Toward a formal theory of embodied cognition
Biosyst. January 30, 2021 DOI: 10.1016/j.biosystems.2021.104356 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper Peer reviewed |
|---|---|
| Citations | 27 |
| Key points | Proposes that information and control theorems can model embodied cognition, yielding the Yerkes-Dodson inverted-U relation and a canonical consciousness model. |
Abstract
Since the reduction of uncertainty associated with cognition necessarily implies existence of a 'dual' information source, the Source Coding and Rate Distortion Theorems of information theory, in combination with the Data Rate Theorem of control theory, are sufficient to model much of embodied cognition. An iterated Morse Function 'free energy' allows construction of a 'higher entropy' analog constrained to obey the approximate dynamics of the Onsager gradient model of nonequilibrium thermodynamics. However, since palindromes are unlikely, there is no time reversal symmetry and hence no Onsager reciprocal relations. The group symmetry-breaking associated with physical phase transitions then emerges in terms of groupoids associated with equivalence classes of high probability developmental paths. The formalism is agnostic regarding representation and the Yerkes-Dodson inverted-U relation emerges directly, as do stochastic versions of the underlying dynamics, and a canonical approach to an embodied consciousness model. Contrary to the Western cultural fixation with individual salience at the expense of context, for cognition against and in the real world, context is often central, rather than simply being a screen against which some culturally-specific cartoon is supposed to play.