Evolutionary Dynamics of Reputation-Based Voluntary Prisoner's Dilemma Games

TL;DR

This work studies how reputation-based exit decisions affect cooperation in a reputation-conditioned voluntary Prisoner’s Dilemma. Using a four-strategy framework (C, D, CE, DE) with monitoring cost $c$ and exit payoff $\epsilon$, it shows that well-mixed populations sustain cooperation only when $\epsilon > c$, yielding a continuum of mixed equilibria; in structured networks, topology enables multiple exit-incentive–dependent pathways, including local cyclic dominance and persistent oscillations, mediated by conditional exiters. The results highlight how exit incentives and interaction topology jointly shape cooperative outcomes in distributed systems, with implications for designing robust multi-agent platforms and for extending reputation-based exit to multiplayer settings. The approach provides a principled way to incorporate reputation into participation decisions and reveals rich dynamical regimes beyond traditional voluntary participation models.

Abstract

Cooperation underlies many natural and artificial systems. While voluntary participation can sustain cooperation without informational assumptions, real interactions are rarely anonymous, leaving the joint effects of participation and reputation insufficiently understood. We propose a reputation-based voluntary Prisoner's Dilemma in which agents incur a monitoring cost to inspect opponents and decide whether to exit an interaction for a fixed incentive to avoid exploitation or to default to cooperation or defection. We show that reputation-conditioned exit generates multiple coexistence pathways that sustain cooperation across population structures. In well-mixed populations, cooperation persists through stable mixed coexistence, whereas in structured populations, exit-incentive-dependent regimes emerge, including local cyclic dominance and persistent oscillations. Together, these results extend voluntary participation frameworks and underscore the role of exit-incentive design in cooperative multi-agent systems.

Figures (5)

• Figure 1: Reputation-based exit sustains cooperation through stable coexistence when exit incentives exceed monitoring costs. Phase portraits show four-strategy replicator dynamics on a two-dimensional simplex projection. (A) When monitoring costs exceed exit benefits ($c > \epsilon$), defection is the unique asymptotically stable equilibrium (black filled circle). (B) When exit benefits exceed monitoring costs ($c < \epsilon$), defection loses global stability and trajectories converge to a one-dimensional manifold of mixed equilibria $E_7$ (black line) involving cooperation, defection, and exit strategies ($CE$ and $DE$). Black filled circles denote representative stable equilibria on this manifold, whereas gray circles indicate unstable equilibria. Streamlines indicate the direction of selection, and color encodes evolutionary speed, with warmer colors indicating faster dynamics. Parameters are fixed at $b=1.5$ and $c=0.4$; the exit incentive is $\epsilon=0.2$ in (A) and $\epsilon=0.6$ in (B).
• Figure 2: Structured populations enable the coexistence of cooperation and reputation-based exit strategies through multiple pathways. (A) Schematic of the three degree-regular network topologies: regular lattice, regular small-world network, and random regular graph. (B, C) Phase diagrams showing the stationary composition of strategies as functions of the exit incentive $\epsilon$ and monitoring cost $c$ for the regular lattice (left), regular small-world network (middle), and random regular graph (right). Results are shown for (B) a weak dilemma ($b = 1.02$) and (C) a strong dilemma ($b = 1.5$). Each labeled region (e.g., $C+D+CE$ or $C+CE+DE$) indicates the subset of strategies that persist with nonzero stationary frequency. Regions labeled $O$ denote parameter regimes exhibiting persistent oscillations with coexistence of all four strategies.
• Figure 3: Adjusting exit incentives reshapes dominance among cooperators, defectors, and exit strategies in structured populations. Shown are the stationary fractions of strategies as functions of the exit incentive $\epsilon$ for the regular lattice (left), regular small-world network (middle), and random regular graph (right). Results are shown for (A) weak dilemma strength ($b=1.02$) and (B) strong dilemma strength ($b=1.5$). Vertical dashed lines indicate transitions between coexistence regimes identified in Fig. 2 and labeled above each panel. The monitoring cost is fixed at $c=0.4$ in all panels. Light and dark red denote $C$ and $CE$, respectively, while light and dark blue denote $D$ and $DE$.
• Figure 4: Conditional exiters mediate exit-incentive–dependent coexistence routes via local cyclic dominance. Representative snapshots of spatial strategy configurations for exit incentives $\epsilon = 0.1, 0.4, 0.5$, and 0.9 (top to bottom). For each row, the first and last panels show the initial and final configurations, while the intermediate panels (second to fourth columns) correspond to selected intermediate time steps chosen for illustration. Simulations are performed on a square lattice with dilemma strength fixed at $b=1.5$ and monitoring cost at $c=0.4$. Colors denote strategies: $C$, light red; $D$, light blue; $CE$, red; $DE$, dark blue.
• Figure 5: Structured populations can generate persistent oscillatory dynamics. Time series of strategy fractions for (A) a complete graph (well-mixed population), (B) a regular lattice, (C) a regular small-world network, and (D) a random regular graph. To illustrate representative dynamics across population structures, the dilemma strength is set to $b=1.5$ for the complete graph and to $b=1.02$ for the structured populations. The monitoring cost and exit incentive are fixed at $c=0.4$ and $\epsilon=0.9$, respectively. Time axes differ across panels for visualization purposes.

Theorems & Definitions (14)

• Theorem 3.1
• proof
• Theorem 3.2
• proof
• Theorem 3.3
• proof
• Theorem 3.4
• proof
• Theorem 3.5
• proof