Toward a General Theory of Inductive Inference

Abstract

This paper outlines an information processing scheme for using real (possibly “fuzzy”) observations to predict the results of other experiments within the context of a “theory.” The approach is based on the dichotomy between observer and system. It includes the explicit (geometrical) construction of “theories” from a group defining observational “degrees of freedom.” Empirical evidence is also analyzed fundamentally. Both observations and predictions involve the use of a “communication channel” linking system with observer. The quantum or classical nature of the theory is apparent in terms of channel distortion. The MEP provides the inference for the appropriate channel probability. The scheme allows the inference of stochastic processes and accommodates particle physics.