On Optimality of Private Information in Bayesian Routing Games

Abstract

We study an information design problem in transportation networks, in the presence of a random state that affects the travel times on the links. An omniscient system planner -- aiming at reducing congestion -- observes the network state realization and sends private messages to the users -- who share a common prior on the network state but do not observe it directly -- in order to nudge them towards a socially desirable behavior. The desired effect of these private signals is to correlate the users' selfish decisions with the network state and align the resulting Bayesian Wardrop equilibrium with the system optimum flow. Our main contribution is to provide sufficient and necessary conditions under which optimality may be achieved by a fair private signal policy in transportation networks with injective link-path incidence matrix and affine travel time functions.