Modeling and Simulation of Network Event Propagation Interference based on mSIR-CA

Jie Zang, Xu Ren, Yan Wang, Ping Xiao, Zeng R X, Chen, S Zhang, Ma X Y, Y Zhou, Ye Q Y, Xia Y X, Dou Y L, Liu Y S, Zhang H T, Xu H L, Li Y Q Wang X W, Qiu C C, J Wang, Wang Fu L, P Bolzern, P Colaneri, De Nicolao, G, H Ma Y, Zhao Y Y, Qiang Y R, Chen F J Wei D Z, Lin L N, J Shi, Yang, L Lin, Chen F J Xie J L, T Qin, Z Shen, H Liu, J Cao H, Li2022 · DOI

TL;DR

This paper introduces mSIR-CA, a simulation framework that models how multiple news events or "public opinions" compete for attention on the internet. By combining Cellular Automata (CA) with an expanded Susceptible-Infected-Recovered (SIR) model, the researchers demonstrate that network events are never truly independent; the presence, timing, and intensity of a secondary event can significantly "squeeze" and limit the spread of a target event.

What Problem Does This Paper Address?

Most traditional models of information spread—like those used to track rumors or viral marketing—treat a single event as if it exists in a vacuum. In reality, the internet is a crowded "attention economy." When a major news story breaks, it doesn't just spread; it competes with every other trending topic for a user's limited time and interest.

The authors identify a gap in existing research: while we understand how a single event propagates, we have less data on interference. How does a new, secondary event disrupt the trajectory of an existing one? Understanding this "interference law" is crucial for government departments and organizations trying to manage public opinion or mitigate the spread of misinformation.

Methods

The authors propose the mSIR-CA model, which integrates two classic concepts:

  1. Cellular Automata (CA): A grid-based system where each "cell" represents an internet user. The state of a user changes based on the states of their neighbors.
  2. SIR Model: Borrowed from epidemiology, this tracks individuals as they move from Susceptible (ready to hear news) to Infected (spreading the news) to Recovered (immune/lost interest).

The "m" (Multiple) Innovation: Standard SIR models only have one "infection" type. The mSIR-CA model introduces complex state transition rules for multiple events (Event A and Event B).

  • Neighborhood Selection: The authors chose the Moore neighborhood (the 8 cells surrounding a central cell). They argued that the Von Neumann neighborhood (4 neighbors) was too simple, while the "Extended Moore" was too complex and caused events to spread unrealistically fast.
  • State Transitions: They defined a hierarchy of states, including "Offline," "Online/Susceptible," "Affected by Event A/B," and finally "Propagating" or "Immune." Crucially, once a user becomes immune to one event in this model, they are stabilized and won't be affected by the competing event, simulating the finite nature of human attention.

Datasets

This paper is a theoretical modeling and simulation study; it does not use a specific real-world dataset (like a Twitter or Weibo crawl). Instead, it uses simulated environments to test the model's logic:

  • Matrix Size: A 300x300 grid representing 90,000 potential users.
  • Initial Conditions: An initial online user density ($D_{to}$) of 0.70.
  • Iterations: Each experiment was run for 2,000 iterations to observe the full lifecycle of the events.

Key Insights and Ideas

  • Information Competition is a Zero-Sum Game: The simulation confirms that the existence of "Event B" directly reduces the maximum propagation range of "Event A." In a network with finite users, events must "fight" for territory.
  • The "First-Mover" Advantage: Timing is the most critical factor in interference. If Event B starts significantly later than Event A, its ability to interfere is negligible because Event A has already "captured" the susceptible population.
  • Intensification as a Strategy: By increasing the "propagation parameters" (intensity) of a secondary event mid-way through the simulation, the authors could "squeeze" the space of the target event. This suggests that the best way to stop a rumor might not be to debunk it, but to provide a more compelling, high-intensity alternative narrative.

Key Results

The authors conducted three main experimental scenarios:

  1. Fair Competition: When two events start simultaneously with equal intensity, they split the network's capacity, significantly lowering the reach of both compared to a single-event scenario.
  2. Timing Effects: The later an interference event occurs, the lower its impact. The curves in the paper show that late-arriving events fail to break the "momentum" of the target event.
  3. Strength/Intensity Effects: By strengthening the propagation capacity of Event B, the researchers could actively shrink the "exclusive space" of Event A. Quantitatively, the simulation showed that a high-intensity interference event could limit a target event's spread to a fraction of its potential reach.

Limitations and Future Work

  • Grid vs. Graph: The use of a 2D grid (Cellular Automata) is a simplification. Real-world social networks are "scale-free" or "small-world" graphs where some users (influencers) have thousands of connections while others have few. The authors acknowledge other models (like BA scaleless networks) in their introduction but stick to CA for this framework.
  • Platform Specifics: The paper claims "platform versatility," but the rules for how a user reacts on TikTok might differ wildly from LinkedIn. The model currently treats "cyberspace" as a uniform environment.
  • Human Psychology: The model assumes a somewhat mechanical transition between states. Future work could integrate more nuanced psychological factors, such as "confirmation bias," where a user's existing beliefs make them more or less susceptible to specific types of interference.

Summary updated

Abstract

Network events are not completely independent in the communication process, and they will be affected by other events in the same communication process. In order to study the change of propagation law caused by the interference between network events, this paper first designs the interference framework of network event propagation according to the diffusion characteristics of network events and the view that different events in the same diffusion process will have an impact. In the framework, the mSIR-CA (multiple SIR-cellular automata) model is proposed by combining the cellular automata theory with the complex state transition rules which are improved based on SIR model and suitable for multiple event propagation, and the mSIR-CA model is used to simulate the event interference propagation under different conditions. The simulation results show that interference events can have a significant impact on the target event diffusion, and the whole process follows its inherent law, which can provide a new theoretical basis for the relevant departments to deal with public opinion and other application scenarios.
BibTeX
@misc{zang2022modeling,
  author = {Jie Zang and Xu Ren and Yan Wang and Ping Xiao and Zeng R X and Chen and S Zhang and Ma X Y and Y Zhou and Ye Q Y and Xia Y X and Dou Y L and Liu Y S and Zhang H T and Xu H L and Li Y Q Wang X W and Qiu C C and J Wang and Wang Fu L and P Bolzern and P Colaneri and De Nicolao and G and H Ma Y and Zhao Y Y and Qiang Y R and Chen F J Wei D Z and Lin L N and T Chen and J Shi and Yang and L Lin and Chen F J Xie J L and T Qin and Z Shen and H Liu and J Cao H and Li},
  title = {{Modeling and Simulation of Network Event Propagation Interference based on mSIR-CA}},
  year = {2022},
  doi = {10.18178/wcse.2022.06.044},
}
Figure