Cross-Platform Narrative Prediction: Leveraging Platform-Invariant Discourse Networks

TL;DR

The paper tackles the challenge of cross-platform narrative diffusion by reframing it as a cross-platform proximity problem using platform-invariant discourse networks. It constructs a unified representation where users are linked through shared narrative engagement, enabling prediction of narrative emergence across platforms with neighbor activity alone. Empirical results on 5.7M posts across four platforms show high predictive accuracy (AUC up to $0.88$) using only $2.9\%$ of active users, and retrospective deployment demonstrates early detection of high-impact narratives days before mainstream emergence. The approach offers practical benefits for real-time monitoring and proactive intervention in fragmented information ecosystems, without relying on platform-specific signals or diffusion parameterizations.

Abstract

Online narratives spread unevenly across platforms, with content emerging on one site often appearing on others, hours, days or weeks later. Existing cross-platform information diffusion models often treat platforms as isolated systems, disregarding cross-platform activity that might make these patterns more predictable. In this work, we frame cross-platform prediction as a network proximity problem: rather than tracking individual users across platforms or relying on brittle signals like shared URLs or hashtags, we construct platform-invariant discourse networks that link users through shared narrative engagement. We show that cross-platform neighbor proximity provides a strong predictive signal: adoption patterns follow discourse network structure even without direct cross-platform influence. Our highly-scalable approach substantially outperforms diffusion models and other baselines while requiring less than 3% of active users to make predictions. We also validate our framework through retrospective deployment. We sequentially process a datastream of 5.7M social media posts occurred during the 2024 U.S. election, to simulate real-time collection from four platforms (X, TikTok, Truth Social, and Telegram): our framework successfully identified emerging narratives, including crises-related rumors, yielding over 94% AUC with sufficient lead time to support proactive intervention.

Figures (4)

• Figure 1: Traditional user networks (left) fragment users by platform, while discourse networks (right) connect users through shared narrative engagement regardless of platform. Purple regions show cross-platform narrative overlap; node colors indicate platforms.
• Figure 2: Cumulative gain (operator yield curve) comparing Discourse and Fused networks on the 7-day prediction window. The dashed line indicates random performance.
• Figure 3: UMAP projections of two separate narratives before (left) and after (right) applying our LLM-based claim extraction. Prior to normalization, key elements of the narratives are muddied across platforms, with content blending rather than separating cleanly. After claim extraction, coherent, separate narratives emerge, showing that platform-specific linguistic noise exerts less influence once the narrative is normalized.
• Figure 4: Complementary cumulative distribution function (CCDF) of cross-degree across networks. The discourse network shows substantially higher typical connectivity, with a median cross-degree of 9 and mean of 14.9 ($\sigma=21.2$), compared to the fused network where most users have $\leq 1$ cross-tie despite a few extreme hubs. The discourse network distribution thus reflects more widespread cross-platform ties among ordinary users, while the fused network is dominated by a small set of super-connectors.