- main fields: Economics of Networks | Applied Game Theory
- secondary fields: Economics of Information | Political Economy
This paper studies learning based on information obtained through social or professional networks. Building on the framework first proposed by DeGroot (1974), agents repeatedly update their beliefs by weighting the information acquired from their peers. The innovation lies in the introduction of dynamically updated weights. This allows agents to weight a contact with poor information little at first, but more later on, if that contact has in the meantime gathered better information from other, more knowledgeable agents. The main finding is that individuals' social influence will depend on both their popularity (as captured by eigenvector centrality) and their expertise (as captured by information precision) in a simple and intuitively appealing way. It is moreover shown that even completely uninformed agents can contribute to social learning, and that under some network structures, providing certain agents with better information could actually lead society to worse assessments. The paper also studies how the relationship between expertise and popularity in a network affects the learning process.