Semantics and Thermodynamics

ABSTRACT: Inferring models from given data leads through many different changes in representation. Most are subtle and profitably ignored. Nonetheless, any such change affects the semantic content of the resulting model and so, ultimately, its utility. A model's semantic structure determines what its elements mean to an observer that has built and uses it. In the search for an understanding of how large-scale thermodynamic systems might themselves take up the task of modeling and so evolve semantics from syntax, the present paper lays out a constructive approach to modeling nonlinear processes based on computation theory. It progresses from the microscopic level of the instrument and individual measurements, to a mesoscopic scale at which models are built, and concludes with a macroscopic view of their thermodynamic properties. Once the computational structure of the model is brought into the analysis it becomes clear how a thermodynamic system can support semantic information processing.