Social network effects on information aggregation

Abstract

In this thesis, we investigated how sociometric information can be used to improve different methods of aggregating dispersed information. We specifically compared four different approaches of information aggregation: vanilla opinion poll, opinion polls where sociometric data is inferred from the population's own perception of social connectivity, opinion polls where sociometric data is obtained independent of the populations beliefs and data aggregation using market mechanisms. On comparing the entropy of the error of between the prediction of each of these different methods with the truth, preliminary results suggest that sociometric data does indeed improve the enterprise of information aggregation. The results also raise interesting questions about the relevance and application of different kinds of sociometric data as well as the somewhat surprising efficiency of information market mechanisms.