Agent-Based Modeling in Network Formation

Abstract

This project studies network formation as the outcome of decentralized strategic interactions among heterogeneous agents. It introduces an agent-based modeling (ABM) framework that represents agents as automata with limited information and local decision rules. By lifting restrictive analytical assumptions, such as homogeneity, global interaction, and perfect information, the ABM approach allows network structures to emerge from micro-level behavior. This framework is applied to the distance-based utility model (DBUM), where agents optimize over a parameter space of decay, link cost, and myopic radius across 120 simulated configurations. Computational results confirm and extend analytical predictions. The decay parameter emerged as the primary determinant of network configuration: high decay values favor hub-dominated structures, while intermediate and low values produce complete networks when costs are low. The link cost determined the overall connectivity: low link formation cost was a necessary condition for complete network formation and high cost reliably produced empty networks. The myopic radius, on the other hand, had negligible effect on the distribution of emerging network topologies, which means that strategically relevant information is concentrated in the local neighborhood of an agent. These findings demonstrate that ABM can successfully reproduce and extend analytical predictions, and reveal dynamic properties that can not be tracked by closed-form methods.