Identity diversification and homogenization: evidence from frequent estimates of similarity of self-authored, self-descriptive text

For more than a decade, individuals composed and edited self-authored self-descriptions as social media biographies. Did these identities become more diverse over time because of a “rise in individualism” and increasing tolerance or did they become more homogenous through social learning, conformity, and fear of isolation? We analyzed longitudinal and cross-sectional Twitter bio samples with a variety of lexical and semantic methods for the 2012–2022 interval. We show that longitudinally, users diversified on lexical and semantic levels. On a cross-sectional sample—representing the state of the platform at any time point—we again observed a trend of diversification at the lexical level, but a trend of diversification reversed toward re-homogenization on the semantic level. Further, by focusing on local maxima and minima of identity similarity we identified “coordination shocks”—temporally confined intervals where similar users became overactive on the platform and drove short-term deviations from longer-term trends.