Redeeming Falsifiability?

Abstract

We revisit Popper's falsifiability criterion. A tester hires a potential expert to produce a theory, offering payments contingent on the observed performance of the theory. In our model, instead of knowing the true data-generating process, the expert knows the state-of-the-art belief over data-generating processes. A non-expert does not. We argue that if the expert can, moreover, acquire additional information to refine this knowledge, falsifiability does have the power to distinguish between experts and non-experts and to identify valuable theories, capitalizing on experts' ability to acquire and refine knowledge.