Preprint / scholarly article

From Rigidity to Insight: A Framework for Verifiable AI Metacognition

K. Takahashi

Published
DOI
10.5281/zenodo.16908339

Full text PDF (Zenodo)

Abstract

We provide a verifiable, neuro-inspired mechanism that enables an AI agent to move from rigid, over-fitted beliefs to adaptive, life-long learning by intentionally perturbing its own cognitive states.

Keywords

  • AI
  • large language models
  • AI safety
  • metacognition
  • reinforcement learning
  • free energy principle
  • predictive processing
  • autonomous systems
  • self-correction
  • optimization
  • philosophy of mind

Identifiers and source records