Learning to Teach Fairness-aware Deep Multi-task Learning

TL;DR

This paper introduces L2T-FMT, a dynamic teacher-student framework for fairness-aware multi-task learning that adaptively balances fairness and accuracy, outperforming existing methods on real datasets.

Contribution

It proposes a flexible, dynamic approach to fairness in multi-task learning using a teacher-student model that learns to select objectives adaptively.

Findings

L2T-FMT improves fairness by 12-19%.

L2T-FMT enhances accuracy by up to 2%.

Reduces trade-off weights from 2T to T.

Abstract

Fairness-aware learning mainly focuses on single task learning (STL). The fairness implications of multi-task learning (MTL) have only recently been considered and a seminal approach has been proposed that considers the fairness-accuracy trade-off for each task and the performance trade-off among different tasks. Instead of a rigid fairness-accuracy trade-off formulation, we propose a flexible approach that learns how to be fair in a MTL setting by selecting which objective (accuracy or fairness) to optimize at each step. We introduce the L2T-FMT algorithm that is a teacher-student network trained collaboratively; the student learns to solve the fair MTL problem while the teacher instructs the student to learn from either accuracy or fairness, depending on what is harder to learn for each task. Moreover, this dynamic selection of which objective to use at each step for each task reduces the number of trade-off weights from 2T to T, where T is the number of tasks. Our experiments on three real datasets show that L2T-FMT improves on both fairness (12-19%) and accuracy (up to 2%) over state-of-the-art approaches.