Integrated information and dimensionality in continuous attractor dynamics.

Area of Science:

• Neuroscience
• Theoretical Physics
• Cognitive Science

Background:

• Integrated Information Theory (IIT) posits consciousness arises from integrated information in neuronal dynamics.
• Current IIT formulations face practical and theoretical hurdles in empirically measuring integrated information from neuronal signals.
• Simultaneous observation of all system elements is practically infeasible for current IIT measurements.

Purpose of the Study:

• To propose a novel framework for assessing integrated information in continuous dynamical systems.
• To address practical limitations in empirically testing Integrated Information Theory (IIT).
• To introduce topological dimensionality of attractor dynamics as a measurable indicator of integrated information.

Main Methods:

• Utilized delay embedding to reconstruct attractor dynamics from partial neuronal observations.
• Proposed topological dimensionality of shared attractor dynamics as a proxy for integrated information.
• Demonstrated the framework's invariance to coordinate transformations, signifying a robust measure.

Main Results:

• Topological dimensionality effectively indicates integrated information in continuous attractor dynamics.
• Delay embedding allows reconstruction of dynamics and dimensionality from incomplete data.
• The proposed method aligns with recent neural recording findings from awake and anesthetized animals.

Conclusions:

• Topological dimensionality offers a viable and experimentally tractable approach to evaluating integrated information.
• This framework extends IIT to continuous dynamical systems, significantly relaxing empirical evaluation constraints.
• The study provides a crucial methodological advancement for testing theories of consciousness with real neural data.