Dimitri- Basili-Gilboa eds
Cognitive Processes and Economic Behavior
By using an iterative notion of common knowledge, Rubinstein drew attention to the (possibly puzzling) fact that ‘almost common knowledge’ of a relevant game might not be enough for the efficient, but risky, joint course of actions to be played as a Nash Equilibrium (NE) of the game. Monderer-Samet (1989) argued in support of an alternative notion of approximate common knowledge, viewed as common p-beliefs, when p is sufficiently close to one. With this alternative idea they were able to prove a general result in which the above mentioned action profile could be played, with sufficiently high probability, when players are ‘approximately’ best repliers.