What Is the Epistemic Import of Derivational Robustness?

Brian McLoone (Auburn University)

Derivational robustness occurs if and only if there is agreement among each model in an ensemble of models concerning the truth of some proposition. We show that a recent and prominent Bayesian defense of robustness analysis faces substantial obstacles. In doing so, we argue against the widely-held view that measurement robustness—i.e., agreement among multiple measurements of a quantity—has epistemic credentials that are similar to those of derivational robustness.