Bridging the Narrative Divide: Cross-Platform Discourse Networks in Fragmented Ecosystems

Patrick Gerard, Hans W. A. Hanley, Luca Luceri, Emilio Ferrara · 2025

Read Bridging the Narrative Divide: Cross-Platform Discourse Networks in Fragmented Ecosystems by Patrick Gerard, Hans W. A. Hanley, Luca Luceri, Emilio Ferrara (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-2505-21729.txt.

Introduction

Patrick Gerard and three co-authors map how political stories hop between social platforms, in a 2025 arXiv paper. Talk online now spreads across split, walled services, and each keeps its own data. That makes the path of a story hard to trace. The team's platform-free method rebuilds the hidden web of users across services by finding shared stories and who takes part in them.

Tested on tasks like spotting influence drives and guessing a user's stance, the method beat older tools while using far less data. Applied to Truth Social and X, it found a tiny mixed group of bridge users. That group, under half a percent of users, seeds nearly seventy percent of the stories that migrate. The work gives watchers a way to see leaks coming.

Catalog record

CREATORPatrick Gerard, Hans W. A. Hanley, Luca Luceri, Emilio Ferrara
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE104 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-21729

Browse the full library · About the Tower of Babel

Full text

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

Bridging the Narrative Divide: Cross-Platform Discourse Networks in Fragmented Ecosystems Patrick Gerard, Hans W. A. Hanley, Luca Luceri, Emilio Ferrara Abstract Political discourse has grown increasingly fragmented across different social platforms, making it challenging to trace how narratives spread and evolve within such a fragmented information ecosystem. Reconstructing social graphs and information diffusion networks is challenging, and available strategies typically depend on platform-specific features and behavioral signals which are often incompatible across systems and increasingly restricted. To address these challenges, we present a platform-agnostic framework that allows to accurately and efficiently reconstruct the underlying social graph of users’ cross-platform interactions, based on discovering latent narratives and users’ participation therein. Our method achieves state-of-the-art performance in key network–based tasks: information operation detection , ideological stance prediction , and cross-platform engagement prediction —while requiring significantly less data than existing alternatives and capturing a broader set of users. When applied to cross-platform information dynamics between Truth Social and X (formerly Twitter), our framework reveals a small, mixed-platform group of bridge users , comprising just 0.33% of users and 2.14% of posts, who introduce nearly 70% of migrating narratives to the receiving platform. These findings offer a structural lens for anticipating how narratives traverse fragmented information ecosystems, with implications for cross-platform governance, content moderation, and policy interventions. Code — https://tinyurl.com/CANE2025 Data — https://tinyurl.com/CANE2025 Introduction Political communication today unfolds across a fragmented digital landscape, shaped by the dynamics of ideologically distinct platforms ( Zhang et al. 2025 ) . During the January 6th, 2021, Capitol insurrection, participants engaged in parallel but intersecting conversations across Parler, X (formerly Twitter), and Gab ( Sipka, Hannak, and Urman 2022 ; Vishnuprasad et al. 2024 ) , which were later amplified on mainstream platforms via news reporting ( Zulli, Coe, and Isaacs 2023 ; Luceri, Cresci, and Giordano 2021 ) . More recently, the emergence of Truth Social, a platform created by U.S. President Donald Trump, has introduced a new locus of politically salient discourse ( Zhang et al. 2025 ; Gerard, Botzer, and Weninger 2023 ) , with narratives frequently originating there before diffusing outward ( Shah et al. 2024 ) . This fragmentation became particularly salient during the 2024 U.S. presidential election, when narratives, unverified claims, and emotionally charged rhetoric moved rapidly across ecosystem boundaries ( Minici et al. 2024 ) . Yet despite growing concern, we still lack systematic tools for tracing how discourse migrates between these siloed communities, or for identifying the structures that enable it. Most traditional network approaches rely on platform-specific behaviors such as reposts, mentions, or follower ties ( Cinelli et al. 2021 ) , which are often incompatible across ecosystems and increasingly inaccessible due to API restrictions ( Tromble 2021 ) . More recent content-based methods infer user relationships through linguistic similarity ( Ng, Cruickshank, and Carley 2022 ; Luceri et al. 2024a ) , but typically depend on pairwise message comparisons or token-level co-occurrence ( Cinus et al. 2025 ) . As we show in this work, such approaches fail to capture persistent alignment, generate unstable user communities, and generalize poorly across platforms. These limitations obscure the structural architecture of discourse, particularly the aspects that allow narratives to move between fragmented online communities. To overcome these limitations and understand cross-platform dynamics during the 2024 U.S. Presidential election, we model users not through interaction patterns or surface-level text similarity, but as distributions over latent narratives. Our discourse-centered framework links users through shared participation in semantic clusters of ideas, capturing persistent alignment even without direct ties. This approach disentangles user similarity from platform-specific behavior, allowing for robust and scalable network construction across sparse and fragmented ecosystems. Empirically, we find that this method offers several advantages. We test its effectiveness across both standard user representation tasks and a new benchmark for cross-platform engagement prediction. In both intra- and inter-platform settings, our method matches or exceeds the performance of existing approaches while requiring significantly less data. It captures a broader set of users, identifies stable communities across platforms, and supports large-scale modeling of discourse. Our method, without relying on platform-specific metadata, not only improves generalization and diffusion modeling across siloed ecosystems but also performs competitively within single-platform contexts. Applied to posts from Truth Social and X during the 2024 U.S. Presidential election, our framework reveals not only when and where political narratives migrated between ecosystems but also how and through whom this diffusion occurred. Utilizing our framework, we concretely ask: 1. How do political narratives migrate across fragmented platforms like X and Truth Social during the 2024 U.S. Presidential election, and what structural pathways enable their movement? 2. Do shared discourse communities emerge across X and Truth Social during the 2024 election cycle, and if so, how are they positioned within the broader structure of political communication? 3. Are certain users disproportionately associated with cross-platform narrative diffusion during the 2024 U.S. Presidential election, and what structural roles do they occupy within the discourse network? Our analysis reveals that political narratives migrated between X and Truth Social in a highly structured manner. Many of these narratives migrated between X and Truth Social through a small but structurally distinct set of users, comprising just 0.33% of the total. These users accounted for only 2.14% of all posts, yet served as the initial cross-platform carriers for nearly 70% of narratives that successfully migrated. These bridge users were embedded at the intersection of fragmented discourse communities and consistently appeared early in the life cycle of narratives that later gained traction across ecosystems. Critically, these insights do not rely on retrospective interaction patterns or explicit behavioral ties. Instead, they emerge from a structural perspective made possible by our modeling approach. By linking users through shared participation in latent discourse clusters, our platform-agnostic framework uncovers connective structures that remain hidden in traditional interaction-based or semantic similarity-based graphs. This shift opens new avenues for studying how influence operates across media environments and positions cross-platform discourse networks as a scalable and generalizable lens for understanding narrative diffusion. Related Work Network-Based User Representations. A central challenge in social media analysis is how to represent users in a way that reflects their relationships, ideological alignment, and interactions. Traditional methods rely heavily on platform-specific signals, including follower networks ( Bollen et al. 2011 ) , reposts and mentions ( Cinelli et al. 2021 ) , hashtags ( Alieva, Ng, and Carley 2022 ; Burghardt et al. 2024 ) , or URLs ( Tardelli et al. 2024a ) . These signals, however, face increasing limitations due to restrictive data policies ( Tromble 2021 ) and fragmentation of user activities across platforms ( Minici et al. 2024 ) . Semantic approaches, which construct graphs from linguistic similarity of user content ( Ng, Cruickshank, and Carley 2023 ; Luceri et al. 2024a ) , attempt to bypass such limitations. However, these methods typically rely on pairwise comparisons and often suffer from computational complexity and sensitivity to minor linguistic variations ( Luceri et al. 2024a ) . Thus, current methods rarely provide platform-agnostic and scalable solutions. Behavioral Modeling with Social Graphs. Once constructed, social graphs are typically evaluated by their utility in downstream tasks, including ideological stance prediction and coordination detection. Ideological stance prediction methods often combine content and

Read the complete text