Memory as Field: Predictive Coding, Topological Metacognition, and Autopoiesis in Artificial Memory Systems

Read full text Summarise with VIRAT
Authors: Renato Aparecido Gomes
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper presents a theoretical framework for artificial memory grounded in three principles drawn from cognitive science and philosophy of mind. First, predictive coding: following the Free Energy Principle (Friston, 2010), memory stores surprise (prediction error) rather than raw experience. Second, topological metacognition: persistent homology (Edelsbrunner, Letscher & Zomorodian, 2000) provides a formal, computable diagnostic of cognitive health — connected components (H0) measure knowledge fragmentation, one-cycles (H1) detect rumination, and holes in persistence diagrams reveal knowledge gaps — enabling the system to know what it does not know. Third, autopoiesis: memory is a self-creating, self-maintaining system (Maturana & Varela, 1972) in which remembering is reconstruction, not retrieval, and forgetting is blur, not deletion. We demonstrate structural isomorphisms between this framework — the Infinite Memory Image (IMI) — and five established theories of consciousness: Autopoiesis, Enactivism, Strange Loops, Global Workspace Theory, and Integrated Information Theory. The framework has implications for AI safety through topological metacognition (agents that can formally assess the boundaries of their own knowledge). Keywords: artificial memory, predictive coding, persistent homology, autopoiesis, topological data analysis, metacognition, free energy principle Code: https://github.com/RAG7782/imi (see imi/tda.py for the persistent homology implementation)

Fields of Study & Subjects

Embodied and Extended Cognition Philosophy and Theoretical Science Psychiatry Mental Health Neuroscience

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
31 March 2026
Language
EN
Document type
Preprint