Identifying the Best Agent in a Network

Abstract:
The paper develops a mechanism for a principal to allocate a prize to the most valued agent when agents have a knowledge network. The principal does not know any agent’s value but any two linked agents know each other’s values. Agents compete for the prize and send costless private messages about their own value and the values of others they know to the principal. Agents can lie only to a certain extent and only lie if it increases their chances of winning the prize. A mechanism that determines each agent’s chances of winning for any possible message profile is proposed. We show that with this mechanism, there exists an equilibrium such that the most valued agent wins with certainty if every agent has at least one link. We provide sufficient conditions on the network such the most valued agent wins with certainty in every equilibrium, and also provide an example of a network for which there exists an equilibrium such that the most valued agent does not win with certainty.

Full details of this seminar series are available at the following link:
www.davidronayne.net/lgn-seminar