David Lewis in the lab: experimental results on the emergence of meaning

David Lewis famously pioneered a game-theoretic approach in which meaning is possible in signaling games if the agents can converge to a signaling system. In a signaling system success is guaranteed for the receivers reliably infer the correct state of the world from the message sent by the sender. Yet how likely are such outcomes? We employ a series of laboratory experiments to address this question. Specifically, we rely on the methodology of experimental economics to design an experiment that can precisely assess whether learning from experience can lead to a signaling system of a Lewis signaling game. We study a simple signaling game with two states, two messages and two acts and explore two treatments in which nature is biased (i.e. the two states are not equiprobable). We then investigate a slightly more complicated variant of this game where there are three acts, three messages and three states. We find that by and large our experimental results fall in line with the fairly robust qualitative predictions of the theoretical literature.