Derivation Protocol v1.0¶

Status: Canonical. Applies to all variants authored or admitted on or after 16 April 2026. URL: /derivation-protocol/v1/ — permanent. Changes produce a new version at a new URL.

What this document is¶

A variant is a domain-specific reading of Split-Domain Cognition. It takes the principle and shows what the principle means, in a specific domain, in language a practitioner of that domain recognises without first learning the AI-pipeline vocabulary.

This document specifies how a variant is derived and how it is admitted. Every accepted variant has been authored by someone who answered the seven questions for their domain, and reviewed against the four tests.

Plate 02Seven questions answered in writing; four tests applied in writing. A pass on all four — or no canonical status.

The seven questions¶

A variant is authored by answering seven questions about a domain. The questions are the bridge between the general principle and the specific practice.

1. What is the language work in this domain?¶

Identify what is being described, interpreted, generated, or articulated. What material do the domain's practitioners spend most of their attention attending to? (Studio: student work. Curation: the work being considered. Animal ethics: the practice or product under review. Editorial: the submission. Hiring: the candidate's evidence.)

2. What is the judgement work in this domain?¶

Identify the verdicts being issued. What does the domain have to decide? What are the outcomes that close a case? (Studio: assessment. Curation: inclusion, exclusion, positioning. Animal ethics: whether a practice is ethically acceptable. Editorial: publish / revise / reject. Hiring: decision.)

3. Where are the two collapsing in current practice?¶

Describe, with concrete examples, how the language layer and the judgement layer are currently fused. What phrases carry the fusion? What forms enable it? What institutional arrangements reward it? ("Welfare" in animal ethics. "Strong piece" in critique. "Promising candidate" in hiring. Each phrase does evaluative work under cover of descriptive work.)

4. What is the empirical cost of the collapse?¶

Name what the domain loses when the two are fused. The cost must be specific to the domain. (Editorial: author calibration. Curation: audience re-entry. Animal ethics: rights violations passing as welfare questions. Hiring: candidates cannot learn from rejection. Studio: students cannot learn what is being taught.)

5. What does the separation look like in this domain?¶

Describe what the three layers would be, on the domain's own terms. In the domain's language: what is the descriptive artefact? What is the explicit criterion? What is the narration of the verdict?

6. What is the relationship between this variant and the author's existing output?¶

Every variant has a landing in the author's practice — a tool, a published position, a partnership, an artwork, a grant, a conversation log. The variant names these relationships, not to justify itself but to place itself.

7. What is the deeper philosophical grounding, if any?¶

Some domains carry philosophical commitments that ground why the SDC move matters at all in that domain. Where such a grounding exists, the variant names it. Where none is available, the variant is honest about working from the principle alone.

The four validation tests¶

A variant is admitted to the canon when all four tests pass. Review is conducted against this document's version (v1.0 as of this URL).

1. The recognition test¶

A practitioner in the domain, reading the variant, recognises their own practice in it. The practitioner need not agree with the variant's argument. The practitioner need only recognise that the variant is talking about the practice they actually do — not a caricature, not an outside projection.

2. The specificity test¶

The variant could not be written about a different domain with only the nouns changed. If the argument would hold with "studio" replaced by "editorial" and the concrete examples swapped, the variant has not yet reached the domain's own grain.

3. The cost test¶

The variant names a cost that the domain itself acknowledges, at least when pressed. A variant that names a cost the domain would not recognise as a cost is making a claim from outside the domain, which is legitimate but should be flagged.

4. The stake test¶

The variant's author has genuine stake in the domain — professional, lived, ethical, or intellectual. A variant in a domain where the author has no stake is outside the scope of the canon. Intellectual-only variants are permissible but must be flagged as such at the top of the variant.

Variant document structure¶

The variant is written in a structure that renders the seven answers visible to the domain. The skeleton below has worked well; it is a starting template, not a mandate.

Opening. A paragraph naming the domain and the collapse it faces, concretely. Legible to a practitioner without any SDC vocabulary.
The language work. What the domain attends to, in the domain's terms.
The judgement work. What the domain decides, in the domain's terms.
The collapse, with examples. The concrete fusion.
The cost. What is lost when the collapse is accepted as normal.
The separation. What the three layers would be, in the domain's language.
Philosophical grounding, if any.
Relationship to the author's existing output.
Closing. What the variant makes possible.

Typical length: 2,000–4,000 words. Shorter if the domain is narrow. Longer risks becoming a generic treatise on SDC with the domain's specificity lost.

States and lifecycle¶

Variants move through explicit states:

Proposed. Named only; no development yet.
Noted. Argument sketched as a pre-variant note; not yet a full variant.
Under Review. Full variant drafted and submitted; four tests being applied.
Canonical. Admitted to the canon. Has a stable URL and a version number.
Archived. Withdrawn or superseded. Reason recorded; URL remains.

A variant's state is visible on its page. External projects may adopt a variant in any state; the declaration records which state was adopted. (A project adopting a Proposed variant is making a different kind of claim than a project adopting a Canonical one.)

How this document relates to other canonical documents¶

SDC in brief — the principle stated plainly, for practitioners who have not yet read the long form.
Preamble — the default SDC refuses, which the four tests operationalise.
This document — the policy governing variant authoring and admission.
Each variant — a specific reading produced by running the seven questions and passing the four tests.

Versioning¶

This is v1.0 of the derivation protocol. Subsequent versions will be published at their own URLs (/derivation-protocol/v1-1/, /derivation-protocol/v2/, etc.). Past versions will remain live.

Past variants are reviewed against the protocol version current at the time of their admission. A protocol change does not automatically re-adjudicate past acceptances. A project can state: "This variant was reviewed against derivation-protocol v1.0" and the claim will remain checkable in perpetuity.