Good's compromise: Comments on I. J. Good

Abstract

I. J. Good and Karl Popper have at least one major disagreement. Good believes that we can and must use subjective probability in science. Popper thinks that the use of such judgements are harmful for the growth of science. In Good's paper an attempt is made to show how our sub jective judgements of the probability of theories can effectively be used in science to evaluate theories. Good's general approach to showing the usefulness of probability judgements is to argue that the making of such judgements is necessary in order to evaluate theories. Then in order to develop a theory of their usefulness we only have to show how they can be improved and applied. I will argue that such judgements are not needed and that Good has failed to account for the growth of knowledge. Good's defense of his position rests on the claim that subjective prob ability judgements are needed. There seem to be three parts to Good's claim. The first is that we will inevitably use probability simply because our brains operate in a particular way. The second is that we need to be critical of the way they are used if we want our judgements to be coherent and intellectually honest. The third is that their honest use will lead to success in evaluating theories and acting which could not be obtained if we did not honestly recognize the necessity of using them. According to Good, an honest objectivism leads to subjectivism. If we suppose that Good is right and that the decision to accept a theory or carry out an action is based on probability estimates we quickly face a problem. The probability estimate is itself a theory of the likelihood of truth or success. If a probability estimate is accepted, and the acceptance of all theories involves estimates of probability, then we must have an estimate of the probability of the probability estimate being right. We clearly cannot do this. So at some point we decide to accept a probability estimate without an estimation of the probability of that estimate being right. At some point we decide to accept a theory as true without any probability judgement. Good recognizes the problem of making judgements about our judge