Learning from Shared News: When Abundant Information Leads to Belief Polarization*.

The article investigates how belief polarization can arise from learning via shared news within social networks, focusing on the interplay between selective news sharing, misperception of shared information, and information quality. It develops a theoretical model where agents receive unbiased firsthand signals about a binary state and share these signals selectively with friends—some dogmatic agents share only signals supporting one state, while normal agents share all signals. Crucially, agents misperceive the sharing process by underreacting to friends’ silence, treating it too much as absence rather than suppression of news, which combined with sufficiently low information quality, can cause beliefs to diverge and polarize even without fake news or media bias. The model shows that polarization can worsen as friend networks expand and that news aggregators, by increasing effective information quality through summarization, can mitigate polarization. Applying the theory to U.S. climate change opinion data, the article suggests that modest misperception and small imbalances in friend networks, alongside low-quality information possibly due to skepticism campaigns, explain observed partisan polarization, which accelerated with the rise of social media.