摘要翻译：
我们给出了一个分类器去偏方法的实证研究，表明去偏方法在实践中往往不能泛化样本外，实际上会使公平性变得更差而不是更好。对去偏治疗效果的严格评估需要广泛的交叉验证，而不是通常所做的。我们证明了这种现象可以解释为偏差-方差权衡的结果，通过施加公平性约束，方差的增加是必要的。后续实验验证了估计方差强烈依赖于受保护类的基率的理论预测。考虑到公平性--性能权衡证明了部分去偏实际上可以在实际中对样本外数据产生更好的结果这一违反直觉的概念是合理的。
---
英文标题：
《Debiasing classifiers: is reality at variance with expectation?》
---
作者：
Ashrya Agrawal and Florian Pfisterer and Bernd Bischl and Francois
Buet-Golfouse and Srijan Sood and Jiahao Chen and Sameena Shah and Sebastian
Vollmer
---
最新提交年份：
2021
---
分类信息：

一级分类：Computer Science 计算机科学
二级分类：Machine Learning 机器学习
分类描述：Papers on all aspects of machine learning research (supervised, unsupervised, reinforcement learning, bandit problems, and so on) including also robustness, explanation, fairness, and methodology. cs.LG is also an appropriate primary category for applications of machine learning methods.
关于机器学习研究的所有方面的论文（有监督的，无监督的，强化学习，强盗问题，等等），包括健壮性，解释性，公平性和方法论。对于机器学习方法的应用，CS.LG也是一个合适的主要类别。
--
一级分类：Computer Science 计算机科学
二级分类：Computers and Society 计算机与社会
分类描述：Covers impact of computers on society, computer ethics, information technology and public policy, legal aspects of computing, computers and education. Roughly includes material in ACM Subject Classes K.0, K.2, K.3, K.4, K.5, and K.7.
涵盖计算机对社会的影响、计算机伦理、信息技术和公共政策、计算机的法律方面、计算机和教育。大致包括ACM学科类K.0、K.2、K.3、K.4、K.5和K.7中的材料。
--
一级分类：Economics 经济学
二级分类：Econometrics 计量经济学
分类描述：Econometric Theory, Micro-Econometrics, Macro-Econometrics, Empirical Content of Economic Relations discovered via New Methods, Methodological Aspects of the Application of Statistical Inference to Economic Data.
计量经济学理论，微观计量经济学，宏观计量经济学，通过新方法发现的经济关系的实证内容，统计推论应用于经济数据的方法论方面。
--

---
英文摘要：
We present an empirical study of debiasing methods for classifiers, showing that debiasers often fail in practice to generalize out-of-sample, and can in fact make fairness worse rather than better. A rigorous evaluation of the debiasing treatment effect requires extensive cross-validation beyond what is usually done. We demonstrate that this phenomenon can be explained as a consequence of bias-variance trade-off, with an increase in variance necessitated by imposing a fairness constraint. Follow-up experiments validate the theoretical prediction that the estimation variance depends strongly on the base rates of the protected class. Considering fairness--performance trade-offs justifies the counterintuitive notion that partial debiasing can actually yield better results in practice on out-of-sample data.
---
PDF下载：
-->