Abstract

Colin Howson has recently argued that accuracy arguments for probabilism fail because they assume a privileged ‘coding’ in which TRUE is assigned the value 1 and FALSE is assigned the value 0. I explain why this is wrong by first showing that Howson’s objections are based on a misconception about the way in which degrees of confidence are measured, and then reformulating the accuracy argument in a way that manifestly does not depend on the coding of truth-values. Along the way, I will explain how to formulate the laws of probability and rational expectation in a scale-invariant way, and how to properly understand the values of the credence functions that we use to represent rational degrees of confidence