A Model Ensemble-Based Post-Processing Framework for Fairness-Aware Prediction

TL;DR

This paper introduces a versatile post-processing ensemble framework that improves fairness in machine learning predictions across different tasks without heavily compromising accuracy.

Contribution

It presents a model-agnostic ensemble-based post-processing method that enhances fairness in diverse predictive tasks, a novel approach in fairness-aware machine learning.

Findings

Effectively improves fairness metrics across tasks

Maintains predictive accuracy with minimal loss

Applicable to various models and fairness definitions

Abstract

Striking an optimal balance between predictive performance and fairness continues to be a fundamental challenge in machine learning. In this work, we propose a post-processing framework that facilitates fairness-aware prediction by leveraging model ensembling. Designed to operate independently of any specific model internals, our approach is widely applicable across various learning tasks, model architectures, and fairness definitions. Through extensive experiments spanning classification, regression, and survival analysis, we demonstrate that the framework effectively enhances fairness while maintaining, or only minimally affecting, predictive accuracy.