AI for Social Good, Interpretability, Adversarial Robustness