Inferential rules for confirmatory robustness

Abstract

This paper analyses when, and to what extent, the robustness of a result yields confirmation. I develop two inferential rules that specify how modellers and experimenters should update their conditional probabilities when new derivational or experimental information becomes available. While similar rules apply to derivational and experimental robustness, they are insufficient on their own to generate empirical confirmation from derivational robustness. That requires suitable indirect confirmation relations linking model results to empirical evidence. I examine several such relations and show how robustness can increase confirmation by demonstrating that both an empirically validated result and a model prediction depend on the same components, while certain false auxiliaries are irrelevant. When empirical confirmation arises from derivational robustness, it does so by strengthening these indirect links rather than by securing a high absolute probability for a robust theorem. The resulting account clarifies both the scope and limits of robust confirmatory reasoning.