Abstract
Access to resources strongly constrains the decisions we make. While we mightwish to offer every student a scholarship, or schedule every patient forfollow-up meetings with a specialist, limited resources mean that this is notpossible. When deploying machine learning systems, these resource constraintsare simply enforced by varying the threshold of a classifier. However, thesefinite resource limitations are disregarded by most existing tools for fairmachine learning, which do not allow the specification of resource limitationsand do not remain fair when varying thresholds. This makes them ill-suited forreal-world deployment. Our research introduces the concept of"resource-constrained fairness" and quantifies the cost of fairness within thisframework. We demonstrate that the level of available resources significantlyinfluences this cost, a factor overlooked in previous evaluations.