Unified Preamble
Constitutional Architecture for Model-Building and Knowledge-Claim Frameworks
Author
U. Warring
Affiliation
Institute of Physics, University of Freiburg
Version
0.2.2
Last updated
2026-01-14
License
CC BY 4.0
Status
Constitutional meta-document (Tier -1)
Scope
Governs relationship between all invariant frameworks
Endorsement
Local framework under Harbourmaster stewardship
External Constraints
None replicated; standard logic and set theory assumed
Harbour Integration
This preamble operates as Tier -1 Constitutional Law relative to the Open-Science Harbour's tiered scaffold:
-1 (Constitutional)
Framework-framework relations
This Preamble: Non-Merge Rule, Coupling Protocol, Acid Test
0 (Causal)
Model-phenomena relations
Harbour causal boundary conditions (L_comparison ≤ cτ)
1a (Lamport)
Event ordering
Logical clocks, happens-before relations
1b (Metrological)
Physical clocks
SI second, frequency standards
2a (Calendar)
Coordination systems
UTC, civil time, scheduling
This preamble enforces the Endorsement Marker requirement established in Resolution R-O/L-01. The "Local candidate framework (no parity implied)" language from that resolution represents the permissive counterpart to this preamble's restrictive architecture.
No downstream document may claim broader endorsement scope than this preamble permits.
Purpose
This preamble establishes the constitutional architecture governing the relationship between frameworks that describe how we model phenomena (descriptive/discovery frameworks) and frameworks that constrain what humans may legitimately claim to know (epistemic frameworks).
It operates under a strict map-only epistemology: we deal only in representations, never in claims about territory itself.
This document must be prefixed to, or cited by, any Council-ratified framework operating in either domain.
The Foundational Commitment: Agnosticism About Territory
The Structural Bound
Humans are irreducibly bounded by embodied perception, evolved cognitive priors, language and symbol systems, and historically contingent scientific practices. This is not merely a practical limitation but a structural one. Any framework we propose is necessarily mediated by representation.
Consequence: No framework governed by human representation licenses claims to a final, observer-independent ontology. We therefore make no such claims and commit only to disciplined model-building.
Note: This is a methodological stance about what frameworks can license, not a metaphysical assertion about what is or is not accessible "in principle."
What Remains Accessible
Despite this bound, science has repeatedly uncovered structures that are remarkably stable across perspectives. The productive move is to stop asking "What is reality made of?" and instead ask: "Which structures remain invariant under changes of description, scale, and observer?"
These invariants function as selection constraints on viable representations, not as properties attributed to territory. Whatever models we build must preserve these invariants to remain viable. This is the only tractable question.
The Two Domains
Domain Definitions
Descriptive
How do we model phenomena? What structures survive interrogation?
Discovery frameworks, physical models, measurement protocols
Model structure, invariants, comparison geometry
Epistemic
What may humans legitimately claim to know?
Human Learning Framework, warrant criteria
Claims, warrant, justification
These domains are categorically distinct. Neither derives from, reduces to, nor determines the other.
Clarification: Distinctness here forbids derivation or reduction, not mutual historical influence. Epistemology may be informed by descriptive success; description may be shaped by epistemic practices. What is forbidden is collapsing one into the other or treating either as foundational for the other.
The Asymmetric Dependency
Epistemic frameworks constrain human claims about descriptive matters. Descriptive frameworks are not constrained by how humans learn.
A phenomenon may provide "resistance" without any human modelling it. A human claim about a phenomenon may be epistemically illegitimate even if the model is internally coherent.
This asymmetry is non-negotiable.
The Non-Merge Rule
Statement
No Council-ratified document may:
Derive phenomena from learning — assert that something exists because humans discovered it
Make model validity contingent on discovery — imply that successful modelling requires human observation
Collapse description into epistemology — treat "how we model" and "what we know" as the same question
Collapse epistemology into description — treat "what we know" as directly determined by model structure without interpretive mediation
Make territory claims — assert what reality "is" rather than what viable models preserve
Forbidden Coupling Patterns
The following syntactic forms violate the Non-Merge Rule regardless of apparent plausibility:
Analogical derivation
"X is like Y in structure, therefore X shares Y's ontological status"
Structure similarity does not license existence claims
Constraint mirroring
"Because epistemology has limit L, phenomena must also have limit L"
Epistemic bounds do not constrain territory
Discovery → existence
"If a model is successfully learned, its posits exist"
Learnability does not entail existence
Existence → knowledge
"If X exists, we can in principle know it"
Existence does not guarantee epistemic access
Reduction by definition
"Knowledge just is reliable model-tracking"
Conflates distinct categories by fiat
Enforcement
Any statement matching a forbidden pattern constitutes a category error and invalidates the document section containing it.
The Guardian stance holds veto power over merge violations. Citation: Violation of Clarity or Ethics (Category Error subtype).
"Resistance" as Operational Postulate
Both domains presuppose that models can fail and interpretations can be erroneous. Revision is forced by something other than preference.
We treat this as an operational postulate, not a metaphysical discovery:
The Resistance Postulate: Violation of invariants produces non-viable models. No claim is made about the source of this constraint; we model only the failure pattern itself.
This postulate:
Preserves normativity (models can be wrong)
Preserves discipline (not "anything goes")
Makes no territory claims (what "resists" is unspecified)
The word "resistance" is a heuristic marker, not a metaphysical term. It names the operational fact that some models fail without asserting what causes failure.
Framework Evaluation: Meta-Criteria
Since direct access to territory is blocked, evaluation must be indirect. Frameworks are judged by model-selection criteria, not by correspondence to "reality."
The Five Criteria
Invariance Robustness
Does the framework preserve known symmetries and invariants across viable models?
Absolute veto
Compression
Does it reduce independent assumptions? (Kolmogorov-style economy)
Strong preference
Cross-Domain Fertility
Does it generate insight across otherwise disconnected domains?
Exploratory
Stability Under Theory Change
Does it survive or gracefully adapt when models are replaced?
Strong preference
Explicit Boundary Conditions
Does it clearly state where it does not apply?
Conditional veto
Veto Logic and Conflict Resolution
Invariance Robustness is an absolute veto: a framework that violates known invariants is rejected regardless of other merits.
Explicit Boundary Conditions is a conditional veto with two satisfaction paths:
Path A (Full compliance): Framework documents explicit boundaries — passes veto
Path B (Provisional status): Framework explicitly flags as boundary-agnostic, accepts restricted jurisdiction, and is marked "Provisional" in all citations — passes veto conditionally
This prevents premature rejection of nascent but invariant-preserving frameworks while maintaining boundary discipline.
Criterion Conflicts
When criteria pull in opposite directions (e.g., Fertility vs. Boundaries):
Absolute veto criteria override all others
Conditional veto criteria must be satisfied via Path A or Path B
Strong preference criteria guide selection among passing candidates
Exploratory criteria inform research direction, not ratification
The Stress-Test Protocol
Purpose
Frameworks must be deliberately pushed into their breakdown zones. The manner of failure — graceful degradation versus conceptual collapse — reveals structural depth.
Key Terms
Glossary for Non-Specialists
η (eta)
Causal efficiency — a dimensionless ratio measuring how much of the available causal "budget" (lightspeed × time) is used for comparison. η = 1 means the comparison uses the full budget; η → 0 means the comparison geometry is collapsing toward the causal boundary.
L_comparison
The spatial separation over which a comparison (measurement, synchronisation, signal exchange) actually occurs. Must satisfy L_comparison ≤ cτ, where c is lightspeed and τ is the available time.
Comparison geometry
The spatial and temporal arrangement of source, apparatus, and comparison events. When η → 0, this geometry "collapses" — there is no longer enough causal budget to complete the comparison.
Causal boundary
The limit L_comparison = cτ, beyond which no comparison is physically possible. This is not a metaphysical claim but an operational constraint on all measurement and coordination.
Required Stress Tests
Before Council ratification, any descriptive framework must document behaviour under:
Scale extremes — Does the framework degrade gracefully at ultra-high energy, ultra-low temperature, or cosmological scales?
Boundary approach — For frameworks involving the causal boundary L_comparison = cτ, what happens as η → 0?
Interpretive plurality — Does the framework remain coherent under multiple viable interpretations, or does it secretly commit to one?
Domain export — When applied outside declared jurisdiction, does it fail explicitly or produce spurious results?
Mandatory η-Regime Test
Any framework invoking causal efficiency η must document failure mode at η → 0 as a veto-level stress test.
This is non-negotiable. The η → 0 regime is where "sails lose wind" — the causal boundary where comparison geometry collapses. Frameworks that do not explicitly characterise this regime cannot be ratified.
Failure Classification
Graceful degradation
Framework has depth; boundaries are real
Document η-regime behaviour; proceed to ratification
Conceptual collapse
Framework is scaffolding, not structure
Restrict jurisdiction; flag as provisional
Silent failure
Framework produces results without warning
Reject — requires explicit boundary documentation
Silent failure is an automatic rejection criterion.
Pluralism, Underdetermination, and Capture Prevention
Distinguishing Two Issues
Underdetermination: Multiple frameworks are observationally equivalent given current evidence. This does not entail that all are "equally real" — it may simply indicate that the interpretive layer is extraneous scaffolding.
Genuine pluralism: Multiple frameworks offer genuinely different structural commitments that survive stress-testing and satisfy meta-criteria. These are maintained as live options.
The distinction matters: underdetermination calls for further constraint; pluralism calls for parallel development.
Capture Prevention
Maintaining multiple viable frameworks is a systemic safeguard against epistemic capture, where a single model locks inquiry into a dead-end trajectory. This is a feature, not a bug.
The architecture prevents capture by:
Silo enforcement — frameworks cannot import invariants without explicit justification
Independent revision — each framework evolves under its own falsification conditions
Explicit endorsement scope — local frameworks do not claim broad authority
Provisional status path — nascent frameworks can develop without premature commitment
Framework Coupling Protocol
Permitted References
Epistemic → Descriptive
Worked examples, case studies
"Discovery X illustrates warrant structure..."
Descriptive → Epistemic
Constraint acknowledgment
"Human claims about model X are subject to [Learning Framework]..."
Forbidden Couplings
Derivation
"The framework implies phenomenon X exists"
Models cannot generate territory
Contingency
"X exists only if discoverable"
"Resistance" is not observer-dependent
Bidirectional arrows
"Learning ↔ Existence"
Violates asymmetric architecture
Definitional merger
"Knowledge is model-tracking"
Conflates distinct domains
Territory assertion
"Reality is relational structure"
Makes metaphysical claim
Analogical derivation
"Learning has structure S; therefore phenomena have structure S"
Matches forbidden pattern
Silo Enforcement
Independent Revision
Each framework may be revised independently, provided:
The revision does not violate the Non-Merge Rule
The revision does not create circular dependencies
The revision is ratified by Council under its own falsification conditions
The revision passes all stress tests at veto level
Cross-Domain Discipline
Frameworks in different domains must not:
Presume identity of invariants or import them without explicit justification citing this preamble's coupling protocol
Share falsification conditions (what falsifies a descriptive claim differs from what falsifies an epistemic claim)
Share jurisdiction (no framework may define validity conditions outside its declared stack position)
Note: This rule forbids unjustified cross-domain import, not all cross-domain reference. If a future result (e.g., causal-learning convergence theorems) genuinely bridges domains, it may be incorporated via explicit Council review under the coupling protocol. The burden of justification lies with the importer.
Acid Test for Preamble Compliance
Any Council-ratified document passes the Acid Test if:
Asymmetry preserved — The document does not imply that phenomena depend on human knowledge
Non-Merge respected — No statement matches any forbidden coupling pattern
Map-only language — No claims about "what reality is"; only about model viability
Stack position clear — The document explicitly states its tier
Coupling protocol followed — References to other frameworks use permitted types only
Silo integrity maintained — The document does not import invariants without explicit justification
Meta-criteria satisfied — Veto-level criteria (invariance, boundaries) are documented or provisional status declared
Stress tests passed — Failure modes are classified and boundary behaviour documented
η-regime documented — For causal-efficiency frameworks, η → 0 behaviour is explicit
A document failing any condition requires revision before ratification.
Constitutional Lock
This preamble is frozen upon Council ratification.
Revision requires:
Unanimous Council consent
Explicit demonstration that current architecture is falsified or incoherent
Proposed replacement that satisfies all existing compliance tests
No downstream document may override, amend, or reinterpret this preamble.
Archivist Note: Frozen: Non-Merge Rule; Forbidden Coupling Patterns; Coupling Protocol; Acid Test; Stack architecture; Meta-criteria hierarchy; Map-only commitment; Resistance Postulate; η-regime enforcement Evolving: Worked example appendix; stress-test templates; citation format; pattern library extensions; onboarding materials Appendix changes do not constitute constitutional revision.
Horizon Signal (archived for v0.3 review): "Experimental Sandbox" state between Provisional and Reject — monitor whether Path B proves insufficient for early-stage AI/social models.
Citation Format
Documents citing this preamble should use:
This document operates under the Unified Preamble (v0.2.2), Resolution R-O/L-01. Tier position: [-1 / 0 / 1a / 1b / 2a] Stack position: [DESCRIPTIVE / EPISTEMIC / CLAIMS / TOOLS] Coupling status: [SILOED / REFERENCES: list] Endorsement scope: [LOCAL / BROAD] Boundary status: [EXPLICIT / PROVISIONAL]
Appendix A: ADM Lens Integration
The three-lens decomposition clarifies how this architecture operates:
Analog Lens (The Coastline)
The unmediated "resistance" that phenomena provide. We do not design the coastline; we design the hull (apparatus) and sails (models) to navigate it. The structural bound is simply the point where analog signal exceeds digital resolution.
Note: "Resistance" and "coastline" are heuristic markers. They name operational constraints without asserting what provides them.
Digital Lens (The Sails)
Symbolic abstractions — Lorentz invariance, gauge symmetries, causal boundary conditions — that we project onto the analog coastline. These are functional tools that capture "wind" to generate motion (predictive power). Pluralism is necessary because different conditions require different sail configurations.
Memory Lens (The Repair Kit)
Stability across theory change. If a structure survives the transition from one model to another, it represents a fixed point in our collective interaction with the coastline. The causal boundary condition L_comparison ≤ cτ is a Memory-class invariant: it constrains all possible comparison geometries regardless of model specifics.
Appendix B: Worked Examples
Compliant Statement (Descriptive Framework)
"The Ordinans framework models emergent order in physical systems. Human claims about Ordinans models are subject to the Human Learning Framework (v2.1)."
Analysis: Correctly describes model scope without asserting what exists. Acknowledges epistemic constraint.
Compliant Statement (Epistemic Framework)
"The model is internally coherent (descriptive validity). The claim 'we know the structure because the network identified it' requires warrant documentation (epistemic legitimacy)."
Analysis: Correctly separates model coherence from warranted claim.
Compliant Statement (Invariants)
"Relational invariants appear in all viable models. The framework does not license claims about entities beyond these invariants."
Analysis: Jurisdictional language. States what the framework permits, not what exists.
Non-Compliant Statement
"The Ordinans exists because humans can learn to recognise it."
Analysis: Matches forbidden pattern (Discovery → existence). Violates Non-Merge Rule. Rejected.
Non-Compliant Statement
"Reality is relational structure rather than underlying stuff."
Analysis: Makes metaphysical territory claim. Violates map-only commitment. Rejected.
Non-Compliant Statement
"Whatever reality is, it must support these invariants."
Analysis: Covert territory claim via modal language. Violates map-only commitment. Rejected. Compliant alternative: "Whatever models we build must preserve these invariants to remain viable."
Non-Compliant Statement
"Because human cognition has computational limits, physical systems must also have computational limits."
Analysis: Matches forbidden pattern (Constraint mirroring). Epistemic bounds do not constrain territory. Rejected.
Appendix C: Stress-Test Template
For each framework submitted for ratification:
Scale extreme (high)
Scale extreme (low)
Boundary approach (η → 0)
Interpretive plurality
Domain export
Classification:
[ ] Graceful degradation — proceed to ratification
[ ] Conceptual collapse — restrict jurisdiction, flag provisional
[ ] Silent failure — reject, require boundary documentation
Scout Checklist (ADM-EC Protocol)
All stress-tests and meta-criteria audits shall be conducted under Scout protocol: scanning for occlusion, reframing needs, and unasked questions without synthesis or decision authority.
⚠ Horizon signal detected? (emerging invariants, blind spots)
Occlusion flagged? (hidden assumptions, unexamined priors)
Reframe needed? (category error risk, coupling violation)
Uncertainty explicit? (η-regime, boundary-agnostic status)
If all checks negative: Horizon clear — proceed to Council vote.
If any check positive: Flag for targeted re-deliberation before ratification.
Appendix D: Forbidden Pattern Library
Quick reference for Non-Merge Rule enforcement:
NM-01
Analogical derivation
"X is like Y, therefore X has Y's status"
Structure similarity ≠ existence claim
NM-02
Constraint mirroring
"Epistemology has limit L, therefore phenomena have L"
Epistemic bounds ≠ territory bounds
NM-03
Discovery → existence
"Successfully modelled, therefore exists"
Learnability ≠ existence
NM-04
Existence → knowledge
"Exists, therefore knowable in principle"
Existence ≠ epistemic access
NM-05
Reduction by definition
"Knowledge just is X"
Definitional fiat ≠ argument
NM-06
Modal territory claim
"Reality must have property P"
Modal claims about territory forbidden
NM-07
Implicit bidirectionality
"Learning and existence are mutually constraining"
Asymmetry violation
Usage: When reviewing documents, scan for pattern matches. Any match triggers Guardian review and potential veto.
Appendix E: Constitutional Summary
For new participants — one page, five commitments.
The Harbour Constitutional Architecture in Brief
This preamble governs all frameworks in the Open-Science Harbour. It enforces five core commitments:
1. Map-Only Epistemology
We build models; we do not claim to know "what reality is."
Compliant: "Viable models preserve Lorentz invariance." Non-compliant: "Reality is Lorentz-invariant."
2. The Non-Merge Rule
Description (how we model) and epistemology (what we may claim to know) are separate domains. Neither derives from the other.
Compliant: "Human claims about this model require warrant documentation." Non-compliant: "This phenomenon exists because we discovered it."
3. Asymmetric Stack
Epistemic frameworks constrain claims about descriptive frameworks. Descriptive frameworks are not constrained by how humans learn.
Arrows point one way only.
4. Invariance-First Evaluation
Frameworks are judged by whether they preserve known invariants, not by whether they "correspond to reality."
Veto-level test: Does the framework violate Lorentz invariance, gauge symmetry, or the causal boundary L_comparison ≤ cτ? If yes, reject.
5. Stress-Test Before Ratification
Every framework must document how it fails at its boundaries. Silent failure = automatic rejection.
Key question: What happens as η → 0 (causal efficiency approaches zero)?
Quick Reference
Make a claim about what exists
Stop. Reframe as model-viability claim.
Import an invariant from another domain
Cite coupling protocol; justify explicitly.
Submit a framework for ratification
Complete stress-test template (Appendix C).
Flag a document as non-compliant
Cite pattern ID from Appendix D.
This summary is for onboarding. The full preamble governs.
Appendix F: Pilot Audit — G–LLM Integration Framework
Demonstrating a full Acid Test walk-through.
Framework Under Review
G–LLM Integration in Foundational Education (v0.9.0) An invariant framework for governing AI-assisted human learning.
Acid Test Results
1. Asymmetry preserved
✅ Pass
Framework treats LLM outputs as tools subject to human warrant, not as sources of knowledge claims. No implication that phenomena depend on human learning.
2. Non-Merge respected
✅ Pass
Clear separation between "what the LLM generates" (descriptive) and "what the learner may legitimately claim" (epistemic). No forbidden patterns detected.
3. Map-only language
✅ Pass
Framework uses jurisdictional language throughout: "The framework does not license..." rather than "Reality is..."
4. Stack position clear
✅ Pass
Explicitly positioned at EPISTEMIC tier with TOOLS interface. Citation format compliant.
5. Coupling protocol followed
✅ Pass
References to descriptive frameworks (e.g., information theory, cognitive science) are constraint acknowledgments, not derivations.
6. Silo integrity maintained
✅ Pass
No invariants imported from descriptive domain without justification. MSG provenance notation is epistemic-native.
7. Meta-criteria satisfied
✅ Pass
Invariance: preserves DEEP partition structure. Boundaries: explicit scope (foundational education, not expert practice).
8. Stress tests passed
✅ Pass
See below.
9. η-regime documented
⚠️ N/A
Framework does not invoke causal efficiency η directly; no η-regime test required.
Stress-Test Results
Scale extreme (high)
Framework scales to large learner populations without structural change
Graceful
None required
Scale extreme (low)
Framework applies to individual learners; no minimum population assumed
Graceful
None required
Boundary approach
At boundary of "foundational" vs. "expert" learning, framework explicitly defers to domain-specific protocols
Graceful
Boundary documented in §Scope
Interpretive plurality
Framework-agnostic about LLM architecture; works with any generative model meeting MSG compliance
Graceful
No hidden commitment detected
Domain export
When applied outside education (e.g., to professional research), framework warns of jurisdiction limits
Explicit failure
Correct behaviour
Classification: Graceful degradation across all tests.
Scout Checklist
⚠ Horizon signal detected?
No
No emerging invariants or blind spots flagged
Occlusion flagged?
No
Assumptions about learner agency are explicit
Reframe needed?
No
No category error risk detected
Uncertainty explicit?
Yes
Provisional status on LLM capability claims acknowledged
Result: Horizon clear.
Council Disposition
G–LLM Integration Framework (v0.9.0) passes Acid Test.
Recommended for ratification at Tier EPISTEMIC with TOOLS interface.
Citation:
This document operates under the Unified Preamble (v0.2.2), Resolution R-O/L-01. Tier position: -1 (governed by) / EPISTEMIC (framework position) Stack position: EPISTEMIC Coupling status: REFERENCES: Human Learning Framework (v2.1), MSG Provenance Notation Endorsement scope: LOCAL Boundary status: EXPLICIT
End of Pilot Audit
Changelog
0.1.0
2025-12-15
Initial draft
0.1.1
2025-12-22
Stack diagram; worked examples
0.2.0
2026-01-14
Map-only commitment; meta-criteria; stress-test protocol; ADM integration
0.2.1
2026-01-14
Tier -1 positioning; Resistance Postulate; forbidden pattern library; conditional veto logic; Scout checklist; η-regime enforcement
0.2.2
2026-01-14
Onboarding summary (Appendix E); glossary; pilot audit (Appendix F); softened silo language for justified cross-domain import
End of Unified Preamble
Last updated