Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Why do people think liberals drink lattes? How social media afforded self-presentation can shape subjective social sorting

Samantha C. Phillips, Kathleen M. Carley, Kenneth Joseph

TL;DR

This paper investigates how social media self-presentation, via public bios, can foster subjective social sorting that aligns non-political identifiers with political identities. Using a bottom-up analysis of bios from roughly 50 million X users across five timepoints (2016–2018), the authors construct co-occurrence networks of identifiers and apply Bayesian-adjusted log-odds comparisons to quantify alignment along left/right and pro-/anti-establishment axes. They find that about 9.2% of non-political identifiers show significant alignment with political identities, with strong patterns for religion, military, activism, family, and gender signals, and notable temporal shifts linked to the rise of anti-establishment right identities during the Trump era. The results imply that self-presentation on social media can contribute to perceived polarization and offer guidance for depolarization strategies and algorithmic interventions, while recognizing limitations related to exposure measurement and causality.

Abstract

Social sorting, the alignment of social identities, affiliations, and/or preferences with partisan groups, can increase in-party attachment and decrease out-party tolerance. We propose that self-presentation afforded by social media profiles fosters subjective social sorting by shaping perceptions of alignments between non-political and political identifiers. Unlike previous work, we evaluate social sorting of naturally occurring, public-facing identifiers in social media profiles selected using a bottom-up approach. Using a sample of 50 million X users collected five times between 2016 and 2018, we identify users who define themselves politically and generate networks representing simultaneous co-occurrence of identifiers in profiles. We then systematically measure the alignment of non-political identifiers along political dimensions, revealing alignments that reinforce existing associations, reveal unexpected relationships, and reflect online and offline events. We find that while most identifiers bridge political divides, social sorting of identifiers along political lines is occurring to some degree in X profiles. Our results have implications for understanding the role of social media in facilitating (the perception of) polarization and polarization mitigation strategies such as bridging interventions and algorithms.

Why do people think liberals drink lattes? How social media afforded self-presentation can shape subjective social sorting

TL;DR

This paper investigates how social media self-presentation, via public bios, can foster subjective social sorting that aligns non-political identifiers with political identities. Using a bottom-up analysis of bios from roughly 50 million X users across five timepoints (2016–2018), the authors construct co-occurrence networks of identifiers and apply Bayesian-adjusted log-odds comparisons to quantify alignment along left/right and pro-/anti-establishment axes. They find that about 9.2% of non-political identifiers show significant alignment with political identities, with strong patterns for religion, military, activism, family, and gender signals, and notable temporal shifts linked to the rise of anti-establishment right identities during the Trump era. The results imply that self-presentation on social media can contribute to perceived polarization and offer guidance for depolarization strategies and algorithmic interventions, while recognizing limitations related to exposure measurement and causality.

Abstract

Social sorting, the alignment of social identities, affiliations, and/or preferences with partisan groups, can increase in-party attachment and decrease out-party tolerance. We propose that self-presentation afforded by social media profiles fosters subjective social sorting by shaping perceptions of alignments between non-political and political identifiers. Unlike previous work, we evaluate social sorting of naturally occurring, public-facing identifiers in social media profiles selected using a bottom-up approach. Using a sample of 50 million X users collected five times between 2016 and 2018, we identify users who define themselves politically and generate networks representing simultaneous co-occurrence of identifiers in profiles. We then systematically measure the alignment of non-political identifiers along political dimensions, revealing alignments that reinforce existing associations, reveal unexpected relationships, and reflect online and offline events. We find that while most identifiers bridge political divides, social sorting of identifiers along political lines is occurring to some degree in X profiles. Our results have implications for understanding the role of social media in facilitating (the perception of) polarization and polarization mitigation strategies such as bridging interventions and algorithms.
Paper Structure (31 sections, 2 equations, 2 figures, 12 tables)

This paper contains 31 sections, 2 equations, 2 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Data
  4. Methods
  5. Identifier extraction
  6. Identifier categorization
  7. Alignment of identifiers with political identities
  8. Ethics
  9. Results
  10. Overview of political identification on X between 2016 and 2018
  11. Comparison of activity and influence
  12. Bias and variance of alignments over time
  13. Alignments of individual identifiers
  14. Religion
  15. Occupations/industry; military
  16. ...and 16 more sections

Figures (2)

  • Figure 1: Number of users with political identities in their bio per time period. Division of identifiers selected to maximize clarity.
  • Figure 2: Mean (top row) and standard deviation (bottom row) of differences in weighted log-odds between Right and Left (left column) and Pro- and Anti-establishment (right column). Highlighted types of identifiers correspond to discussed results.