John D.
Norton


home >> research >> induction and confirmation: formal approaches  
The probability calculus is not the universal logic of induction; there is no such thing. An axiom system disassembles the probability calculus into distinct notions about induction, which it is urged, may be invoked independently to tailor a logic of induction to the problem at hand. The probability calculus fails as the inductive logic of certain indeterministic systems.  "Probability Disassembled," British Journal for the Philosophy of Science, 58 (2007), pp. 141171. Download.  
Forming the dual is a familiar operation in logic and mathematics. Truth is the dual of falsity; and (A or B) is the dual of (A and B). Here I develop the corresponding notion for additive measures, such as probability measures. The resulting dual additive measures are degrees of disbelief and turn out to obey their own peculiar calculus. An ignorance state is conveniently characterized as one that is selfdual.  "Disbelief as the Dual of Belief," International Studies in the Philosophy of Science, 21(2007), pp. 231252. Download.  
What if, like me, you don't think that the probability calculus is the One, True Logic of Induction? Then you want to know what other logics are possible. Here I map out a large class of inductive logics that originate in the idea that the inductive support B affords A, that is "[AB]," is defined in terms of the deductive relations among propositions. I demonstrate some very general properties for these logics. In large algebras of propositions, for example, inductive independence is generic in all of them. A nogo result forces all the logics to supplement the deductive relations among propositions with intrinsically inductive structures.  "Deductively Definable Logics of Induction," Journal
of Philosophical Logic. 39 (2010), pp. 617654. Download. For a less formal development, see "What Logics of Induction are There?" in Goodies. 

Nontrivial calculi of inductive inference are shown to be incomplete. That is, it is impossible for a calculus of inductive inference to capture all inductive truths in some domain, no matter how large, without resorting to inductive content drawn from outside that domain. Hence inductive inference cannot be characterized merely as inference that conforms with some specified calculus. 
"A Demonstration of the Incompleteness of Calculi of Inductive
Inference" Draft "The Ideal of the Completeness of Calculi of Inductive Inference: An Introductory Guide to its Failure" Draft 

Here is an illustration of how it is possible in a principled way to devise a strict weakening of the probability calculus that is nonadditive and not prone to the Bayesian's notorious problem of the priors. It represents an interesting technical exercise, but I regard it as superceded by the analyses of the later papers above.  "The Theory of Random Propositions," Erkenntnis,
41, 1994, pp. 325352. 
