|M.Sc Student||Wagner Jonatan|
|Subject||Informativity vs.Conformity: Opinion Sharing in|
|Department||Department of Industrial Engineering and Management||Supervisors||Professor Aharon Ron Lavi|
|Professor Itai Arieli|
We study opinion sharing (“voting”) processes on acyclic graphs. Individuals sequentially express their opinion, and care both about expressing their true opinion as well as about being similar to their graph neighbors. We define and analyze the informativity of the resulting game, which is the proportion of voters who express (in equilibrium) their true opinion. Our main interest is in understanding how the informativity depends on the voting order. We show that the natural “diffusion” order in which a node votes (only) after at least one of its neighbors votes has poor informativity when the number of leaves in the graph is large. We then show a slight but careful modification of this orderthat in many cases significantly improves informativity. A key technical building block in our analysisis the case of star graphs. This case turns out more complicated than one would initially expect, with some surprising results.