Abstract

In this paper I defend the possibility of robustness analysis as confirmatory. Given that models are highly idealized, multiple models with different sets of idealizations are constructed to show that some result is not dependent on those idealizations. This method of robustness analysis has been criticized since, no matter how many false models agree, all of them are false and lack confirmatory power. I argue that this line of criticism makes an assumption that a model is confirmatory only if it ontically represents its target. I draw on work about explanations to motivate a challenge to this assumption, and argue that this assumption needs bolstering.