Can Models Have Skill?

Abstract

Climate scientists and climate modelers often speak of determining whether or not a model has “skill.” This seems to imply that climate models themselves are epistemic experts. In this paper, I take a careful look at how model skill is evaluated in climate science with an eye to determining the extent to which we ought to view the idea of model skill as a step in the direction toward post-human science. I begin by consider the paradigm of verification and validation which comes from the world of engineering. Those climate scientists sometimes speak of verifying and validating their models, I argue that this paradigm is unsuitable for climate science. I then consider the question of whether or not there are general norms with which climate model skill can be established. William Goodwin has criticized my earlier work on the grounds that it “makes it unlikely that the legitimacy of [computer models] can be reconstructed, or rationalized, in terms of generally recognized, ahistorical evidential norms” (J Gen Philos Sci 46(2):339–350, 2015, p. 342). Nevertheless, I double down on my earlier claims in this paper. I concede that “a philosopher who hopes to address the epistemological concerns that have been raised about the reliability of climate models isn’t going to be able to tell a normatively grounded story that will secure the unassailable reliability of the results of climate modeling” (p. 342). I argue that when we look on the work of those who are in the business of modeling highly complex non- linear systems, the best we are ever going to be able to do is to arrive at a situation where “a simulation modeler could explain to his peers why it was legitimate and rational to use a certain approximation technique to solve a particular problem” by appealing to “very context specific reasons and particular features” (p. 344). If this is right, it suggests the prospects of science failing to “remain human” continue to be bleak.