This paper deals with the notion of objective randomness in classical deterministic theories. After the introduction, section 2 establishes an important distinction between a strictly metaphysical thesis of determinism (as characterized in the Montague-Earman definition, for example), and the doctrine of determinism, which can be roughly characterized as a methodological set of principles. The doctrine of determinism is associated with the idea that probability assignments can only reflect our ignorance of facts, and it also grounds the (ontological) thesis of separability: A system or process can be characterized completely in terms of the properties that a system has when in a given state, independently of the properties of other systems, The key notions of "completeness" and "independence" are only briefly discussed, as they are examined more in depth elsewhere.
Section three examines attempts to characterize a notion of objective randomness in ergodic theory. The characterization can be seen to be equivalent to the formulation of a notion of a "physically impossible process". One way of expressing this idea is grounded on the thesis of the "coarse graining" of our measuring instruments. This leads to the problem of distinguishing "objective" from "apparent" randomness. The problem seems to be intrinsic to any attempt of characterizing the required notion of physical impossibility (and thus of objective randomness) in terms of an ideal observer. The alternative of trying to characterize physical impossibility in terms of a theory of algorithms suffers from the difficulty that it is not clear what would be the required notion of (non-epistemic) computability.
The "coarse graining" approach, as well as the alternative in terms of a theory of algorithms seem to share the usual confusion between a strictly metaphysical thesis of determinism and the (methodological) doctrine of determinism. In section four an alternative approach is suggested. It is noticed that a denial of the thesis of separability is compatible with a strict deterministic theory, and thus that at least some classical systems (the "statistical" ones) can be described in terms of non-separable states. We do not have to think that the impossibility of preparing a classical state is too only possible explanation for physical (objective) randomness in deterministic theories. This proposal is an elaboration of an idea of Blatt (1959), although Blatt was still trying to understand objective randomness within an ideal-observer framework.[S.M.]
Classical extension of quantum-correlated separable states
