Partisan biases in social information use

Preferential learning from like-minded others can help individuals acquire adaptive knowledge and socially appropriate behaviour, but it can also reinforce echo chambers and fuel polarization. Ingroup bias is well-documented in the social transmission of opinions, attitudes, and values. However, important questions about its role in the integration of social information when forming factual beliefs are outstanding. We present a naturalistic yet controlled experiment showing that social information is most impactful when provided by ingroup rather than outgroup sources. Participants predicted the 2020 US elections by state and could adjust their predictions after observing the prediction of a Democrat or a Republican. Adjustments were largest when observing fellow-partisans, and when social information favoured the participants’ party. Exploratory analyses reveal that these partisan biases are driven by Republican participants. Our findings help understand the variation of social information use along political orientations and its consequences for belief polarization in increasingly fragmented populations.