Neural representation in active inference: Using generative models to interact with-and understand-the lived world

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Biology

Background:

• Active inference provides a normative framework for brain function.
• Organisms utilize generative models to minimize prediction errors, quantified by variational free energy.
• Generative models are crucial for adaptive behavior and understanding the world.

Purpose of the Study:

• To explore neural representation through the lens of active inference.
• To analyze the role of generative models in adaptive navigation and cognitive processes.
• To elucidate the link between generative models, belief dynamics, and neural representation.

Main Methods:

• Conceptual analysis of active inference and generative models.
• Examination of variational free energy minimization in biological systems.
• Exploration of different types of generative models (action-perception, planning, imagination).

Main Results:

• Generative models are learned for adaptive navigation, not solely for understanding.
• Diverse generative models support various cognitive functions, from action-perception to planning.
• Generative models and belief dynamics are linked to neural representation and cognitive capabilities.

Conclusions:

• Generative models play diverse roles in cognition and the evolution of neural representation.
• The type of generative model influences an agent's cognitive abilities within its ecological niche.
• Open questions remain regarding the evolution of generative models and advanced cognition.