Calibration: Modelling the measurement process

Calibration procedures establish a reliable relation between the final states ('indications') of a measurement process and features of the objects being measured ('outcomes'). This article analyzes the inferential structure of calibration procedures. I show that calibration is a modelling activity, namely the activity of constructing, deriving predictions from, and testing theoretical and statistical models of a measurement process. Measurement outcomes are parameter value ranges that maximize the predictive accuracy and mutual coherence of such models, among other desiderata. This model-based view of calibration clarifies the source of objectivity of measurement outcomes, the nature of measurement accuracy, and the close relationship between measurement and prediction. Contrary to commonly held views, I argue that measurement standards are not necessary for calibration, although they are useful in maintaining coherence across large networks of measurement procedures.