Prediction and the Problem of Induction

1 The Problem of Knowledge Epistemology is commonly defined as the study of the origins and nature of knowledge, where knowledge is true, justified belief. This is a strong characterization of the subject matter of epistemology. It is so strong that skeptics have mounted good arguments to the conclusion that we know very little, if anything at all. Even if the skeptics are right, there remains the fact that many false, and unjustified, beliefs provide reliable predictions that are approximately true. Meteorology, nowadays, is remarkably accurate although forecasts are not often exactly true. The daytime high is forecast to be 15C, but it reaches 16C. Should epistemologists ignore meteorology as an interesting object of study? I think not. This book is about the epistemology of science in a broader sense of epistemology than commonly conceived in modern philosophical circles. In search of a broader conception of the problem, let me remind you of Hume’s problem of induction. Suppose we consider our belief that “All billiard balls will move when struck at any time.” Although this belief is not true (billiard balls glued to the table will not move when struck), it does yield reliable predictions in ordinary circumstances. The epistemological problem here is: What are the intelligent processes by which such predictions are made and what explains their reliability (to the extent that they are reliable)? Epistemologists have always sought a general answer to this question. It’s no good spinning a different yarn for each case. What’s wanted is some general explanation that would immediately cover any other kind of example. It is not tied to the peculiarities of billiard balls, or the particular characteristics of the predictor. It has to be subject-neutral. Classically, logic is the tool supposed to provide such subject-neutral answers to such epistemological questions. The method of enumerative induction is the simplest and most naive answer to the question about how reliable predictions are made: If past experience has yielded a constant conjunction of cases in which objects with property A also have property B, with no exceptions, then we draw the conclusion that all As are Bs and predict that the next A will be a B. Notice that the answer is formulated in a subject-neutral way. In our example, A is the property of being a struck billiard ball, and B is the property of being of moving. But the answer also covers other examples, and this is an attractive feature. Hume’s critique of induction shows that the method of enumerative induction does not yield justified true belief, where ‘justified’ means ‘justified with certainty’. Consider any inference of the form: