SI based disease model over signed network

Cong Wan, Cong Wang, Yanxia LvarXiv 2018 · arXiv

Abstract

This paper investigates the dynamics of disease spreading within signed social networks, where relationships are categorized as either friendly (positive) or hostile (negative). By integrating the Susceptible-Infected (SI) model with structural balance theory through a novel energy function, the authors reveal how social networks can develop a "self-immunity" mechanism. This mechanism naturally creates social barriers that isolate infected clusters, effectively halting the epidemic's progress through the strategic rewiring of relationship signs.

Introduction

Traditional epidemiological models often treat networks as simple maps of connections, ignoring the qualitative nature of social bonds. In reality, individuals interact more frequently and closely with friends than with enemies. Furthermore, social networks are dynamic; they evolve to minimize psychological tension, a concept known as structural balance (e.g., "the friend of my friend is my friend"). This study addresses the critical interplay between disease transmission and network evolution, exploring how a network's drive toward stability influences the reach of an epidemic.

Methodology

The researchers developed a model combining three core frameworks:

  1. SI Model: Nodes are either healthy (state 1) or permanently infected (state -1).
  2. Structural Balance Theory: Based on Heider's work, triangles of nodes seek balance, preferring configurations with either three positive edges or one positive and two negative edges.
  3. Energy Function: A total energy equation $E = \alpha \cdot E_e + (1 - \alpha) \cdot E_n$ guides the network's evolution. $E_e$ represents structural energy (unbalanced triangles), and $E_n$ represents spreading energy (alignment between disease states and edge signs).
Structural balance in triangles: (a) and (b) are balanced; (c) and (d) are unbalanced.

The model was tested using Monte Carlo simulations on a complete signed network of 20 nodes. At each step, the system attempts to flip an edge sign or spread the disease, accepting changes that lower the total energy.

Key Insights: Social "Self-Immunity"

The study's most significant finding is the Structure and Spreading Theorem. The authors demonstrate that a network reaches a perfectly balanced state ($E = -1$) when it bifurcates into two distinct clusters: one containing only infected nodes and the other only susceptible nodes. In this state, all intra-cluster links are positive, while all inter-cluster links are negative.

This evolution acts as a form of social "self-immunity." As the network minimizes energy, it naturally "isolates the virulent virus" by transforming edges between healthy and infected groups into negative (hostile) links. These negative edges function as functional barriers or "firewalls," effectively insulating the healthy population from the disease.

Key Results

The final extent of the infection is primarily driven by the initial infection ratio ($\rho$) and the spreading probability ($\lambda$). Interestingly, the initial distribution of positive and negative edges ($\sigma$) has a negligible impact on the final outcome, as the evolutionary process effectively overrides the starting configuration.

Sensitivity analysis showing the number of infected nodes relative to initial infection ratio, positive edge ratio, and energy weight.

The parameter $\alpha$ plays a crucial role in determining the network's priority. If the network prioritizes social harmony (high $\alpha$) over disease prevention, it may inadvertently facilitate the spread of the virus by flipping edges to positive to satisfy structural balance requirements, thereby opening new transmission paths.

Conclusion and Future Work

While the current model provides elegant theoretical insights into the protective role of social hostility, it is limited by its small scale and the simplicity of the SI model. Future research will focus on scaling the simulations to larger, sparser real-world datasets like Epinions and Slashdot, and extending the framework to include SIS and SIR models where recovery and immunity are possible.

Summary updated

Abstract

Signed network is a kind of network that associates each edge a positive or negative sign which could express friendly or unfriendly relationship between individuals. When diseases spreading over the signed network, some unfriendly edges may refuse to spread. Moreover, the signed network is dynamic, according to the structure balance theory which is the most important theory in the study of signed network, edges in the signed network will flip their signs over time. How does disease spreading interact with signed network evolving becomes a challenging issue. In this paper, we propose an energy function to describe the disease spreading and network evolving together, and we introduce the notion of Structure and Spreading Balanced. We extend the structure balance theory of Cartwright and propose a Structure and Spreading theorem. Finally, we carry out Monte-Carlo simulations on complete signed network to validate our theorem. In the experiment, we find that the signed network has self-immunity during disease spreading, which can be used explain the phenomenon of isolating the virulent virus by isolation.
BibTeX
@misc{wan2018si,
  author = {Cong Wan and Cong Wang and Yanxia Lv},
  title = {{SI based disease model over signed network}},
  year = {2018},
  eprint = {1803.08040},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1803.08040},
}
Figure