Post-Disaster Resource Redistribution and Cooperation Evolution Based on Two-Layer Network Evolutionary Games

Technical Summary: Post-Disaster Resource Redistribution and Cooperation Evolution Based on Two-Layer Network Evolutionary Games

Problem Statement
Large-scale disasters create severe resource scarcity and infrastructure paralysis, hindering effective recovery. While shelters serve as critical organizational units for resource allocation and victims often cooperate spontaneously for survival, the dynamic interplay between organizational-level resource redistribution and individual-level victim cooperation remains poorly understood. Existing literature typically focuses on either optimizing resource allocation through mathematical models (focusing on efficiency, cost, or fairness) or analyzing cooperation mechanisms using evolutionary game theory in isolation. A significant gap exists in modeling the asymmetric cross-layer feedback between shelter organizations and the victims they house, particularly under conditions of scarcity where incentives and punishments must be calibrated to prevent free-riding and sustain collective action.

Methodology
The study proposes a structured two-layer network evolutionary game model to capture the coupling between shelters (upper layer) and victims (lower layer).

1. Network Topology:

   • Upper Layer: Shelters are modeled as nodes in a scale-free network (Barabási–Albert model) representing organizational interactions.
   • Lower Layer: Victims within each shelter are distributed on an $L \times L$ two-dimensional lattice, representing local interactions.
   • Coupling: The layers are coupled such that the payoff of the upper layer influences the lower layer, and the cooperation level of the lower layer feeds back into the upper layer's payoff structure.

2. Game Dynamics:

   • Upper Layer (Shelters): A continuous-strategy Public Goods Game (PGG). Shelters contribute a portion of their available resources ($s_i = \theta_i y_i$) to a common pool. The total group payoff is enhanced by a factor $r$. Shelters receive government subsidies ($\sigma$) for contributions but face punishment ($\gamma_1$) if the victim cooperation level within their shelter falls below a threshold $\beta$.
   • Lower Layer (Victims): A binary-strategy evolutionary game (Cooperate $C$ or Defect $D$). Victims receive a per-capita payoff derived from the shelter's upper-layer outcome. Cooperators pay a cost $\kappa$, while defectors face a punishment $\gamma_2$ that is dynamic: $\gamma_2 = \gamma_0 + \alpha(1 - \theta_i)$. This means punishment intensifies if the shelter's contribution ($\theta_i$) is low, linking organizational performance to individual enforcement.

3. Update Rules:

   • Strategy updates in both layers follow the Fermi rule, where players imitate neighbors with higher payoffs based on selection intensity $k$.
   • The study employs Monte Carlo simulations on two network types: a theoretical scale-free network and a real-world spatial network constructed from the geographic coordinates of shelters in Beijing.

Key Contributions

• Novel Framework: The paper introduces a coupled two-layer model that integrates continuous resource redistribution strategies at the organizational level with discrete cooperative behaviors at the individual level, explicitly modeling asymmetric cross-layer feedback.
• Mechanism Analysis: It systematically investigates the non-linear effects of three institutional parameters: the public goods enhancement factor ($r$), government subsidies ($\sigma$), and punishment mechanisms (baseline $\gamma_0$, execution coefficient $\alpha$, and shelter punishment $\gamma_1$).
• Targeted Intervention Strategy: The study evaluates the efficacy of random versus structure-based (targeted) punishment deployment, specifically focusing on high-degree "hub" shelters.

Key Results

• Threshold Effects of Incentives:
  • Enhancement Factor ($r$): Moderate values of $r$ promote cooperation by amplifying returns. However, excessive $r$ leads to diminishing marginal benefits, causing shelters to reduce inputs and victims to free-ride, ultimately lowering overall cooperation.
  • Subsidies ($\sigma$): Similar to $r$, moderate subsidies support cooperation, but excessive subsidies weaken intrinsic motivation, leading to a decline in victim cooperation as they rely on guaranteed payoffs rather than mutual aid.
• Role of Punishment:
  • Dominance of Baseline Punishment: The victim-baseline punishment coefficient ($\gamma_0$) is the decisive factor in suppressing defection. A transition from low to high cooperation occurs when $\gamma_0$ exceeds a critical threshold (approx. 0.9).
  • Execution and Shelter Punishment: The execution coefficient ($\alpha$) ensures the effectiveness of punishment, while the shelter punishment coefficient ($\gamma_1$) plays a supplementary role.
  • Combined Measures: Only a combination of all three penalty types sustains high cooperation levels; single measures are insufficient.
• Spatial Deployment Efficiency:
  • Targeted vs. Random: Targeted punishment applied to highly connected (hub) shelters significantly enhances system-wide cooperation more efficiently than random deployment. This approach leverages network centrality to spread cooperative norms at a lower intervention cost.
  • Asymmetric Coupling: The upper layer (shelters) exerts a direct and dominant influence on the lower layer. Conversely, the lower layer's feedback is more moderate and delayed. High shelter input initially suppresses victim cooperation (reducing the need for mutual aid) but eventually promotes it through sustained supervision and punishment.
• Real-World Validation: Simulations on the Beijing shelter network confirm the generality of the model. While geographical constraints in the real network slow the diffusion of cooperation compared to synthetic scale-free networks, the evolutionary trends and the effectiveness of the mechanisms remain consistent.

Significance and Claims
The paper claims to provide practical insights for policymakers designing post-disaster management strategies. It argues that effective recovery requires a balance between top-down coordination and grassroots collaboration. Specifically, the study suggests that:

1. Calibrated Incentives: Policies should avoid "benefit overflow" by implementing moderate, conditional subsidies rather than indiscriminate increases in aid intensity.
2. Enforceable Sanctions: A reliable punishment system with high execution coefficients is essential to eliminate the payoff advantage of defection, particularly in the early, uncertain stages of disaster response.
3. Structural Targeting: Under resource constraints, targeting interventions (punishment or incentives) at highly connected organizational hubs is more cost-effective for achieving rapid, system-wide cooperation than random distribution.

The authors conclude that while their model simplifies complex social factors (such as reputation or information asymmetry) and assumes static networks, it successfully demonstrates how calibrated incentives and enforceable sanctions can foster robust cooperation across organizational and individual levels in post-disaster environments.