Anmerkungen:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments September 21, 2021 erstellt
Beschreibung:
Wisdom of crowds can be used to improve forecasting accuracy by integrating multiple perspectives and averaging out error. Small crowd selection can improve crowd wisdom by balancing the tradeoff between removing poorly performing individuals and adding many judges to mediate error. Previous crowd selection methods rely on non-parametric approaches, which do not consider all possible subsets of the crowd. We propose a new parametric method for small crowd selection by decomposing the Expected Squared Error of aggregated judgment and introduce an efficient algorithm to consider more possible subsets. We test the performance of our method against previous methods and conclude that performance depends on the environmental context and the statistical properties of individuals’ judgments. Using our parametric structure, we analyze which methods perform the best in different environments, providing insights into how to choose the best crowd selection method for any given crowd