Analysis of Surprisingly Popular Voting for Opinion Aggregation on Social Networks

This paper investigates how the structure of social networks affects an opinion aggregation method, surprisingly popular (SP) voting. SP leverages respondents’ metacognition: estimation of the rarity of their opinions among those of all respondents. Since a respondent’s metacognition is considered to be shaped by people around them, we can better understand metacognition and SP voting performance by focusing on the structure of social networks. We analyzed the effect of respondents’ referring to connected individuals on SP voting performance when they predict other respondents’ opinions. We also analyzed the effect of the structure of social networks on respondents’ metacognition and on SP voting performance. Simulation experiments using various social networks revealed that respondents’ referring to connected individuals is important in SP voting performance and that the structure of social networks plays an important role in respondents’ metacognition and SP voting performance. Even if respondents refer to connected individuals, their metacognition can differ from the actual distribution of opinions across all respondents depending on the structure of the social networks, which greatly affects SP voting performance.