Preprint / scholarly article

A Category-Theoretic Framework for a Self-Organizing World Model in Artificial Intelligence

K. Takahashi

Published
DOI
10.5281/zenodo.16417130

Full text PDF (Zenodo)

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

Identifiers and source records