting (though a little wordy) analysis, I here provide his four key points for success:
- Do not be envious of the success of other players.
- Do not be the first to defect.
- Do not be too clever.
- Reciprocate with both cooperation and defection.
As mentioned previously, the book applies these four strategies in a variety of disciplines. In biology, for example, Axelrod runs an evolutionary version of the IPD, with strategies that do well surviving to “reproduce” in future generations while strategies that fail to do so slowly die out and go extinct. As expected, TIT-FOR-TAT, within a few hundred generations, establishes itself as a firm leader, occupying over 14% of all ecosystem spaces. What is perhaps more interesting is seeing how other strategies fared and why they failed to TFT’s simplicity. Some programs began extremely successful, jumping up to 8% within just 50 generations. However, these programs did so well by preying on the weak programs, such as RANDOM. When these weak “prey” died out from being overexploited, their “predators” also died out, unable to compete with the robustness of TFT and programs of the like.
Perhaps the most interesting aspect of Axelrod’s Cooperation book, however, is how we can see these strategies applied in our everyday lives. As we go on through our college careers and see our peers go on to great things, we must remember — do not be envious of the success of other “players”. As we are left with the choice to leave our friends behind and be selfish, we must realize — do not be the first to defect.As we grow into the people we wish to be, we must always remember the harm of overconfidence and overthinking — do not be too clever. And finally, as we are left time after time, hurt or helped by the people around us and we must make our choices about them — reciprocate with both cooperation and defection.
This post was made with Chapter 6 (Games) and Chapter 7 (Evolutionary Game Theory) in mind. Some of the topics that are connected include the Prisoner’s Dilemma, Evolutionary Stable Strategies, and Nash Equilibria.