Sender-receiver games David Lewis (convention) Crawford & Sobel Set of states, set of messages, set of acts Sender has private information about state of world selects a message receiver can choose an act on the basis of the message Sender & receiver has common interests (Lewis) Crawford & Sobel drop that assumption act that is 'right' for each state in that if R does it, both S and R get payoff of 1 So if sender accurately transmits state, receiver will do the right act But there is no initial meaning attached to (elements of) messages Game of incomplete information... Nature chooses a state, sender chooses message, receiver chooses action Blume, DeJong, Kim & Sprinkle 1998 economics lab version of S-R game "Experimental evidence on the emergence of meaning" Replicator dynamics (Hofbauer & Zigmund) Simon Huttinger UCI Salzburg Herrnstein 1970 "On the law of Effiect" Roth Erev "learning in Extensive Form Games" Alan Beggs 2005 on the convergence of reinforcement learning Babbling Equilibria with all Zero eigenvalues of the jacobian http://crookedtimber.org/2004/08/03/babbling-taxis/ games with too few or too many signals...