Algorithmic Contagion: A Network-SEIR Analysis of Xenophobic Disinformation Diffusion During Indonesia’s 2024 Election

Abstract

The 2024 Indonesian General Election was marked by a sudden, coordinated surge in xenophobic narratives targeting Rohingya refugees. This study investigates the diffusion mechanics of this viral hate, testing the hypothesis that algorithmic architectures on platforms such as TikTok and X (formerly Twitter) accelerate radicalization through specific epidemiological pathways. We employed a Stochastic Network SEIR (Susceptible-Exposed-Infectious-Recovered) model to analyze the Indo-Elect-24 dataset, comprising 2.4 million interaction events across a network of 10.2 million nodes. Unlike traditional aggregate models, we utilized a heterogeneous adjacency matrix to identify super-spreader nodes. Parameters were estimated using Bayesian inference via Markov Chain Monte Carlo sampling to quantify uncertainty. The model achieved a high goodness-of-fit (RMSE = 0.042; R-squared = 0.91). We found the Basic Reproduction Number (R0) for anti-Rohingya narratives was significantly higher on TikTok (R0 = 5.42 [95% CI: 5.12–5.72]) compared to X (R0 = 2.81 [95% CI: 2.65–2.97]). Crucially, the Exposed compartment revealed an Algorithmic Latency period where passive consumption drives radicalization before active sharing. Network analysis identified that 8.2% of nodes accounted for 64.8% of total transmission. In conclusion, the study confirms that hate speech functions as a bio-engineered pathogen with pandemic-level virality, driven by algorithmic amplification rather than organic social consensus.