Recent years have witnessed an enormous advance in the area of Machine Learning, reflected by the popularity of Artificial Inteligence systems. For most of the history of machine learning research, the main goal was the development of machine learning algorithms that led to more accurate models, but it is now very clear that there are many other important areas to develop. We want models to be fair to unprivileged groups in society, to not reveal private information used in the model training, to provide comprehensible explanations to humans in order to help identifying causal relationships, among many relevant goals other than simply improving model accuracy. In this work, we explore possible new relationships between fairness, privacy and Quantitative Inforamtion Flow. The first exploration is an analysis of papers that explore the impact of privacy-enhancing mechanisms on Machine Learning fairness notions. Our second exploration is the possibility of dividing a fixed local differencial privacy budget between variables with varying degrees of sensitivity. Finally, we explore modeling both local differential privacy parameters within the Quantitative Information Flow framework.