Refugees of the Digital Space: Platform Migration from TikTok to RedNote

TL;DR

The paper investigates how ban-induced migration from TikTok to RedNote unfolds as a cross-cultural platform migration problem under algorithmic governance. It adopts a multi-method framework combining sentiment analysis, topic modeling, and an entropy-weighted influence score to compare high- and low-influence users across three migration phases: Pre-Ban, Refugee Surge, and Stabilization. A key methodological contribution is the entropy-based scoring, where $S_i = \sum_j w_j x_{ij}$ with $w_j = d_j / \sum_j d_j$ and $d_j = 1 - H_j$, $H_j = - (1/\log n) \sum_i p_{ij} \log(p_{ij})$, enabling robust cross-phase, cross-group comparisons of content strategy and affective expression. Findings reveal that while core topics remain stable, high-influence users adopt more culturally resonant or commercially oriented content, political discourse is selectively activated, and sentiment patterns diverge by influence level; together these patterns reflect how platform migration is shaped by structural affordances and user agency. The study advances theories of platform society, affective publics, and transnational digital environments, with implications for policy and platform governance in migratory digital publics.

Abstract

In January 2025, the U.S. government enacted a nationwide ban on TikTok, prompting a wave of American users -- self-identified as ``TikTok Refugees'' -- to migrate to alternative platforms, particularly the Chinese social media app RedNote (Xiaohongshu). This paper examines how these digital migrants navigate cross-cultural platform environments and develop adaptive communicative strategies under algorithmic governance. Drawing on a multi-method framework, the study analyzes temporal posting patterns, influence dynamics, thematic preferences, and sentiment-weighted topic expressions across three distinct migration phases: Pre-Ban, Refugee Surge, and Stabilization. An entropy-weighted influence score was used to classify users into high- and low-influence groups, enabling comparative analysis of content strategies. Findings reveal that while dominant topics remained relatively stable over time (e.g., self-expression, lifestyle, and creativity), high-influence users were more likely to engage in culturally resonant or commercially strategic content. Additionally, political discourse was not avoided, but selectively activated as a point of transnational engagement. Emotionally, high-influence users tended to express more positive affect in culturally connective topics, while low-influence users showed stronger emotional intensity in personal narratives. These findings suggest that cross-cultural platform migration is shaped not only by structural affordances but also by users' differential capacities to adapt, perform, and maintain visibility. The study contributes to literature on platform society, affective publics, and user agency in transnational digital environments.

Figures (7)

• Figure 1: Temporal trend of post frequency on RedNote. Colored background shading enhances the visual segmentation of these phases. The timeline reflects three distinct stages of user migration and interaction, offering a macroscopic view of shifting engagement.
• Figure 2: Audience Engagement Level of Content Created by High vs. Low-Influence Users Per Phase
• Figure 3: User trajectory across platform migration phases: Each phase-Pre-Ban, Refugee Surge, and Stabilization-is divided into three user types based on activity level: Inactive(users with fewer than 3 posts in a phase), Low influence(bottom 75% by influence score), and High influence(top 25%). The Sankey diagram visualizes user transitions between these categories across phases, revealing patterns of drop-off, influence emergence, and stabilization.
• Figure 4: Topic Proportion by Influence Level
• Figure 5: Top categories among high-influence users based on engagement entropy.