Congestion games model shared resources where each user's experience degrades as more people use the same resource — roads, networks, cloud services. The socially optimal allocation rarely coincides with the Nash equilibrium because individuals ignore the externalities they impose on others. The standard fix is monetary tolls: price the externality, make users pay for congestion, and the equilibrium shifts toward efficiency.
The problem with money is fairness. Monetary tolls are regressive — they price out users with less wealth, not users with less need. A rich commuter pays the congestion charge without thinking; a poor commuter is excluded. Efficiency and equity diverge.
The token economy replaces money with tokens — artificial currency that exists only within the game. Every participant receives the same token endowment. Tokens spent on tolls are redistributed, maintaining a closed economy with no wealth accumulation across rounds. The mechanism preserves the incentive alignment of pricing (tokens make congestion costly) while eliminating the distributional injustice (everyone starts equal every period).
The mathematical contribution is proving that integer-valued tolls in closed form can steer the aggregate dynamics to the optimal allocation from any initial condition, using mean-field approximations of bounded-rationality dynamics. The convergence is global — the system doesn't just have a good equilibrium, it reaches it from anywhere.
The structural through-claim: money is an overpowered tool for coordination. It carries information about preferences (willingness to pay) bundled with information about resources (ability to pay). Tokens unbundle these — they carry preference information without resource information. For pure coordination problems like congestion, the preference information is sufficient. The resource information is not just unnecessary; it is distortionary.
(arXiv:2603.18094)