Infinite Regression in Strategic Decision Making : An Application of Rice's Theorem

A model of decision-making in a strategic environment is developed, where boundedly rational players care about both implementation complexity costs and computational complexity costs. The choice of strategies is carried out by an algorithm, which itself is seen as the choice of a higher level algorithm. And so on up a hierarchy of algorithms. It is shown that the choice of ‘optimal’ algorithms at each level in this infinite hierarchy is not, in general, computationally feasible