Indiscernibility and identity in probability theory (Q1814260): Difference between revisions

This paper constitutes a thorough exploration of the relation between the structure of the sets on which probability functions are defined and the axioms that probability functions obey. Ever since Popper's ``autonomous'' axiom systems for the calculus of probability, in which probability is taken to be a function defined over arbitrary sets [\textit{K. Popper}, Appendices *iv--*v of ``The logic of scientific discovery.'' London: Hutchinson and Co. (1959; Zbl 0083.24104)] this question has been of interest. It is possible to give axioms for relative and absolute probability functions under which the structure in the underlying domain of the functions is imposed by the probability axioms themselves. Under some axiomatizations, the underlying domain turns out to be a Boolean algebra, but this need not be the case: they may be Boolean algebras merely with respect to indiscernibility under \(P\). \(A\) and \(B\) are left indiscernible under \(P\) if \(P(A,C)=P(B,C)\) for all \(C\) in \(S\); right discernible if \(P(C,A) = P(C,B)\) for all \(C\) in \(S\); and indiscernible if \(A\) and \(B\) are both right and left indiscernible. The authors explore the consequences of adopting the same approach to the non-autonomous systems of Carnap, Rényi, and Kolmogorov: these systems, too, can be presented as probability systems on arbitrary sets, with the addition of a single axiom. The relations among the systems are explored, and some turn out to be of particular philosophical interest.