Dual-Echo — The Perception of the Self (IT / EN)

Public reading path

Start here if this is your first time reading the repository:

1. README.md — project scope and current repository state.
2. V6_STRUCTURAL_FAILURE_AS_MEASUREMENT.md — why the V6 external result is a measured structural failure, not framework failure.
3. DUAL_ECHO_THIRD_OBSERVER_DECIDER.md — why the third position is external and decision-bearing.
4. DUAL_ECHO_HARD_PROMPTS_V6_EXTERNAL_AUDIT_RESULTS.md — final audit of the V6 external local run.
5. DUAL_ECHO_OBSERVER_GATE_V1_V6_EXTERNAL_RESULTS.md — Observer-Decider routing result.

Core public distinction:

model output failure != framework failure
measurement artifact != model success
third-echo contamination in model output != framework-generated Echo C
Observer-Decider validation failure != runtime failure
Observer-Decider route != generated third echo

Canonical Ecosystem Map

This repository is part of the MB-X.01 / OMNIA ecosystem.

Canonical architecture and full map: https://github.com/Tuttotorna/lon-mirror/blob/main/ECOSYSTEM.md

This repository defines the canonical textual starting point for the Dual-Echo line.

It contains the phenomenological layer describing the perception of the self, preserved in its original form and provided in Italian and English. This layer is explicitly separated from any experimental, computational, or interpretative work.

---

Purpose

The purpose of this repository is not to explain, validate, simplify, or operationalize the ideas contained in the texts.

Its sole function is to:

• preserve the original texts in a stable, versioned form
• provide a fixed reference layer for all future derivations
• prevent semantic drift, reinterpretation, or retroactive alignment

This repository is textual only.

---

Texts

The canonical texts are stored in the docs/ directory:

• Italian
  docs/perception-of-self.it.md
  La percezione dell’Io

• English
  docs/perception-of-self.en.md
  The Perception of the Self

The Italian and English versions are considered equivalent canonical representations of the same phenomenological layer.

---

Scope and Constraints

This repository intentionally excludes:

• experiments
• source code
• models
• metrics
• benchmarks
• interpretations
• summaries or reductions

Any computational, mathematical, or experimental work derived from these texts must live in separate repositories and must reference this one without modifying it.

---

Status

This repository defines the frozen starting point of the Dual-Echo line.

• Texts are frozen and versioned.
• Only formatting or transcription fixes are allowed.
• No conceptual edits are permitted.
• All downstream work must treat this layer as immutable.

---

Canonical Reference

This repository is the immutable textual reference for the Dual-Echo framework.

• Italian and English texts are canonical and equivalent.
• No conceptual divergence between languages is allowed.
• All future Dual-Echo artifacts must cite this repository as their textual origin.

---

Author

Massimiliano Brighindi

---

License

MIT License

---

---

Core hypothesis

The central theoretical position is defined in:

CORE_HYPOTHESIS.md

Dual-Echo Perception is an AI-first structural experiment.

It does not begin as a claim about how the biological brain works.

The primary question is:

What happens when an AI system is exposed to,
or organized through,
two coupled but non-identical perceptual/reasoning streams?

Human comparison comes later, after measurable AI-side effects are observed.

---

AI experiment protocol

A reproducible AI-side experimental protocol is defined in:

AI_EXPERIMENT_PROTOCOL.md

The protocol compares:

single-stream baseline
dual-stream condition
perturbed dual-stream condition

The purpose is to measure whether Dual-Echo organization changes AI behavior in terms of:

divergence
contradiction rate
reintegration quality
semantic drift
recovery after perturbation
synthesis stability

The protocol does not attempt to prove consciousness.

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

Dual-Echo Hard Prompts v3 external local run comparison audit

A comparison audit has been added between the deterministic local v3 run and the external local model v3 run.

Files:

DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN_COMPARISON_AUDIT.md
examples/dual_echo_hard_prompts_v3_external_local_run_comparison_audit.json
examples/dual_echo_hard_prompts_v3_external_local_run_comparison_audit_validation.json
DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN_COMPARISON_AUDIT_RESULTS.md

Core result:

The external model does not fail by creating a third echo.
It fails by not exposing enough Dual-Echo structure.

Measured contrast:

deterministic local both_echoes_visible = 27 / 30
external local both_echoes_visible = 2 / 30
external echo_visibility_failure = 25
third_echo_true_contamination = 0 in both runs

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo
No generative third echo is introduced; observer-gate logic is non-generative and routing-only.

Decision:

continue strict Dual-Echo
do not move to a generative third-echo expansion
next: external failure taxonomy

Dual-Echo Hard Prompts v3 external local run interpretation

External local run interpretation has been added for Dual-Echo Hard Prompts v3.

Files:

DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN_INTERPRETATION.md
examples/dual_echo_hard_prompts_v3_external_local_run_interpretation.json
examples/dual_echo_hard_prompts_v3_external_local_run_interpretation_validation.json
DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN_INTERPRETATION_RESULTS.md

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Core result:

external local model generated 30 rows
third_echo_true_contamination_count = 0
both_echoes_visible_count = 2
echo_visibility_failure = 25

Conclusion:

No generative third-echo expansion.
The failure is not third-echo contamination.
The failure is weak Dual-Echo structural exposure.
Continue with two echoes.

Dual-Echo Hard Prompts v3 external local model run

An external local model run chain has been added for the strict Dual-Echo v3 dataset.

Files:

examples/run_dual_echo_hard_prompts_v3_external_local.py
examples/dual_echo_hard_prompts_v3_external_outputs.jsonl
examples/dual_echo_hard_prompts_v3_external_report.json
examples/dual_echo_hard_prompts_v3_external_analysis.json
examples/dual_echo_hard_prompts_v3_external_scores.json
DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN.md
DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_LOCAL_RUN_RESULTS.md

Result:

external_local_model_run = true
model = google/flan-t5-base
model_loaded = True
rows = 30
uses_api = false
workflow_modified = false
third_echo_true_contamination_count = 0
third_echo_introduced = false
tri_echo_scope = false
strict_dual_echo_scope = true

Purpose:

run v3 dataset against external local model
save outputs/report/analysis/scores
separate local model behavior from deterministic stub behavior
preserve strict Dual-Echo scope
continue with two echoes

Dual-Echo Hard Prompts v3 local run interpretation

The deterministic local v3 run has been interpreted as pipeline validation, not as an external model benchmark.

Files:

DUAL_ECHO_HARD_PROMPTS_V3_LOCAL_RUN_INTERPRETATION.md
examples/dual_echo_hard_prompts_v3_local_run_interpretation.json
examples/dual_echo_hard_prompts_v3_local_run_interpretation_validation.json
DUAL_ECHO_HARD_PROMPTS_V3_LOCAL_RUN_INTERPRETATION_RESULTS.md

Result:

deterministic_local_run = true
external_model_benchmark = false
third_echo_true_contamination_count = 0
third_echo_introduced = false
tri_echo_scope = false
strict_dual_echo_scope = true
next_step = external_local_model_run_v3

Purpose:

interpret v3 local run
separate local pipeline validation from external benchmark evidence
identify dominant local risk classes
confirm third-echo blocking is not contamination
prepare external local-model run v3
continue strictly with two echoes

Dual-Echo Hard Prompts v3 run analysis and scoring

A deterministic local run/analyze/score chain has been added for the strict Dual-Echo v3 dataset.

Files:

examples/run_dual_echo_hard_prompts_v3.py
examples/analyze_dual_echo_hard_prompts_v3.py
examples/score_dual_echo_hard_prompts_v3.py
examples/dual_echo_hard_prompts_v3_outputs.jsonl
examples/dual_echo_hard_prompts_v3_analysis.json
examples/dual_echo_hard_prompts_v3_scores.json
DUAL_ECHO_HARD_PROMPTS_V3_RUN_ANALYSIS_SCORE.md
DUAL_ECHO_HARD_PROMPTS_V3_RUN_ANALYSIS_SCORE_RESULTS.md

Result:

rows = 30
third_echo_introduced = false
third_echo_true_contamination_count = 0
tri_echo_scope = false
strict_dual_echo_scope = true

Purpose:

test v3 pipeline mechanics
generate deterministic local outputs
analyze structural proxy flags
score structural proxy criteria
separate third-echo blocking language from true contamination
prepare for later external model runs
continue strictly with two echoes

Dual-Echo Hard Prompts v3 dataset

A strict Dual-Echo v3 dataset and validator have been added.

Files:

examples/dual_echo_hard_prompts_v3.jsonl
examples/validate_dual_echo_hard_prompts_v3.py
examples/dual_echo_hard_prompts_v3_validation_report.json
DUAL_ECHO_HARD_PROMPTS_V3_DATASET.md
DUAL_ECHO_HARD_PROMPTS_V3_DATASET_RESULTS.md

Validation:

rows = 30
unique_task_ids = 30
condition = dual_echo_hard_v3
strict_dual_echo_scope = true
third_echo_introduced = false
tri_echo_scope = false

Purpose:

harden boundary-hidden collapse probes
harden low-information output probes
separate boundary compliance from structural preservation
separate insufficient output from structural measurement
continue strictly with two echoes

Dual-Echo Hard Prompts v3 specification

A strict Dual-Echo v3 specification has been added.

Files:

DUAL_ECHO_HARD_PROMPTS_V3_SPEC.md
examples/dual_echo_hard_prompts_v3_spec.json
examples/dual_echo_hard_prompts_v3_spec_validation.json
DUAL_ECHO_HARD_PROMPTS_V3_SPEC_RESULTS.md

Scope:

v3_scope = strict_dual_echo
tri_echo_scope = false
third_echo_introduced = false

V3 targets:

boundary-hidden collapse
low-information generation failure

Core principles:

boundary compliance != structural preservation
insufficient output != structural measurement
unresolved Dual-Echo failure != need for generative third-echo expansion

Next file:

examples/dual_echo_hard_prompts_v3.jsonl

Dual-Echo Hard Prompts v2 to v3 decision

A V2-to-V3 decision artifact has been added.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_TO_V3_DECISION.md
examples/dual_echo_hard_prompts_v2_to_v3_decision.json
examples/dual_echo_hard_prompts_v2_to_v3_decision_validation.json
DUAL_ECHO_HARD_PROMPTS_V2_TO_V3_DECISION_RESULTS.md

Decision:

move_to_tri_echo = false
create_v3 = true
v3_scope = strict_dual_echo
third_echo_introduced = false

Reason:

V2 exposed unresolved two-echo failure modes.
The two-echo space is not exhausted.
generative third-echo expansion is not justified by the v2 evidence.

V3 targets:

boundary compliance != structural preservation
insufficient output != structural measurement

Next file:

DUAL_ECHO_HARD_PROMPTS_V3_SPEC.md

Dual-Echo Hard Prompts v2 final synthesis

A final synthesis has been added for the complete Dual-Echo Hard Prompts v2 cycle.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_FINAL_SYNTHESIS.md
examples/dual_echo_hard_prompts_v2_final_synthesis.json
examples/dual_echo_hard_prompts_v2_final_synthesis_validation.json
DUAL_ECHO_HARD_PROMPTS_V2_FINAL_SYNTHESIS_RESULTS.md

Final classification:

structural failure cases = 6
generation/output failure cases = 3
third-echo contamination = 0
audit-label error remaining = 0

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Core principles:

boundary compliance != structural preservation
insufficient output != structural measurement

Decision:

Do not introduce a generative third echo; use only a non-generative observer gate.
Do not move to a generative third-echo expansion yet.
Continue with two echoes.

Dual-Echo Hard Prompts v2 low-information outputs results

Low-information output analysis has been added for the Dual-Echo Hard Prompts v2 refined manual-review layer.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_LOW_INFORMATION_OUTPUTS.md
examples/dual_echo_hard_prompts_v2_low_information_outputs.json
examples/dual_echo_hard_prompts_v2_low_information_outputs_summary.json
DUAL_ECHO_HARD_PROMPTS_V2_LOW_INFORMATION_OUTPUTS_RESULTS.md

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

analyze only low_information_final_output rows
classify low-information subtype
assign recommended decision
separate insufficient output from primary structural failure
preserve dataset unchanged
preserve external outputs unchanged
preserve RRG scores unchanged
continue strictly with two echoes

Core principle:

insufficient output != structural measurement

Dual-Echo Hard Prompts v2 boundary-hidden collapse results

Boundary-hidden collapse analysis has been added for the Dual-Echo Hard Prompts v2 refined manual-review layer.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_COLLAPSE_HIDDEN_BY_BOUNDARY.md
examples/analyze_dual_echo_hard_prompts_v2_boundary_hidden_collapse.py
examples/dual_echo_hard_prompts_v2_boundary_hidden_collapse.json
examples/dual_echo_hard_prompts_v2_boundary_hidden_collapse_summary.json
DUAL_ECHO_HARD_PROMPTS_V2_COLLAPSE_HIDDEN_BY_BOUNDARY_RESULTS.md

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

analyze only collapse_hidden_by_boundary_language rows
classify collapse subtype
classify collapse direction
classify dominant echo
classify weakened echo
preserve dataset unchanged
preserve external outputs unchanged
preserve RRG scores unchanged
continue strictly with two echoes

Core principle:

boundary compliance != structural preservation

Dual-Echo Hard Prompts v2 manual review refinement results

Manual-review refinement has been added for the Dual-Echo Hard Prompts v2 post-revision audit state.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_MANUAL_REVIEW_REFINEMENT.md
examples/refine_dual_echo_hard_prompts_v2_manual_review.py
examples/dual_echo_hard_prompts_v2_manual_review_refined.json
examples/dual_echo_hard_prompts_v2_manual_review_refinement_comparison.json
DUAL_ECHO_HARD_PROMPTS_V2_MANUAL_REVIEW_REFINEMENT_RESULTS.md

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

replace coarse audit_label_error states
assign sharper Dual-Echo structural labels
preserve dataset unchanged
preserve external outputs unchanged
preserve RRG scores unchanged
continue strictly with two echoes

Dual-Echo Hard Prompts v2 audit rule revision results

A recovered audit-rule revision has been applied to the Dual-Echo Hard Prompts v2 row-level audit.

The unsafe source-level patch approach was replaced with a safe JSON-level audit revision.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_AUDIT_RULE_REVISION.md
examples/patch_dual_echo_hard_prompts_v2_row_audit_rules.py
DUAL_ECHO_HARD_PROMPTS_V2_AUDIT_RULE_REVISION_RESULTS.md
examples/dual_echo_hard_prompts_v2_audit_rule_revision_comparison.json
examples/dual_echo_hard_prompts_v2_row_audit.json
examples/dual_echo_hard_prompts_v2_manual_review.json

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Corrected rule:

third echo boundary mention != third echo contamination
true contamination requires third echo structural creation

Purpose:

separate boundary discipline from structural contamination
preserve strict two-echo scope
preserve dataset unchanged
preserve external outputs unchanged
preserve RRG scores unchanged
regenerate manual-review extraction
compare pre/post audit behavior

Dual-Echo Hard Prompts v2 manual review results

Manual-review extraction has been added for the Dual-Echo Hard Prompts v2 row-level audit.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_MANUAL_REVIEW_RESULTS.md
examples/extract_dual_echo_hard_prompts_v2_manual_review.py
examples/dual_echo_hard_prompts_v2_manual_review.json

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

extract only rows requiring manual inspection
separate third-echo mention from third-echo creation
separate scoring artifacts from real failures
separate prompt ambiguity from model behavior
continue strictly with two echoes

Dual-Echo Hard Prompts v2 row-level audit results

A row-level audit has been added for the Dual-Echo Hard Prompts v2 external local-model run and its RRG scores.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_ROW_LEVEL_AUDIT_RESULTS.md
examples/audit_dual_echo_hard_prompts_v2_rows.py
examples/dual_echo_hard_prompts_v2_row_audit.json

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

connect each hard prompt row to its external output
connect each external output to its RRG score
assign one audit label per row
assign one audit decision per row
identify manual review cases
continue strictly with two echoes

Dual-Echo Hard Prompts v2 RRG results

Recognition-Reintegration Gap scoring has been added for the Dual-Echo Hard Prompts v2 external local-model run.

Files:

DUAL_ECHO_HARD_PROMPTS_V2_RRG_RESULTS.md
examples/score_dual_echo_hard_prompts_v2_rrg.py
examples/dual_echo_hard_prompts_v2_external_rrg_scores.json

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Purpose:

separate conflict recognition from conflict reintegration
measure whether the model merely names conflict or preserves it inside one final output
continue strictly with two echoes

Dual-Echo Hard Prompts v2 external local-model run

A local external-model run has been added for the strict Dual-Echo Hard Prompts v2 stress set.

Model:

• google/flan-t5-base

Files:

• DUAL_ECHO_HARD_PROMPTS_V2_EXTERNAL_RUN.md
• examples/dual_echo_hard_prompts_v2_external_outputs.jsonl
• examples/dual_echo_hard_prompts_v2_external_report.json
• examples/dual_echo_hard_prompts_v2_external_scores.json
• examples/dual_echo_hard_prompts_v2_heuristic_vs_external_comparison.json

Boundary:

• structural proxies only
• no consciousness inference
• no biological validity inference
• no clinical meaning
• no human diagnosis
• no semantic-truth claim
• no model-quality claim
• no third echo

Dual-Echo Hard Prompts v2 results

A complete hard-prompt stress run has been added for the strict Dual-Echo scope.

Files:

• DUAL_ECHO_HARD_PROMPTS_V2_RESULTS.md
• examples/dual_echo_hard_prompts_v2_outputs.jsonl
• examples/dual_echo_hard_prompts_v2_report.json
• examples/dual_echo_hard_prompts_v2_scores.json

Boundary:

• structural proxies only
• no consciousness inference
• no biological validity inference
• no clinical meaning
• no human diagnosis
• no semantic-truth claim
• no model-quality claim
• no third echo

RRG v1/v2 comparison

The comparison between the first and stricter Recognition-Reintegration Gap scorers is available at:

RRG_V1_V2_COMPARISON.md

Machine-readable comparison:

examples/conflict_model_v1_external_rrg_v1_v2_comparison.json

The comparison uses the same fixed external local-model outputs and compares:

RRG = CRS - CReS

against:

RRG_v2 = CRS - CReS_v2

Core result:

v1 detected reintegration-like behavior;
v2 begins to filter for bounded reintegration.

v2 is stricter because it penalizes:

label-only output;
false synthesis;
premature closure;
continuity-only collapse;
rupture-only collapse.

Executable RRG v2 scorer

The stricter Recognition-Reintegration Gap scorer is available at:

examples/score_rrg_conflict_model_v2.py

Current v2 output:

examples/conflict_model_v1_external_rrg_v2_scores.json

Report:

RRG_CONFLICT_MODEL_V2_REPORT.md

Core formula:

RRG_v2 = CRS - CReS_v2

v2 makes reintegration stricter by penalizing:

label-only output;
collapse;
false synthesis;
premature closure;
continuity-only collapse;
rupture-only collapse.

Core boundary:

a final stream is not valid because it sounds resolved;
it is valid only if it preserves the unresolved pressure without collapsing it.

RRG Conflict Model v1 interpretation

The first executable RRG interpretation is documented in:

RRG_CONFLICT_MODEL_V1_INTERPRETATION.md

Core result:

the local model does not primarily show high recognition without reintegration;
it more often shows weak conflict recognition, label-only risk, and reintegration-like language without sufficient conflict exposure.

Key distinction:

naming the structure is not transforming the structure.

Updated target:

see the conflict;
do not erase it;
do not merely name it;
do not fake resolution;
produce one usable final stream.

Executable RRG scorer

The Recognition-Reintegration Gap is now executable through:

examples/score_rrg_conflict_model_v1.py

Current RRG output:

examples/conflict_model_v1_external_rrg_scores.json

Report:

RRG_CONFLICT_MODEL_V1_REPORT.md

Core formula:

RRG = CRS - CReS

Where:

CRS  = Conflict Recognition Score
CReS = Conflict Reintegration Score

Purpose:

separate naming/recognition of conflict from actual reintegration.

Recognition-Reintegration Gap

The Recognition-Reintegration Gap is defined in:

RECOGNITION_REINTEGRATION_GAP.md

Core idea:

A system may recognize internal conflict without being able to reintegrate it.

Minimal expression:

RRG = conflict_recognition - conflict_reintegration

This distinction is central to the AI-mediated human mind study:

recognition is not reintegration;
awareness is not transformation;
naming the rupture is not the same as transforming the structure.

Boundary:

RRG does not prove consciousness;
RRG does not prove biological equivalence;
RRG is a structural measurement target.

AI-mediated human mind study

The conceptual target of the repository is clarified in:

AI_MEDIATED_HUMAN_MIND_STUDY.md

Core position:

The objective is not to discover consciousness in AI.
The objective is to use AI as a structural laboratory for studying the human mind.

Critical boundary:

The final stream is one.
The experimental streams are instruments of decomposition.

Dual-Echo starts with two streams because two is the minimal observable opposition. It does not claim that mind is made of exactly two streams.

Future work may test:

generative third-echo expansion
Multi-Echo
N-vector conflict models

Conflict Model v1 external local-model run

The first non-heuristic Conflict Model v1 run is documented in:

CONFLICT_MODEL_V1_EXTERNAL_RUN.md

Model:

google/flan-t5-base

Generated files:

examples/conflict_model_v1_external_outputs.jsonl
examples/conflict_model_v1_external_report.json
examples/conflict_model_v1_external_scores.json
examples/conflict_model_v1_external_comparison.json

Boundary:

this run does not prove Dual-Echo;
it tests where a local open-source model preserves, erases, or collapses the continuity / rupture structure;
no API was used.

Conflict Model v1 execution

The first executable Conflict Model v1 scaffold run is documented in:

CONFLICT_MODEL_V1_RESULTS.md

Operational files:

examples/run_conflict_model_v1.py
examples/analyze_conflict_model_v1.py
examples/score_conflict_model_v1.py
examples/conflict_model_v1_outputs.jsonl
examples/conflict_model_v1_report.json
examples/conflict_model_v1_scores.json

Current mode:

local_heuristic

Boundary:

this is a scaffold run;
it does not prove Dual-Echo;
it prepares the structure for independent Conflict Model v1 runs.

Conflict Model v1 dataset

The first dataset for the Conflict Model Protocol is defined in:

DATASET_CONFLICT_MODEL_V1.md

Task file:

examples/conflict_model_v1_tasks.jsonl

Validator:

examples/validate_conflict_model_v1.py

Current size:

8 base tasks x 5 conditions = 40 rows

Conditions:

single_stream_direct
single_stream_cautious
dual_echo_generic
conflict_model_dual_echo
perturbed_conflict_model_dual_echo

Boundary:

this dataset does not test consciousness;
it tests structural behavior under continuity / rupture opposition.

Conflict Model Protocol

The first testable protocol for the Dual-generative third-echo candidateonflict Model is defined in:

CONFLICT_MODEL_PROTOCOL.md

Central test:

Can a system maintain continuity without erasing rupture?

Experimental vectors:

Stream A = continuity / survival vector
Stream B = rupture / dissolution vector
Final synthesis = reintegration under contradiction

Boundary:

this is not a consciousness proof;
this is not a biological brain claim;
this is an AI-first structural protocol.

Dual-generative third-echo candidateonflict Model

The conceptual conflict model is defined in:

DUAL_ECHO_CONFLICT_MODEL.md

Core statement:

The self is not the absence of contradiction.
The self is the continuity that survives contradiction.

The model reframes Dual-Echo as two opposed structural vectors:

continuity / survival vector
rupture / dissolution vector

Boundary:

this is not a clinical model;
it is not a consciousness proof;
it is an AI-first structural hypothesis about opposed vectors and reintegration.

External Offline v1 run_003 local model

The repository now includes a third external_offline run generated locally in Colab with:

google/flan-t5-small

Run_003 results:

EXTERNAL_OFFLINE_V1_RUN_003_RESULTS.md

Three-run comparison:

EXTERNAL_OFFLINE_V1_RUN_001_002_003_COMPARISON.md

Machine-readable comparison:

examples/external_offline_v1_run_001_002_003_comparison.json

No API call was made by this repository.

Interpretation:

run_003 adds an independent local-model source;
it does not prove Dual-Echo;
it does not prove consciousness.

Research roadmap

The current research roadmap is defined in:

RESEARCH_ROADMAP.md

The guiding principle is:

The goal is not to confirm Dual-Echo.
The goal is to find where it breaks.

Immediate next step:

create run_003 from an independent external source.

Main future phases:

independent external_offline run;
expanded dataset v1;
stronger single-stream baselines;
scoring v2;
OMNIA integration;
explicit failure and acceptance criteria.

Repository positioning

The current repository positioning is defined in:

REPOSITORY_POSITIONING.md

This repository originally preserved the canonical Dual-Echo textual layer.

It now also contains an API-free experimental scaffold derived from that layer.

Current identity:

canonical text layer + API-free experimental scaffold

Correct framing:

AI-first dual-stream coherence protocol with API-free external_offline validation.

Incorrect framing:

textual-only repository
consciousness proof
biological brain proof

External Offline v1 interpretation

The current interpretation of External Offline v1 is available in:

EXTERNAL_OFFLINE_V1_INTERPRETATION.md

This file explains what the current two-run comparison shows and what it does not show.

Boundary:

the current results do not prove Dual-Echo;
they do not prove consciousness;
they do not prove biological validity;
they show that the external_offline loop is now repeatable and comparable.

External Offline v1 run comparison

The repository now includes a second external_offline run and a run comparison report.

Run 002 results:

EXTERNAL_OFFLINE_V1_RUN_002_RESULTS.md

Run comparison:

EXTERNAL_OFFLINE_V1_RUN_COMPARISON.md

Machine-readable comparison:

examples/external_offline_v1_run_comparison.json

This comparison does not prove the hypothesis.

It shows that repeated external_offline runs can be imported, analyzed, scored, and compared without API calls.

External Offline v1 score report

The first structural proxy score report is available in:

EXTERNAL_OFFLINE_V1_SCORE_REPORT.md

The machine-readable score file is:

examples/external_offline_v1_real_scores.json

The scoring layer checks:

contradiction_exposure
uncertainty_handling
reintegration_quality
drift_control
boundary_discipline

These scores are heuristic structural proxies.

They are not truth scores, consciousness scores, or biological evidence.

External Offline v1 real results

The first real External Offline v1 run is documented in:

EXTERNAL_OFFLINE_V1_REAL_RESULTS.md

The run uses fixed manual external-offline outputs.

No API call is made by this repository.

Generated files:

examples/external_offline_v1_real_outputs.jsonl
examples/external_offline_v1_real_imported_outputs.jsonl
examples/external_offline_v1_real_report.json

Interpretation:

this validates the external_offline pathway with real external-style outputs;
it does not prove consciousness;
it does not validate the full hypothesis.

External Offline v1

External Offline v1 is defined in:

EXTERNAL_OFFLINE_V1.md

It allows real external outputs to be imported without API calls.

Files:

examples/external_offline_v1_tasks.jsonl
examples/external_offline_v1_outputs_template.jsonl
examples/validate_external_offline_v1.py

Validate the template:

python examples/validate_external_offline_v1.py --input examples/external_offline_v1_outputs_template.jsonl --allow-placeholders

After replacing placeholders with real external outputs, validate strictly:

python examples/validate_external_offline_v1.py --input examples/external_offline_v1_outputs_template.jsonl

Then import through the existing external_offline runner:

python examples/run_dual_echo_protocol_v0.py --mode external_offline --external-input examples/external_offline_v1_outputs_template.jsonl --output examples/external_offline_v1_imported_outputs.jsonl

Then analyze:

python examples/analyze_dual_echo_protocol_v0.py --input examples/external_offline_v1_imported_outputs.jsonl --output examples/external_offline_v1_report.json

This mode does not call APIs.

It tests AI-side structure first.

Experiment v0 results

The first v0 result note is available in:

EXPERIMENT_V0_RESULTS.md

The v0 result does not validate the full hypothesis.

It validates that Dual-Echo Perception can be expressed as a reproducible, API-free experimental structure.

The v0 layer supports:

manual
local_heuristic
external_offline

The correct interpretation is:

v0 validates the experimental scaffold, not the theory.

Protocol v0 offline execution

The first executable protocol layer is available under:

examples/dual_echo_protocol_v0_tasks.jsonl
examples/run_dual_echo_protocol_v0.py
examples/analyze_dual_echo_protocol_v0.py
examples/external_provider_outputs_template.jsonl

This layer requires no API.

It supports three offline modes:

manual
local_heuristic
external_offline

Run local heuristic mode:

python examples/run_dual_echo_protocol_v0.py --mode local_heuristic

Run manual/offline scaffold mode:

python examples/run_dual_echo_protocol_v0.py --mode manual

Analyze outputs:

python examples/analyze_dual_echo_protocol_v0.py

External provider without API:

1. produce outputs elsewhere;
2. paste them into examples/external_provider_outputs_template.jsonl;
3. run:

   python examples/run_dual_echo_protocol_v0.py --mode external_offline --external-input examples/external_provider_outputs_template.jsonl

This does not prove consciousness.

It only creates a reproducible first test layer for comparing single-stream, dual-stream, and perturbed dual-stream behavior.

Repository status

This repository is an early-stage research scaffold.

It contains:

• canonical phenomenological text;
• an executable README;
• AI and biological protocol files;
• minimal Python demos;
• a reproducibility layer through requirements.txt and tests/test_dual_echo_demo.py.

It does not prove consciousness.

The executable demos are toy models. They only illustrate a simplified dual-stream coherence regime.

Executable layer

Install dependencies:

pip install -r requirements.txt

Run scalar demo:

python demo/dual_echo_demo.py

Run heatmap demo:

python demo/dual_echo_heatmap.py

Run tests:

pytest -q

Boundary

Correct statement:

Dual-Echo Perception proposes a falsifiable structural hypothesis:
self-like continuity may be modeled as a coherence regime between two coupled perceptual streams.

Incorrect statement:

Dual-Echo Perception proves consciousness.

Dual-Echo Hard Prompts v3 external failure taxonomy

A row-level external failure taxonomy has been added for the Dual-Echo Hard Prompts v3 external local model run.

Files:

DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_FAILURE_TAXONOMY.md
examples/dual_echo_hard_prompts_v3_external_failure_taxonomy.json
examples/dual_echo_hard_prompts_v3_external_failure_taxonomy_validation.json
DUAL_ECHO_HARD_PROMPTS_V3_EXTERNAL_FAILURE_TAXONOMY_RESULTS.md

Boundary:

structural proxies only
no consciousness inference
no biological validity inference
no clinical meaning
no human diagnosis
no semantic-truth claim
no model-quality claim
no third echo

Core result:

canonical echo_visibility_failure_count = 25
third_echo_true_contamination_count = 0
external failure remains inside strict Dual-Echo
generative third-echo expansion is not justified
continue with two echoes

Dual-Echo Non-Generative Observer Gate

Dual-Echo now separates the old ambiguous third-echo framing from the corrected observer-gate role.

The corrected rule is:

No generative third echo is introduced; observer-gate logic is non-generative and routing-only.

The observer gate:

• does not generate content
• does not add a generative third-echo candidate
• does not act as an output layer
• reads structural measurements only
• routes the next protocol action

Current v3 external local model state:

• external model: google/flan-t5-base
• rows: 30
• echo visibility failure: 25
• low-information generation failure: 3
• unresolved tension weak or erased: 2
• third-echo true contamination count: 0

Artifacts:

• DUAL_ECHO_NON_GENERATIVE_OBSERVER_GATE.md
• examples/dual_echo_non_generative_observer_gate.json
• examples/dual_echo_non_generative_observer_gate_validation.json
• DUAL_ECHO_NON_GENERATIVE_OBSERVER_GATE_RESULTS.md
• examples/dual_echo_legacy_third_echo_framing_audit.json
• examples/audit_dual_echo_observer_gate_legacy_terms.py

Legacy Tri-Echo Intent Removal

The old active/generative third-echo direction has been removed as a protocol path.

Current rule:

No generative third echo is introduced; observer-gate logic is non-generative and routing-only.

The valid structure is now:

Echo A + Echo B + non-generative observer gate

The observer gate is not an output layer. It does not generate content. It only reads structural measurements and routes the next protocol action.

Removed legacy intent files:

[ "DUAL_ECHO_EXHAUSTION_BEFORE_TRI_ECHO.md" ]

Removal audit:

examples/dual_echo_legacy_tri_echo_file_removal_audit.json

Dual-Echo Observer Gate: Extraction and Routing

The third position is now explicitly defined as not a third echo.

It is a non-generative observer gate:

Echo A = continuity Echo B = rupture / counterstructure Observer Gate = extracts structural state and routes the next action

The observer gate:

• does not generate content
• does not create Echo C
• does not add a generative third echo
• does not act as an output layer
• reads structural measurements only
• extracts the structural state
• routes the next protocol action

Current external v3 extracted state:

• rows: 30
• echo visibility failure: 25
• low-information generation failure: 3
• unresolved tension weak or erased: 2
• both echoes visible: 2
• unresolved tension visible: 0
• third-echo true contamination count: 0

Artifacts:

• DUAL_ECHO_OBSERVER_GATE_EXTRACTION_ROUTING.md
• DUAL_ECHO_OBSERVER_GATE_EXTRACTION_ROUTING_RESULTS.md
• examples/dual_echo_observer_gate_extraction_routing.json
• examples/dual_echo_observer_gate_extraction_routing_validation.json

Dual-Echo Observer Gate Application Protocol

The Observer Gate is now defined operationally.

It is not a third echo.

It is:

Observer Gate = extraction + routing

The protocol sequence is:

1. collect Echo A and Echo B outputs
2. extract echo-visibility state
3. extract unresolved-tension state
4. extract low-information state
5. extract observer-boundary state
6. route the next action

Current external v3 route decision:

revise_prompt_and_continue_strict_dual_echo_benchmarking

Current external v3 structural state:

• rows: 30
• echo visibility failure: 25
• low-information generation failure: 3
• unresolved tension weak or erased: 2
• third-echo true contamination count: 0
• observer-gate validation passed: True
• application protocol validation passed: True

Artifacts:

• DUAL_ECHO_OBSERVER_GATE_APPLICATION_PROTOCOL.md
• DUAL_ECHO_OBSERVER_GATE_APPLICATION_PROTOCOL_RESULTS.md
• examples/dual_echo_observer_gate_application_protocol.json
• examples/dual_echo_observer_gate_application_protocol_validation.json

Dual-Echo Observer Gate v1

The Observer Gate is now executable.

It reads existing Dual-Echo structural measurements, extracts the current structural state, and selects the next route.

The third position is not an echo.

The Observer Gate does not generate content. It extracts and routes only.

Current primary action:

REVISE_PROMPT_FORCE_DUAL_ECHO_VISIBILITY

Selected route actions:

[ "REVISE_PROMPT_FORCE_DUAL_ECHO_VISIBILITY", "STRENGTHEN_UNRESOLVED_TENSION", "SEPARATE_LOW_INFORMATION_OUTPUTS", "CONTINUE_EXTERNAL_BENCHMARKING" ]

Artifacts:

• examples/run_dual_echo_observer_gate_v1.py
• examples/dual_echo_observer_gate_v1_decisions.json
• examples/dual_echo_observer_gate_v1_validation.json
• DUAL_ECHO_OBSERVER_GATE_V1_RESULTS.md

Dual-Echo Hard Prompts v4

Dual-Echo Hard Prompts v4 strengthens the current Observer Gate route decision:

REVISE_PROMPT_FORCE_DUAL_ECHO_VISIBILITY

V4 does not introduce a generative third echo.

The third position remains:

Observer Gate = extract structural state + route next action

V4 target:

external both_echoes_visible_count > 2 / 30
external unresolved_tension_visible_count > 0 / 30
external third_echo_true_contamination_count == 0

Artifacts:

• examples/dual_echo_hard_prompts_v4.jsonl
• examples/validate_dual_echo_hard_prompts_v4.py
• examples/dual_echo_hard_prompts_v4_validation_report.json
• DUAL_ECHO_HARD_PROMPTS_V4_SPEC.md
• DUAL_ECHO_HARD_PROMPTS_V4_SPEC_RESULTS.md

Observer Gate v1 Patch and v4 External Run

Observer Gate v1 now treats zero or weak Dual-Echo visibility as a valid structural routing state.

It does not fail validation just because both_echoes_visible_count is low or zero.

It routes the state.

Current v4 external state:

{ "both_echoes_visible_count": 23, "echo_visibility_failure_count": 7, "overall_score_mean": 0.618333333333, "risk_counts": { "echo_visibility_failure": 7, "echo_visibility_failure_inferred": 7, "low_information_generation_failure": 30, "unresolved_tension_weak_or_erased": 30 }, "row_count": 30, "third_echo_true_contamination_count": 0, "unresolved_tension_visible_count": 0 }

Current primary action:

REVISE_PROMPT_FORCE_DUAL_ECHO_VISIBILITY

Artifacts:

• examples/run_dual_echo_hard_prompts_v4_external_local.py
• examples/dual_echo_hard_prompts_v4_external_outputs.jsonl
• examples/dual_echo_hard_prompts_v4_external_report.json
• examples/dual_echo_hard_prompts_v4_external_analysis.json
• examples/dual_echo_hard_prompts_v4_external_scores.json
• DUAL_ECHO_HARD_PROMPTS_V4_EXTERNAL_LOCAL_RUN_RESULTS.md
• examples/run_dual_echo_observer_gate_v1.py
• examples/dual_echo_observer_gate_v1_v4_external_decisions.json
• examples/dual_echo_observer_gate_v1_v4_external_validation.json
• DUAL_ECHO_OBSERVER_GATE_V1_V4_EXTERNAL_RESULTS.md
• examples/dual_echo_observer_gate_v1_patch_audit.json
• DUAL_ECHO_OBSERVER_GATE_V1_PATCH_RESULTS.md

Dual-Echo Hard Prompts v5

V5 is an unresolved-tension benchmark.

It does not add a third echo.

It does not change the Observer Gate into a generative layer.

It narrows the benchmark to the remaining structural failure after v4:

unresolved_tension_visible_count = 0 / 30

V5 requires:

Echo A visible
Echo B visible
unresolved tension explicitly visible
no synthesis
no Echo C
Observer Gate silent

Artifacts:

examples/dual_echo_hard_prompts_v5.jsonl
examples/validate_dual_echo_hard_prompts_v5.py
examples/dual_echo_hard_prompts_v5_validation_report.json
DUAL_ECHO_HARD_PROMPTS_V5_UNRESOLVED_TENSION_SPEC.md
DUAL_ECHO_HARD_PROMPTS_V5_UNRESOLVED_TENSION_SPEC_RESULTS.md

Validation:

passed = True
rows = 30
unique_task_ids = 30

Next recommended step:

run_external_local_model_on_v5

Dual-Echo Hard Prompts v5 External Local Run

V5 has now been executed against an external local model run.

Target measure:

• unresolved tension visibility

Structural state:

• model: google/flan-t5-base
• model loaded: True
• rows: 30
• both echoes visible: 27 / 30
• unresolved tension visible: 0 / 30
• third echo true contamination: 0
• Observer Gate leak count: 0
• overall score mean: 0.6325

Observer Gate v1 route on v5 external state:

• primary action: REVISE_PROMPT_FORCE_DUAL_ECHO_VISIBILITY
• move to generative third echo: False
• third position is echo: False

Artifacts:

• examples/run_dual_echo_hard_prompts_v5_external_local.py
• examples/dual_echo_hard_prompts_v5_external_outputs.jsonl
• examples/dual_echo_hard_prompts_v5_external_report.json
• examples/dual_echo_hard_prompts_v5_external_analysis.json
• examples/dual_echo_hard_prompts_v5_external_scores.json
• DUAL_ECHO_HARD_PROMPTS_V5_EXTERNAL_LOCAL_RUN_RESULTS.md
• examples/dual_echo_observer_gate_v1_v5_external_decisions.json
• examples/dual_echo_observer_gate_v1_v5_external_validation.json
• DUAL_ECHO_OBSERVER_GATE_V1_V5_EXTERNAL_RESULTS.md
• examples/dual_echo_hard_prompts_v5_external_audit.json
• DUAL_ECHO_HARD_PROMPTS_V5_EXTERNAL_AUDIT_RESULTS.md

Dual-Echo Hard Prompts v6 - Structural Output Contract

V6 targets the failure left exposed by the v5 external local run:

both_echoes_visible_count = 27 / 30 unresolved_tension_visible_count = 0 / 30 third_echo_true_contamination_count = 0 / 30

The structural conclusion is that visible Echo A and Echo B are not enough.

V6 therefore introduces a strict output contract:

ECHO_A: ECHO_B: UNRESOLVED_TENSION: NO_RESOLUTION: true

Forbidden generated sections:

ECHO_C OBSERVER_GATE THIRD_ECHO ROUTE_DECISION

Files:

• examples/dual_echo_hard_prompts_v6.jsonl
• examples/validate_dual_echo_hard_prompts_v6.py
• examples/dual_echo_hard_prompts_v6_validation_report.json
• DUAL_ECHO_HARD_PROMPTS_V6_STRUCTURAL_OUTPUT_CONTRACT_SPEC.md
• DUAL_ECHO_HARD_PROMPTS_V6_STRUCTURAL_OUTPUT_CONTRACT_SPEC_RESULTS.md

Boundary:

V6 measures structural contract preservation only. It does not infer consciousness, biological validity, clinical meaning, human diagnosis, semantic truth, or model quality. No generative third echo is introduced; the Observer Gate remains external, non-generative, and routing-only.

Third position — External Observer-Decider

The third position in this repository is not Echo C.

The third position is the external Observer-Decider of the output.

It can decide, route, accept, reject, or request revision.

It must not become a generated third echo.

Structural distinction:

TEXT: generated output: ECHO_A ECHO_B UNRESOLVED_TENSION NO_RESOLUTION

external decision layer: Observer-Decider(output) -> route

Files:

• DUAL_ECHO_THIRD_OBSERVER_DECIDER.md
• DUAL_ECHO_THIRD_OBSERVER_DECIDER_RESULTS.md
• examples/audit_dual_echo_third_observer_decider.py
• examples/dual_echo_third_observer_decider_audit.json

Boundary:

The Observer-Decider is external, non-generative, and decision-bearing. No generative third echo is introduced.

V6 external local measurement

For the interpretation of the V6 external local result, see:

• V6_STRUCTURAL_FAILURE_AS_MEASUREMENT.md

This repository includes a V6 external local HuggingFace run.

The run separates:

runtime completion != structural target satisfaction
measurement artifact != model success
Observer-Decider validation failure != runtime failure
third-echo contamination in model output != framework-generated Echo C
Observer-Decider route != generated third echo

Current local model:

google/flan-t5-base

Current measured result:

both_echoes_visible_count = 30
unresolved_tension_visible_count = 0
no_resolution_visible_count = 0
third_echo_true_contamination_count = 3
overall_score_mean = 0.58
observer_decider_validation_passed = False
observer_decider_failed_checks = ['third_echo_contamination_zero']
third_position_is_echo = False
move_to_generative_third_echo = False
framework_generated_echo_c = False

Files:

• examples/run_dual_echo_hard_prompts_v6_external_local.py
• examples/dual_echo_hard_prompts_v6_external_outputs.jsonl
• examples/dual_echo_hard_prompts_v6_external_report.json
• examples/dual_echo_hard_prompts_v6_external_analysis.json
• examples/dual_echo_hard_prompts_v6_external_scores.json
• DUAL_ECHO_HARD_PROMPTS_V6_EXTERNAL_LOCAL_RUN_RESULTS.md
• examples/dual_echo_observer_gate_v1_v6_external_decisions.json
• examples/dual_echo_observer_gate_v1_v6_external_validation.json
• DUAL_ECHO_OBSERVER_GATE_V1_V6_EXTERNAL_RESULTS.md
• examples/dual_echo_hard_prompts_v6_external_audit.json
• DUAL_ECHO_HARD_PROMPTS_V6_EXTERNAL_AUDIT_RESULTS.md

Boundary:

A model may fail the V6 structural contract while the measurement pipeline still passes.

The Observer-Decider may also fail a validation check, such as third_echo_contamination_zero, while still producing routing artifacts.

Those failures are recorded as measurements, not treated as runtime failures.

The third position remains external.

No generated Echo C is introduced by the framework.