Does enforcing fairness mitigate biases caused by subpopulation shift?
Does enforcing fairness mitigate biases caused by subpopulation shift?
复制标题
DOI：
--
复制
发表时间：
2020-11
期刊：
影响因子：
--
通讯作者：
Subha Maity;Debarghya Mukherjee;M. Yurochkin;Yuekai Sun
中科院分区：
文献类型：
--
作者：
Subha Maity;Debarghya Mukherjee;M. Yurochkin;Yuekai Sun
许多算法偏差的例子是由亚群变化引起的。例如，ML模型在训练数据中代表性不足的人口统计组上的表现往往更差。在本文中，我们研究在训练过程中执行算法公平性是否会提高训练模型在目标域中的性能。一方面，我们设想的情况下，强制执行公平性并不能提高目标域的性能。事实上，它甚至可能损害性能。另一方面，我们推导出必要和充分条件下，强制执行算法的公平性导致贝叶斯模型在目标域。我们还说明了我们的理论结果在模拟和真实的数据的实际影响。
Many instances of algorithmic bias are caused by subpopulation shifts. For example, ML models often perform worse on demographic groups that are underrepresented in the training data. In this paper, we study whether enforcing algorithmic fairness during training improves the performance of the trained model in the \emph{target domain}. On one hand, we conceive scenarios in which enforcing fairness does not improve performance in the target domain. In fact, it may even harm performance. On the other hand, we derive necessary and sufficient conditions under which enforcing algorithmic fairness leads to the Bayes model in the target domain. We also illustrate the practical implications of our theoretical results in simulations and on real data.