Machine learning currently plays an increasingly important role in people's lives in areas such as credit scoring, auto-driving, disease diagnosing, and insurance quoting. However, in many of these areas, machine learning models have performed unfair behaviors against some sub-populations, such as some particular groups of race, sex, and age. These unfair behaviors can be on account of the pre-existing bias in the training dataset due to historical and social factors. In this paper, we focus on a real-world application of credit scoring and construct a fair prediction model by introducing latent variables to remove the correlation between protected attributes, such as sex and age, with the observable feature inputs, including house and job. For detailed implementation, we apply Bayesian approaches, including the Markov Chain Monte Carlo simulation, to estimate our proposed fair model.

Most literature in fairness has focused on improving fairness with respect to one single model or one single objective. However, real-world machine learning systems are usually composed of many different components. Unfortunately, recent research has shown that even if each component is "fair", the overall system can still be "unfair". In this paper, we focus on how well fairness composes over multiple components in real systems. We consider two recently proposed fairness metrics for rankings: exposure and pairwise ranking accuracy gap. We provide theory that demonstrates a set of conditions under which fairness of individual models does compose. We then present an analytical framework for both understanding whether a system's signals can achieve compositional fairness, and diagnosing which of these signals lowers the overall system's end-to-end fairness the most. Despite previously bleak theoretical results, on multiple data-sets -- including a large-scale real-world recommender system -- we find that the overall system's end-to-end fairness is largely achievable by improving fairness in individual components.