Maintaining fairness across distribution shift: do we have viable solutions for real-world applications?
Maintaining fairness across distribution shift: do we have viable solutions for real-world applications?
复制标题
保持分配转移的公平性：我们是否有针对实际应用的可行解决方案？
DOI：
10.48550/arxiv.2302.12254
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
A. D'Amour
中科院分区：
文献类型：
--
作者：
Jessica Schrouff;Natalie Harris;Oluwasanmi Koyejo;Ibrahim M. Alabdulmohsin;Eva Schnider;Krista Opsahl;Alex Brown;Subhrajit Roy;Diana Mincu;Christina Chen;Awa Dieng;Yuan Liu;Vivek Natarajan;A. Karthikesalingam;Katherine C. Heller;S. Chiappa;A. D'Amour
在评估机器学习模型时，公平性和鲁棒性通常被视为正交维度。然而，最近的研究揭示了公平性和鲁棒性之间的相互作用，表明公平性属性在分配转移下不一定保持不变。在医疗保健设置中，这可能导致例如，根据“医院a”中选定的度量标准执行公平的模型在“医院B”中部署时显示不公平。虽然一个新兴领域已经出现，以开发可证明的公平和稳健的模型，但它通常依赖于对转变的强烈假设，限制了其对现实世界应用的影响。在这项工作中，我们通过参考因果框架，探讨了最近提出的缓解策略适用的环境。使用皮肤病学和电子健康记录中的预测模型的例子，我们表明现实世界的应用是复杂的，并且经常使这些方法的假设无效。因此，我们的工作突出了技术、实践和工程方面的差距，这些差距阻碍了为现实世界的应用开发健壮公平的机器学习模型。最后，我们讨论了机器学习管道的每个步骤的潜在补救措施。
Fairness and robustness are often considered as orthogonal dimensions when evaluating machine learning models. However, recent work has revealed interactions between fairness and robustness, showing that fairness properties are not necessarily maintained under distribution shift. In healthcare settings, this can result in e.g. a model that performs fairly according to a selected metric in “hospital A” showing unfairness when deployed in “hospital B”. While a nascent field has emerged to develop provable fair and robust models, it typically relies on strong assumptions about the shift, limiting its impact for real-world applications. In this work, we explore the settings in which recently proposed mitigation strategies are applicable by referring to a causal framing. Using examples of predictive models in dermatology and electronic health records, we show that real-world applications are complex and often invalidate the assumptions of such methods. Our work hence highlights technical, practical, and engineering gaps that prevent the development of robustly fair machine learning models for real-world applications. Finally, we discuss potential remedies at each step of the machine learning pipeline.
登录
查看更多内容
DOI：
10.1145/3442188.3445910
发表时间：
2020-06
期刊：
Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
影响因子：
--
作者：
Vedant Nanda;Samuel Dooley;Sahil Singla;S. Feizi;John P. Dickerson
通讯作者：
Vedant Nanda;Samuel Dooley;Sahil Singla;S. Feizi;John P. Dickerson
DOI：
10.1145/3442188.3445865
发表时间：
2019-11
期刊：
Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
影响因子：
--
作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
通讯作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
公平机器学习中来自偏见数据的残余不公平
DOI：
--
发表时间：
2018
期刊：
Proceedings of the 35th International Conference on Machine Learning
影响因子：
--
作者：
Kallus, Nathan;Zhou, Angela
通讯作者：
Zhou, Angela
通过多组件不变学习实现群体感知的分层贝叶斯域适应
DOI：
10.1145/3368555.3384451
发表时间：
2020
期刊：
Inference and Learning
影响因子：
--
作者：
Mhasawade, Vishwali;Rehman, Nabeel Abdur;Chunara, Rumi
通讯作者：
Chunara, Rumi
样本选择偏差下的公平稳健分类
DOI：
10.1145/3459637.3482104
发表时间：
2021
期刊：
30th ACM International Conference on Information & Knowledge Management
影响因子：
--
作者：
Du, Wei;Wu, Xintao
通讯作者：
Wu, Xintao