Consistent Estimators for Learning to Defer to an Expert

TL;DR

This paper introduces a method for learning predictors that can either make decisions or defer to an expert, using a novel reduction to cost-sensitive learning with theoretical guarantees and experimental validation.

Contribution

It proposes a new approach for learning to defer decisions to experts by reducing the problem to cost-sensitive learning with a consistent surrogate loss.

Findings

The method is theoretically consistent.

Effective in various experimental tasks.

Generalizes cross entropy loss.

Abstract

Learning algorithms are often used in conjunction with expert decision makers in practical scenarios, however this fact is largely ignored when designing these algorithms. In this paper we explore how to learn predictors that can either predict or choose to defer the decision to a downstream expert. Given only samples of the expert's decisions, we give a procedure based on learning a classifier and a rejector and analyze it theoretically. Our approach is based on a novel reduction to cost sensitive learning where we give a consistent surrogate loss for cost sensitive learning that generalizes the cross entropy loss. We show the effectiveness of our approach on a variety of experimental tasks.