On the role of interpretability in fuzzy learning and data mining