Note on the Harris-Rapoport Controversy

Abstract

I have been working for four years on the development, interpretation and application of metagame theory, and I would like, if I may, to refer readers of this controversy to my book Games, Metagames and Rationality, which will be published by M. I. T. Press some time in 1970. Here I feel reluctantly obliged to say that both Harris (1969) and Rapoport (1970), in my view, miss the point. A valid interpretation of metagame theory applied to the one-shot game requires remembering two things which they forget. ( 1 ) The theory is an equilibrium theory. It does not predict what players will do, but only what they will do under conditions of stability, i.e., when each somehow succeeds in predicting the other's choice. These stability conditions are not, as Harris always and Rapoport sometimes seem to suppose, a requirement laid down by the rules of the game. They are a requirement laid down by the theoreticirrn because he is interested in certain actual outcomes only-those which fulfill the stability conditions. But stability in a game situation, where what a player chooses will generally depend on what he predicts the other will choose, is not a simple condition. Its complexity is modelled by the infinite tree of metagames based on a given game. ( 2 ) The theory is predictive, tested by oneshot game experiments, and not in the least nosmatiue. The fact that in Prisoner's Dilemma it finds two equilibria-one at mutual cooperation and one at mutual defection, each of which has its own rationale, shows that it is incapable of producing recommendations or advice to the players as to what they should do. It merely predicts that if the outcome is stable, i.e., predicted by both players, it will be one of these two equilibrium outcomes. With reference to ( 2 ) , let us note that in the early stages of its development nearly every scientific theory has tried to be three things at once: it has tried to be Logically necessary (deducible from logic alone), normative (productive by itself of recommendations or guidance as to what we should do) , and empirical. After much mental effort and revolutionary reconceptualization, each mature science has finally decided that it can only be the third of these three things. Astronomy, physics, evolutionary theory, economicseach has followed this course; and I believe that the theory of human conflict and cooperation must follow it too.