Source-linked AI summary
If players are sparse social dilemmas are too: Importance of percolation for evolution of cooperation
Zhen Wang, Attila Szolnoki, Matjaz Perc
TL;DR
The paper asks how population density affects cooperation in spatial social dilemmas on interaction graphs with vacant sites. It studies stochastic pairwise imitation across dilemmas and graph structures, then tests the mechanism with myopic updating. Cooperation is optimally promoted near the graph’s percolation threshold under imitation, whereas myopic updating makes that threshold irrelevant by preventing efficient information spread.
Problem
The paper examines how different population densities affect cooperation in spatial social dilemmas on interaction graphs containing vacant sites.
Method
The study compares stochastic pairwise imitation across three social dilemmas and multiple lattices, using myopic updating as a reverse test of the proposed mechanism.
Results
Under pairwise imitation, the percolation threshold is a universal indicator of the population density that optimally promotes cooperation across the studied dilemmas and interaction graphs.
Takeaways & Limitations
Cooperation is favored when players are connected enough for cooperative information to spread but sufficiently sparse to hinder defector invasion.
Abstract
from arXiv · showhide
Spatial reciprocity is a well known tour de force of cooperation promotion. A thorough understanding of the effects of different population densities is therefore crucial. Here we study the evolution of cooperation in social dilemmas on different interaction graphs with a certain fraction of vacant nodes. We find that sparsity may favor the resolution of social dilemmas, especially if the population density is close to the percolation threshold of the underlying graph. Regardless of the type of the governing social dilemma as well as particularities of the interaction graph, we show that under pairwise imitation the percolation threshold is a universal indicator of how dense the occupancy ought to be for cooperation to be optimally promoted. We also demonstrate that myopic updating, due to the lack of efficient spread of information via imitation, renders the reported mechanism dysfunctional, which in turn further strengthens its foundations.
Results
Under stochastic pairwise imitation, cooperation is maximized at an intermediate population density tied to the interaction graph’s percolation threshold. This pattern extends across social dilemmas, whereas myopic updating removes the density optimum because information cannot spread efficiently.
- Results: Stochastic imitation avoids frozen states and initial-condition sensitivity caused by deterministic best-neighbor updating.The stochastic rule introduces uncertainty into strategy adoption, revealing a consistent intermediate-density optimum.
- Results: An intermediate population density maximizes cooperation across square, honeycomb, triangular, and cubic lattices.For the prisoner’s dilemma, the optimal density shifts with the graph’s percolation threshold; the honeycomb lattice requires the highest density, while the cubic lattice requires the lowest.
- Results: At very low density, disconnected players and graph regions prevent cooperators from forming compact protective clusters.As density rises, interconnectedness enables spatial reciprocity, but higher temptation to defect can continue reducing cooperation.
- Results: For snowdrift and stag-hunt games, the percolation threshold remains a benchmark for cooperation, although stag-hunt dynamics can reach an all-cooperator state above it.The snowdrift game retains an intermediate cooperation maximum, whereas the less severe stag-hunt game can attain all-C outcomes.
- Results: Myopic updating makes population density monotonically affect cooperation and eliminates the percolation-based optimum across the studied dilemmas.In snowdrift dynamics, role separation can emerge, but its cooperation increase is practically negligible compared with imitation.
Discussion
The percolation threshold emerges as the optimal population density for cooperation under pairwise imitation across social dilemmas and diverse lattices. The mechanism depends on uncertainty in strategy adoption and on effective information spread through imitation.
- Discussion: The percolation threshold constitutes the optimal population density for resolving pairwise social dilemmas across all considered social dilemma games and a wide class of lattices.The result is also consistent with earlier findings for public-goods games governed by group interactions.
- Discussion: Some uncertainty in strategy adoption prevents frozen states and reduces dependence on initial conditions, especially at high population density.
- Discussion: Myopic updating removes the decisive effect of the percolation threshold because players can no longer exchange information directly through imitation.
- Discussion: The optimal density amplifies spatial reciprocity: low density disrupts cooperator clusters, whereas high density enables defectors to invade and split them apart.
- Discussion: Percolation links population-density effects in social dilemmas to the formation, protection, and communication of cooperative clusters.
Methods
The study models three spatial social dilemmas on several interaction-graph topologies and evaluates strategy evolution with Monte Carlo updating. Pairwise imitation uses payoff comparisons with uncertainty, while myopic updating provides an alternative rule.
- Methods: The simulations consider spatial prisoner’s dilemma, snowdrift, and stag-hunt games on square, honeycomb, and triangular interaction graphs.
- Methods: Players use binary strategies, with payoffs parameterized by R = 1, P = 0, and game-specific temptation and sucker-payoff choices.
- Methods: Each Monte Carlo step selects a player and a neighbor, compares their game payoffs, and lets the first attempt to impose its strategy on the second.
- Methods: Strategy adoption includes uncertainty K = 0.1 normalized by lattice degree k, so better-performing players are more readily imitated but worse-performing players can still be adopted.
- Methods: System sizes range from L = 200 to 1200 and relaxation times from 10^4 to 10^6 Monte Carlo steps, with results reported as size-independent.
- Methods: Myopic updating is examined as an alternative to payoff-comparison imitation.