@I to @Me: An Anatomy of Username Changing Behavior on Twitter

Paridhi Jain, Ponnurangam KumaraguruarXiv 2014 · arXiv

Abstract

Twitter provides a unique environment where users are identified by a permanent numeric ID but can change their string-based usernames indefinitely. This paper presents the first longitudinal study of this behavior, analyzing a dataset of 8.7 million users. The findings reveal that approximately 10% of users change their handles, with "changers" typically being more active and popular than static users. This behavior is often strategic, aimed at gaining tweet space, attracting followers, or participating in trending events.

Introduction

While research on Online Social Networks (OSNs) frequently examines content evolution and network structure, the temporal dynamics of profile attributes—specifically usernames—remain under-explored. On Twitter, the mutability of usernames introduces several challenges, including identity fragmentation when old links break, username "squatting" by new users on vacated handles, and difficulties in cross-platform identity resolution.

This study seeks to quantify how often these changes happen, who performs them, and the underlying motivations.

Methodology

The researchers employed a two-tier monitoring strategy to capture both broad trends and rapid changes:

  • Macroscopic Scan: Monitoring 8.7 million users (collected via event-monitoring tools) over four distinct scans.
  • Microscopic Scan: Tracking a subset of 10,000 "high-frequency" changers every 15 minutes for 117 days.
  • Correlation & Validation: Using Pearson correlation to link changes to account metrics and conducting qualitative surveys to validate user motivations.

Prevalence and Frequency of Changes

Roughly 10% of the monitored 8.7 million users changed their usernames at least once during the study. For users who change their names multiple times, the behavior is often "bursty."

Distribution of username change frequency among the monitored user base.

Approximately 61% of subsequent changes occur within a 24-hour window of the previous change, suggesting periods of experimentation or immediate tactical needs.

Nature of Username Changes

When users change their handles, they rarely make minor tweaks. Analysis using the Longest Common Subsequence (LCS) shows that 75% of new usernames are significantly different from their predecessors, indicating a desire for a fresh identity rather than simple typo correction.

Cumulative fraction of instances vs. normalized longest common subsequence length, showing high dissimilarity in name changes.

Furthermore, about 36% of users eventually "revisit" a previous handle. This "switching back" is often strategic; 47% of such users saw a gain in followers afterward, suggesting that returning to a known identity can help recover lost popularity.

Strategic Motivations

The data suggests that username changes are frequently driven by utility:

  • Space Optimization: 44% of changes resulted in shorter handles. In a 140-character environment, every character saved provides more room for content and hashtags.
  • Rotational Handles: The researchers identified "rotational" behavior where groups pass high-value usernames between different User IDs to maintain visibility for promotional purposes.
  • Activity Correlation: Changers generally possess higher in-degrees (followers), out-degrees (friends), and tweet counts, indicating that name-changing is a trait of highly engaged users.

Conclusion

Username changing is a significant phenomenon on Twitter, driven by a mix of identity experimentation and strategic optimization. Understanding these patterns is crucial for improving identity resolution systems and developing automated username suggestion tools that align with user history and goals.

Summary updated

BibTeX
@misc{jain2014i,
  author = {Paridhi Jain and Ponnurangam Kumaraguru},
  title = {{@I to @Me: An Anatomy of Username Changing Behavior on Twitter}},
  year = {2014},
  eprint = {1405.6539},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1405.6539},
}
Figure