- TL;DR: In depth review of "fair machine learning" algorithms and proposal for implementing ethical frameworks in this field.
- Keywords: algorithmic fairness, discrimination, fair machine learning, ethics
- Abstract: Machine Learning is becoming more and more accessible for developers to implement in automatic decision making which may involve tasks that can lead to systematic discrimination. Several studies have revealed the ease in which machine learning algorithms can learn to replicate biases from human values when trained on data that contains signal about such biases for a specific task (Boulbaski et al. 2016, Larson et al. 2016). In this work we will divulge specific algorithms and settings for algorithmic fairness while emphasizing on the limitations of the approach taken in the state of the art. We will also contribute with an ethical overview of the concept offered for fairness in the field. Then we will identify key points for future work on fair AI.