This talk will show how standard scientific practices—particularly model formulation—can introduce new epistemic commitments to the scientific community or research program that uses the relevant models. I will demonstrate this claim with loop analysis: a modelling approach advocated by ecologist and population geneticist Richard Levins during the 1970s and early 1980s. While Levins was not the first to advocate feedback analysis, even among evolutionary biologists, his approach was considered idiosyncratic by his peers. I will argue that Levins’ idiosyncratic approach embodies a dialectical view of nature, a view that was highly influenced by his commitment to Marxism.