Are you a player or a strategy?

Traditional game theory assumes that individuals choose strategies.  Evolutionary game theory assumes that individuals are strategies.

Our colleague, Dan Houser, has just published an important new paper (co-authored with Robert Kurzban) that supports the assumption of evolutionary game theory. (The paper is also featured in the Economist.)  In a public goods game, Kurzban and Houser are able to identify three systematic strategies; cooperate, free ride and reciprocate (cooperate more when others do so).  The first surprising result of their paper is that these strategies can be tied to specific individuals.  Some individuals cooperate, others free ride and others reciprocate and the strategies that these individuals "choose" are stable.  (This doesn’t mean that individuals play strategies robotically regardless of context a better analogy may be to think of strategies like personalities – even a quiet person can yell sometimes.)

Strategy choice is so stable that Kurzban and Houser can create very cooperative groups simply by weeding out the free riders.  What is even more surprising, however, is that when individuals are randomly assigned to groups each strategy type earns about the same payoff.  Even though the strategies are very different, no strategy dominates the others in a randomly assigned group – this is exactly what one would predict if individuals are strategies in an evolutionary game. 

Kurzban and Houser’s paper raises a number of interesting questions.  First, if they can pick out free riders why can’t the group members do so – in which case the baseline set of strategies won’t be an ESS.  How much does it take to push people away from their natural strategies?  Is the proportion of the population that plays each strategy the same in different cultures?  And what do the proportions (13% co-operators, 20% free-riders and 63% reciprocators) tell us about conditions in the evolutionary era when such strategies first solidified?  I will take these up with Dan.


Comments for this post are closed