Global Profits, Local Decisions: Why Global Cooperation Falters in Multi-level Games

Jinhua Zhao, Xinguo Yu, Rui Ding, Cuiling Gu, Xianjia Wang

Abstract

Global cooperation is often hindered by the "Tragedy of the Commons," where individual incentives to defect outweigh collective benefits. This paper explores this phenomenon through a multi-level evolutionary game theory lens, modeling interactions across pairwise, local, and global scales. The study reveals that because individuals update their strategies by comparing payoffs with immediate neighbors, global benefits—which are distributed equally—become "invisible" in the decision-making process. Consequently, global cooperation fails to emerge unless supported by strong local or pairwise incentives.

Introduction

The challenge of global cooperation, such as international climate agreements, is frequently framed as a Tragedy of the Commons. While the collective benefits of cooperation are immense, individual actors often find it more advantageous to "defect" and reap the rewards of others' contributions without incurring costs. Traditional game theory models often simplify these interactions into isolated games. However, real-world social and political structures are inherently hierarchical, involving simultaneous direct (pairwise), regional (local), and international (global) interactions.

This research addresses how these nested incentives interact. Specifically, it investigates why high global profit rates often fail to motivate cooperative behavior at the individual level, leading to the failure of global initiatives despite their clear collective value.

Methodology

The researchers utilized a multi-level Public Goods Game (PGG) model set on a periodic 2D square lattice. Each player in the population interacts with four immediate neighbors.

Game Structure

Players participate in three concurrent levels of interaction:

  1. Pairwise: Direct interactions with immediate neighbors.
  2. Local: Interactions within a broader neighborhood, including first and second-order neighbors.
  3. Global: A single interaction involving the entire population.

Strategy and Evolution

The model tested two strategy configurations:

  • Binary Strategy: Players are either pure cooperators or pure defectors across all three levels.
  • Level-Based Strategy: Players can independently choose to cooperate or defect at each specific level (e.g., cooperating locally but defecting globally).

Strategy evolution is governed by the Fermi function. In each round, a player compares their total payoff to that of a randomly selected neighbor. If the neighbor's payoff is higher, the player is likely to imitate the neighbor's strategy, simulating social learning and the propagation of successful behaviors.

The Invisibility of Global Profit

A central finding of the study is the "invisibility" of global profit in the binary strategy setting. Because the global payoff is distributed equally to all members of the population, it does not create a "payoff gap" between neighbors. When a player calculates the difference between their payoff and a neighbor's to decide whether to switch strategies, the global component ($r_g$) cancels out mathematically.

In the level-based setting, this dynamic creates a systematic disadvantage for global cooperators. A global cooperator incurs a cost that a global defector avoids, yet both receive the same global benefit. In local comparisons, the global cooperator always appears less successful than a global defector, making global cooperation an "evolutionary loser."

Key Results and Phase Transitions

The simulation results demonstrate that the ultimate state of the population—whether it reaches a consensus of total cooperation or total defection—is determined almost exclusively by the profit rates of pairwise ($r_p$) and local ($r_l$) interactions.

Phase diagram illustrating the dominance of cooperation (C) or defection (D) as a function of pairwise ($r_p$) and local ($r_l$) profit rates.

As shown in Figure 1, the phase diagram identifies clear tipping points. Cooperation only becomes the dominant strategy when $r_p$ and $r_l$ are sufficiently high to overcome the relative advantage of defection. Notably, increasing the global profit rate ($r_g$) has no impact on these boundaries; simulations showed that even a fourfold increase in $r_g$ resulted in no change to the evolutionary trajectory.

Simulation of cooperator frequency and welfare over time, showing a transition to a fully cooperative state.

Figure 2 illustrates the temporal evolution of a population toward cooperation. While the frequency of cooperators may initially dip, strong local incentives eventually allow cooperative clusters to expand and dominate the lattice. Conversely, in level-based games, strategies involving global cooperation are eliminated early in the process, leaving the population to settle into a state of global defection regardless of the global stakes.

Discussion and Limitations

The study highlights a critical structural barrier to global cooperation: the local nature of human and institutional decision-making. If actors only change their behavior by imitating more successful peers, and if global benefits do not contribute to relative success, then global initiatives are doomed to fail without local reinforcement.

The authors acknowledge several simplifications, such as the use of a regular 2D lattice and homogeneous players. Future work could explore more complex networks or the introduction of sanctions and rewards, which might provide the necessary "payoff gap" to make global cooperation evolutionarily viable.

Summary updated

BibTeX
@article{zhao2026global,
  author = {Jinhua Zhao and Xinguo Yu and Rui Ding and Cuiling Gu and Xianjia Wang},
  title = {{Global Profits, Local Decisions: Why Global Cooperation Falters in Multi-level Games}},
  journal = {Expert Systems with Applications},
  year = {2026},
  doi = {10.1016/j.eswa.2025.129450},
}
Figure