Aggregation of Expert Opinions

Conflicts of interest arise between a decision maker and agents who have information pertinent to the problem because of differences in their preferences over outcomes. We show how the decision maker can extract the information by distorting the decisions that will be taken, and show that only slight distortions will be necessary when agents are "informationally small." We further show that as the number of informed agents becomes large the necessary distortion goes to zero. We argue that the particular mechanisms analyzed are substantially less demanding informationally than those typically employed in implementation and virtual implementation. In particular, the equilibria we analyze are "conditionally" dominant strategy in a precise sense. Further, the mechanisms are immune to manipulation by small groups of agents.