Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

You See It, They Don't: An Exploratory Study of User-to-User Variation in Instagram Comments

Brahmani Nutakki, Manon Lilott Kempermann, Ingmar Weber

Abstract

In March 2025, Meta announced a new AI system to rank the order of the comments shown to Instagram users. With existing research showing how feed personalization systems can lead to increased polarization, the introduction of this new system raises similar questions. This paper presents a small-scale exploratory study examining whether the ranking system produces systematic differences in visible comments shown to different users, particularly for news-related content. Using four sock-puppet accounts varying in gender and political leaning, we collect visible comments on posts from ten news and ten non-news accounts. This collection is repeated twice from two VPN locations to assess location effects. We ask 1) how many visible comments vary across different users, 2) is this variation higher for news accounts than non-news accounts, and 3) can user-attributes like gender, political leaning, and location systematically explain the observed variation. Contrary to our expectations, we find that visible comments on news posts are less likely to vary across users than those on non-news posts. Variation is better explained by account metrics like comment and follower counts than by user attributes. These findings provide an initial glimpse into personalized comment ranking on Instagram and motivate larger, more systematic audits of how comment personalization may shape online discourse. To support further research, we provide the code to collect comments and the data upon request.

You See It, They Don't: An Exploratory Study of User-to-User Variation in Instagram Comments

Abstract

In March 2025, Meta announced a new AI system to rank the order of the comments shown to Instagram users. With existing research showing how feed personalization systems can lead to increased polarization, the introduction of this new system raises similar questions. This paper presents a small-scale exploratory study examining whether the ranking system produces systematic differences in visible comments shown to different users, particularly for news-related content. Using four sock-puppet accounts varying in gender and political leaning, we collect visible comments on posts from ten news and ten non-news accounts. This collection is repeated twice from two VPN locations to assess location effects. We ask 1) how many visible comments vary across different users, 2) is this variation higher for news accounts than non-news accounts, and 3) can user-attributes like gender, political leaning, and location systematically explain the observed variation. Contrary to our expectations, we find that visible comments on news posts are less likely to vary across users than those on non-news posts. Variation is better explained by account metrics like comment and follower counts than by user attributes. These findings provide an initial glimpse into personalized comment ranking on Instagram and motivate larger, more systematic audits of how comment personalization may shape online discourse. To support further research, we provide the code to collect comments and the data upon request.
Paper Structure (18 sections, 5 figures, 1 table)

This paper contains 18 sections, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Data Collection
  3. Sock-Puppet Accounts Creation
  4. Posts Collection
  5. Comments Collection
  6. Empirical Analysis and Results
  7. Descriptive Analysis
  8. Post-level Regression Analysis
  9. Comment-level Regression Analysis
  10. Discussion and Conclusion
  11. Acknowledgments
  12. Ethics Checklist
  13. Appendix
  14. List of Politicians Followed by Sock-Puppet Accounts
  15. Regression Models
  16. ...and 3 more sections

Figures (5)

  • Figure 1: This heatmap shows the average variation proportion of comments across all posts between each pair of crawls. The upper triangle contains values for posts from News Accounts, and the lower triangle for those from Non-News Accounts.
  • Figure 2: Posterior mean predictions (95% HDI) from the beta-binomial regression across categorical and continuous covariates, holding other predictors constant. The categories with ' are the reference categories, and * indicates categories whose credible interval for the odds ratio excludes 1.
  • Figure 3: Forest plot of Bayesian logistic regression coefficients shown as odds ratios for Post-Level Regression. The dashed vertical line at 1 marks “no effect,” and the right panels report diagnostics for each parameter.
  • Figure 4: Forest plot of Bayesian logistic regression coefficients shown as odds ratios for Comment-Level Regression. The dashed vertical line at 1 marks “no effect,” and the right panels report diagnostics for each parameter.
  • Figure 5: Posterior mean predictions for Comment-Level Regression across categorical and continuous covariates, holding other predictors constant. The categories with ' are the reference categories, and * indicates categories whose credible interval for the odds ratio excludes 1.