Zenith Grant Awardee

David Wolpert

Santa Fe Institute

Co-Investigators

Artemy Kolchinsky, Santa Fe Institute

Project Title

Observers as self-maintaining non-equilibrium systems

Project Summary

The concept of an “observer” is central to many fields, including cognitive science, artificial intelligence, quantum physics and statistical physics (e.g., in the form of Maxwell’s demon). In all these fields observers are systems that acquire information about an external system. Traditionally such information has been quantified using Shannon’s measure of syntactic information. In this project we propose a measure of semantic information (i.e. information has “meaning” for the observer) that is grounded in statistical physics. In our approach, information is semantic to the extent that it is used by the observer to maintain itself over time as a coherent entity. We analyze the consequences of this measure for the physics of observation. In our approach, “being an observer” is not a simple yes / no trait. Rather, our definition of semantic information quantifies the degree to which one system observes another. This will allow us to identify the set of observers contained within any physical system, including the appearance (self-organization) and dissolution of observers, and to study how networks and hierarchies of observers relate to flows of information and energy in a larger encompassing system.

Technical Abstract

In many scientific fields, observers are systems that acquire information about an external system. Traditionally this information is quantified using Shannon’s measure of syntactic information. We propose to instead adopt a semantic measure of information, and use it to investigate novel aspects of the physics of observation. Specifically, we suppose that an “observer” is a non-equilibrium system coupled to some observed external system. The observer has some data D that is correlated with the state of the external system. We identify the amount of semantic information in D as how much the observer can slow its convergence to thermodynamic equilibrium during its interaction with the external system by exploiting D. In our approach, “being an observer” is not a simple yes / no trait. Rather, using our definition of semantic information, we quantify the degree to which one system observes another. This allows us to identify those subsystems within an overall system that “most strongly” observe some other subsystem. It also provides a way to consider the self organization and dissolution of observers, and to study how networks and hierarchies of observers relate to flows of information and energy in a larger encompassing system.