Schulz, W., Robertson, C. E., Sjåstad, H., Ashokkumar, A., Oswald, L., Yang, J., Munzert, S., Bavel, J. J. V., & Guess, A. M. (2026). Online Discourse is Falsely Polarized by Self-Selection. https://doi.org/10.31234/osf.io/7z69m_v1
Summary
This paper argues that online political discourse appears more polarized than public opinion actually is — a phenomenon the authors call “false polarization” — and traces its origin to self-selection into posting. Using more than a decade of linked social media behavior, controlled experiments, and a 40-country survey, the authors show that people with stronger or more extreme ideological attitudes are disproportionately likely to post about political issues. The visible discourse is therefore a skewed sample of the underlying population, producing a systematic distortion that makes online conversation seem more divided than it really is. The distinction is important: the polarization is an artifact of who chooses to speak, not necessarily of what people believe.
Key Contributions
- Provides comprehensive, multi-method evidence quantifying the magnitude and scope of false polarization on social media.
- Identifies selective posting (self-selection) as a specific mechanism generating apparent rather than actual polarization.
- Links large-scale behavioral data on over 1.7 million posts to individual-level ideological attitudes.
- Establishes cross-national generalizability through a 40-country survey.
Methods
The evidence draws on three complementary strands: (1) an observational analysis of over a decade of social media behavior linked to individual ideological attitudes, spanning more than 1.7 million posts; (2) a series of controlled experiments; and (3) a large global survey covering 40 countries. The combined sample totals N = 49,222, allowing triangulation across observational, experimental, and cross-national designs.
Findings
- Individuals with stronger ideological attitudes are more inclined to post political content, skewing the visible discourse.
- The observed polarization in online discourse exceeds the actual polarization of opinions in the underlying population.
- These patterns hold consistently across observational data, experiments, and the cross-national survey.
- The distortion is broad in scope, not confined to particular platforms, issues, or countries.
Connections
This work connects to research on the gap between perceived and actual polarization and the misperception of public opinion, complementing studies of affective and perceived polarization such as Voelkel2026-lc and Bergeron-Boutin2026-yh. Its emphasis on how a self-selected minority of active posters shapes visible discourse relates to accounts of who produces and amplifies political content online, including Green2025-ap and Falkenberg2026-ka. The self-selection mechanism also speaks to broader questions of algorithmic and behavioral filtering of exposure examined in Bakshy2015-rn.