Dynamics of Cooperation in Minority Games in Alliance Networks

Xinjie Zhang, Yong Tang, Jason Xiong, Weijia Wang, Yicheng ZhangSustainability 2018 · DOI

TL;DR

This paper explores how cooperation evolves in social systems modeled as "alliance networks" using a "minority game" framework. By simulating players who adapt their strategies and even rewire network connections, the study finds that dynamic network structures and global information lead to more stable and cooperative outcomes compared to static networks and local decision-making.

What Problem Does This Paper Address?

Understanding how cooperation emerges and is sustained in complex social systems is a fundamental challenge across various disciplines. While game theory and network science offer valuable insights, there's a need to bridge these fields to model how cooperation dynamics are influenced by the underlying social structures (alliance networks) and how these structures, in turn, evolve based on player interactions. Specifically, this paper addresses the question of how players making binary decisions in an alliance network, with varying levels of information (local vs. global), influence the network's overall cooperation and stability. The authors aim to fill a gap by integrating the "minority game" model with dynamic alliance networks, allowing for network evolution alongside player strategy adaptation.

Methods

The researchers employ an agent-based simulation approach to model an "evolutionary minority game" played on alliance networks (MGAN). In this model:

  • Players and Networks: A population of N players is arranged on different network topologies: Erdős-Rényi (ER) random networks, scale-free (SF) networks, and small-world (SW) networks. An edge between two players represents a cooperative alliance.
  • Minority Game: Players make binary decisions (A or B) aiming to be in the minority group to receive a positive score (+1). Being in the majority results in a negative score (-1), and ties result in a zero score.
  • Information Levels: Two main strategies are considered:
  • Global Score Strategy: Players have access to global information about the entire population's choices.
  • Local Score Strategy: Players only consider the choices of their immediate neighbors. This is further divided into "local cumulative" (summing scores from all local games) and "local" (single-round local game) strategies.
  • Decision Adaptation: Players dynamically adjust their probability of choosing A or B based on their scores and an "aversion coefficient" ($\alpha$). A lower $\alpha$ encourages changing strategy, while a higher $\alpha$ (e.g., 1.1) signifies stubbornness.
  • Network Rewiring: A key novelty is the introduction of a "rewiring policy." When a player is in the majority, they have the option to break an alliance with a majority neighbor and form a new alliance with a non-neighbor. This allows the network topology to evolve dynamically.

The simulations track various network properties, including average score, player choice ratios, average node degree, betweenness centrality, eigenvector centrality, and clustering coefficient, to assess cooperation and network stability.

Datasets

Not discussed in the paper. The study relies on simulated data generated from the agent-based model on various synthetic network structures.

Key Insights and Ideas

  • Dynamic Networks Foster Cooperation: The most significant insight is that allowing networks to evolve through rewiring (players forming and breaking alliances) leads to better overall systemic performance, higher scores, and increased stability compared to static networks. This suggests that adaptability in social structures is crucial for sustained cooperation.
  • Global Information is Key for Stability: Players with global information (global score strategy) tend to achieve more balanced outcomes and stable decision-making, approaching an equilibrium state. Local information, while mimicking real-world constraints, can lead to imbalances and "collective herding."
  • Competitive Decision-Making Can Improve Cooperation: Even starting from a "poor initial start" (low cooperation), competitive decision-making (as in the minority game) can drive the network towards a state with a higher average degree, indicating improved cooperation.
  • Stubbornness Hinders Cooperation: Conversely, stubbornness in decision-making (high $\alpha$ values) discourages cooperation and leads to poorer outcomes, especially in certain network types like scale-free networks.
  • Network Rewiring as a "Second Chance": The rewiring mechanism acts as a way for players to escape unfavorable majority positions, allowing for the abandonment of disadvantaged alliances and the formation of new, potentially more beneficial ones.

Key Results

  • Higher Average Scores with Rewiring: Simulations consistently showed that the network rewiring strategy resulted in a higher average cumulative score for the population across different network types and strategies.
  • Stability in ER and SW Networks: For Erdős-Rényi (ER) and small-world (SW) networks, both global and local strategies (with rewiring) led to a stable ratio of players choosing side A, generally around 0.45.
  • Imbalance in SF Networks: Scale-free (SF) networks showed more pronounced imbalances with local strategies, where a significant portion of players might consistently choose one side, leading to lower scores. Global strategies, however, helped stabilize SF networks.
  • Average Degree Increases with Competition: The average node degree, a proxy for cooperation, generally increased over time, especially with competitive decision-making and lower aversion coefficients, indicating improved network connectivity and cooperation.
  • Betweenness Centrality Decreases: Average betweenness centrality tended to decrease over time, suggesting players become less critical as intermediaries in the evolving network structure.
  • Eigenvector Centrality Stabilizes: Average eigenvector centrality showed stable trends with minor fluctuations, indicating a consistent influence structure within the population.
  • Clustering Coefficient Varies: The clustering coefficient showed more fluctuation in ER networks compared to SF and SW networks, suggesting different levels of local interconnectedness.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Limited Local Information Scope: The current model only considers local information from immediate neighbors (distance of one). Future work could explore the impact of considering information from neighbors at greater distances.
  • Simplified Cooperation Measurement: While degree, decision balance, and score are used, new metrics could be developed to more comprehensively model cooperation in evolving games.
  • Incomparable Score Strategies: The global and local score strategies are inherently different, making direct comparison of their absolute results challenging. A normalized framework could improve comparability.
  • Binary Decisions vs. Thresholds: The model uses binary decisions. Future research could incorporate the Deffuant model's approach, where cooperation is established only if players' decisions or opinions exceed a certain threshold. This would allow for a more nuanced understanding of cooperation formation.
  • Exploring Other Network Types and Dynamics: The study focused on ER, SF, and SW networks. Investigating other network structures and more complex rewiring rules could yield further insights.

Summary updated

Abstract

Alliance networks are the underlying structures of social systems in business, management, and society. The sustainability and dynamics of a social system rely on the structural evolutions of the topologies. Understanding the evolution sheds light on the dynamics and sustainability of a social system. Minority game models have been successfully applied across social science, economy, management, and engineering. They provide simple yet applicable modeling to articulate the evolutionary cooperation dynamics of competitive players in binary decision situations. By extending the minority games played in alliance networks, the cooperation in structured systems of different network topologies is analyzed. In this model, local and global score strategies are considered with and without cooperation rewiring options. The cooperation level, the score, and the topological properties are investigated. The research uses a numerical simulation approach on random networks, scale-free networks, and small-world networks. The results suggest that the network rewiring strategy leads to higher systemic performance with a higher score and a higher level of stability in decision-making. Competitive decision-making can lead to a higher level of cooperation from a poor initial start. However, stubbornness in decision-making can lead to a poor situation when cooperation is discouraged. Players with local or global information adopt local and global score strategies. The results show that local strategies might lead to imbalance, while a global strategy might achieve a relatively stable outcome. This work contributes to bridge minority games in structured networks to study the cooperation between formation and evolution, and calls for future minority game modeling on social networks.
BibTeX
@inproceedings{zhang2018dynamics,
  author = {Xinjie Zhang and Yong Tang and Jason Xiong and Weijia Wang and Yicheng Zhang},
  title = {{Dynamics of Cooperation in Minority Games in Alliance Networks}},
  booktitle = {Sustainability},
  year = {2018},
  doi = {10.3390/su10124746},
}
Figure