Foundations of Symbolic Languages for Model Interpretability