Source-linked AI summary
Resolving social dilemmas on evolving random networks
Attila Szolnoki, Matjaz Perc
TL;DR
The paper asks whether evolving random interaction networks can promote cooperation in social dilemmas beyond the effects of static networks. It applies strategy-independent link deletions and timed additions, finding that the timescale of additions tunes mechanisms from Red Queen dynamics to group selection. With sufficiently large timescale separation, cooperation dominates across an extensive region covering all major social dilemma types.
Problem
Most evolutionary-game studies on complex networks considered static interaction networks, motivating analysis of whether evolving random networks can promote cooperation through additional dynamical mechanisms.
Method
The model deletes links when players adopt strategies and adds random links every τ full Monte Carlo steps on random networks across prisoner’s dilemma, snowdrift, and stag-hunt games.
Results
With large timescale separation, a powerful group-selection mechanism emerges and full cooperator dominance spans an extensive T −S region covering all major social dilemma types.
Takeaways & Limitations
The link-addition frequency can tune which cooperation-promoting mechanism emerges, including Red Queen dynamics and group selection.
Abstract
from arXiv · showhide
We show that strategy independent adaptations of random interaction networks can induce powerful mechanisms, ranging from the Red Queen to group selection, that promote cooperation in evolutionary social dilemmas. These two mechanisms emerge spontaneously as dynamical processes due to deletions and additions of links, which are performed whenever players adopt new strategies and after a certain number of game iterations, respectively. The potency of cooperation promotion, as well as the mechanism responsible for it, can thereby be tuned via a single parameter determining the frequency of link additions. We thus demonstrate that coevolving random networks may evoke an appropriate mechanism for each social dilemma, such that cooperation prevails even by highly unfavorable conditions.
Introduction. –
The paper investigates whether simple, strategy-independent coevolutionary changes to random interaction networks can generate dynamical mechanisms that promote cooperation and resolve social dilemmas. Link deletions and additions preserve network heterogeneity while making the mechanism tunable through the link-addition timescale τ.
- Introduction. –: Existing studies mostly used static interaction networks, motivating analysis of evolving interaction structures.Complex-network games have often supported cooperation, but the majority of studies considered static networks.
- Introduction. –: Coevolutionary rules can promote cooperation through heterogeneous states and newly emerging dynamical processes.The paper focuses on network adaptations occurring alongside strategy evolution.
- Introduction. –: Link deletions follow strategy adoption, whereas new links are added every τ game iterations, making τ the tunable addition timescale.The balance between these processes can be adjusted by making link additions faster or slower than deletions.
- Introduction. –: The coevolutionary rules largely preserve the initial random topology and heterogeneity while allowing Red Queen and group-selection mechanisms to emerge.The responsible mechanism depends on τ and the governing social dilemma.
- Introduction. –: Strategy-independent network adaptation can resolve the governing social dilemma in favor of collective wellbeing.The claimed effect operates at the macroscopic level of evolutionary games.
Social dilemmas and setup. –
The model combines three two-strategy social dilemmas with a random interaction network whose links adapt after strategy changes and at intervals controlled by τ. The rule preserves game participation while remaining independent of strategy labels.
- Social dilemmas and setup. –: The model covers prisoner’s dilemma, snowdrift, and stag-hunt games using payoff parameters R = 1, P = 0, −1 ≤S ≤1, and 0 ≤T ≤2.The payoff ordering determines which of the three social dilemma types applies.
- Social dilemmas and setup. –: Players begin as cooperators or defectors with equal probability on a random network of N individuals with average degree kavg = 4.Duplicate links are omitted during network construction.
- Social dilemmas and setup. –: When a player adopts a strategy, it retains only the donor link and deletes its other links, reducing its degree to kx = 1.Figure 1 schematically depicts this simultaneous strategy-adoption and link-deletion rule.
- Social dilemmas and setup. –: Every τ full Monte Carlo steps, each individual may form a link to a randomly chosen unconnected player, counteracting link depletion.This addition process is described as aging in the model.
- Social dilemmas and setup. –: The adaptation rule is strategy independent, and kmax = 500 is sufficiently large to avoid influencing the initial random topology.Detached players are relinked so every player maintains at least one neighbor.
Results. –
At τ = 1, evolving random networks sustain cooperation across the snowdrift quadrant through oscillatory dynamics, while increasing τ expands cooperator dominance and changes the underlying mechanism toward group selection. The results link low τ to Red Queen dynamics and high τ to stronger, broader cooperation promotion.
- τ = 1 phase diagram: At τ = 1, cooperators persist throughout the snowdrift quadrant, including a broad mixed-state region with oscillatory solutions, unlike static random networks.Complete cooperator dominance occurs for sufficiently low T, while mixed C + D and oscillatory states occupy a broad region.
- τ = 1 phase diagram: Along the snowdrift diagonal, oscillations begin near r ≅ 0.50 through a continuous transition and terminate at r = 0.732 through a discontinuous transition.Cooperator density follows complete dominance up to r ≅ 0.41, then oscillatory amplitudes grow before the abrupt transition back to a stationary state.
- τ = 1 phase diagram: The oscillatory regime reflects a Red Queen mechanism driven by interplay between cooperator density and network structure.At the discontinuous transition, steady and oscillatory solutions coexist, with the selected attractor depending on the initial cooperator density.
- Increasing τ: At τ = 500, full cooperator dominance spans the entire traditional snowdrift region and weak prisoner’s dilemma, while defectors dominate the stag hunt only for high T and S < −0.6.The mixed and oscillatory phases disappear relative to the τ = 1 diagram.
- Increasing τ: Slower link additions let cooperative domains grow around high-degree players and promote influential hubs, yielding a group-selection mechanism across all three social dilemmas.This replaces the Red Queen mechanism in snowdrift games and predominantly heterogeneity-based promotion at τ = 1.
- Increasing τ: Increasing τ produces cascade-like dormancy in cooperator density, as slow link additions reconnect detached groups and trigger renewed avalanches of strategy adoption.At τ = 500, dormant intervals surpass active phases, supporting the emergence of spontaneous group selection.
Summary. –
Evolving random networks promote cooperation most strongly when link deletions and additions operate on sufficiently separated time scales. Slow additions produce group selection and broad cooperator dominance, while faster additions support alternative mechanisms such as the Red Queen.
- Large time-scale separation enables spontaneous group selection and full cooperator dominance across an extensive T−S region covering all major social dilemmas.Frequent link additions hinder isolated homogeneous groups, shifting cooperation promotion toward network heterogeneity or the Red Queen mechanism.
- In the snowdrift game, the Red Queen mechanism emerges from oscillatory changes in network structure interacting with cooperator density.The oscillatory phase can coexist with steady states, indicating bistability dependent on initial conditions.