Private Allocation of Public Goods

Abstract

We study the problem of designing a truthful mechanism for fair allocation of divisible public goods. We consider a setting with n agents and m items. Each item is associated with a size, and the total size of the allocated items must not exceed the available capacity. All agents can access an allocated item, but agents might have different valuations for different items. To aggregate agents’ preferences in a fair and efficient way, we focus on the notion of core, which incorporates Pareto efficiency and sharing inventive. In public good settings, agents might have the incentive to misreport their preferences and free ride on the items allocated by others. To address this issue, we present an approximately truthful mechanism. Our mechanism solves a convex optimization problem in a differentially private manner to find a fair allocation.