Understanding Fairness in Online Allocation
In many settings, resources are allocated among agents over time without the use of monetary transfers: cloud resources among employees, food among food banks, vaccines among hospitals, etc. The underlying aim is to try and be 'fair' in these allocations...but what exactly do we mean?
Understanding fairness in allocations is one of the most beautiful and relevant topics today, with deep connections to market design, optimization and normative philosophy. Building on a foundational result of Varian's that relates these approaches, I will describe some of our work in (a) understanding market mechanisms for fair online allocation based on artificial credits, (b) the power and limits of optimization approaches, and (c) how modern ideas in control and online decision-making give strong guarantees for stochastic versions of these problems, and how these methods work in practice for food-bank allocation.