Online communities bring together participants from diverse backgrounds and often face challenges in aggregating their opinions. We infer lessons from the experience of individual contributors to Wikipedia articles about U.S. politics. We identify two factors that cause a tendency toward moderation in collective opinion: Either biased contributors contribute less, which shifts the composition of participants, or biased contributors moderate their own views. Our findings show that shifts in the composition of participants account for 80%–90% of the moderation in content. Contributors tend to contribute to articles with slants that are opposite their own views. Evidence suggests that encountering extreme contributors with an opposite slant plays an important role in triggering the composition shift and changing views. These findings suggest that collective intelligence becomes more trustworthy when mechanisms encourage confrontation between distinct viewpoints. They also suggest, cautiously, that managers who aspire to produce content “from all sides” should let the most biased contributors leave the collective conversation if they can be replaced with more moderate voices. 

Greenstein, Shane, Grace Gu, and Feng Zhu. 2021. "Ideology and Composition Among an Online Crowd: Evidence From Wikipedians." Management Science, 67(5): 3067-3086.