Equality of Opportunity in Supervised Learning
Moritz Hardt Eric Price Nathan Srebro
Abstract
We propose a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features. Assuming data about the predictor, target, and membership in the protected group are available, we show how to optimally adjust any learned predictor so as to remove discrimination according to our definition. Our framework also improves incentives by shifting the cost of poor classification from disadvantaged groups to the decision maker, who can respond by improving the classification accuracy.
中文速览
机器学习模型在信贷、招聘、司法等高风险领域越来越普及,但如何定量衡量并消除对受保护群体(如种族、性别)的算法歧视,目前仍缺乏公认方法。本文提出一个叫做"均等机会/均等胜算"(equalized odds / equal opportunity)的公平性准则:要求预测器在控制真实结果之后,与受保护属性统计独立——换句话说,对不同群体的真正阳性率和假阳性率应当一致。在此基础上,作者给出一种通用的后处理方法,可以将任意已训练好的打分模型,通过对不同群体采用不同(甚至随机化的)阈值来低损耗地转换为满足该准则的公平分类器,而无需改动原有的训练流程。理论分析表明,这种后处理不仅能保留贝叶斯最优预测器的精度,还将分类不确定性带来的代价从弱势群体转移到决策者一方,从而激励决策者主动收集更好的特征数据。以FICO信用评分为案例的实验进一步表明,与"人口统计均等"(demographic parity)相比,该方法在满足公平性约束的同时损失的预测效用明显更少,为机器学习公平性提供了一个兼具理论支撑和实际可操作性的解决方案。
原文 arXiv:1610.02413;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1610.02413v1