Just finished reading this wonderful book by Robert Axelrod. I was first introduced to this book by a scientist from Microsoft Research, who gave a talk at IISc, almost entirely focussing on this book.
The crux of the book is what is known as the Iterated Prisoners’ Dilemma, which simulates the way in which fundamentally selfish agents behave. The heart of IPD is a pay-off matrix, which details the pay-offs for mutual co-operation, as well as defection. This presents a long-term incentive for co-operation as well as a short-term incentive for defection. It fits (casts) several real-world problems as an IPD and discusses very interesting scenarios. One of the scenarios that is indeed striking is when during the World War II, there existed a co-operative tie between warring factions, when the two camps used to fire at each other, but deliberately off-target, both indicating the potential for defection as well as the incentive to co-operate. Several interesting scenarios in biology and trade are also highlighted.
Tit for tat
Of all the computer programs that participated in Axelrod’s contest, Tit for Tat (which works by co-operating to begin with and reciprocating the opponent’s previous move in the rest of the moves) was the winner. There was also a second edition of the contest, with all information about the first one being made public, and (not so?) surprisingly, Tit for Tat finished first again! There is a lot of discussion about Tit for Tat in the book.
The Importance of being Nice(st)
Another observation in the results of the contests was that the nicer rules, i.e. the more forgiving rules, mostly performed better. Coincidentally, there was also a paper in Nature, yesterday that discusses the importance of being nice and ‘non-punishing’.
There is a small summary of interesting findings in the book, which I shall try and update soon…
Overall, it’s an amazing book that you would not want to miss out!