Intervention Mechanism Design for Networks With Selfish Users

Abstract: We consider a multi-user network where a network manager and selfish users interact. The network manager monitors the behavior of users and intervenes in the interaction among users if necessary, while users make decisions independently to optimize their individual objectives. In this paper, we develop a framework of intervention mechanism design, which is aimed to optimize the objective of the manager, or the network performance, taking the incentives of selfish users into account. Our framework is general enough to cover a wide range of application scenarios, and it has advantages over existing approaches such as Stackelberg strategies and pricing. To design an intervention mechanism and to predict the resulting operating point, we formulate a new class of games called intervention games and a new solution concept called intervention equilibrium. We provide analytic results about intervention equilibrium and optimal intervention mechanisms in the case of a benevolent manager with perfect monitoring. We illustrate these results with a random access model. Our illustrative example suggests that intervention requires less knowledge about users than pricing.