MRCF: Meta-Recursive Cognitive Framework
A philosophical and methodological framework for preserving human authority, semantic depth, and recursive inquiry in human-AI cognition.
What MRCF Is
MRCF is FERZ's philosophical and methodological framework for structured human-AI inquiry. It treats dialogue with AI not as prompt optimization, but as a recursive cognitive process in which language, attention, authority, and meaning can either deepen or collapse.
MRCF is a FERZ-defined framework developed by Edward Meyman. It is documented as Part I of the Meta-Recursive Cognition Corpus (Zenodo).
The framework is designed to preserve human intellectual agency while using AI as a thought amplifier. Its concern is not whether AI can produce an answer, but whether the human remains capable of asking deeper questions, retaining authority over judgment, and resisting semantic flattening across successive cycles of interaction.
MRCF is not prompt engineering and not a runtime authorization mechanism. Prompt engineering optimizes a single exchange; a runtime authorization boundary gates actions before execution. MRCF concerns the recursive reasoning process itself: whether inquiry deepens, and whether the human retains authority.
Where MRCF Sits
Within FERZ, MRCF sits beside, not inside, the runtime authorization stack. Deterministic governance answers whether an action was authorized before execution. MRCF addresses a different question: whether the humans designing, interpreting, and revising governance systems are preserving semantic depth and intellectual authority while working with AI.
AI Changes Cognition, Not Just Outputs
AI systems do not only answer questions. They shape the questions humans ask next. They influence vocabulary, confidence, attention, and the perceived location of authority.
A human-AI dialogue can deepen inquiry, or it can flatten it. It can sharpen judgment, or it can quietly relocate judgment into the machine. MRCF names that risk and provides a method for resisting it.
Language, Thought, and Authority Co-Evolve
MRCF begins from the premise that language and thought co-evolve. When a human reasons with AI, the exchange is not a neutral sequence of prompts and responses. Each cycle can alter the human's conceptual frame, question-space, and sense of agency.
The central risk is not only wrong answers. It is authority drift: the gradual transfer of intellectual initiative from the human to the system. The corresponding discipline is cognitive sovereignty: the preservation of human judgment as the locus of meaning, responsibility, and final authority.
Framework Elements
Seven elements govern how recursive human-AI inquiry can deepen thought rather than flatten it.
Recursive inquiry
Each cycle should make the next question more precise, not merely produce another answer.
Linguistic precision
Word choice is not cosmetic. It determines what distinctions remain available to thought.
Anti-semantic flattening
The framework resists premature simplification when complexity is necessary for understanding.
Intellectual agency
AI may amplify reasoning, but it must not become the locus of judgment.
Emergent questioning
Recursive dialogue should generate questions neither party would have produced in a single pass.
Philosophical courage
The user must remain willing to engage ambiguity, contradiction, and difficult abstraction without outsourcing judgment.
Contextual calibration
Recursive depth must be matched to cognitive load, domain stakes, and the user's capacity to retain agency.
Authority Retention and Meta-Validation
CARP: Cognitive Authority Retention Protocol
CARP, the Cognitive Authority Retention Protocol, is the discipline within MRCF that keeps the human as the locus of judgment. It treats authority retention as an active practice rather than an assumption.
CARP asks whether the user entered the interaction with a position, whether the interaction altered that position through reason rather than deference, whether the user can explain the final judgment independently, and whether the system's contribution remained assistive rather than dispositive.
MRVP: Meta-Recursive Validation Protocol
MRVP, the Meta-Recursive Validation Protocol, asks whether a framework can be examined using its own method without collapsing into circular affirmation. In MRCF, self-application is not proof by itself. It is a stress test: whether recursive inquiry exposes limits, contradictions, and refinement points rather than merely confirming the framework's premise.
MRCF in Practice
MRCF is applied wherever reasoning is the work and the stakes reward depth over speed.
These are applications of MRCF, not its boundary. The framework is broader than any single use case.
What MRCF Is Not
Not a runtime authorization mechanism
It does not sit in the execution path and gate live actions.
Does not issue verdicts
It does not issue ALLOW, DENY, or ABSTAIN.
Does not produce an authorization artifact
It emits no replayable record of a governed decision.
Does not prove pre-execution authorization
It does not establish that a governed action was permitted before execution.
Its subject is the human-AI reasoning process itself: how inquiry deepens or collapses, how authority is retained or delegated, and how meaning survives recursive interaction.
Published Foundation
MRCF is documented as Part I of a consolidated research corpus published on Zenodo, which serves as its stable citation target. The record supports authorship and prior-art posture without presenting MRCF as a patented system.
Canonical corpus. Meyman, E. (2026). The Meta-Recursive Cognition Corpus: Canonical Edition (2025 Research Lineage, v2.0). Zenodo. doi.org/10.5281/zenodo.18665681. MRCF is Part I.
Research paper. Meyman, E. (2025). Recursive Cognition Framework: Language-Thought Co-Evolution in Human-AI Collaboration. SSRN. doi.org/10.2139/ssrn.5284311
Defensive publication. Recursive Cognition Framework with Meta-Validation and Authority Retention Protocols. IP.com Prior Art Database, IPCOM000276466D.
Common Questions
What is MRCF?
MRCF, the Meta-Recursive Cognitive Framework, is FERZ's philosophical and methodological framework for structured human-AI inquiry. It treats dialogue with AI as a recursive cognitive process in which language, attention, authority, and meaning can either deepen or collapse, and it works to preserve human intellectual agency.
Is MRCF a governance or authorization mechanism?
No. MRCF sits beside, not inside, the runtime authorization stack. It does not issue ALLOW, DENY, or ABSTAIN and does not produce an authorization artifact. Its subject is the human-AI reasoning process, not runtime authorization.
What is cognitive sovereignty?
Cognitive sovereignty is the preservation of human judgment as the locus of meaning, responsibility, and final authority. MRCF treats it as an active discipline that resists authority drift, the gradual transfer of intellectual initiative from the human to the system.
What is semantic flattening?
Semantic flattening is the premature simplification of concepts when complexity is necessary for understanding. MRCF resists it, preserving the distinctions and depth that recursive inquiry depends on across successive cycles of interaction.
How is MRCF publicly documented?
MRCF is documented as Part I of the Meta-Recursive Cognition Corpus on Zenodo, alongside an earlier research paper and a defensive publication. This page presents that record as authorship and prior-art documentation, not as a patent-rights claim.
MRCF and the Other Lane
MRCF is the human-cognition lane. Deterministic governance and the Five Tests Standard define the runtime authorization boundary in the other lane: whether an action was authorized before execution.
Apply MRCF to a Complex Reasoning Problem
MRCF is a discipline for reasoning with AI while keeping human judgment as the locus of meaning and authority. Bring a hard problem and work it recursively, or read the published foundation.
FERZ, Inc. (2026). MRCF: Meta-Recursive Cognitive Framework. https://ferz.ai/methodologies/mrcf
BibTeX
@misc{ferz_mrcf_2026,
author = {{FERZ, Inc.}},
title = {{MRCF: Meta-Recursive Cognitive Framework}},
year = {2026},
howpublished = {\url{https://ferz.ai/methodologies/mrcf}}
}