Methodologies

Two domains: runtime authorization and recursive human-AI cognition.

FERZ methodologies address two related domains. The first is runtime authorization: the point before execution where a consequential AI action is evaluated, permitted or blocked, and recorded in a tamper-evident authorization artifact. The second is recursive human-AI cognition: whether a person retains authority, semantic depth, and the capacity to ask better questions while reasoning with AI. Governance infrastructure enforces the first. MRCF examines the second.

FERZ methodologies span two related problems: how consequential AI actions are governed before execution, and how humans preserve authority, meaning, and judgment while reasoning with AI. Deterministic governance supplies the runtime authorization boundary. MRCF examines recursive human-AI cognition: the point at which AI assistance can either deepen inquiry or flatten it.

The governance problem is an enforcement concern, not a monitoring one. It is distinct from identity and access management, which authorizes principals to reach resources rather than AI actions to execute, and from policy engines, which evaluate rules without binding the result to a tamper-evident, independently reconstructable record of the governed action. The deterministic governance framework, the authorization boundary, and the verdict space are formalized across the FERZ research corpus, including On the Impossibility of Observability-Based Authorization.

Choose the right lane

Governance infrastructure

Governing AI actions before execution

For consequential AI actions, governance requires a runtime authorization boundary: a point before execution where the proposed action is evaluated, permitted or blocked, and recorded in a tamper-evident authorization artifact.

Start with deterministic governance →

Human-AI cognition

Preserving human authority and meaning

For human reasoning with AI, the problem is not only answer quality. It is whether the human retains authority, semantic depth, and the capacity to ask better questions across recursive dialogue.

Read MRCF →

Primary methodologies

Deterministic Governance

Pre-execution authorization for consequential AI actions. Identical governed state produces identical verdicts and rationale, with a tamper-evident authorization artifact sufficient for independent replay.

What deterministic governance means →

AI Capsule™

Machine-readable representation of governance constraints, source references, interfaces, invariants, assumptions, and validation rules. Capsules support reconstruction and review; they are not themselves runtime authorization boundaries.

Explore AI Capsule →

MRCF

A philosophical and methodological framework for recursive human-AI inquiry, cognitive sovereignty, anti-semantic flattening, and authority retention.

Explore MRCF →

The governance infrastructure methods inform FERZ product families including LASO(f), DELIA, CausaCore, and Constitutional Blockchain Architecture. MRCF operates at a different layer: the human reasoning process that shapes doctrine, policy, and governance design.

Supporting research

Semantic Condensation Methodology

FERZ has also published research on semantic condensation for large-document processing under AI context-window constraints. SCM is a supporting research method for preparing complex technical and regulatory material for analysis. It is not a runtime authorization boundary.

Read the SSRN publication

Why observability is not governance

Monitoring, logging, and after-the-fact review show that an AI system did something. They do not establish that the effect-bearing action was permitted before it released. The governing control is the authorization boundary: the point at which the action is allowed to proceed, recorded in a tamper-evident authorization artifact. If that artifact cannot be produced, the action fails closed.

The verdict space is ALLOW, DENY, and ABSTAIN. An ABSTAIN blocks execution pending authorized human override; escalation is the consequence of ABSTAIN, not a fourth verdict. Provenance is a test, not a verdict.

The current FERZ standard is the Five Tests Standard: Stop, Ownership, Replay, Escalation, and Provenance. Provenance is origin, not truth; it is a test, not a verdict.

Frequently asked questions

What is the difference between the two FERZ methodology lanes?

One lane governs consequential AI actions before they execute, using a runtime authorization boundary and tamper-evident authorization artifacts. The other, MRCF, examines human-AI cognition: whether a person retains authority, semantic depth, and the capacity to ask better questions across recursive dialogue.

Is an AI Capsule a runtime authorization boundary?

No. An AI Capsule is a machine-readable representation of constraints, source references, interfaces, invariants, assumptions, and validation rules. It supports reconstruction and review. The authorization boundary is the control point that evaluates a proposed action before execution and emits ALLOW, DENY, or ABSTAIN.

What is the verdict space?

ALLOW, DENY, and ABSTAIN. An ABSTAIN blocks execution pending authorized human override. Escalation is the consequence of ABSTAIN, not a fourth verdict. Provenance is a test, not a verdict.

Why is observability not sufficient for governance?

Monitoring and logging show that an AI system did something. They do not establish that the action was permitted before it released. The governing control is the authorization boundary, recorded in a tamper-evident authorization artifact. If that artifact cannot be produced, the action fails closed.

Is MRCF a governance or authorization mechanism?

No. MRCF does not issue ALLOW, DENY, or ABSTAIN and does not produce an authorization artifact. It is a philosophical and methodological framework for recursive human-AI inquiry, cognitive sovereignty, and resistance to semantic flattening.

Does this page describe FERZ products?

No. This page describes FERZ methodologies. Products such as LASO(f), DELIA, CausaCore, and Constitutional Blockchain Architecture are implementation families that apply these ideas in different enforcement or governance contexts.

Need to determine which methodology applies to your environment? FERZ can help distinguish authorization, representation, rule codification, and human-AI reasoning problems before they are collapsed into a single generic AI governance initiative.

Request architecture discussion

References

The FERZ research corpus is the authoritative source for the concepts on this page.

Cite this page

FERZ, Inc. (2026). Methodologies. https://ferz.ai/methodologies

BibTeX
@misc{ferz2026methodologies,
  author = {{FERZ, Inc.}},
  title  = {Methodologies},
  year   = {2026},
  howpublished = {\url{https://ferz.ai/methodologies}}
}