💻 computer science

Higher-order Rumor Propagation Dynamics in Hypergraph Networks with Hyperedge Selection and Information Overload

This paper proposes a Hyper-ISDR model on hypergraph networks that integrates group-level interactions, hyperedge selection strategies, and information overload to analyze the competitive dynamics of rumor and debunking propagation, deriving stability thresholds and stochastic persistence criteria validated by simulations and empirical data.

Original authors: Lixiang Liu, Hongxing Han, Xiaokun Wu, Keke Shang, Shanshan Chen

Published 2026-09-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lixiang Liu, Hongxing Han, Xiaokun Wu, Keke Shang, Shanshan Chen

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the digital age, information rarely travels from one person to another in a simple, straight line. Instead, it moves through complex webs of social groups, where a single post can ripple through a family chat, a professional network, and a public forum simultaneously. Scientists who study how ideas spread have long used mathematical models to map these flows, often treating social connections like a series of one-on-one handshakes. However, real life is messier than a handshake. People often share news within a group of friends, a community, or a topic-based circle where everyone hears the message at once. This "group effect" changes how fast and how far a story travels. Furthermore, the modern internet is saturated with so much content that people often become overwhelmed, tuning out or ignoring messages simply because there are too many of them. Understanding how rumors and the facts that debunk them compete in this crowded, group-oriented environment is crucial for predicting how misinformation might take hold or fade away.

A team of researchers has built a new mathematical framework to capture this complexity, moving beyond simple one-to-one connections to model how information spreads through entire groups at once. They created a system that treats social networks as collections of "hyperedges," which are essentially groups of people connected by a shared interest or discussion, rather than just pairs of friends. In their model, they tracked four types of people: those who know nothing about a story, those spreading a rumor, those spreading the truth to correct it, and those who have stopped engaging with the topic entirely. The researchers wanted to see how these groups interact when two forces are at play: the way groups activate to share information, and the way people get tired of hearing too much.

The study introduced two distinct ways a group can share a message. In one scenario, a person who knows a secret might broadcast it to every single group they belong to at the exact same time, like shouting in every room they enter. In the other scenario, they might pick just one group to tell, focusing their attention locally. The researchers found that when people broadcast to all their groups simultaneously, rumors spread much more easily and persist longer, even if the people spreading them are not the most popular individuals. This "broadcast" style of sharing lowers the barrier for a rumor to become widespread, making it harder to stop once it starts.

To make the model more realistic, the team also included a factor for information overload. They observed that when too many people are shouting out rumors or corrections at the same time, the average person becomes less likely to listen or react. The researchers described this as a dampening effect: as the volume of active information rises, the effectiveness of each new message drops. Their simulations showed that while this overload does not stop a rumor from starting if the conditions are right, it does significantly reduce how many people eventually believe it. It acts as a natural brake on the intensity of the spread, preventing the system from reaching the absolute maximum possible saturation, even when the rumor is highly contagious.

The researchers tested their ideas using computer simulations on synthetic networks that mimic the structure of real social media, where some people belong to many groups and others belong to few. They also applied their model to real-world data from two specific rumor events that occurred on Chinese social media platforms. One involved a false story about a popular writer's work, and the other concerned a rumor about housing fees in Shanghai. By feeding the actual hourly spread of these stories into their equations, they found that their model could accurately reproduce the general shape of how these rumors grew, peaked, and faded. The model successfully captured the broad trends of the events, showing that it can describe the macroscopic evolution of how misinformation moves through a population.

However, the study also highlighted the limits of prediction. While the model could trace the overall rise and fall of a rumor, it could not predict every sudden spike or local fluctuation that happens in real time. These short-term bursts are often driven by unpredictable human behavior or external news cycles that a mathematical formula cannot foresee. To address this, the researchers added a layer of randomness to their equations, simulating the natural noise and uncertainty of human interaction. This stochastic version of the model showed that while the average path of a rumor is predictable, the actual path will always wobble around that average, creating a range of possible outcomes rather than a single fixed line.

The findings suggest that the structure of our online groups matters more than we might think. If a social network is dominated by large groups where many people interact at once, or if the distribution of group sizes is uneven, rumors are more likely to survive and thrive. Conversely, if the network is made up mostly of small, tight-knit circles, the spread is slower and less persistent. The research also confirmed that the way people choose to share information—whether they broadcast to everyone or target specific groups—fundamentally alters the stability of the system. By understanding these dynamics, we gain a clearer picture of why some stories vanish quickly while others become entrenched, and how the sheer volume of information we consume can paradoxically protect us by making us less responsive to every new claim.

Technical Summary: Higher-order Rumor Propagation Dynamics in Hypergraph Networks with Hyperedge Selection and Information Overload

1. Problem Statement

Rumor propagation in online social systems exhibits nonlinear, heterogeneous, and competitive diffusion patterns, particularly during public emergencies. Existing models often rely on pairwise interactions (ordinary graphs), which fail to capture group-level dynamics such as discussion groups, topic communities, and collective exposure events. Furthermore, current research faces three specific gaps:

Activation Mechanisms: Many models assume a single hyperedge activation rule, whereas real-world diffusion involves both broad broadcasting across multiple groups and localized transmission within selected groups.
Information Overload: The cumulative effect of active disseminators (both rumor spreaders and debunkers) causing cognitive fatigue or selective avoidance (attenuation) has not been clearly distinguished in terms of its impact on propagation thresholds versus long-term prevalence.
Stochastic vs. Deterministic Analysis: The relationship between reproduction thresholds, equilibrium stability, and stochastic extinction/persistence in heterogeneous hypergraphs has rarely been examined within a unified analytical framework.

2. Methodology

The authors propose a Hyper-Ignorant-Spreader-Debunker-Removed (Hyper-ISDR) model on heterogeneous hypergraph networks.

2.1 Model Framework

Network Structure: The system is modeled as a hypergraph $H=(V, E)$ , where hyperedges connect more than two nodes to represent group interactions. Nodes are categorized into four compartments: Ignorant ( $I$ ), Spreader ( $S$ ), Debunker ( $D$ ), and Removed ( $R$ ).
Mixed Hyperedge Activation: The model introduces a weighted combination of two activation strategies:
- Reactive Process (RP): An active node activates all incident hyperedges simultaneously (broad multi-hyperedge broadcasting).
- Contact Process (CP): An active node activates one incident hyperedge at each step (localized group-based transmission).
- A path indicator function $e_p(k)$ interpolates between these regimes based on weights $\omega_{RP}$ and $\omega_{CP}$ .
Information Overload Attenuation: An attenuation function $g(A(t)) = \exp(-\beta A(t))$ is introduced, where $A(t)$ is the aggregate density of active individuals ( $S+D$ ). This function nonlinearly reduces the effective propagation intensity as information exposure increases.
Dynamical Systems:
- Deterministic Model: Derived via a heterogeneous mean-field approximation, resulting in a system of Ordinary Differential Equations (ODEs) describing the macroscopic evolution of densities across hyperdegree classes.
- Stochastic Model: An extension using Stochastic Differential Equations (SDEs) with multiplicative perturbations to account for random fluctuations in user activity and behavioral heterogeneity.

2.2 Theoretical Analysis

Thresholds: The authors derive basic reproduction numbers for rumor spreading ( $R_0^S$ ) and debunking ( $R_0^D$ ).
Stability: They establish conditions for the local and global asymptotic stability of the Rumor-Free Equilibrium (RFE), the Debunker-Boundary Equilibrium (DBE), and the Coexistence Equilibrium (Ec).
Stochastic Criteria: Extinction and persistence criteria are derived for the SDE system, showing how noise intensities modify the deterministic thresholds.

3. Key Contributions

Unified Activation Mechanism: The formulation of a heterogeneous mean-field Hyper-ISDR model with a mixed RP–CP activation mechanism, allowing the representation of both multi-hyperedge broadcasting and localized transmission within a single framework.
Overload Attenuation Function: The introduction of an overload-induced attenuation function that decouples the effects of information overload on propagation thresholds from its effects on post-threshold prevalence.
Comprehensive Stability Analysis:
- Derivation of basic reproduction numbers and threshold conditions for the global stability of the RFE and DBE in the deterministic system.
- Characterization of the existence of coexistence equilibria.
- Establishment of global positive solution existence and derivation of extinction/persistence criteria for the stochastic system under multiplicative perturbations.
Empirical Validation: Application of the model to empirical rumor-event data (Short Essay Rumor and Housing Pension Rumor) to evaluate its ability to reproduce macroscopic spreading patterns and transient fluctuations.

4. Results

Activation Weights: Increasing the RP weight ( $\omega_{RP}$ ) lowers the threshold for persistent rumor spreading and increases the final rumor prevalence. This is because the RP strategy allows high-hyperdegree nodes to activate multiple paths simultaneously.
Hyperedge Structure: Networks dominated by large hyperedges or those with bimodal hyperedge-size distributions sustain higher levels of rumor and debunking activity compared to small-hyperedge-dominant distributions.
Information Overload: The overload parameter $\beta$ does not alter the basic propagation thresholds ( $R_0$ ). Instead, it reduces the post-threshold prevalence (endemic level) by suppressing the effective propagation intensity.
Stochastic Effects: The proposed mixed RP–CP mechanism results in higher stochastic thresholds compared to pure CP-based models. The noise intensity has a weak effect on the proposed thresholds but strongly affects pure CP-based persistence thresholds.
Empirical Fit: The deterministic model successfully captures the main macroscopic evolution patterns (growth, peak, decay) of real rumor events. The stochastic extension captures transient fluctuations and short-term variations that the deterministic model misses, providing an uncertainty range rather than a precise pointwise fit.

5. Significance and Claims

The paper claims to provide a more realistic mathematical framework for analyzing competitive information diffusion in complex social systems. By integrating higher-order interactions (hypergraphs), mixed activation strategies, and information overload effects, the model offers a nuanced understanding of how rumor persistence is shaped by network topology and user behavior.

The authors modestly state that while the model reproduces macroscopic trends and transient fluctuations, it is not intended for event-level predictive accuracy. The significance lies in clarifying the mechanisms by which hyperedge activation rules, structural heterogeneity, and random fluctuations influence the stability and persistence of competitive rumor-debunking dynamics. The study suggests that controlling rumor spread may require targeting not just transmission rates, but also the structural activation patterns (e.g., limiting broad broadcasting) and managing information load to mitigate cognitive fatigue.