Extreme physical information as a principle of universal stability

Abstract

The EPI principle for finding a scientific law arises as follows. A coherent universe requires stable scientific laws. Each such law is a dual system AB consisting of a theoretical system A, in a state a, that interacts with a system B that may be an observer. Both A and B are assumed to be real systems. The interaction is via a probe particle, which carries information about a to B, and in doing so perturbs the total system AB in two ways-(1) It perturbs the information flow from A such that a is observed by B as y = a + x, where x is an unknown number obeying a frequency law p(x). And (2) the law is likewise perturbed, as δp(x). The law, a characteristic of the total system effect AB, is to be found. The above requirement of a stable law requires that the change I-J in information about a from A to B remain invariant to the perturbations, i.e., δ(I-J)/δp(x) = 0. This is mathematically equivalent to I-J = extremum, and is the EPI principle that may be used to find p(x). Interestingly, the extremum is usually a minimum, meaning that, despite the unknown perturbations x and δp(x), the output information I ≃ J. That is, observation tends to agree with reality, as demanded of a coherent universe. Moreover, the entire observation-interaction procedure has physical reality, meaning that the output physical law is created on the spot.' In applications of EPI, information functional I is always of one known form (Fisher's). Also, the epistemic nature of EPI allows a degree of prior knowledge about a to be used to form J (a). In descending order of accuracy in the resulting outputs p(x), these forms of prior knowledge are called (a) abduction (highest quality, with perfect outputs), (b) deduction (next highest) and (c) induction (lowest, giving merely smooth outputs). Numerous applications of EPI are given. © 2009 Springer US.