Institute for Mathematical Behavioral Sciences

This work explores one approach to bring together evolutionary game theory and rational actor theory and thereby create a formal model of how preferences in populations of rational actors will form in the presence of selective forces. Game theory can not judge or predict preferences, but only determine the strategies actors will choose, given their preferences. Therefore, these preferences must be set by some other theory. The most common method is to assume the existence of a selection mechanism. From this springs the intuition of directness: The actors that survive the selection mechanism must be those that have utility for outcomes directly in proportion to the fitness of those outcomes. This work correctly formalizes the model underlying this intuition and shows that this intuition is very wrong. Instead I derive analytical solutions, confirmed by numerical simulations, of the preferences that will emerge in populations of rational actors. The equilibrium utility structures, called manipulative preferences due to their ability to force opponents to take favorable actions, are stable, and attractive, and can not be invaded (even by direct preferences). Moreover, they increase the aggregate fitness level (welfare) of the population, compared to direct preferences. In addition, I show how the base game to this model is also the solution to a different but connected problem of how Principals should choose Agents, when their Agents are faced by game theoretic rather than decision theoretic tasks. The fallacy of directness has an analogous error in the Principal-Agent literature that optimal Agents should mirror their Principals, that is, have the same preference structure. I show how the solutions of the previous model also solve for the utility structure of optimal Agents when they are faced with game theoretic tasks. This has application in Principal-Agent models, spatial voting models, and ideas of representation by sampling in democratic theory.