Abstract
Motivated by the emergence of decentralized machine learning (ML) ecosystems,we study the delegation of data collection. Taking the field of contract theoryas our starting point, we design optimal and near-optimal contracts that dealwith two fundamental information asymmetries that arise in decentralized ML:uncertainty in the assessment of model quality and uncertainty regarding theoptimal performance of any model. We show that a principal can cope with suchasymmetry via simple linear contracts that achieve 1-1/e fraction of theoptimal utility. To address the lack of a priori knowledge regarding theoptimal performance, we give a convex program that can adaptively andefficiently compute the optimal contract. We also study linear contracts andderive the optimal utility in the more complex setting of multipleinteractions.