A growing body of literature in fairness-aware ML (fairML) aspires to mitigate machine learning (ML)-related unfairness in automated decision making (ADM) by defining metrics that measure fairness of an ML model and by proposing methods that ensure that trained ML models achieve low values in those measures. However, the underlying concept of fairness, i.e., the question of what fairness is, is rarely discussed, leaving a considerable gap between centuries of philosophical discussion and recent adoption of the concept in the ML community. In this work, we try to bridge this gap by formalizing a consistent concept of fairness and by translating the philosophical considerations into a formal framework for the training and evaluation of ML models in ADM systems. We derive that fairness problems can already arise without the presence of protected attributes, pointing out that fairness and predictive performance are not irreconcilable counterparts, but rather that the latter is necessary to achieve the former. Moreover, we argue why and how causal considerations are necessary when assessing fairness in the presence of protected attributes. We achieve greater linguistic clarity for the discussion of fairML and propose general algorithms for practical applications.

翻译：公平感知机器学习（fairML）领域日益增长的文献旨在通过定义衡量机器学习模型公平性的指标，并提出确保训练模型在这些指标上达到较低值的方法，来缓解自动化决策中与机器学习相关的不公平现象。然而，公平的基本概念——即公平是什么的问题——鲜有讨论，导致数百年的哲学探讨与机器学习社区近期对该概念的采纳之间存在显著鸿沟。在本研究中，我们尝试通过形式化一致的公平概念，并将哲学思考转化为自动化决策系统中机器学习模型训练与评估的形式化框架，来弥合这一鸿沟。我们推导出，公平问题在缺乏受保护属性时已可能出现，指出公平性与预测性能并非不可调和的对立面，相反，后者是实现前者的必要条件。此外，我们论证了在存在受保护属性时评估公平性为何及如何需要因果推理。我们为公平机器学习的讨论实现了更高的语言清晰度，并提出了适用于实际应用的通用算法。