Collective wisdom in polarized groups

Abstract

The potential for groups to outperform the cognitive capabilities of even highly skilled individuals, known as the “wisdom of the crowd”, is crucial to the functioning of democratic institutions. In recent years, increasing polarization has led to concern about its effects on the accuracy of electorates, juries, courts, and congress. While there is empirical evidence of collective wisdom in partisan crowds, a general theory has remained elusive. Central to the challenge is the difficulty of disentangling the effect of limited interaction between opposing groups (homophily) from their tendency to hold opposing viewpoints (partisanship). To overcome this challenge, we develop an agent-based model of collective wisdom parameterized by the experimentally-measured behaviour of participants across the political spectrum. In doing so, we reveal that differences across the political spectrum in how individuals express and respond to knowledge interact with the structure of the network to either promote or undermine wisdom. We verify these findings experimentally and construct a more general theoretical framework. Finally, we provide evidence that incidental, context-specific differences across the political spectrum likely determine the impact of polarization. Overall, our results show that whether polarized groups benefit from collective wisdom is generally predictable but highly context-specific.

Communities of Knowledge in Trouble

“…Urgent theoretical questions for future investigation are (a) which cues take precedence when multiple, conflicting cues are present, as when the political activist group America’s Frontline Doctors spoke out against COVID-mitigation measures while dressed to signal their status as practicing physicians and hence members of the medical-scientific community (Frenkel & Alba, 2020); (b) how individuals navigate the fact that they belong to multiple, partially overlapping communities (Rabb, Bowers, et al, 2022), as when members of one partisan group live in an area heavily populated by the other; and (c) whether the presence of mutual distrust, a contingent fact about the United States during COVID-19 (Iyengar et al, 2019) that likely exacerbated the problem (Druckman et al, 2021), or network separation is necessary for communities to come to opposite beliefs. Advances in hybrid modeling/behavioral experiment designs (e.g., Bak-Coleman et al, 2022) are likely to provide traction on these questions.…”