Integrated world modeling theory expanded: Implications for the future of consciousness

Area of Science:

• Neuroscience and Artificial Intelligence
• Theoretical Consciousness Studies

Background:

• Existing theories of consciousness, Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT), offer valuable but distinct perspectives.
• The free energy principle and active inference (FEP-AI) provide a unifying computational framework for understanding brain function.

Purpose of the Study:

• To introduce and elaborate on Integrated World Modeling Theory (IWMT) as a synthetic theory of consciousness.
• To bridge insights from IIT and GNWT using the FEP-AI framework.
• To explore novel computational and mechanistic principles for understanding consciousness.

Main Methods:

• Review of philosophical principles and neural systems underpinning IWMT.
• Description of predictive processing models and machine learning architectures (autoencoders, turbo-codes, graph neural networks).
• Exploration of probabilistic graphical models, flow networks, and game theory for estimating integrated information.

Main Results:

• IWMT integrates IIT and GNWT within the FEP-AI framework, offering a unified perspective on consciousness.
• Novel computational approaches are proposed for estimating integrated information and addressing critiques of existing theories.
• Connections are drawn between probabilistic modeling, quantum mechanics, and the generation of discrete conscious experience.

Conclusions:

• IWMT provides a robust, unifying model for consciousness research, expanding on current theoretical frameworks.
• Future directions include refining computational methods, exploring attentional selection, and understanding the evolutionary origins of consciousness.
• While not solving the Hard Problem, IWMT offers a promising path toward understanding the physical substrates and dynamics of conscious experience.