Stud. Hist. Phil. Mod. Phys. http://doi.org/bvzm (2016)

When Duncan Haldane published, in 1988, the model that would eventually win him a share of the 2016 Nobel Prize in Physics, it was clear to him that the model was “unlikely to be directly physically realizable” (even though it was implemented in experiment a quarter of a century later). Models that, unlike idealizations or approximations, do not directly represent physical systems often prove influential — the topological Haldane model is but a case in point. For all that, though, 'toy models' have so far received little attention from philosophers of science, argues Joshua Luczak, who sets out to change that.

Luczak defines toy models, roughly, as models that do not perform a representational function. Instead, they fulfil other functions, such as elucidating aspects of a theory, testing consistency between concepts and, not least, acting as 'hypothesis-generating tools'. Idealizations or approximations might do that as well, but always in relation to a target system they represent. Toy models, in contrast, lead to insights through arguments from analogy. These distinctly different routes to employing models might hold important lessons about how we are making discoveries.