This document is normative for level separation in matrix.
It does not resolve epistemic tensions. It makes them explicit and enforceable.
Violations of level separation are treated as structural failures, not content errors.
Most epistemic failures do not arise from false statements, but from unmarked transitions between levels of analysis.
Typical collapses include:
- description → evaluation
- evidence → conclusion
- model output → decision
- agreement → truth
- compliance → legitimacy
matrix exists to make such transitions explicit,
inspectable, and contestable.
matrix operates with three strictly separated levels.
These levels are not metaphysical claims. They are operational constraints.
What lives here:
- statements about the world
- data, measurements, observations
- models and model outputs
- policies, rules, procedures
- decisions and actions
Typical forms:
- “X happened.”
- “The model outputs Y.”
- “Policy P requires Z.”
Key property: Object-level items do not justify themselves.
They may be true or false, effective or ineffective, but they carry no inherent authority.
What lives here:
- claims about object-level items
- admissibility judgments
- scope limitations
- validity or applicability assessments
- interpretations of evidence or outputs
Typical forms:
- “This claim is admissible.”
- “This evidence supports (but does not prove) X.”
- “This model output is valid under assumptions A–C.”
Key property: Meta-level statements confer conditional status, never absolute authority.
They remain revisable.
What lives here:
- constraints on how meta-level judgments may be formed
- STOP rules
- level-transition rules
- non-negotiable structural limits
Typical forms:
- “This transition is illegitimate.”
- “Further inference would be self-sealing.”
- “STOP: authority is being assumed without basis.”
Key property: Meta²-level rules do not argue. They intervene.
Transitions between levels are:
- permitted only when explicitly marked
- never automatic
- always contestable
-
Object → Meta
“The data shows X, therefore X is true.” -
Meta → Object
“This is admissible, therefore it must be done.” -
Object → Meta²
“This model is accurate, so STOP rules don’t apply.” -
Meta² → Object
“This is structurally illegitimate, so the content is false.”
All of these are category errors.
Level separation itself is a normative imposition.
It cannot be derived from facts. It cannot be proven without circularity.
From a strict foundational perspective, this makes it unzulässig.
However:
Without this imposition:
- everything becomes self-justifying,
- no STOP is possible,
- no boundary can be enforced.
The system therefore carries this break openly.
It is not resolved. It is managed.
Runs operate within this level structure.
A run may:
- expose an illegitimate transition,
- demonstrate how a collapse occurs,
- or reserve space for such an analysis.
Runs do not redefine levels. They test failures against them.
STOP rules always operate at Meta² level.
They do not:
- refute content,
- decide correctness,
- or resolve disagreement.
They prevent further progression when progression itself becomes illegitimate.
This document does not claim that:
- object-level truth is inaccessible,
- meta-level evaluation is sufficient,
- meta²-level constraints are final.
It claims only that unmarked level collapse is epistemically destructive.
Any contribution to matrix must be readable as:
- clearly object-level,
- clearly meta-level,
- or clearly meta²-level.
If this cannot be determined, the contribution is structurally invalid.
Level separation does not guarantee correctness. It guarantees inspectability.
Inspectability is the minimum condition under which disagreement, revision, and responsibility remain possible.