Preprint / scholarly article
A Category-Theoretic Framework for a Self-Organizing World Model in Artificial Intelligence
- Published
- DOI
- 10.5281/zenodo.16417130
Abstract
This paper proposes a category-theoretic framework for building a self-organizing world model in AI from the agent's own validated inferences and experience. It formalizes domains as categories, cross-domain reasoning as functors, and learning as an active-inference process that reduces knowledge fragmentation.
Keywords
- category theory
- world models
- large language models
- active inference
- free energy principle
- analogical reasoning
- knowledge representation
- category-theoretic