Effect of Distribution Replication Code
Code to replicate the paper "The accuracy benefits of group discussion depend on initial belief distribution" by Joshua Becker, Abdullah Almaatouq, and Emőke-Ágnes Horvát.
Data from Gurcay et al. (2015) uploaded to this repository with permission from authors.
Abstract: A critical problem in social science is to determine how social processes impact a group’s accuracy in estimation tasks such as financial forecasting, medical diagnoses, and hiring assessments. While prior research suggests that communication improves accuracy, these analyses assume that the effect of communication is the same for all tasks. In contrast, we find substantial heterogeneity in the effect of communication across different tasks. Our re-analysis of 608 experimental trials from 4 previously published experiments shows that the statistical distribution of pre-communication estimates, which varies between tasks, substantially impacts the effect of communication on the accuracy of the aggregate (i.e., mean) belief. When communication is limited to sharing numeric estimates, communication consistently increased the accuracy of the mean belief, but the initial distribution moderated the strength of that effect. However, in experiments where people engaged in open-ended discussion, communication sometimes increased accuracy and sometimes decreased accuracy, depending on initial belief distribution. These results call for revised theoretical models of social belief formation.