Abstract

In “Idealizations and the Testing of Theories by Experimentation” Ronald Laymon develops a strategy of scientific confirmation known as “confirmation as improvability,” where theories are confirmed on the basis of accuracy in predicting. “Confirmation as improvability” must rely on some sort of curve fitting procedure in order to show how well a scientific theory’s predictions fit data. Generally, there are two sorts of curve fitting procedures: curve fitting procedures that do not incorporate a mathematical notion of simplicity and curve fitting procedures that do incorporate mathematical simplicity. Curve fitting procedures that do not incorporate mathematical simplicity are vulnerable to “the overfitting problem.” “Confirmation as improvability” relies on curve fitting procedures that do not incorporate mathematical simplicity and thus, “confirmation as improvability” is vulnerable to “the overfitting problem.” I show “confirmation as improbability’s” vulnerability to “the overfitting problem” and then show how “confirmation as improvability” could function more effectively after adopting a curve fitting procedure that incorporates mathematical simplicity.