Idealization, scientific realism, and the improvement model of confirmation

That many of our most successful scientific theories involve one or more idealizations poses a challenge to traditional models of theory confirmation. One popular response amongst scientific realists is the “improvement model of confirmation”: if tightening up one or more of the idealizations leads to greater predictive accuracy, then this supports the belief that the theory’s inaccuracy is a result of its idealizations and not because it is wrong. In this article I argue that the improvement model is deeply flawed and that therefore idealizations continue to undermine “success-to-truth” arguments for scientific realism.