Optimising Equal Opportunity Fairness in Model Training
Optimising Equal Opportunity Fairness in Model Training
复制标题
优化模型训练中的机会均等公平性
DOI：
10.48550/arxiv.2205.02393
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Lea Frermann
中科院分区：
文献类型：
--
作者：
Aili Shen;Xudong Han;Trevor Cohn;Timothy Baldwin;Lea Frermann
现实世界的数据集通常编码刻板印象和社会偏见。这种偏见可以被训练过的模型隐含地捕捉到，从而导致有偏见的预测，并加剧现有的社会偏见。现有的去偏见方法，如对抗性训练和从表示中删除受保护的信息，已被证明可以减少偏见。然而，公平标准和培训目标之间的脱节使得很难从理论上对不同技术的有效性进行推理。在这项工作中，我们提出了两个新的训练目标，直接优化广泛使用的机会均等标准，并表明它们在减少偏差的同时保持两个分类任务的高性能是有效的。
Real-world datasets often encode stereotypes and societal biases. Such biases can be implicitly captured by trained models, leading to biased predictions and exacerbating existing societal preconceptions. Existing debiasing methods, such as adversarial training and removing protected information from representations, have been shown to reduce bias. However, a disconnect between fairness criteria and training objectives makes it difficult to reason theoretically about the effectiveness of different techniques. In this work, we propose two novel training objectives which directly optimise for the widely-used criterion of equal opportunity, and show that they are effective in reducing bias while maintaining high performance over two classification tasks.
DOI：
10.24963/ijcai.2019/199
发表时间：
2019-08
期刊：
--
影响因子：
--
作者：
Yongkai Wu;Lu Zhang;Xintao Wu
通讯作者：
Yongkai Wu;Lu Zhang;Xintao Wu
DOI：
--
发表时间：
2020-12
期刊：
ArXiv
影响因子：
--
作者：
Yuji Roh;Kangwook Lee;Steven Euijong Whang;Changho Suh
通讯作者：
Yuji Roh;Kangwook Lee;Steven Euijong Whang;Changho Suh