The Psychology of Hesitation: Modeling Rumor Spreading in Multilayer Social Networks

TL;DR

As social media platforms become increasingly interconnected, rumors no longer stay confined to a single "silo." This paper proposes the SIHR rumor spreading model, adding a crucial "Hesitator" (H) state to the traditional SIR framework. By analyzing this in a multilayer network context, the authors demonstrate why rumors spread faster across platform boundaries and provide strategic insights on how to break these chains of toxic information.

Background: Why Single-Layer Models Fail

Historically, rumor modeling was borrowed from epidemiology—treating a rumor like a virus (SIR: Susceptible-Infectious-Recovered). However, humans are more complex than biological hosts. We don't always believe or reject information immediately; we hesitate.

Furthermore, we live in a multilayer cyberspace. You might see a rumor on Facebook, ignore it, then see it again on Twitter from a different friend. This "repeated propagation" in multilayer structures (Interdependent Networks) creates a reinforcement effect that single-layer models cannot explain.

Methodology: The SIHR Framework & Multilayer Topology

The authors propose a new state transition model. Instead of moving directly from Ignorant to Spreader, users often pass through a Hesitator (H) phase.

1. The SIHR Transition Logic

• Ignorant (I) → Hesitator (H): Initial exposure at rate $\alpha$.
• Hesitator (H) ↔ Spreader (S): A dynamic loop. Hesitators become Spreaders at rate $\mu$ if persuaded; Spreaders become Hesitators at rate $\omega$ if they lose interest or become dubious.
• Spreader/Hesitator → Stifler (R): Realizing the truth and stopping the spread at rate $\gamma$.

2. Multilayer Architecture

The paper utilizes a bilayer network (Small-World + Scale-Free) to mimic the real world. In this setup, nodes have intra-layer edges (friendships on one platform) and inter-layer edges (the same person switching between accounts).

Fig 1: The architecture of multilayer networks showing how rumors traverse across platform boundaries.

Experiments: What Makes Rumors Explode?

The authors used mean-field equations to simulate the dynamics. The results confirm a terrifying reality: multilayer networks have a "double average degree" effect, leading to significantly higher rumor peaks.

Fig 2: Evolution in combined networks showing the transient peaks of Spreaders and Hesitators.

Key Findings from Parameter Analysis:

• Average Degree is King: The more connected the layers (Higher $<k>$), the faster the peak is reached.
• The Persuasion Shift: Increasing the rate $\mu$ (Hesitator to Spreader) causes the rumor activity to increase sharply.
• Scale-Free Influence: The presence of "Hub nodes" (users with thousands of followers) acts as an accelerator, especially when these hubs span across multiple networks.

Critical Insight: How to Stop the Spread

The authors don't just point out the problem; they suggest Security Strategies:

1. Decoupling Layers: Breaking the interdependence between networks (e.g., limiting cross-platform automated sharing) can split the growth curve.
2. Alertness Training: Reducing the $\alpha$ rate by improving public "safety consciousness" and media literacy.
3. Hub Monitoring: Monitoring and intervening at high-degree nodes is more effective than broad-spectrum censorship.

Conclusion & Future Outlook

The SIHR model provides a more realistic lens for understanding the "infodemic" age. By identifying the Hesitator state, it opens the door for interventions before someone becomes an active spreader.

Future Work: The authors suggest that the next frontier is using AI and Big Data to detect these threats in the "early stage" (first-order phase transition) before the damage is manifested across the entire multilayer structure.