Peter Vanderschraaf (University of California, Merced)
Conventions are often analyzed via an equilibrium-in-beliefs approach or a dynamical limit approach. A purely equilibrium-in-beliefs analysis can only explain the origins of a convention with reference to higher-order coordinated beliefs, threatening a vicious regress. A dynamical limit analysis at best explains the origins of basic conventions of one-shot games, while many important conventions are history-dependent conventions of repeated games. I argue that a new theory of convention is called for where coordinated beliefs and dynamical updating interact. I show by example how a convention can be both an equilibrium-in-beliefs and the limit of a dynamical inductive learning process.
