How Morphological Computation Shapes Integrated Information in Embodied Agents

Explore how morphological computation shapes integrated information in embodied agents through this conference talk. Delve into the Integrated Information Theory's quantitative approach to consciousness and its application to neural networks. Learn about a technique combining methods to examine information flows among body, brain, and environment in embodied agents. Discover the implications of this framework in a simple experimental setup, where optimal policies for goal-directed behavior are determined using the "planning as inference" method. Investigate the antagonistic relationship between morphological computation and integrated information under changing morphological circumstances. Understand the importance of measuring information flow to and from the brain in determining its influence on agent behavior. The talk covers an introduction to morphological computation and integrated information, the sensori-motor loop, methods for measuring information flows, the experimental setup, and results.