The Relevance Criterion of Confirmation

Abstract

The relevance criterion of confirmation is the principle of which the basic form is that an hypothesis (h) is confirmed by an observation (b) in relation to a body of background knowledge or belief (k) if and only if what is observed would have been more likely to occur, given the hypothesis together with the background knowledge or belief, than it would have been given that background knowledge or belief alone. Formally, b confirms h in relation to k if and only if Pr(b,k.h)>Pr(b,k). In an earlier article (Mackie, I963) I called this the inverse principle, and argued that reliance on it is implicit within various types of solution that have been suggested for the paradox of the ravens. Mr Hooker and Mr Stove have more recently (Hooker and Stove, 1967) given convincing reasons for calling it, instead, the relevance criterion, showing that it can be traced back to what Keynes called the 'favourable relevance' of an hypothesis to an observation statement.