As a methodologist of science, John Maynard Keynes scrutinised the classical induction theory of Bernoulli and Laplace and proposed a new set of principles of induction in A Treatise on Probability (1921). He took up the same subject, after the publication of The General Theory of Employment, Interest and Money (1936), in his debate with Jan Tinbergen on econometrics. Keynes's criticism of econometrics is generally considered to be done on the ground of the principles of induction which he formulated in the former book. However, this standard interpretation seems odd, because those principles depended heavily on his theory of probability and he radically changed this analysis of probability in his mature economic theory. If he kept the theory of induction intact in his criticism of econometrics, then he was curiously absent-minded. I do not think that Keynes was such an inconsistent thinker, and I believe that his later theory of induction should be understood from the perspective of a much modern type of theory, i.e., something like a Kuhnian paradigm-bound view of science, or one of its more sophisticated variants. It is founded on the concept of inter-personal probability or 'expectation', and it is also constructed to be amenable to the interplay of 'theory and practice'.
