Interpretability and Causality in Machine Learning