Predictive models are often limited by their strong focus on prediction accuracy, leading to potential for shortcut learning and limited out-of-set generalization. Recent interpretability methods have ...