Title: Metawisdom of the Crowd: How Choice Within Aided Decision Making Can Make Crowd Wisdom Robust

Abstract: Quality information can improve individual judgments but make group decisions less accurate; if individuals attend to the same information, the predictive diversity that underlies crowd wisdom may be lost. We explore this tension within the context of decision support systems that provide the choice of decision aids and before then primary judgments. We argue that whenever a set of decision aids induce diverse errors, this structure leads to higher group accuracy because aid choice will exhibit predictive diversity itself. In two experiments—the prediction of inflation (N=1907, pre-registered) and a tightly controlled bean-count estimation task (N=1198)—we find strong evidence for this.