Communication and Common Interest

Explaining the maintenance of communicative behavior in the face of incentives to deceive, conceal information, or exaggerate is an important problem in behavioral biology. When the interests of agents diverge, some form of signal cost is often seen as essential to maintaining honesty. Here, novel computational methods are used to investigate the role of common interest between the sender and receiver of messages in maintaining cost-free informative signaling in a signaling game. Two measures of common interest are defined. These quantify the divergence between sender and receiver in their preference orderings over acts the receiver might perform in each state of the world. Sampling from a large space of signaling games finds that informative signaling is possible at equilibrium with zero common interest in both senses. Games of this kind are rare, however, and the proportion of games that include at least one equilibrium in which informative signals are used increases monotonically with common interest. Common interest as a predictor of informative signaling also interacts with the extent to which agents' preferences vary with the state of the world. Our findings provide a quantitative description of the relation between common interest and informative signaling, employing exact measures of common interest, information use, and contingency of payoff under environmental variation that may be applied to a wide range of models and empirical systems.Author Summary: How can honest communication evolve, given the many incentives to deceive, conceal information, or exaggerate? In recent work, it has often been supposed that either common interest between the sender and receiver of messages must be present, or special factors (such as a special cost for dishonest production of signals) must be in place. When talk is cheap, what is the minimum degree of common interest that will suffice to maintain communication? We give new quantitative measures of common interest between communicating agents, and then use a computer search of signaling games to work out the relationship between the degree of common interest and the maintenance of signaling that conveys real information. Surprisingly, we find that informative signaling can in some cases be maintained with zero common interest. These cases are rare, and we also find that the degree of common interest is a good predictor of whether informative signaling is a likely outcome of an interaction. The upshot is that two agents with highly incompatible preferences may still find ways to communicate, but the more they see eye-to-eye, the more likely it is that communication will be viable.