Social contagion under hybrid interactions

Xincheng Shu, Man Yang, Zhongyuan Ruan, Qi XuanChaos 2024 · DOI

Sure, here is a blog-style summary of the academic paper "Social contagion under hybrid interactions":

TL;DR

This paper introduces a novel model for social contagion that combines two common mechanisms: threshold-based adoption and game-theoretic imitation. By simulating this hybrid model, the researchers found that the dynamics of contagion are significantly influenced by the payoff of adoption and the "rationality" of individuals, leading to complex phenomena like double phase transitions. This work offers a more nuanced understanding of how information and behaviors spread through social networks.

What Problem Does This Paper Address?

Social contagion, the spread of ideas, behaviors, or diseases through a network, is a fundamental phenomenon in social systems. Existing models often focus on one of two primary mechanisms:

  • Threshold models: These assume individuals adopt a behavior only when a certain proportion of their neighbors have already adopted it (e.g., peer pressure).
  • Game-theoretic models: These assume individuals are rational and make decisions based on maximizing their own utility, often involving payoffs for adoption or non-adoption.

However, real-world social interactions are rarely so simplistic. Individuals often exhibit a mix of both threshold-driven behavior and rational decision-making based on perceived benefits and costs. Previous research has largely overlooked models that simultaneously incorporate these two distinct, yet often coexisting, interaction forms. This paper aims to bridge this gap by proposing and analyzing a general model that integrates these hybrid interaction dynamics.

Methods

The authors developed a computational model that simulates social contagion on a network. The key innovation is the introduction of two types of nodes:

  • Type-0 nodes: These follow the threshold mechanism. They adopt a new behavior only if the fraction of their neighbors who have already adopted it exceeds a predefined threshold.
  • Type-1 nodes: These are imitators driven by game theory. They adopt a behavior based on the perceived payoff of adoption versus cost, influenced by the strategies of their neighbors. Their decision to imitate is governed by a Fermi function, which quantifies their "rationality" (controlled by a parameter $\beta$). A higher $\beta$ means more rational decision-making, while a lower $\beta$ implies less rational or even anti-rational behavior.

The model uses a random Erdős-Rényi (ER) network structure. Initially, one node is an adopter, and the rest are susceptible. The system evolves over time, with susceptible nodes transitioning to adopters based on their type and the states of their neighbors. The researchers used simulations to explore various scenarios, including positive and negative adoption payoffs, and employed a susceptibility measure to identify phase transitions. They also developed an analytical framework using approximate master equations (AMEs) to complement their simulation results.

Datasets

Not discussed in the paper. The study relies on simulated network data generated by their model.

Key Insights and Ideas

  • Hybridity is Crucial: The interplay between threshold-based and imitation-based adoption is key to understanding complex contagion dynamics, going beyond single-mechanism models.
  • Payoff Matters: The sign of the adoption payoff (benefit > cost vs. benefit < cost) dramatically alters the contagion process.
  • Rationality's Dual Role:
    • With positive payoffs, higher rationality among imitators can accelerate adoption and lead to double phase transitions (two distinct points where the system's behavior changes dramatically).
  • With negative payoffs, higher rationality can actually inhibit spread, while decreased rationality can lead to faster propagation of "irrational" behaviors.
  • Double Phase Transitions: For positive payoffs and specific parameter regimes, the model exhibits two distinct tipping points, suggesting a more complex transition to widespread adoption than simple models predict.
  • Imitator Clusters: The formation of large, connected clusters of imitators can act as a catalyst for widespread adoption, especially when they are exposed to early adopters.

Key Results

  • Positive Payoffs (b > c):
    • Increasing the fraction of rational imitators ($r$) can promote adoption and lead to double phase transitions, particularly when the threshold parameter ($\phi$) is high.
  • The speed of contagion is directly regulated by the rationality parameter ($\beta$): higher $\beta$ leads to faster spread.
  • Negative Payoffs (b < c):
    • Highly rational imitators ($\beta \to +\infty$) act as "blocked nodes," hindering spread.
  • Lower rationality ($\beta < +\infty$) allows for the spread of negative behaviors, with smaller $\beta$ leading to faster contagion.
  • Network Structure: The average degree ($z$) of the network influences contagion. For $r > 0$, increasing $z$ can initially hinder spread (due to threshold requirements) but then promote it by increasing the likelihood of imitators being exposed to adopters.

Limitations and Future Work

  • Mean-Field Approximation: The analytical framework (AMEs) relies on mean-field approximations, which can miss certain phenomena, particularly the first tipping point observed in simulations, especially in fragmented networks ($z < 1$).
  • Network Topology: The study primarily focuses on random ER networks. Future work could explore how different network structures (e.g., scale-free, community structures) affect the hybrid contagion dynamics.
  • Parameter Heterogeneity: While the model allows for different types of nodes, the parameters within each type (e.g., individual thresholds, payoffs) are assumed to be uniform. Extending the model to include heterogeneity in these parameters would offer a more realistic representation of social systems.
  • Dynamic Payoffs/Thresholds: The current model assumes fixed payoffs and thresholds. Investigating how these parameters change dynamically over time could reveal new insights.

Summary updated

BibTeX
@inproceedings{shu2024social,
  author = {Xincheng Shu and Man Yang and Zhongyuan Ruan and Qi Xuan},
  title = {{Social contagion under hybrid interactions}},
  booktitle = {Chaos},
  year = {2024},
  doi = {10.1016/j.chaos.2024.115687},
}
Figure