Information Signaling With Concurrent Monetary Incentives in Bayesian Congestion Games

The uncertainty held by a system's users can cause ineffective decision-making. Nowhere is this more apparent than in transportation networks, where drivers' uncertainty over current road/traffic conditions can negatively alter their routing choices. To alleviate this, an informed system operator may signal information to uninformed users to persuade them into taking more preferable actions (e.g., Google/Apple maps providing live traffic updates). In this work, we study public signalling mechanisms in the context of Bayesian congestion games. We observe the phenomenon that though revealing information can reduce system cost in some settings, in others, it can induce worse performance than not signalling at all. However, we find an important relationship between information signalling and monetary incentives: by utilizing both mechanisms concurrently, the system operator can guarantee that revealing information does not worsen performance. We prove these findings in a general class of Bayesian congestion games. To understand the magnitude at which information signalling can affect system performance, we put a deeper focus in the class of parallel networks with polynomial latency functions and analytically characterize bounds on the change in system cost from signalling. Finally, we consider the problem of solving for optimal signals with and without the concurrent use of monetary incentives. We construct solvable optimization problems whose solutions give optimal signalling policies even when the signalling policy is limited in its support; we then quantify the benefit of these and other signalling mechanisms in numerical examples.