Reducing Social Media Misinformation Doesn’t Change Political Beliefs, Study Finds

Does misinformation on Facebook and Instagram actually alter user beliefs? A large-scale academic study partnering with Meta during the 2020 presidential election revealed that while reducing exposure to untrustworthy sources by roughly 70% succeeded in changing feed diets, it did not measurably shift political attitudes or beliefs among users.

Feeds Changed, But Beliefs Remained Static

UC Berkeley Political Science Ph.D. student Olivier Bergeron-Boutin tested that exact question on Facebook and Instagram during the 2020 presidential election.

The researchers successfully altered algorithmic delivery to reduce feed exposure to content from sources that repeatedly spread misinformation. The intervention achieved its technical objective: participants assigned to the intervention group were exposed to roughly 70% less content from untrustworthy sources over a three-month period surrounding the election. However, their political beliefs and attitudes did not measurably change. This held true even among users who were previously exposed to the highest volumes of content from those specific sources.

The Concentration of Untrustworthy Content

The study also uncovered significant structural realities about how unverified information distributes across major social networks. Exposure to untrustworthy sources was far from evenly spread across Facebook and Instagram populations. Most users on both platforms saw relatively little content originating from these domains.

Instead, a small, highly active cohort of users accounted for the vast majority of exposure.

Photo: ls.berkeley.edu

Nyhan added that reducing exposure to content from those sources did not trigger the effects that many speculated it would have on people’s beliefs, attitudes and behaviors.

“Stories don’t spread evenly through a population,” Tangherlini said. He emphasized that narratives flow across social networks that are already predisposed to receive, retell, and refine them, functioning essentially like an echo chamber.

Implications for Platform Regulation and Political Polarization