Characterizing Network Structure of Anti-Trans Actors on TikTok CCS: Human-centered computing Collaborative and social computing CCS: Computing methodologies Natural language processing CCS: Networks Network analysis CCS: Information systems Retrieval tasks and goals

Maxyn Leitner, Rebecca Dorn, Fred Morstatter, Kristina Lerman · 2025

Read Characterizing Network Structure of Anti-Trans Actors on TikTok CCS: Human-centered computing Collaborative and social computing CCS: Computing methodologies Natural language processing CCS: Networks Network analysis CCS: Information systems Retrieval tasks and goals by Maxyn Leitner, Rebecca Dorn, Fred Morstatter, Kristina Lerman (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16507.txt.

Introduction

Maxyn Leitner and co-authors map networks of anti-trans actors on TikTok in a 2025 study. The platform has helped trans and nonbinary creators find community, but it has also let hostile actors spread hate speech with ease. The study asks how the two camps differ in network shape and how much they clash.

The team built a labeling scheme for trans-related views with help from expert annotators from the community. A search-aided classifier then sorted videos into pro, anti, and neutral groups. Network study found many cross-camp contacts, which points to direct targeting of trans users. Readers get evidence for why stronger moderation tools are needed.

Catalog record

CREATORMaxyn Leitner, Rebecca Dorn, Fred Morstatter, Kristina Lerman
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16507

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Characterizing Network Structure of Anti-Trans Actors on TikTok CCS: Human-centered computing Collaborative and social computing CCS: Computing methodologies Natural language processing CCS: Networks Network analysis CCS: Information systems Retrieval tasks and goals Maxyn Leitner, Rebecca Dorn, Fred Morstatter, Kristina Lerman Abstract . Content Warning : Trans-antagonistic Rhetoric and Terminology The recent proliferation of short form video social media sites such as TikTok has been effectively utilized for increased visibility, communication, and community connection amongst trans/nonbinary creators online. However, these same platforms have also been exploited by right-wing actors targeting trans/nonbinary people, enabling such anti-trans actors to efficiently spread hate speech and propaganda. Given these divergent groups, what are the differences in network structure between anti-trans and pro-trans communities on TikTok, and to what extent do they amplify the effects of anti-trans content? In this paper, we collect a sample of TikTok videos containing pro and anti-trans content, and develop a taxonomy of trans related sentiment to enable the classification of content on TikTok, and ultimately analyze the reply network structures of pro-trans and anti-trans communities. In order to accomplish this, we worked with hired expert data annotators from the trans/nonbinary community in order to generate a sample of highly accurately labeled data. From this subset, we utilized a novel classification pipeline leveraging Retrieval-Augmented Generation (RAG) with annotated examples and taxonomy definitions to classify content into pro-trans, anti-trans, or neutral categories. We find that incorporating our taxonomy and its logics into our classification engine results in improved ability to differentiate trans related content, and that Results from network analysis indicate many interactions between posters of pro-trans and anti-trans content exist, further demonstrating targeting of trans individuals, and demonstrating the need for better content moderation tools. 1. Introduction TikTok’s distinctive algorithmic and community dynamics make it a compelling case for studying the dual role of social media in fostering community-building and enabling targeted harassment, especially for marginalized groups like trans and nonbinary people. Platforms like TikTok allow trans/nonbinary creators to connect, share affirmations of identity, and disseminate essential resources about healthcare ( Aldridge et al. 2024 ) and support systems ( Buss et al. 2022 ) . Simultaneously, these platforms facilitate the proliferation of anti-trans rhetoric, weaponized by malign actors to spread hate speech and amplify harmful stereotypes ( Srouji 2022 ) . This work contributes to the understanding and modeling of these dynamics through three key innovations. First, we introduce a comprehensive taxonomy of sentiment toward trans and nonbinary communities, grounded in sociological and gender studies frameworks. This taxonomy addresses critical gaps in prior frameworks by including underexamined subcategories such as transmisogyny, anti-transmasculinity, and exorsexism, enabling a detailed and intersectional understanding of trans-related sentiment. This enables a more granular analysis of online discourse and its impact on marginalized communities. Second, we develop a novel classification pipeline to identify pro-trans, anti-trans, and neutral content on TikTok, introducing a state-of-the-art classification pipeline that combines Retrieval-Augmented Generation (RAG) with curated annotated examples and taxonomy definitions, setting a new benchmark for classifying nuanced trans-related sentiment. Our RAG system is armed with two key resources: a curated set of annotated examples and the definitions from extant literature relating to our taxonomy. These resources not only guide the LLM’s contextual understanding but also significantly improve classification accuracy, especially in distinguishing nuanced pro-trans and anti-trans sentiments. Finally, we apply our classification and taxonomy to analyze the reply network structures of pro-trans and anti-trans actors on TikTok. By comparing these network dynamics, we reveal the mechanisms by which anti-trans content proliferates and the structural characteristics that amplify its reach. This analysis informs potential strategies to mitigate online harms while empowering trans and nonbinary communities in digital spaces. By combining sociological theory, novel computational techniques, and network analysis, this paper contributes a robust framework for studying the dynamics of marginalized communities and their adversaries in online ecosystems. These contributions advance the field of computational social science and offer actionable strategies for combating online harassment. 2. Background and Related Work 2.1. Anti-Trans Rhetoric Online Due to its rapid proliferation and popularity across generational and political lines, TikTok makes for an incredibly effective social media space to analyze anti-trans and pro-trans content, as well as the ways in which they interact. To begin, TikTok has served as a battle ground on which progressive views, particularly with respect to trans/non-binary people, have been scrutinized and exposed to attack from right wing actors ( Hubrig 2023 ; Carless 2023 ) . Right Wing content is incredibly popular on TikTok, and is often formatted in an attempt to appeal to the platform’s younger user base ( Weimann and Masri 2023 ; Albertazzi and Bonansinga 2024 ) . In concert with this, a combination of TikTok’s content moderation approaches and recommendation algorithms often results in users being served increasingly right wing content, and anti-trans rhetoric serves as effective entry into that pipeline ( Boucher 2022 ; Little et al. 2021 ) . 2.2. Trans and Gender Diverse TikTok Users Structural forces such as content recommendation systems work at odds with trans/non-binary users, who report a lack of agency over being shown content that maligns their identities and convictions in ways that affect their beliefs about themselves ( Ellen Simpson et al. 2021 ; Claudiu Gabriel Ionescu and Monica Licu 2023 ) . In the face of this opposition, trans/non-binary users persist and form folk theories in order to resist prevailing anti-trans narratives. Where anti-trans actors are motivated to falsely report queer content as inappropriate as a way of training TikTok’s content recommendation algorithms ( Dawson 2024 ) , queer and trans users build and test theories of how to influence said algorithms in the other direction, towards greater visibility and acceptance of their identities ( Karizat et al. 2021 ) . Transfeminine users have reported an awareness of how having their content promoted by the algorithm both opens them up to greater risk of attack from anti-trans actors, but also greater opportunity, moderating the former through strategies DeVito calls algorithmic trapdoors ( Michael Ann DeVito and Michael A. DeVito 2022 ) . Given this landscape, TikTok is an important and illustrative platform on which to investigate the differences in network structure between anti-trans and pro-trans communities. 2.3. Perspectives of Trans Sentiment Previous work in classifying trans sentiment has mainly focused on the detection of anti-trans sentiment ( Lu and Jurgens 2022 ; Chakravarthi et al. 2022 ) . However, these attempts fail to distinguish details in which trans people are targeted, and either presume or fail to analyze the identity of the source of anti-trans content, at the risk of visibilizing some forms of anti-trans sentiment over others. Given the painted heterogeneity of trans identities and experiences, previous taxonomy studies are limited both in the diversity of their subject pool and the specificity of anti-trans sentiment characterized ( Chang and Chung 2015 ) . 3. Theoretical Grounding for a Trans Sentiment Taxonomy We derive a taxonomy characterizing sentiment towards trans and gender diverse individuals in order to better understand and analyze online harms and social dynamics surrounding trans people. This work follows taxonomy design guidance ( Kundisch et al. 2021 ) , including specifying the taxonomy’s goal and purpose, outlining qualitative techniques used, and evaluating the

Read the complete text