Integrated information in the thermodynamic limit

Area of Science:

• Neuroscience
• Computational Neuroscience
• Theoretical Neuroscience

Background:

• Information integration is key to biological, neural, and cognitive systems.
• Integrated Information Theory (IIT) quantifies integration but is computationally expensive for large systems.
• Understanding how integration scales and is maintained in dynamic environments is crucial.

Purpose of the Study:

• To develop scalable measures for information integration within large neural networks.
• To investigate the scaling properties and environmental interactions of integrated systems.
• To explore how systems maintain integration in the face of environmental changes.

Main Methods:

• Revised assumptions of Integrated Information Theory (IIT).
• Applied kinetic Ising models and mean-field approximations to large neural networks.
• Analyzed divergent tendencies of system components at critical points.

Main Results:

• Demonstrated how modified IIT measures scale in large neural networks.
• Showed that information integration diverges in the thermodynamic limit at critical points.
• Utilized integration measures to delineate system-environment boundaries and model adaptive integration.

Conclusions:

• Scalable IIT measures enable analysis of large-scale neural systems.
• Information integration exhibits critical points and divergence, offering insights into system organization.
• Adaptive mechanisms can preserve system integration through environmental interactions.