"An Information-Theoretic Approach for Modeling Minimally Cognitive Agents"

Abstract


The importance of entropy and the closely related (Shannon)
information is well established in physics. However, Ashby's Law of
Requisite Variety (1956) and its later rediscovery and extension by
Touchette and Lloyd (2000, 2004), as well as the considerations of the
Landauer Principle (Landauer 1961; Bennett and Landauer 1985) show
that the relevance of information extends beyond the physical into the
computational realm.

In the relation between low-level computation and high-level cognition
in organisms, many questions have remained open because of the
complexity of organismic information processing. In recent years,
however, it has become increasingly clear that it is not essential to
have a detailed picture of cognitive mechanisms to understand the core
requirements for organismic information processing. Instead, one can
use informational principles to model many relevant aspects of the
cognitive dynamics of organisms in a minimalistic fashion (Polani
2009). Specifically, by modeling Shannon information flows in agents
one obtains invariants, variational principles, as well as adaptive
and evolutionary drivers which help to understand the cognitive
dynamics of the agent. A "cartoonish" way of putting this would be:
"If physics is about the dynamics of energy flows, cognition is about
the dynamics of (Shannon) information flows".

In my talk, I will give an overview over the general methodology and a
few "appetizer" examples for cognitive phenomena that can be modeled
using Shannon information.