Fairness Violations and Mitigation under Covariate Shift
Fairness Violations and Mitigation under Covariate Shift
复制标题
DOI：
10.1145/3442188.3445865
复制
发表时间：
2019-11
期刊：
影响因子：
--
通讯作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
中科院分区：
文献类型：
--
作者：
Harvineet Singh;Rina Singh;Vishwali Mhasawade;R. Chunara
我们研究的问题，学习公平的预测模型，看不见的测试集分布不同的训练集。针对数据分布变化的稳定性是负责任地部署模型的重要任务。领域适应文献解决了这个问题，尽管稳定性的概念限于预测精度。我们确定了稳定模型的充分条件，无论是在预测精度和公平性方面，可以学习。使用因果图描述的数据和预期的变化，我们指定了一种方法，利用数据中的条件独立性的特征选择，以估计测试集的准确性和公平性指标。我们表明，对于特定的公平性定义，所得到的模型满足最坏情况下的最优性的形式。在医疗保健任务的背景下，我们说明了这种方法在做出更公平的决定的优势。
We study the problem of learning fair prediction models for unseen test sets distributed differently from the train set. Stability against changes in data distribution is an important mandate for responsible deployment of models. The domain adaptation literature addresses this concern, albeit with the notion of stability limited to that of prediction accuracy. We identify sufficient conditions under which stable models, both in terms of prediction accuracy and fairness, can be learned. Using the causal graph describing the data and the anticipated shifts, we specify an approach based on feature selection that exploits conditional independencies in the data to estimate accuracy and fairness metrics for the test set. We show that for specific fairness definitions, the resulting model satisfies a form of worst-case optimality. In context of a healthcare task, we illustrate the advantages of the approach in making more equitable decisions.