How to stop worrying about model microfoundationality: lessons from multiscale cancer modeling

Abstract Microfoundational or bottom-up models are models that aim to reproduce the high-level behavior of a system by modeling only interactions between lower-level entities. One argument for privileging microfoundational models in biology rests on the belief that they have unique virtues (e.g. explanatory value, modularity, etc.) that make them superior to non-microfoundational models. This paper challenges this reductionist ideal. By analyzing multiscale cancer models, I show that the five virtues identified by Epstein and Forber (Synthese 190:203-218, 2013) are not exclusive to microfoundational models. My analysis of middle-out, multiscale models of tumor progression shows that they too embody these virtues. Because they incorporate both higher- and lower-level processes, I contend that non-microfoundational models are often more virtuous and therefore better suited to modeling complex biological phenomena. Ultimately, my analysis suggests that we have no good a priori reasons to privilege microfoundational models in biology and instead ought to be pluralists with respect to biological modeling.