BRIAN SKYRMS Signals: Evolution, Learning, and Information

In 1969, David Lewis used the nascent theory of games to argue that semantic meaning can be conventional. Imagine a situation in which the world can be in one of two states. An agent, the sender, observes the actual state and then sends a message to a second agent, the receiver. The receiver cannot directly observe the state but does perform some action after observing the sender’s message. Both players are paid off if and only if the receiver performs an action appropriate for the state of the world. Lewis noted that if, for example, signal A is sent only in state one, and the receiver responds to signal A by performing the action appropriate for state one, then it looks like signal A has come to mean something like ‘state one has occurred’ or ‘perform the action appropriate for state one’. However, because Lewis was limited to techniques available at the time, he was unable to provide a thorough account of how agents adopt conventions. In single chapters of earlier books, Skyrms ([1996], [2004]) introduced an approach for joining Lewis’s signalling game with modern approaches to game theory, namely, those based on evolutionary dynamics and theories of Brit. J. Phil. Sci. 64 (2013), 883–887