Partner-switching games with initial-network-dependent interaction patterns
Hsuan-Wei Lee, Lehigh University
Abstract: According to prior research on partner-switching games, altering partners could stabilize cooperation. Yet the role of per-edge (interactive diversity, ID) vs. per-node (interactive identity, II) strategies under a single co-evolutionary rule remains deficiently understood. In this talk, we propose an edge-based game as a natural extension of a well-studied node-based model in adaptive networks. By varying only the initial distribution of strategies, our framework can transition from interactive identity to interactive diversity without altering the update rules. We show that, in contrast to node-based scenarios, allowing each edge to carry its own strategy label grants defectors an exclusive "learning-by-rewiring" mechanism, thereby significantly enlarging the parameter region in which cooperation thrives. We find that cooperation in the edge-based (ID) setup can emerge even from low initial cooperation, because defectors may incrementally increase their cooperative ratio by severing disadvantageous edges. This effect is absent in node-based (II) games, which constrain each node to a single global strategy. Moreover, numerical evidence suggests that higher initial cooperation does not guarantee higher final cooperation in ID games, underscoring the non-monotonic and homophily-driven nature of cooperative cluster formation. Our results thus clarify how partner-switching, when combined with interactive diversity, yields complex threshold behaviors unseen in the conventional node-based setting. Finally, we use pair approximation to illustrate why accurate theoretical predictions require richer local information in edge-based models, where each directed edge may evolve independently.
Tea and refreshments available at 3:00 p.m. in the Assmus Conference Room (CU 212).
