Computing agents' reputation within a network

Abstract

We propose a model of information transmission and reputation building within a social network that exploits portfolio theory and option structures. The network aims to estimate an unknown parameter through multiple communication rounds. At every communication round, estimates of different agents' abilities are shared, avoiding the repetition of information. These estimates are interpreted as financial assets driven by a compound Poisson process. After every communication round, agents construct a fictitious portfolio of options whose underlying is the vector of shared estimates. The portfolio's weights are exploited to aggregate the information received in the communication round. Sufficient conditions for reaching consensus or polarization are provided.