Social Contagion of Altruistic Punishment in Heterogeneous Networked Populations

 Altruistic punishment—incurring personal cost to sanction norm violators—is key to sustaining cooperation, yet its emergence in large societies remains puzzling. Using agent-based modeling on scale-free networks with 10,000 nodes, we explored how the spread of punitive behavior depends on network topology and memory length. Unlike prior work assuming homogeneous mixing, our model incorporates two distinct agent types: "moralists" who punish defectors unconditionally, and "strategic punishers" who punish only after observing others doing so. Simulations ran for 5,000 generations under a public goods game with punishment costly to punishers (cost=2) and to defectors (fine=4). Results show that strategic punishers act as social catalysts: a single strategic punisher in a high-degree node (degree > 50) triggers cascades of punishment imitation, reaching 78% of the population within 150 generations. Moralists alone fail to spread beyond clustered neighborhoods unless the clustering coefficient drops below 0.2. We derived a critical threshold: when the ratio of imitation probability to punishment cost exceeds 1.3, altruistic punishment becomes globally stable. Introducing reputation memory (agents remember last 10 interactions) accelerates contagion by 40% but also increases vulnerability to "second-order free riders" who imitate punishment without contributing to the public good. Network rewiring experiments show that dynamic networks (agents can sever ties with defectors) produce bistable outcomes: either full cooperation or complete collapse of punishment, depending on initial seed placement. Statistical analysis of 1,000 runs revealed that punishment contagion follows a logistic growth pattern with a characteristic lag phase of 50 generations. Interestingly, intermediate levels of noise (5–10% random strategy mutations) promote robustness by preventing fixation of non-punishing strategies. Our findings reconcile conflicting experimental results in behavioral economics by showing that the spatial structure of social networks fundamentally alters the cost-benefit calculus of punishment. Policy implications include designing decentralized norm enforcement systems (e.g., community-led monitoring) that leverage high-influence individuals.