Composition Science Atlas Research Framework

Purpose

This document defines how research from architecture, perception science, neuroscience, typography, painting, film, music, cartography, industrial design, human factors, linguistics, biology, systems engineering, and other disciplines enters the Composition Science knowledge system.

Its purpose is not merely to standardize reports. It is to make findings comparable without erasing the differences between disciplines.

The framework must allow future research agents to determine:

1. What a discipline has learned.
2. What kind of evidence supports that knowledge.
3. Whether the evidence is independent or derivative.
4. Which underlying composition principles may be shared across domains.
5. Where translation between domains is valid, partial, or misleading.
6. What contradictions reveal about boundary conditions.
7. What should be promoted into the theory registry.
8. What remains uncertain and should be studied next.

Research State Snapshot

Theory Version

Not yet assigned. The current body of work is pre-unification and should be treated as a collection of candidate principles rather than a settled theory.

Knowledge Base Version

Atlas Framework 1.0.

Highest Confidence Areas

• Human perceptual limits constrain every visual composition system.
• Domain context changes the usefulness of otherwise valid principles.
• Stable identifiers and evidence traceability are necessary for cumulative research.
• Contradictions and failures carry as much scientific value as confirming evidence.

Lowest Confidence Areas

• Whether a small set of universal composition laws exists.
• Whether principles can be quantitatively transferred between media.
• Whether a “composition genome” will prove analytically useful rather than merely descriptive.
• How confidence should be aggregated across heterogeneous evidence types.

Largest Remaining Unknown

Can cross-disciplinary similarities be reduced to common causal mechanisms, or are many of them only metaphorical similarities created by broad language such as hierarchy, rhythm, balance, and flow?

Active Research Streams

• Visual perception and attention
• Color science
• Typography
• Architecture and spatial cognition
• Product and industrial design
• Familiarity, convention, and intuitive interaction
• Web component foundations
• Artist and movement composition analysis

Recently Invalidated Ideas

None formally registered. However, the assumption that convergence across multiple disciplines is inherently strong evidence is rejected by this framework.

Priority Changes

Cross-disciplinary synthesis should not wait until all discipline research is complete. Translation, contradiction detection, and evidence lineage should be recorded from the beginning.

Key Findings

• Atlas should be organized as an evidence graph, not only as a folder hierarchy or encyclopedia.
• Every discipline should be studied through a common question set, but discipline-specific methods and vocabulary must be preserved.
• Cross-disciplinary agreement must be tested for shared ancestry before it counts as independent convergence.
• “Universal principles” should begin as hypotheses and earn promotion through causal explanation, boundary testing, prediction, and replication.
• Translation records are first-class research artifacts. They are not informal analogies.
• Contradictions should usually produce boundary conditions, subtypes, or rejected translations rather than being averaged away.
• The composition genome should initially be treated as an exploratory model, not a scientific law.

1. Governing Model

Project Atlas contains six connected layers.

Disciplines
    ↓
Source Evidence
    ↓
Domain Findings
    ↓
Cross-Discipline Translations
    ↓
Candidate Principles and Boundary Conditions
    ↓
Theory, Predictions, Experiments, and Applications

Each layer answers a different question.

No layer may silently substitute for another. A respected design tradition is not empirical evidence. A correlation is not a mechanism. A metaphor is not a validated translation. A useful heuristic is not automatically a universal principle.

2. Canonical Record Types and Stable Identifiers

The REP stable identifier system remains authoritative. Atlas adds subtypes without replacing it.

Recommended Atlas concept subtypes:

The older OBS-, EVD-, and LAW- labels in the Composition Science template should be normalized in future revisions:

• Observation records should be journal entries or evidence records.
• EVD- should become EV-.
• A proposed law should begin as HY-PRIN- and may become TH- only after governance review.

3. Discipline Research Protocol

Every discipline study must answer the same core questions. These questions make comparison possible, but researchers may add domain-specific questions where necessary.

3.1 Discipline Definition

• What does this discipline study?
• What practical or theoretical problems is it trying to solve?
• What counts as success inside the discipline?
• Which schools, subfields, or traditions disagree about its purpose?

3.2 Native Constraints

• What physical constraints shape the discipline?
• What biological and perceptual constraints apply?
• What technological constraints apply?
• What social, cultural, economic, legal, or institutional constraints apply?
• Which constraints are permanent, and which are historically contingent?

3.3 Native Units and Variables

• What does the discipline measure?
• Which variables are directly measurable?
• Which constructs are inferred?
• Which terms lack stable operational definitions?
• At what spatial, temporal, or organizational scale does it operate?

3.4 Foundational Principles

• Which principles are repeatedly treated as fundamental?
• Which are supported by experiments?
• Which emerged from craft tradition or expert practice?
• Which may be stylistic conventions presented as laws?
• Which have known counterexamples?

3.5 Mechanisms

• Why is each principle believed to work?
• Is the proposed mechanism perceptual, cognitive, physical, social, cultural, economic, or technical?
• Does the evidence establish causation, correlation, prediction, or only description?
• What alternative explanations remain plausible?

3.6 Failure and Boundary Conditions

• Where does the principle fail?
• What populations, cultures, tasks, media, environments, or expertise levels change the result?
• Is failure caused by the principle being wrong, or by applying it outside its native scope?
• Are there nonlinear thresholds, reversals, or tradeoffs?

3.7 Historical Lineage

• Where did the idea originate?
• Which later disciplines borrowed it?
• Are multiple sources truly independent, or do they descend from one original claim?
• Has the idea changed meaning during transmission?

3.8 Composition Relevance

• What aspects of organization, attention, sequence, relationship, meaning, or action does the field illuminate?
• Which Atlas concepts might it support or challenge?
• What is lost when its ideas are translated into visual or interface design?

4. Evidence Evaluation Model

A single star score is too crude for Atlas. Evidence must be evaluated across separate dimensions.

4.1 Evidence Dimensions

Each evidence record should include:

Each dimension should be rated separately as:

• Strong
• Moderate
• Weak
• Unknown
• Not applicable

A final confidence judgment may be assigned, but the dimensional profile must remain visible.

4.2 Evidence Type Classification

Evidence records must identify their type:

• Controlled experiment
• Field experiment
• Observational study
• Meta-analysis or systematic review
• Formal model or proof
• Engineering test
• Standards or regulatory evidence
• Historical artifact
• Long-running professional practice
• Expert testimony
• Case study
• Failure report
• Ethnographic observation
• Qualitative interview
• Simulation
• Derived analysis
• Secondary synthesis

Different evidence types answer different questions. Controlled experiments may establish causal relationships but lack ecological validity. Historical practice may establish durability but not causation. Expert consensus may identify useful hypotheses but can preserve inherited error.

4.3 Independence Test

Before citing cross-disciplinary convergence, researchers must check:

1. Did the disciplines rely on the same original researcher or theory?
2. Did one discipline explicitly borrow the principle from another?
3. Are the sources analyzing the same dataset or examples?
4. Are they using the same broad term for different mechanisms?
5. Are apparent replications actually repeated citations of one finding?
6. Did similar practical constraints independently produce the same solution?

Convergence is strongest when distinct disciplines reach compatible findings through different methods, datasets, histories, and practical pressures.

5. Universal Principle Lifecycle

No principle begins as a law.

Stage 0 — Repeated Observation

Similar patterns appear in one or more disciplines.

Stage 1 — Candidate Concept

The pattern receives a provisional definition. Ambiguous umbrella terms should be split where possible. For example, “hierarchy” may contain perceptual salience, semantic importance, organizational rank, and action priority.

Stage 2 — Cross-Discipline Hypothesis

A testable claim proposes a shared mechanism.

Example:

HY-PRIN-007: Increasing perceptual differentiation between task-relevant classes reduces visual search time until added differentiation begins producing competing salience.

This is more useful than “contrast improves hierarchy” because it identifies variables, outcomes, and a likely boundary.

Stage 3 — Translation Testing

The mechanism is examined across multiple domains. Researchers record what remains invariant and what must change.

Stage 4 — Boundary Mapping

Known failures, reversals, tradeoffs, user differences, and context dependencies are documented.

Stage 5 — Predictive Validation

The principle makes predictions not already contained in its source evidence. Those predictions are tested through experiments, cases, or prospective observation.

Stage 6 — Theory Candidate

The principle may be proposed for inclusion in the theory registry when it has:

• A stable operational definition
• Multiple credible evidence records
• At least one independent line of confirmation
• Documented counterevidence
• Known boundary conditions
• A plausible mechanism
• At least one successful prediction or falsification attempt
• Clear distinction from neighboring principles

Stage 7 — Governed Theory Record

Only research governance may promote the principle to TH- status. Theory records remain revisable.

6. Translation Framework

Cross-disciplinary translation is the core scientific activity of Atlas, but also its greatest source of false confidence.

6.1 Translation Record

Every translation must document:

• Source discipline and concept
• Target discipline and concept
• Shared abstract structure
• Source mechanism
• Proposed target mechanism
• Variables preserved
• Variables transformed
• Variables lost
• Scale differences
• Temporal differences
• Human task differences
• Evidence supporting the mapping
• Evidence challenging the mapping
• Known limits
• Testable predictions
• Translation confidence

6.2 Translation Classes

Only mechanistic and well-tested functional equivalence should materially strengthen a universal principle. Structural analogy is valuable for hypothesis generation. Metaphorical resemblance should never be counted as confirmation.

6.3 Example Translation Record

identifier: HY-TRAN-ARCH-UI-001
source_concept: architectural circulation
target_concept: interface navigation
translation_class: functional_equivalence
shared_problem: enabling purposeful movement through a structured environment
preserved_variables:
  - destination visibility
  - route choice
  - landmarks
  - transition points
transformed_variables:
  - physical distance becomes interaction cost
  - rooms become information states
lost_variables:
  - bodily locomotion
  - gravity
  - full-scale spatial memory
confidence: medium
status: active hypothesis

The translation is promising, but not literal. Navigation interfaces may inherit some wayfinding mechanisms while differing substantially in embodiment, scale, persistence, and reversibility.

7. Contradiction Resolution Protocol

Contradictions must not be averaged into a vague compromise.

When two credible findings disagree, classify the disagreement.

For every contradiction:

1. Preserve both claims and their evidence IDs.
2. Identify whether the terms and outcome variables are equivalent.
3. Test for hidden boundary conditions.
4. Generate competing hypotheses.
5. Specify evidence that would discriminate between them.
6. Update confidence without deleting the losing claim.
7. Record whether the contradiction narrows, splits, or invalidates a principle.

8. Universal Principle Registry

The registry is the governed index of candidate and accepted principles.

Each entry must include:

identifier: HY-PRIN-XXX
title: null
canonical_definition: null
status: proposed | active | contested | theory-candidate | deprecated
mechanism: null
independent_evidence_lines: []
disciplines: []
supporting_evidence: []
challenging_evidence: []
translations: []
boundary_conditions: []
predictions: []
experiments: []
applications: []
confusable_concepts: []
confidence_profile: {}
last_reviewed: YYYY-MM-DD

Initial Candidate Families

These are research territories, not accepted laws:

• Differentiation and contrast
• Grouping and segmentation
• Hierarchy and priority signaling
• Rhythm and recurrence
• Balance and distribution
• Flow, sequence, and path
• Scale and proportion
• Affordance and action possibility
• Feedback and state visibility
• Familiarity and learned convention
• Predictability and expectation
• Coherence and consistency
• Variety and novelty
• Tension and release
• Density and compression
• Redundancy and error tolerance
• Progressive disclosure
• Landmarking and orientation
• Figure-ground organization
• Information scent and anticipatory cues

Researchers should be willing to split, merge, rename, or reject these families.

9. Translation Matrix

The translation matrix is a view generated from translation records, not an independent source of truth.

Every cell must eventually link to one or more translation IDs. Blank cells are useful: they reveal missing research or concepts that may not transfer.

10. Composition Genome

The composition genome is an exploratory representation of how principles combine in a work, style, discipline, or system.

It should not imply that composition traits are biologically inherited, fixed, independent, or reducible to a single scalar value.

10.1 Genome Node Requirements

Each node must specify:

• Operational definition
• Observable indicators
• Measurement method
• Scale and unit
• Context
• Interactions with other nodes
• Evidence basis
• Known confounds
• Reliability
• Validity

10.2 Possible Node Categories

• Attention distribution
• Salience concentration
• Repetition interval
• Variation rate
• Spatial density
• Temporal density
• Symmetry
• directional bias
• segmentation strength
• transition abruptness
• hierarchy depth
• information redundancy
• novelty frequency
• predictability
• path constraint
• recovery support

10.3 Prohibited Early Uses

Until validated, genome scores should not be used to:

• Rank artistic quality
• Claim universal aesthetic superiority
• Diagnose user experience from appearance alone
• Compare incomparable media without normalization
• Produce false numerical precision
• Replace direct usability or perception testing

The first goal is descriptive and hypothesis-generating. Predictive use must be earned empirically.

11. Scientific Workflow

Select discipline or uncertainty
        ↓
Review existing Atlas records and prior REPs
        ↓
Create discipline profile
        ↓
Collect source evidence and historical lineage
        ↓
Separate observations from interpretations
        ↓
Extract domain-native findings
        ↓
Identify mechanisms and boundary conditions
        ↓
Create or update translation records
        ↓
Test evidence independence
        ↓
Map support and contradictions to candidate principles
        ↓
Generate falsifiable predictions and experiments
        ↓
Update registries, website views, and research backlog
        ↓
Produce REP and handoff instructions

11.1 Intake Rule

No new report is considered incorporated merely because it exists in the repository. Incorporation requires:

• Valid metadata
• Stable identifiers
• Evidence records
• Explicit confidence
• Mappings to existing concepts or proposed new concepts
• Open questions
• Recommended registry changes
• A completed REP or documented reason why research remains active

11.2 Synthesis Rule

Research agents may propose but may not silently alter canonical theory. They must state:

• What should change
• Why
• Which evidence supports the change
• What existing records are affected
• What uncertainty remains

11.3 Revision Rule

All changes must preserve:

• Prior identifiers
• Contradictory evidence
• Deprecated interpretations
• Revision history
• Supersession links

12. Prioritizing New Disciplines

Disciplines should not be selected only because they are interesting. Use a portfolio strategy.

12.1 Priority Dimensions

• Foundational relevance to perception, cognition, action, or organization
• Methodological rigor
• Independence from disciplines already studied
• Ability to challenge current assumptions
• Availability of primary evidence
• Potential for measurable variables
• Applicability across media
• Neglected populations or contexts
• Value to current engineering or design decisions

12.2 Recommended Research Waves

Wave 1 — Human constraints and causal foundations

• Vision science
• Auditory perception
• Cognitive psychology
• Attention research
• Memory and learning
• Motor control
• Human factors and ergonomics
• Psychophysics

Wave 2 — Disciplines with mature composition practice

• Architecture
• Typography
• Cartography
• Information visualization
• Industrial design
• Film editing and cinematography
• Music theory and perception
• Painting and graphic composition

Wave 3 — Meaning, culture, and coordination

• Linguistics and pragmatics
• Semiotics
• Anthropology
• Sociology
• Behavioral economics
• Rhetoric
• Narrative theory
• Organizational design

Wave 4 — Complex adaptive and engineered systems

• Systems engineering
• Control theory
• Information theory
• Network science
• Ecology
• Evolutionary biology
• Safety engineering
• Resilience engineering

The waves may run in parallel. Their purpose is to maintain balance between biological constraints, craft knowledge, cultural meaning, and formal systems.

13. Repository and Website Structure

A file hierarchy remains useful for stewardship, while the website should expose graph relationships.

/composition-science
  /governance
  /disciplines
  /evidence
  /hypotheses
  /theory
  /translations
  /boundaries
  /experiments
  /genome
  /registries
  /research-journal
  /research-packages
  /generated

Recommended generated website views:

• Discipline explorer
• Principle registry
• Evidence graph
• Translation matrix
• Contradiction map
• Boundary-condition explorer
• Research confidence dashboard
• Open-question queue
• Genome explorer
• Theory change history

Markdown files remain canonical. Website pages should be generated from them and must not become an independent editable knowledge source.

14. Quality Gates

A discipline study is not complete until it passes these gates.

Evidence Gate

• Primary sources were sought.
• Important claims link to evidence IDs.
• Counterexamples and competing viewpoints were reviewed.
• Source lineage and independence were examined.

Concept Gate

• Native terminology is defined.
• Broad concepts are operationalized or marked ambiguous.
• Observation, interpretation, hypothesis, and theory remain distinct.

Translation Gate

• Mappings identify what is preserved, transformed, and lost.
• Metaphors are labeled as metaphors.
• Domain transfer confidence is explicit.

Falsification Gate

• The strongest conclusion has at least one stated failure condition.
• Evidence that would disprove or narrow the claim is specified.
• Negative findings are retained.

Handoff Gate

• A new agent can reconstruct the research.
• Registry updates are explicit.
• Open questions and next actions are prioritized.
• The REP completion checklist is satisfied.

15. Initial Hypotheses Created by This Framework

HY-PRIN-001 — Cross-Domain Constraint Convergence

Hypothesis

When independent disciplines face the same underlying human perceptual or cognitive constraint, they will tend to evolve structurally similar solutions even without direct intellectual transfer.

Predictions

• Similar solutions will appear in historically disconnected traditions.
• The strongest commonalities will correspond to stable human constraints.
• Differences will correlate with medium, task, culture, technology, or scale.

Supporting Evidence

Not yet registered.

Counter Evidence

Possible widespread convergence caused by shared cultural transmission rather than common constraint.

Confidence

Low. High-value research hypothesis.

HY-PRIN-002 — Translation Loss

Hypothesis

Every cross-disciplinary translation loses or transforms variables, and unrecorded translation loss is a major source of false universal principles.

Prediction

Translations that explicitly document lost variables will make more accurate domain predictions than translations based only on shared vocabulary.

Confidence

Medium.

HY-PRIN-003 — Boundary-First Generalization

Hypothesis

A principle becomes more transferable when its boundary conditions are known, even if documenting those limits lowers its apparent universality.

Prediction

Boundary-rich principles will outperform broad heuristics in prospective design decisions.

Confidence

Medium.

HY-PRIN-004 — Principle Interaction Dominance

Hypothesis

Composition outcomes are often governed more by interactions among principles than by the isolated strength of any one principle.

Prediction

Genome models containing interaction terms will predict perception and task outcomes better than independent trait scores.

Confidence

Low-Medium.

16. Observations

JR-OBS-001

Observation

Many disciplines use recurring terms such as hierarchy, rhythm, balance, contrast, and flow.

Interpretation

These terms may indicate shared compositional structures, but their breadth creates a high risk of false equivalence.

Confidence

High for the observation; Medium for the interpretation.

JR-OBS-002

Observation

The current Composition Science template and REP specification use partially different identifier systems.

Interpretation

Identifier divergence will create duplicate records and weaken traceability unless normalized.

Confidence

High.

JR-OBS-003

Observation

A simple evidence-level score combines source quality, replication, transferability, and mechanism into one number.

Interpretation

A single score hides important weaknesses and can produce unjustified confidence.

Confidence

High.

17. Evidence

EV-CS-001

Citation

Composition Science Markdown Template v1.

Summary

Defines standard Composition Science metadata, observations, evidence, candidate laws, open questions, next actions, revision history, and agent instructions.

Supports

• DF-CS-ATLAS-001
• HY-PRIN-002

Challenges

• The template's EVD- and LAW- labels conflict with the REP canonical identifier model.

EV-CS-002

Citation

Research Execution Package Specification v2.

Summary

Defines the REP as the canonical research handoff and establishes identifiers, metadata, mandatory sections, theory impact, evidence traceability, quality metrics, research debt, and completion criteria.

Supports

• DF-CS-ATLAS-001
• The requirement for stable, traceable, executable research artifacts

Challenges

• The REP does not yet define detailed cross-disciplinary translation or evidence-independence rules.

18. Open Questions

1. What operational definition should distinguish a principle, mechanism, pattern, heuristic, and law?
2. How should evidence confidence be aggregated without producing false precision?
3. What graph schema best represents evidence lineage, translation, contradiction, and supersession?
4. Which initial principle family should be used to test the complete workflow?
5. How can cultural variation be incorporated without treating culture as noise?
6. What methods can measure composition traits consistently across static, interactive, spatial, and temporal media?
7. Which existing research streams have already generated findings ready for registry extraction?
8. Should the Composition Science base template be revised to adopt REP identifiers directly?
9. What governance process promotes HY-PRIN- records into TH- records?
10. Which confidence dimensions should be required versus optional for different evidence types?

19. Recommended Next Research

Highest-Value Next Step

Run a pilot synthesis on one candidate principle across three methodologically distinct disciplines.

Recommended pilot:

Differentiation for perceptual search and priority signaling across:

1. Vision science and psychophysics
2. Typography or information visualization
3. Aviation or medical human factors

This candidate is preferable to a broad concept such as “balance” because it can be operationalized using measurable variables such as target-distractor similarity, search time, error rate, salience, and signal detectability.

The pilot should produce:

• One discipline profile per field
• A minimum of ten strong evidence records
• A source-lineage map
• At least three domain findings
• Two or more translation records
• One candidate principle record
• Documented contradictions and boundaries
• At least one falsifiable experiment proposal
• A completed REP

20. Research Backlog

Critical

• Normalize identifiers between the Composition Science template and REP specification.
• Define machine-readable schemas for evidence, translation, principle, and boundary records.
• Select and execute the first cross-disciplinary pilot.
• Establish theory promotion governance.

High

• Inventory existing project files and extract candidate principles and evidence.
• Build the discipline registry.
• Build the source-lineage model.
• Define confidence profiles by evidence type.
• Define contradiction and boundary-condition records.

Medium

• Prototype generated website views.
• Explore graph storage options while retaining Markdown as canonical.
• Develop genome node measurement standards.
• Create linting and validation rules for metadata and IDs.

Deferred

• Automated principle scoring
• Automated aesthetic quality prediction
• Full genome visualization
• Prescriptive design generation from principle records

21. Suggested Specialized Research Agents

• Evidence Lineage Agent: traces intellectual ancestry and duplicated citations.
• Discipline Research Agent: conducts deep native-domain research.
• Translation Agent: proposes and challenges cross-domain mappings.
• Falsification Agent: searches for counterexamples and alternative explanations.
• Registry Curator: normalizes terminology, IDs, and relationships.
• Experimental Design Agent: converts candidate principles into measurable tests.
• Accessibility and Population Agent: identifies excluded users and population boundaries.
• Historical Methods Agent: distinguishes durable practice from inherited convention.

No agent should both propose and approve a theory change without independent review.

22. Parallel Research Opportunities

The following can proceed independently:

• Discipline registry design
• Evidence schema design
• Translation schema design
• Pilot source collection
• Repository inventory
• Website information architecture
• Confidence-model research
• Governance model research

Their outputs should converge through a shared REP and stable identifier model.

23. Risks

False Universality

Broad terms may conceal different mechanisms.

Citation Echo

Many apparent confirmations may trace back to one source.

Prestige Bias

A respected discipline or expert tradition may be treated as stronger evidence than its methods justify.

Measurement Reductionism

Measurable variables may crowd out meaningful but difficult-to-measure phenomena.

Cultural Flattening

Claims derived from narrow populations may be mislabeled as universal.

Repository Entropy

Uncontrolled terminology and duplicate identifiers may make the knowledge base untrustworthy.

Premature Automation

Automated scoring or synthesis may amplify flaws before schemas and governance mature.

Genome Reification

Exploratory traits may be mistaken for objective natural categories.

24. Cross-Discipline Opportunities

• Use psychophysics to operationalize concepts inherited from visual arts.
• Use human factors failure research to challenge aesthetic heuristics.
• Use architecture and cartography to enrich digital wayfinding models.
• Use music and film to study composition over time rather than only in static layouts.
• Use linguistics and semiotics to distinguish perception from meaning.
• Use information theory cautiously to model density, redundancy, and signal differentiation.
• Use ecology and systems engineering to study resilience, interaction effects, and adaptation.
• Use anthropology to test whether proposed universals are culturally situated conventions.

25. Knowledge Relationships

EV-CS-001 ─supports─> DF-CS-ATLAS-001
EV-CS-002 ─supports─> DF-CS-ATLAS-001
DF-CS-ATLAS-001 ─creates─> HY-PRIN-001
DF-CS-ATLAS-001 ─creates─> HY-PRIN-002
DF-CS-ATLAS-001 ─creates─> HY-PRIN-003
DF-CS-ATLAS-001 ─creates─> HY-PRIN-004
HY-TRAN-* ─supports/challenges─> HY-PRIN-*
CN-BOUND-* ─constrains─> HY-PRIN-* or TH-*
EX-* ─tests─> HY-* or TH-*
RP-* ─proposes updates to─> registries

26. Repository Updates

Recommended additions:

• Add this document to /governance or /frameworks.
• Create empty canonical registries for disciplines, principles, evidence, translations, boundaries, and experiments.
• Add schema examples and validation rules.
• Revise the Composition Science Markdown Template to use REP-compatible identifiers.
• Add a project index linking current research streams to the new record types.

27. Website Updates

Generate an Atlas landing page that answers:

• What is Composition Science?
• Which disciplines are represented?
• Which principles are currently proposed?
• What evidence supports each principle?
• Where do disciplines agree or disagree?
• What remains uncertain?
• What research should happen next?

The site should visibly distinguish:

• Observation
• Evidence
• Hypothesis
• Theory
• Translation
• Application

28. AI Consumption Notes

Agents consuming this framework must:

1. Treat all initial principle families as hypotheses.
2. Inspect source lineage before claiming independent convergence.
3. Preserve native disciplinary meaning before translating terms.
4. State what is lost or transformed in every translation.
5. Keep confidence multidimensional.
6. Never infer universality from frequency of citation alone.
7. Prefer causal and predictive claims over purely metaphorical resemblance.
8. Record counterevidence and boundary conditions in the same artifact as supporting evidence.
9. Propose registry changes explicitly rather than modifying theory silently.
10. Produce a REP when a research cycle reaches a meaningful stopping point.

29. Handoff Instructions

The next agent should:

1. Read this framework, the Composition Science template, and REP v2.
2. Create initial registry files using the schemas defined here.
3. Select the differentiation/search pilot.
4. Conduct research in vision science, information visualization or typography, and high-stakes human factors.
5. Trace the lineage of overlapping principles.
6. Create evidence, finding, translation, boundary, and hypothesis records.
7. Attempt to falsify the shared principle.
8. Produce the first pilot REP.
9. Recommend changes to this framework based on actual execution friction.

30. Research Journal

JR-2026-07-21-001

Objective

Convert the proposed Atlas model into an executable cross-disciplinary research framework.

Work Completed

• Reviewed the Composition Science Markdown Template v1.
• Reviewed the Research Execution Package Specification v2.
• Reconciled the two artifact models.
• Defined discipline intake, evidence evaluation, translation, contradiction, principle lifecycle, genome safeguards, workflow, quality gates, and research priorities.

Major Decision

Cross-disciplinary convergence will not count as independent evidence until source lineage and shared ancestry have been examined.

Remaining Uncertainty

The framework has not yet been tested against a complete research pilot.

Next Decision Point

After the first pilot REP, revise the framework based on observed failures, unnecessary complexity, and missing record types.

31. Appendix A — Minimum Discipline Report Skeleton

---
identifier: RP-[AREA]-[NUMBER]
title: [Discipline] Composition Research Package
research_area: Composition Science
discipline: [Discipline]
author_agent: [Agent]
version: 1.0
confidence: [Rating]
completion: [Percent]
priority: [Priority]
status: draft
---

# Executive Summary
# Original Objective
# Scope
# Discipline Definition
# Native Constraints
# Native Variables and Measures
# Foundational Principles
# Mechanisms
# Evidence Registry
# Domain Findings
# Historical Lineage
# Counterevidence and Failures
# Boundary Conditions
# Cross-Discipline Translations
# Candidate Principle Impacts
# Hypothesis Registry
# Failed Assumptions
# Open Questions
# Recommended Next Research
# Repository Updates
# AI Consumption Notes
# Handoff Instructions
# Research Journal
# Completion Checklist

32. Completion Checklist

• Purpose and scope defined
• REP metadata included
• Research state snapshot included
• Canonical record types defined
• Discipline protocol defined
• Evidence evaluation defined
• Evidence independence addressed
• Translation protocol defined
• Contradiction protocol defined
• Principle lifecycle defined
• Genome safeguards defined
• Workflow defined
• Quality gates defined
• Initial hypotheses recorded
• Open questions recorded
• Research backlog prioritized
• Risks documented
• Repository and website impacts documented
• Handoff instructions included
• Revision history included

Revision History

Agent Instructions

When creating or modifying this document:

1. Separate observation from interpretation.
2. Never strengthen a conclusion beyond the available evidence.
3. Preserve contradictory findings.
4. Prefer measurable variables over subjective descriptions.
5. Reference hypotheses, evidence, theory records, translations, and genome nodes whenever possible.
6. Use REP-compatible stable IDs.
7. Record assumptions and confidence explicitly.
8. Inspect evidence ancestry before claiming independent support.
9. Keep the YAML header valid.
10. Do not delete revision history; append to it.
11. Treat metaphors as hypothesis generators, not validation.
12. Revise this framework after real research pilots expose weaknesses.