Communication and Games

The role of costless, pre-play communication (called "cheap talk" in the economics literature) on agents' behavior in strategic settings has been the focus of a substantial (and increasing) body of literature. Most researchers focus on either the game-theoretic implications of costless communication or on experimental results usually demonstrating that communication can serve to (1) break symmetry in some games or (2) reassure partners as to an intended action. Some questions that remain unanswered are: "What are the minimal foundations for agents to use communication? How do (or _can_) agents use communication when playing games?"

Specifically, this line of research intends to examine the role of communication in both the outcome and the dynamics of game play, and how this behavior compares to both traditional game-theoretic analysis and empirical findings. However, the manner in which communication and interaction are modeled are compared in search of foundational answers to questions of impact; by what mechanisms does communication operate to alter game play?

In Miller and Moser (2003), agents may bilaterally exchange a sequence of a priori meaningless messages prior to playing a Stag Hunt game (a classic game which provides a good setting to investigate issues of coordination, efficiency and risk). We examine the role of communication in achieving coordination. In particular, does endowing agents with the ability to communicate lead to more favorable outcomes? To pursue this question we employ an adaptive model of strategically communicating agents (Miller et al. [7]) playing the Stag Hunt game.

We find that communication plays a key role in the ability of agents to reach and maintain superior coordination. In the absence of communication, agents tend to get trapped at the inferior coordination point. However, once agents reach a particular strategic threshold, sending even a priori meaningless messages serves to increase the likelihood that the population will coordinate on the superior outcome. While the system spends the majority of its time with well-coordinated behavior, it is not static; such periods are often punctuated by brief transitions in which the system switches to the alternative coordination point. We analyze the various mechanisms that account for this dynamic behavior and find that there are a few critical pathways by which the system transitions from one coordination point to another. Communication plays a critical, yet short-lived, role in one key pathway.

Our analysis suggests that giving agents the ability to communicate even a priori meaningless messages may promote the emergence of a rich, and often robust, "ecology" of behaviors that allows agents to achieve new, and in this case superior, outcomes.