You have just trained your ML model, performance on the test set is good, you might even start to believe in what it has learnt through some interpretability methods, but can you safely trust it will act as expected under any circumstances ?
Let’s put your model to a battery of tests and see how it handles the unexpected.