by E Sullivan · Cited by 6 — I argue that ML models used in science function as highly idealized toy models. If we treat ML models as a type of highly idealized toy model, ...