Supercooperation in Evolutionary Games on Correlated Weighted Networks

In this work we study the behavior of classical two-person, two-strategies evolutionary games on a class of weighted networks derived from Barab'asi-Albert and random scale-free unweighted graphs. Using customary imitative dynamics, our numerical simulation results show that the presence of link weights that are correlated in a particular manner with the degree of the link endpoints, leads to unprecedented levels of cooperation in the whole games’ phase space, well above those found for the corresponding unweighted complex networks. We provide intuitive explanations for this favorable behavior by transforming the weighted networks into unweighted ones with particular topological properties. The resulting structures help to understand why cooperation can thrive and also give ideas as to how such supercooperative networks might be built.

💡 Research Summary

This paper investigates how the introduction of degree‑correlated edge weights influences the emergence of cooperation in classical two‑person, two‑strategy evolutionary games played on complex networks. Starting from two well‑known families of unweighted scale‑free graphs – the Barabási‑Albert preferential‑attachment model and a random configuration‑model network – the authors assign to each edge a weight w_{ij} that depends on the degrees k_i and k_j of its endpoints according to w_{ij}= (k_i k_j)^{α}. The exponent α controls the strength and sign of the correlation: positive α amplifies connections between high‑degree nodes (hubs), while negative α emphasizes links between low‑degree nodes.

The study focuses on three canonical games: the Prisoner’s Dilemma (PD), the Snowdrift Game (SG), and the Stag‑Hunt (SH). For each game the authors employ two standard imitation dynamics – the deterministic replicator rule and the stochastic Fermi update – and run extensive Monte‑Carlo simulations on networks of size N=10 000 with average degree ⟨k⟩≈4. Each parameter configuration (game payoff, α, update rule) is simulated for 10 000 generations, and results are averaged over 30 independent network realizations.

The main empirical finding is that when α is positive and of moderate magnitude (approximately 1 ≤ α ≤ 2), the overall fraction of cooperators dramatically exceeds that observed on the corresponding unweighted graphs. In the PD, for example, the cooperation level rises from roughly 0.2 in the unweighted case to about 0.7 under α≈1.5. Similar boosts are observed for SG and SH, with cooperation fractions often surpassing 0.8 across large regions of the payoff parameter space. By contrast, negative α values either produce negligible improvements or even depress cooperation, indicating that the direction of the weight‑degree correlation is crucial.

To understand the mechanism, the authors transform the weighted network into an equivalent unweighted one by applying a threshold θ: only edges with w_{ij} ≥ θ are retained, all others are removed. The resulting “thresholded” graphs preserve the original degree distribution but acquire a distinctive topology: strong hub‑hub links form a dense core, while low‑degree nodes attach to this core in a star‑like fashion. This hybrid structure simultaneously exhibits high clustering, positive assortativity, and a pronounced core‑periphery organization.

In the imitation dynamics, the edge weight directly modulates the probability that a strategy is copied across that edge. Consequently, a cooperative hub that adopts cooperation exerts a disproportionately large influence on its many neighbors, rapidly converting peripheral nodes to cooperation. Conversely, if a hub defects, the low‑weight peripheral links make it difficult for defection to spread outward, because the probability of copying a defective strategy across a weak edge is small. This asymmetry creates a self‑reinforcing loop that stabilizes cooperation in the core and propagates it outward, a phenomenon the authors term “supercooperation.”

The authors also conduct a control experiment in which the same set of edge weights is randomly reassigned, destroying the systematic degree‑weight correlation while keeping the weight distribution unchanged. In this randomized scenario, cooperation levels revert to those of the unweighted baseline, confirming that the correlation—not merely the presence of heterogeneous weights—is the key driver of the observed effect.

Beyond the computational results, the paper discusses practical implications. In real‑world systems such as inter‑firm trade networks, scientific collaboration graphs, or online platforms, the intensity of interaction (e.g., transaction volume, communication frequency) often correlates with the activity level or “degree” of the participants. Designing or encouraging such positive weight‑degree correlations could therefore be a low‑cost strategy to foster cooperation, resilience, and collective welfare without restructuring the underlying network topology.

In summary, the study makes three substantive contributions: (1) it demonstrates that degree‑correlated edge weights can dramatically reshape the evolutionary dynamics of cooperation on scale‑free networks; (2) it provides a clear topological interpretation by mapping weighted graphs to unweighted cores with high assortativity and clustering; and (3) it offers actionable insight for engineering “supercooperative” networks in social, economic, and biological contexts, suggesting that targeted weighting schemes may be more practical than wholesale rewiring for promoting collective good.