Dynamics of Information Exchange in Endogenous Social Networks

Most individuals form their opinions about the quality of new products, social trends and political issues via their interactions in social and economic networks. While the role of social networks as a conduit for information is as old as humanity, recent social and technological developments, such as blogging and online social networking, have added further to the complexity of network interactions and created exciting new opportunities for businesses. Despite the ubiquity of social networks and their importance in communication, we know relatively little about how opinions form and how information is spread in such networks. A solid understanding of the dynamics of information exchange and the role of new technologies in large interconnected societies is essential for tomorrow’s businesses, as it has a profound impact to a wide range of their activities.

In the first part of the talk, I describe a novel game-theoretic framework of information exchange among individuals, which are embedded in a social network. A new product, whose unknown quality is captured by an underlying state of the world, is introduced into the market. Individuals form beliefs about the realization of the state and engage in exchange of information with their peers in the social network. Their beliefs evolve dynamically by incorporating their private information as well as the information they receive from their peers, until they take an irreversible action (purchase the product). The timing of the action is endogenous, i.e., individuals not only choose which action to take but also when to take it. In particular, delaying taking an action leads to potentially better decisions, as the individual obtains more information from the social network. However, this additional information comes at a cost, since future payoffs are discounted. Our results provide tight characterizations of conditions under which asymptotic learning occurs, i.e., the fraction of agents that take the correct action converges to one, as the society grows large. Moreover, I extend the framework by incorporating network formation. In particular, I assume that individuals can access different sources of information at a known cost and study how this active pursuit for information impacts asymptotic learning. The model illustrates an interesting dimension of strategic behavior: individuals delay taking an action not in the anticipation of a markdown but in the hope of acquiring more information.

In the second part, I describe recent work on optimal pricing in a market with local network effects. A divisible good (service) is offered to a set of individuals. Consumers derive positive utility from their peers’ consumption of the good. I first characterize the optimal pricing for a monopolist that can perfectly price discriminate the consumers. I show that the optimal prices are proportional to their Bonacich centrality, a measure of their influence in the network. Then, I consider a monopolist that cannot price discriminate, but has full knowledge of the network structure. Interestingly, I show that the optimal pricing algorithm for the monopolist takes the form of a simple index policy based on a centrality ranking of the consumers. Finally, I conclude with discussing directions for future research.

Based on joint work with Prof. Daron Acemoglu (MIT Economics), Prof. Asu Ozdaglar (MIT EECS) and Ozan Candogan (MIT EECS)