Information Design Under Uncertainty for Vehicle-to-Vehicle Communication

The emerging technology of Vehicle-to-Vehicle (V2V) communication aims to improve road safety by allowing vehicles to share information about the world. However, information design is in general a non-trivial problem, and is only made more difficult by uncertainty about the world or agents. In this work, using an existing model of V2V communication with endogenous accident probability, we study an information designer's optimization problem under uncertainty about the "danger level" (the sensitivity of accident probability to agent behavior). First, we consider an information designer who does not know the danger level designing for agents who do; second, an informed designer designing for uninformed agents. In both cases, we present a simple characterization of the worst-case (i.e. largest accident probability) outcome that is possible under the uncertainty. When an information designer is uncertain about the world, the worst case occurs with the largest danger level. By contrast, when agents are uninformed, the worst case is caused by agents' beliefs being the lowest danger level. Both of these results simplify the optimization problem, allowing an optimal signaling policy to be more easily determined.