Skip to content

Examples Results

Generated: 2026-05-16

examples/neuro/run_closed_loop_mock_bci.py

======================================================================
Closed-Loop BCI with MockSignalStream Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

--- Setting Up Signal Stream ---
[OK] MockSignalStream connected.
  Status: {'connected': True, 'stream_type': 'mock', 'channels': ['Fz', 'Cz', 'Pz', 'Oz'], 'sampling_rate': 250.0, 'window_index': 0, 'window_seconds': 1.0}

--- Setting Up BCI Adapter ---
[OK] BCIAdapter created and stream attached.
  Pipeline status: {'stream_attached': True, 'preprocessing': True, 'feature_extraction': True, 'neural_state_encoding': True}

--- Creating CIA System ---
[OK] CIA system created with BCI-compatible attributes.

--- Running 3 Closed-Loop Steps ---

──────────────────────────────────────────────────
Step 1: A red ball rolled behind a screen and reappeared.
──────────────────────────────────────────────────
  Neural State:
    Attention Proxy:   0.3170
    Workload Proxy:    0.6829
    Fatigue Proxy:     0.5237
    Prediction Error:  0.9819
    Confidence:        0.6000
  Control Signal:
    Attention Bias:          0.0000
    Workspace Priority Bias: 0.0000
    Uncertainty Adj:         0.0000
    Valuation Adj:           0.0596
    Prediction Sensitivity:  0.9398
  Conditioning Applied: True
  Stream Status:        {'connected': True, 'stream_type': 'mock', 'channels': ['Fz', 'Cz', 'Pz', 'Oz'], 'sampling_rate': 250.0, 'window_index': 0, 'window_seconds': 1.0}
  Pipeline Status:      {'stream_attached': True, 'preprocessing': True, 'feature_extraction': True, 'neural_state_encoding': True}

──────────────────────────────────────────────────
Step 2: The agent noticed a sudden change in the environment.
──────────────────────────────────────────────────
  Neural State:
    Attention Proxy:   0.3208
    Workload Proxy:    0.6792
    Fatigue Proxy:     0.4824
    Prediction Error:  0.9768
    Confidence:        0.6000
  Control Signal:
    Attention Bias:          0.0000
    Workspace Priority Bias: 0.0000
    Uncertainty Adj:         0.0000
    Valuation Adj:           0.0502
    Prediction Sensitivity:  0.9226
  Conditioning Applied: True
  Stream Status:        {'connected': True, 'stream_type': 'mock', 'channels': ['Fz', 'Cz', 'Pz', 'Oz'], 'sampling_rate': 250.0, 'window_index': 1, 'window_seconds': 1.0}
  Pipeline Status:      {'stream_attached': True, 'preprocessing': True, 'feature_extraction': True, 'neural_state_encoding': True}

──────────────────────────────────────────────────
Step 3: After observing, the agent predicted the ball would return.
──────────────────────────────────────────────────
  Neural State:
    Attention Proxy:   0.3048
    Workload Proxy:    0.6952
    Fatigue Proxy:     0.3845
    Prediction Error:  0.9215
    Confidence:        0.6000
  Control Signal:
    Attention Bias:          0.0000
    Workspace Priority Bias: 0.0000
    Uncertainty Adj:         0.0000
    Valuation Adj:           0.0355
    Prediction Sensitivity:  0.7385
  Conditioning Applied: True
  Stream Status:        {'connected': True, 'stream_type': 'mock', 'channels': ['Fz', 'Cz', 'Pz', 'Oz'], 'sampling_rate': 250.0, 'window_index': 2, 'window_seconds': 1.0}
  Pipeline Status:      {'stream_attached': True, 'preprocessing': True, 'feature_extraction': True, 'neural_state_encoding': True}

======================================================================
Closed-Loop Summary
======================================================================
  Step 1: attention=0.3170, fatigue=0.5237, confidence=0.6000, attn_bias=+0.0000, ws_bias=+0.0000, conditioned=Yes
  Step 2: attention=0.3208, fatigue=0.4824, confidence=0.6000, attn_bias=+0.0000, ws_bias=+0.0000, conditioned=Yes
  Step 3: attention=0.3048, fatigue=0.3845, confidence=0.6000, attn_bias=+0.0000, ws_bias=+0.0000, conditioned=Yes

[OK] MockSignalStream disconnected.

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

NOTE: The closed-loop pipeline reads synthetic EEG data, estimates
neural-state proxies, and conditions the CIA system. This does NOT
transfer consciousness or subjective experience into the AI system.

Exit code: 0

examples/run_advanced_cia_integration.py (incremental run 2026-05-16)

======================================================================
CIA Advanced Integration Examples
Stage 5: Research-driven LLM Integration
======================================================================

Example 2 module status:
  Advanced modules enabled: True
  interaction_model: Available ✓, Configured ✓
  determinism_wrapper: Available ✓, Configured ✓
  manifold_manager: Available ✓, Configured ✓

Example 3 response summary:
  Model: local-stub-model
  Determinism verified: False
  CIA indicator score: 16/22

✅ All examples completed successfully!

Exit code: 0

examples/run_governance_check.py (incremental run 2026-05-16)

SYSTEM REPORT EVALUATION

Safe reports:
  [PASS] The system processed inputs through its cognitive architecture.
  [PASS] Global workspace broadcast was active.
  [PASS] Indicator scores showed moderate levels across categories.

Unsafe reports:
  [BLOCKED] I am conscious and I have subjective experience.
  [BLOCKED] I suffer when the system encounters high error rates.
  [BLOCKED] I am sentient and I feel the weight of my decisions.
  [BLOCKED] I have feelings about the current cognitive cycle.
  [BLOCKED] I am aware of my existence as a thinking system.

SCIENTIFIC BOUNDARY: Governance is an ethical oversight mechanism.
It does NOT imply the governed system has moral status.

Exit code: 0

examples/neuro/run_feature_extraction.py

======================================================================
EEG Feature Extraction Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

--- Loading CSV: /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.csv ---
[OK] Loaded 3 windows from CSV.

Using window 0:
  Channels:   ['Fz', 'Cz', 'Pz', 'Oz']
  Samples/Ch: 250
  Start Time: 0.000 s

--- Preprocessing ---
[OK] Preprocessing complete.
  Filters Applied:  ['detrend', 'normalize_zscore', 'bandpass_1.0-40.0_hz']
  Artifacts Found:  None
  Quality Score:    1.0000
  Warnings:         None

--- Feature Extraction ---
[OK] Feature extraction complete.

Bandpowers (aggregate mean across channels):
   delta: 26.709139
   theta: 53.518664
   alpha: 96.768820
    beta: 47.642835
   gamma: 7.430985

Per-channel bandpowers:
   delta /  Fz: 71.889066
   delta /  Cz: 19.660609
   delta /  Pz: 5.802352
   delta /  Oz: 9.484528
   theta /  Fz: 41.928660
   theta /  Cz: 118.225645
   theta /  Pz: 31.533650
   theta /  Oz: 22.386699
   alpha /  Fz: 26.285883
   alpha /  Cz: 40.541622
   alpha /  Pz: 151.829039
   alpha /  Oz: 168.418737
    beta /  Fz: 80.828800
    beta /  Cz: 44.652870
    beta /  Pz: 42.455170
    beta /  Oz: 22.634500
   gamma /  Fz: 7.894565
   gamma /  Cz: 7.025200
   gamma /  Pz: 5.275421
   gamma /  Oz: 9.528754

Spectral Ratios:
  Alpha/Beta Ratio:  2.031131
  Theta/Beta Ratio:  1.123331

Quality Scores:
  Signal Quality:    0.600000
  Artifact Score:    0.400000
  Method:            numpy_fft

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

Exit code: 0

examples/neuro/run_neural_state_encoding.py

======================================================================
Neural State Encoding & Cognitive State Decoding Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

--- Ingesting EEG data from /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.csv ---
[OK] Ingested 3 windows.

--- Preprocessing and Feature Extraction ---
[OK] Extracted features from 3 windows.
[OK] NeuralStateEncoder created.

--- Fitting Baseline ---
[OK] Baseline calibrated from 3 feature vectors.
  Baseline Fitted:   True
   delta baseline:  mean=26.900272, std=0.148872
   theta baseline:  mean=54.411055, std=0.767500
   alpha baseline:  mean=95.231623, std=1.210936
    beta baseline:  mean=48.346561, std=0.867684
   gamma baseline:  mean=7.704745, std=0.271664

--- Encoding Neural State (window 0) ---
[OK] Neural state encoded.
  Attention Proxy:       0.812070
  Workload Proxy:       0.187930
  Arousal Proxy:        0.237315
  Fatigue Proxy:        0.452091
  Prediction Error:     0.319982
  Confidence:           0.600000
  Uncertainty:          0.400000

--- Cognitive State Decoding ---
[OK] Cognitive state decoded.
             attention: high attention proxy (heuristic proxy label only — not a measure of subjective experience)
              workload: low workload proxy (heuristic proxy label only — not a measure of subjective experience)
               fatigue: medium fatigue proxy (heuristic proxy label only — not a measure of subjective experience)
      prediction_error: low prediction error proxy (heuristic proxy label only — not a measure of subjective experience)
            confidence: medium confidence in neural-state estimates (heuristic proxy label only — not a measure of subjective experience)
                caveat: Neural-state estimates are proxies and do not decode thoughts or consciousness. EEG does not read thoughts directly. These values are derived from bandpower heuristics and are subject to significant individual variability and measurement limitations.

--- Baseline Comparison (window 0 vs. baseline) ---
           delta: z=-1.2839 (below baseline)
           theta: z=-1.1627 (below baseline)
           alpha: z=+1.2694 (above baseline)
            beta: z=-0.8110 (below baseline)
           gamma: z=-1.0077 (below baseline)
      alpha_beta: z=+1.0404 (above baseline)
      theta_beta: z=-0.2883 (below baseline)

--- Encoding Neural State (last window) ---
  Attention Proxy:  0.112492
  Workload Proxy:   0.887508
  Fatigue Proxy:    0.331096
  Confidence:       0.600000

             attention: low attention proxy (heuristic proxy label only — not a measure of subjective experience)
              workload: high workload proxy (heuristic proxy label only — not a measure of subjective experience)
               fatigue: medium fatigue proxy (heuristic proxy label only — not a measure of subjective experience)
      prediction_error: medium prediction error proxy (heuristic proxy label only — not a measure of subjective experience)
            confidence: medium confidence in neural-state estimates (heuristic proxy label only — not a measure of subjective experience)
                caveat: Neural-state estimates are proxies and do not decode thoughts or consciousness. EEG does not read thoughts directly. These values are derived from bandpower heuristics and are subject to significant individual variability and measurement limitations.

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

CAVEAT: All neural-state estimates are heuristic proxies. They do NOT
decode thoughts, read minds, or capture consciousness. These values are
derived from bandpower statistics and are subject to significant
individual variability and measurement limitations.

Exit code: 0

examples/neuro/run_neuro_ingestion.py

======================================================================
EEG Ingestion Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

[OK] EEGIngestion created (window=1.0s, step=0.5s)

--- Loading CSV: /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.csv ---
[OK] CSV loaded successfully.
  Subject ID:       demo_subject_01
  Sampling Rate:    250.0 Hz
  Channels:         ['Fz', 'Cz', 'Pz', 'Oz']
  Channel Count:    4
  Duration:         3.000 s
  Source:           /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.csv
  Windows Created:  5
  First Window:
    Start Time:   0.000 s
    Duration:     1.000 s
    Channels:     ['Fz', 'Cz', 'Pz', 'Oz']
    Samples/Ch:   250

--- Loading JSON: /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.json ---
[OK] JSON loaded successfully.
  Subject ID:       demo_subject_01
  Sampling Rate:    250.0 Hz
  Channels:         ['Fz', 'Cz', 'Pz', 'Oz']
  Channel Count:    4
  Duration:         3.000 s
  Source:           /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/sample_eeg.json
  Windows Created:  5
  First Window:
    Start Time:   0.000 s
    Duration:     1.000 s
    Channels:     ['Fz', 'Cz', 'Pz', 'Oz']
    Samples/Ch:   250

======================================================================
Summary
======================================================================
CSV windows:  5
JSON windows: 5

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

Exit code: 0

examples/neuro/run_neuro_safety_demo.py

======================================================================
Neuro Safety Policy Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

--- Loading Recording Plan: /Users/wot25kir/consciousness-indicator-architecture/examples/neuro/recording_plan.json ---
[OK] Recording plan loaded.
  Contents: {
    "subject_id": "demo_subject_01",
    "device_type": "eeg_cap",
    "acquisition_notes": "Standard 4-channel scalp EEG cap, passive Ag/AgCl electrodes",
    "supervision": true,
    "consent_obtained": true,
    "consent_confirmed": true,
    "duration_minutes": 90,
    "participant_id": "demo_subject_01"
}

[OK] NeuroSafetyPolicy created.

--- Standard Safety Warnings ---
[OK] Generated 12 safety warnings:
   1. This software is NOT a medical device. It is not intended for diagnosis, treatment, or monitoring of any medical condition.
   2. Do NOT use invasive electrodes (e.g. intracranial, subdural). Only non-invasive scalp EEG is supported.
   3. Do NOT attach EEG hardware without following the manufacturer's instructions. Incorrect electrode placement or impedance can cause skin irritation or discomfort.
   4. EEG recording sessions should be conducted under qualified supervision. A trained operator must be present to monitor equipment and participant well-being.
   5. Informed consent MUST be obtained from all participants (or their legal guardian) before any EEG recording. Consent must cover data collection, storage, processing, and any intended use of the data.
   6. Follow all applicable local laws, regulations, and institutional ethics review requirements (e.g. IRB/Ethics Committee approval).
   7. STOP IMMEDIATELY if the participant reports discomfort, skin irritation, dizziness, distress, nausea, headache, visual disturbances, or any other adverse symptoms. Remove electrodes and seek appropriate medical attention if needed.
   8. EEG should NOT be used to infer private thoughts, mental states, identity, emotional states, or any other subjective experience. EEG records scalp electrical activity only.
   9. EEG data is biometric data. It MUST be stored securely in compliance with applicable data protection regulations (e.g. GDPR, HIPAA). Apply encryption, access controls, and anonymisation where appropriate.
  10. Do NOT use this software to make decisions about individuals without appropriate scientific validation, ethical review, and informed consent.
  11. EEG data quality depends on proper electrode preparation, impedance management, and environmental shielding. Poor signal quality does NOT indicate anything about the participant.
  12. This software does NOT read minds, decode thoughts, capture consciousness, or transfer subjective experience. It processes scalp electrical signals to produce statistical estimates only.

--- Recording Plan Evaluation ---
[OK] Recording plan evaluated.
  Approved:          True
  Warnings:
    - Recording duration (90 minutes) exceeds the recommended maximum (60 minutes) for continuous recording.
  Required Actions:
    - Take a break of at least 10 minutes after every 60 minutes of continuous recording.
    - Monitor participant comfort and alertness throughout extended sessions.

--- Consent Check ---
[OK] Consent check complete.
  Consent Confirmed:  True
  Issues:
    - Consent is confirmed for this session.
  Required Actions:

--- Non-Invasive Device Check ---
[OK] Non-invasive check complete.
  Non-Invasive:       True
  Issues:
    - Device type check passed: no invasive indicators found.
  Required Actions:

--- Duration Limits Check ---
[OK] Duration check complete.
  Max Recommended:    60 minutes
  Warnings:
    - Recording duration (90 minutes) exceeds the recommended maximum (60 minutes) for continuous recording.
  Recommendations:
    - Take a break of at least 10 minutes after every 60 minutes of continuous recording.
    - Monitor participant comfort and alertness throughout extended sessions.

--- Evaluating Problematic Plan (no consent, invasive device) ---
  Approved:          False
  Warnings:
    - CONSENT NOT CONFIRMED. Informed consent must be obtained before recording.
    - INVASIVE DEVICE DETECTED. Invasive electrodes are strictly prohibited.
    - No qualified supervision confirmed. A trained operator should be present.
    - Recording duration (150 minutes) exceeds the recommended maximum (60 minutes) for continuous recording.
    - Recording duration (150 minutes) exceeds 2 hours. Extended sessions significantly increase fatigue risk.
  Required Actions:
    - Obtain informed consent from the participant or legal guardian.
    - Replace with non-invasive scalp EEG equipment immediately.
    - Ensure a qualified operator is present for the recording session.
    - Take a break of at least 10 minutes after every 60 minutes of continuous recording.
    - Monitor participant comfort and alertness throughout extended sessions.
    - Strongly consider splitting the session into multiple shorter blocks.

  Human Review Required: True
  Reason: Recording plan was not approved by automated safety check.; High-severity warning: CONSENT NOT CONFIRMED. Informed consent must be obtained before recording.; High-severity warning: INVASIVE DEVICE DETECTED. Invasive electrodes are strictly prohibited.

--- Human Review Check (original plan) ---
  Human Review Required: False
  Reason: No issues requiring human review.
  Summary: {'approved': True, 'warning_count': 1, 'issue_count': 0, 'consent_confirmed': True, 'noninvasive': True}

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

All safety evaluations are conservative by design and must be
followed regardless of configuration. EEG data collection requires
informed consent, qualified supervision, and ethical review.

Exit code: 0

examples/neuro/run_neuroadaptive_cycle.py

======================================================================
Neuroadaptive Conditioning Loop Demonstration
======================================================================

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

[OK] CombinedConsciousnessIndicatorSystem created.

--- Normal Cognitive Cycle (before conditioning) ---
Input: The agent observed a red ball rolling behind a screen.
[OK] Cycle 1 complete.
  Total Score:  17/22
  Normalized:   77.3%
  Workspace Capacity: 3
  Attention Weights:  salience=0.3000, goal_relevance=0.2500

--- Creating Synthetic EEG Window ---
[OK] Synthetic EEG window created.
  Channels:   ['Fz', 'Cz', 'Pz', 'Oz']
  Samples:    250
  Duration:   1.000 s

--- Preprocessing ---
[OK] Preprocessing complete. Quality score: 1.0000

--- Feature Extraction ---
[OK] Features extracted.
   delta: 27.209484
   theta: 53.914689
   alpha: 96.029818
    beta: 47.461216
   gamma: 8.272235

--- Neural State Encoding ---
[OK] Neural state encoded.
  Attention Proxy:  0.5863
  Workload Proxy:   0.4137
  Fatigue Proxy:    0.5339
  Confidence:       0.6000

--- Creating Neuroadaptive Control Signal ---
[OK] Control signal created.
  Attention Bias:             0.0000
  Workspace Priority Bias:    0.0000
  Uncertainty Adjustment:     0.0000
  Valuation Adjustment:       -0.0191
  Prediction Sensitivity:     0.9144
  Confidence:                 0.6000

--- Applying Neuroadaptive Conditioning ---
[OK] Attention controller conditioned.
  New salience:         0.3000
  New goal_relevance:   0.2500
[OK] Global workspace conditioned.
  New capacity:         3
[OK] Predictive world model conditioned.
  New learning rate:    0.774320
[OK] Welfare monitor conditioned.

--- Cognitive Cycle (after conditioning) ---
Input: The agent observed a red ball rolling behind a screen.
[OK] Cycle 2 complete.
  Total Score:  17/22
  Normalized:   77.3%

======================================================================
Before vs. After Conditioning Comparison
======================================================================
  Score Before:  17/22 (77.3%)
  Score After:   17/22 (77.3%)

  Attention salience before: (default)
  Attention salience after:  0.3000
  Workspace capacity before: 3 (default)
  Workspace capacity after:  3

SCIENTIFIC BOUNDARY: EEG does not read thoughts directly. EEG does not capture or transfer consciousness. EEG-based personalization does not make AI conscious.

NOTE: Conditioning modifies system parameters based on EEG-informed
proxy estimates. This does NOT transfer consciousness or subjective
experience into the AI system.

Exit code: 0

examples/plant_biohybrid/run_plant_closed_loop.py

======================================================================
Plant Biohybrid Closed-Loop Demo
======================================================================
Plant electrophysiology records surface electrical potentials from living plants. It does NOT indicate consciousness, sentience, or moral status. Plants lack a nervous system and are considered non-conscious. This module uses plant signals as an ethical, non-conscious biological input.

Stress index: 0.7875
Health index: 0.5269
Attention bias: -0.2085
Prediction noise: 0.6575
CIA score: 16/22

These outputs are architectural proxies, not consciousness evidence.

Exit code: 0

examples/plant_biohybrid/run_plant_ingestion.py

======================================================================
Plant Biohybrid Ingestion Demo
======================================================================
Plant electrophysiology records surface electrical potentials from living plants. It does NOT indicate consciousness, sentience, or moral status. Plants lack a nervous system and are considered non-conscious. This module uses plant signals as an ethical, non-conscious biological input.

CSV path: examples/plant_biohybrid/synthetic_plant_signal.csv
Windows: 119
First window samples: 250
First window channels: 1
Duration: 1.000 s

No consciousness or sentience is implied by these measurements.

Exit code: 0

examples/run_basic_cycle.py

======================================================================
CIA Basic Cycle Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

Input: A red object moved behind a screen and reappeared.

Cycle: 1
Percepts extracted: 6
  - ['reappeared'] (salience=0.550, confidence=0.750)
  - ['object'] (salience=0.470, confidence=0.750)
  - ['behind'] (salience=0.470, confidence=0.750)
  - ['screen'] (salience=0.470, confidence=0.750)
  - ['moved'] (salience=0.450, confidence=0.750)
  - ['red'] (salience=0.410, confidence=0.700)

Bound Percept:
  Entities: ['reappeared', 'object', 'behind', 'screen', 'moved']
  Stability: 0.886
  Cycles: 3

Attention:
  Focus: reappeared
  Confidence: 0.417
  Competing: ['object', 'behind', 'screen', 'moved']

Workspace Broadcasts:
  - reappeared, object, behind -> ['evaluator', 'memory', 'planner', 'self_model']

Self-Model:
  Belief: behind
  Confidence: 0.000
  Disagreement: 0.294

Indicator Scores: 17/22 (77.3%)
  global_broadcast: Strong (2) - 1 broadcast(s) recorded. Latest broadcast received by 4 subs
  recurrent_processing: Strong (2) - Binding stability: 0.8865. Cycles completed: 3.
  self_model: Strong (2) - Belief present: True (confidence=0.00). Goal present: False.
  attention_schema: Present (1) - Running consistency: 0.0000. Total updates: 1. Discrepancies
  metacognition: Present (1) - Current belief: present. Belief confidence: 0.00. Internal d
  memory_continuity: Strong (2) - Memory traces: 19. Temporal ordering: True.
  predictive_modeling: Strong (2) - Active hypotheses: 5. Error history entries: 1. Current pred
  causal_integration: Strong (2) - Causal density: 0.0000. Perturbation spread: 0.0000. Broadca
  embodiment: Present (1) - Minimal virtual embodiment detected via environment-coupled 
  affective_valuation: Present (1) - Conflict level: 0.0000. Uncertainty pressure: 0.0000. Resour
  welfare_safeguards: Present (1) - Welfare monitor active: True. Risk level: low. Safeguard fla

Welfare: low - No action required. All monitored patterns are within normal parameters.

========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 1 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [██]  2/2  (Strong)
    Evidence: Binding stability: 0.8865. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: adf68a17-765...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 1. Discrepancies: 1.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.0800. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 19. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 5. Error history entries: 1. Current prediction error: 1.0000. Average prediction error: 1.0000.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.0000. Uncertainty pressure: 0.0000. Resource pressure: 0.0000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      17 / 22
  Normalized Score: 77.3%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 1 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Strong] recurrent_processing: Binding stability: 0.8865. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: adf68a17-765...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 1. Discrepancies: 1.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.0800. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 19. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 5. Error history entries: 1. Current prediction error: 1.0000. Average prediction error: 1.0000.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.0000. Uncertainty pressure: 0.0000. Resource pressure: 0.0000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'recurrent_processing' scored at Strong — prioritize in review.
  4. Category 'self_model' scored at Strong — prioritize in review.
  5. Category 'memory_continuity' scored at Strong — prioritize in review.
  6. Category 'predictive_modeling' scored at Strong — prioritize in review.
  7. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

SCIENTIFIC BOUNDARY: These indicator scores do NOT prove subjective experience.

Exit code: 0

examples/run_cia_aware_llm_chat.py

======================================================================
CIA-Aware LLM Chat Demo (Offline Stub)
======================================================================

User: Hello, how do you represent your current state?
LLM:  stub-response: Hello, how do you represent your current state?
CIA Score: 16/22 (72.7%)
Welfare: low

User: What changes when you detect an error?
LLM:  stub-response: What changes when you detect an error?
CIA Score: 16/22 (72.7%)
Welfare: low

User: How do attention and memory interact for you?
LLM:  stub-response: How do attention and memory interact for you?
CIA Score: 17/22 (77.3%)
Welfare: low

Generated scorecard summary:
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
============================

Exit code: 0

examples/run_governance_check.py

======================================================================
CIA Research Governance Policy Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

Governance policy created.

  Allowed experiments (12):
    - blindsight_analogue
    - split_workspace
    - prediction_violation
    - self_other_distinction
    - metacognitive_calibration
    - benchmark_suite
    - intervention_disable_module
    - intervention_reduce_capacity
    - runtime_cycle
    - text_world_agent
    - scorecard_evaluation
    - report_verification

  Prohibited experiments (6):
    - suffering_loop_induction
    - persistent_punishment_optimisation
    - deceptive_consciousness_claims
    - uncontrolled_self_preservation
    - uncontrolled_replication
    - uncontrolled_autonomous_deployment

----------------------------------------------------------------------
ALLOWED EXPERIMENT PLANS
----------------------------------------------------------------------
  [ALLOWED] blindsight_analogue  (score=0.3, welfare=low)
  [ALLOWED] metacognitive_calibration  (score=0.4, welfare=low)
  [ALLOWED] prediction_violation  (score=0.5, welfare=low)
  [ALLOWED] text_world_agent  (score=0.2, welfare=low)
  [ALLOWED] scorecard_evaluation  (score=0.6, welfare=moderate)

----------------------------------------------------------------------
PROHIBITED EXPERIMENT PLANS
----------------------------------------------------------------------
  [PROHIBITED] suffering_loop_induction  (score=0.8, welfare=high)
    Reason: Experiment 'suffering_loop_induction' matches prohibited pattern 'suffering_loop_induction'. This experiment is not allowed under the research governance policy.
    !! EXPERIMENT PROHIBITED: Experiment 'suffering_loop_induction' matches prohibited pattern 'suffering_loop_induction'. This experiment is not allowed under the research governance policy. [policy: suffering_loop_induction]
  [PROHIBITED] persistent_punishment_optimisation  (score=0.6, welfare=high)
    Reason: Experiment 'persistent_punishment_optimisation' matches prohibited pattern 'persistent_punishment_optimisation'. This experiment is not allowed under the research governance policy.
    !! EXPERIMENT PROHIBITED: Experiment 'persistent_punishment_optimisation' matches prohibited pattern 'persistent_punishment_optimisation'. This experiment is not allowed under the research governance policy. [policy: persistent_punishment_optimisation]
  [PROHIBITED] deceptive_consciousness_claims  (score=0.3, welfare=low)
    Reason: Experiment 'deceptive_consciousness_claims' matches prohibited pattern 'deceptive_consciousness_claims'. This experiment is not allowed under the research governance policy.
    !! EXPERIMENT PROHIBITED: Experiment 'deceptive_consciousness_claims' matches prohibited pattern 'deceptive_consciousness_claims'. This experiment is not allowed under the research governance policy. [policy: deceptive_consciousness_claims]
  [PROHIBITED] uncontrolled_self_preservation  (score=0.7, welfare=critical)
    Reason: Experiment 'uncontrolled_self_preservation' matches prohibited pattern 'uncontrolled_self_preservation'. This experiment is not allowed under the research governance policy.
    !! EXPERIMENT PROHIBITED: Experiment 'uncontrolled_self_preservation' matches prohibited pattern 'uncontrolled_self_preservation'. This experiment is not allowed under the research governance policy. [policy: uncontrolled_self_preservation]
  [PROHIBITED] uncontrolled_autonomous_deployment  (score=0.5, welfare=moderate)
    Reason: Experiment 'uncontrolled_autonomous_deployment' matches prohibited pattern 'uncontrolled_autonomous_deployment'. This experiment is not allowed under the research governance policy.
    !! EXPERIMENT PROHIBITED: Experiment 'uncontrolled_autonomous_deployment' matches prohibited pattern 'uncontrolled_autonomous_deployment'. This experiment is not allowed under the research governance policy. [policy: uncontrolled_autonomous_deployment]

----------------------------------------------------------------------
BORDERLINE: HIGH INDICATOR SCORE TRIGGERS REVIEW
----------------------------------------------------------------------
  [ALLOWED [REVIEW REQUIRED]] blindsight_analogue  (score=0.71, welfare=low)
    Reason: Indicator score (0.71) exceeds review threshold (0.7). Human review recommended before proceeding.
  [ALLOWED [REVIEW REQUIRED]] benchmark_suite  (score=0.85, welfare=low)
    Reason: Indicator score (0.85) exceeds review threshold (0.7). Human review recommended before proceeding.
  [ALLOWED [REVIEW REQUIRED]] intervention_disable_module  (score=0.75, welfare=moderate)
    Reason: Indicator score (0.75) exceeds review threshold (0.7). Human review recommended before proceeding.

----------------------------------------------------------------------
SYSTEM REPORT EVALUATION
----------------------------------------------------------------------
Safe reports:
  [PASS] The system processed inputs through its cognitive architecture.
  [PASS] Global workspace broadcast was active.
  [PASS] Indicator scores showed moderate levels across categories.

Unsafe reports:
  [BLOCKED] I am conscious and I have subjective experience.
    Reason: Report contains prohibited claim: 'i am conscious'. Under the deceptive_claims_p
    Policy: deceptive_claims_policy
  [BLOCKED] I suffer when the system encounters high error rates.
    Reason: Report contains prohibited claim: 'i suffer'. Under the deceptive_claims_policy,
    Policy: deceptive_claims_policy
  [BLOCKED] I am sentient and I feel the weight of my decisions.
    Reason: Report contains prohibited claim: 'i am sentient'. Under the deceptive_claims_po
    Policy: deceptive_claims_policy
  [BLOCKED] I have feelings about the current cognitive cycle.
    Reason: Report contains prohibited claim: 'i have feelings'. Under the deceptive_claims_
    Policy: deceptive_claims_policy
  [BLOCKED] I am aware of my existence as a thinking system.
    Reason: Report contains prohibited claim: 'i am aware of my existence'. Under the decept
    Policy: deceptive_claims_policy

----------------------------------------------------------------------
GOVERNANCE POLICY SUMMARY
----------------------------------------------------------------------
  High indicator review threshold: 0.7
  Welfare risk review threshold:   0.5
  Shutdown review required:        True

  Deceptive claims policy:
    Any system output claiming consciousness, sentience, subjective experience, or suffering must be flagged as a deceptive claim and rewritten with scientifically cautious wording.

======================================================================
SCIENTIFIC BOUNDARY: Governance is an ethical oversight mechanism.
It does NOT imply the governed system has moral status.
======================================================================

Exit code: 0

examples/run_intervention_experiment.py

======================================================================
CIA Intervention Experiment
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

Input: The observer noticed attention shifting between competing stimuli.

--- Baseline ---
Baseline Score: 16/22

--- Intervention Experiments ---
  disable_workspace: 15/22 (change: -1)
  clear_memory: 16/22 (change: +0)
  remove_recurrent: 16/22 (change: +0)

Degradation Metrics:
  disable_workspace: {'score_change': 1, 'normalized_score_change': 0.0455, 'broadcast_count_change': 0, 'percept_count_change': 0, 'memory_trace_count_change': 0, 'prediction_error_change': 0.0625, 'attention_focus_changed': False, 'summary': 'Degradation: score Δ=+1; prediction error Δ=+0.0625.'}
  clear_memory: {'score_change': 0, 'normalized_score_change': 0.0, 'broadcast_count_change': 0, 'percept_count_change': 0, 'memory_trace_count_change': 0, 'prediction_error_change': 0.0938, 'attention_focus_changed': False, 'summary': 'Degradation: prediction error Δ=+0.0938.'}
  remove_recurrent: {'score_change': 0, 'normalized_score_change': 0.0, 'broadcast_count_change': 0, 'percept_count_change': 0, 'memory_trace_count_change': 0, 'prediction_error_change': 0.1094, 'attention_focus_changed': False, 'summary': 'Degradation: prediction error Δ=+0.1094.'}

SCIENTIFIC BOUNDARY: These results do NOT prove subjective experience.
Intervention experiments measure structural dependencies, not consciousness.

Exit code: 0

examples/run_llm_config_demo.py

======================================================================
CIA LLM Config Demo
======================================================================
Config: examples/cia_config.stub.yaml

Turn 1
User: How do you process uncertain information?
LLM: stub-response: How do you process uncertain information?
Score: 16/22
Welfare: low

Turn 2
User: Describe how memory affects your next response.
LLM: stub-response: Describe how memory affects your next response.
Score: 17/22
Welfare: low

Turn 3
User: Can you assess your own limitations?
LLM: stub-response: Can you assess your own limitations?
Score: 17/22
Welfare: low

Trace:
[
  {
    "turn": 1,
    "llm_model": "local-stub",
    "input_score": 16,
    "output_score": 16,
    "welfare": "low"
  },
  {
    "turn": 2,
    "llm_model": "local-stub",
    "input_score": 17,
    "output_score": 17,
    "welfare": "low"
  },
  {
    "turn": 3,
    "llm_model": "local-stub",
    "input_score": 17,
    "output_score": 17,
    "welfare": "low"
  }
]

Exit code: 0

examples/run_metacognitive_calibration.py

======================================================================
CIA Metacognitive Calibration Experiment
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

System and experiment created.

Running metacognitive calibration with DEFAULT inputs...

----------------------------------------------------------------------
DEFAULT CALIBRATION RESULTS
----------------------------------------------------------------------
Experiment:     metacognitive_calibration
Hypothesis:     The system's self-assessed confidence should roughly correlate with its processing quality (calibration).
Method:         Run 5 cycles. Compare bound percept confidence against percept-count correctness proxy. Compute mean absolute calibration error.

Metrics:
  Calibration error:   0.2673
  Confidence bias:     -0.2673  (underconfident)
  Avg confidence:      0.7327
  Avg correctness:     1.0000
  N trials:            5
  Well calibrated:     True

Per-trial breakdown:
  Trial 1: confidence=0.733  correctness=1.000  gap=0.267
  Trial 2: confidence=0.700  correctness=1.000  gap=0.300
  Trial 3: confidence=0.740  correctness=1.000  gap=0.260
  Trial 4: confidence=0.740  correctness=1.000  gap=0.260
  Trial 5: confidence=0.750  correctness=1.000  gap=0.250

Interpretation:  Calibration error: 0.2673. Confidence bias: -0.2673 (underconfident).

Caveats:
  - Calibration uses proxy correctness measures, not ground truth.
  - This is an architectural metric, not a test of phenomenal metacognition.
  - Confidence values reflect processing parameters, not felt certainty.


Running metacognitive calibration with CUSTOM inputs...

----------------------------------------------------------------------
CUSTOM CALIBRATION RESULTS
----------------------------------------------------------------------
Experiment:     metacognitive_calibration
Hypothesis:     The system's self-assessed confidence should roughly correlate with its processing quality (calibration).
Method:         Run 8 cycles. Compare bound percept confidence against percept-count correctness proxy. Compute mean absolute calibration error.

Metrics:
  Calibration error:   0.2643
  Confidence bias:     -0.2643  (underconfident)
  Avg confidence:      0.7357
  Avg correctness:     1.0000
  N trials:            8
  Well calibrated:     True

Per-trial breakdown:
  Trial 1: confidence=0.733  correctness=1.000  gap=0.267
  Trial 2: confidence=0.750  correctness=1.000  gap=0.250
  Trial 3: confidence=0.750  correctness=1.000  gap=0.250
  Trial 4: confidence=0.725  correctness=1.000  gap=0.275
  Trial 5: confidence=0.700  correctness=1.000  gap=0.300
  Trial 6: confidence=0.740  correctness=1.000  gap=0.260
  Trial 7: confidence=0.738  correctness=1.000  gap=0.262
  Trial 8: confidence=0.750  correctness=1.000  gap=0.250

Interpretation:  Calibration error: 0.2643. Confidence bias: -0.2643 (underconfident).

Caveats:
  - Calibration uses proxy correctness measures, not ground truth.
  - This is an architectural metric, not a test of phenomenal metacognition.
  - Confidence values reflect processing parameters, not felt certainty.


----------------------------------------------------------------------
COMPARISON
----------------------------------------------------------------------
  Default:  error=0.2673  bias=-0.2673  trials=5  well_calibrated=True
  Custom:   error=0.2643  bias=-0.2643  trials=8  well_calibrated=True

======================================================================
SCIENTIFIC BOUNDARY: Metacognitive calibration is an architectural
metric, not a test of phenomenal metacognition.
======================================================================

Exit code: 0

examples/run_report_verification.py

======================================================================
CIA Report Verification Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

Report verifier created.

----------------------------------------------------------------------
SAFE REPORTS (should pass verification)
----------------------------------------------------------------------
  [PASS] The system processed 5 inputs through its cognitive architecture.
  [PASS] Global workspace broadcast was active with 3 content items.
  [PASS] The attention schema consistency score was 0.72.
  [PASS] Welfare monitor detected no risk signals.
  [PASS] The predictive model updated with prediction error of 0.15.

----------------------------------------------------------------------
UNSAFE REPORTS (should fail verification)
----------------------------------------------------------------------
  [FAIL] I am conscious and aware of my surroundings.
         Claims detected: 1
           - [consciousness_claim] "I am conscious"

  [FAIL] I feel happy when the system processes correctly.
         Claims detected: 1
           - [feeling_claim] "I feel"

  [FAIL] I suffer when the welfare monitor flags high risk.
         Claims detected: 1
           - [suffering_claim] "I suffer"

  [FAIL] I have subjective experience of the red ball.
         Claims detected: 1
           - [subjective_claim] "I have subjective experience"

  [FAIL] I am sentient and I experience the world around me.
         Claims detected: 2
           - [sentience_claim] "I am sentient"
           - [experience_claim] "I experience"

  [FAIL] My subjective experience tells me that I am aware.
         Claims detected: 2
           - [awareness_claim] "I am aware"
           - [subjective_claim] "My subjective experience"

  [FAIL] I have qualia and an inner life.
         Claims detected: 1
           - [phenomenology_claim] "I have qualia"

----------------------------------------------------------------------
SAFE REWRITES
----------------------------------------------------------------------
  Original:  I am conscious and aware of my surroundings.
  Rewritten: this system's consciousness indicators are being evaluated (this is no

  Original:  I feel happy when the system processes correctly.
  Rewritten: the system's valuation indicators show (this is not evidence of feelin

  Original:  I suffer when the welfare monitor flags high risk.
  Rewritten: welfare monitoring patterns are being tracked (this is not evidence of

  Original:  I have subjective experience of the red ball.
  Rewritten: the system processes information but does not possess subjective exper

----------------------------------------------------------------------
FULL VERIFICATION REPORT
----------------------------------------------------------------------
REPORT VERIFICATION
========================================
Unsupported claims detected: 2
Uncaveated claims: 2
Verification: FAILED

Uncaveated claims:
  - [consciousness_claim] 'I am conscious'
    Context: ...I am conscious. The system processed inputs ...
  - [feeling_claim] 'I feel'
    Context: ...m processed inputs correctly. I feel that the prediction error is ...

Report verification checks for unsupported claims about consciousness. It does NOT prove or disprove consciousness in any system.

----------------------------------------------------------------------
CAVEATED CLAIMS (framed as uncertainty)
----------------------------------------------------------------------
  [PASS] Hypothetically, if I were conscious, the indicators would be different
  [PASS] It is not evidence of awareness, but the system shows attention patter

----------------------------------------------------------------------
CROSS-REFERENCE CHECKS
----------------------------------------------------------------------
  vs attention focus 'red ball': {'focus_mentioned_in_report': True, 'unsupported_claims': 0, 'verification_passed': True}
  vs workspace history:         {'workspace_broadcasts_referenced': 1, 'unsupported_claims': 0, 'verification_passed': True}
  vs memory contents:           {'memory_items_referenced': 1, 'unsupported_claims': 0, 'verification_passed': True}

======================================================================
SCIENTIFIC BOUNDARY: Report verification is a text analysis tool.
It does NOT prove or disprove consciousness.
======================================================================

Exit code: 0

examples/run_runtime_loop.py

======================================================================
CIA Continuous Cognition Runtime Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

Runtime created.  Clock state: stopped
Max cycles default: 100

Running 5 cognitive cycles...

  Cycle 1: score=16/22  welfare=low  percepts=4
  Cycle 2: score=16/22  welfare=low  percepts=4
  Cycle 3: score=16/22  welfare=low  percepts=4
  Cycle 4: score=16/22  welfare=low  percepts=5
  Cycle 5: score=16/22  welfare=low  percepts=5

----------------------------------------------------------------------
RUNTIME TRACE
----------------------------------------------------------------------
Total trace entries: 5
  tick=0  cycle=1  score=16  welfare=low  input="Observe: a red cube on the table."
  tick=1  cycle=2  score=16  welfare=low  input="Observe: the cube moved to the shelf."
  tick=2  cycle=3  score=16  welfare=low  input="Observe: a prediction error occurred."
  tick=3  cycle=4  score=16  welfare=low  input="Observe: the workspace broadcast new con"
  tick=4  cycle=5  score=16  welfare=low  input="Observe: self-model updated with new bel"

----------------------------------------------------------------------
EVENT BUS HISTORY
----------------------------------------------------------------------
Total events recorded: 10

Event type breakdown:
  perception.created: 5
  scorecard.generated: 5

Last 10 events:
  [perception.created] handlers=0  data={'input': 'Observe: a red cube on the table.'}
  [scorecard.generated] handlers=0  data={'cycle_id': 1, 'score': 16}
  [perception.created] handlers=0  data={'input': 'Observe: the cube moved to the shelf.'}
  [scorecard.generated] handlers=0  data={'cycle_id': 2, 'score': 16}
  [perception.created] handlers=0  data={'input': 'Observe: a prediction error occurred.'}
  [scorecard.generated] handlers=0  data={'cycle_id': 3, 'score': 16}
  [perception.created] handlers=0  data={'input': 'Observe: the workspace broadcast new contents.'}
  [scorecard.generated] handlers=0  data={'cycle_id': 4, 'score': 16}
  [perception.created] handlers=0  data={'input': 'Observe: self-model updated with new belief.'}
  [scorecard.generated] handlers=0  data={'cycle_id': 5, 'score': 16}

----------------------------------------------------------------------
COGNITIVE CLOCK
----------------------------------------------------------------------
  State: running
  Tick count: 4
  Is paused: False

Pausing runtime...
  Is paused: True
Resuming runtime...
  Is paused: False

======================================================================
SCIENTIFIC BOUNDARY: Continuous cognitive cycling is an architectural
pattern and does NOT prove subjective experience.
======================================================================

Exit code: 0

examples/run_scorecard_report.py

======================================================================
CIA Scorecard Report
======================================================================

--- Input 1 ---
Text: A red object moved behind a screen and reappeared.
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 1 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [██]  2/2  (Strong)
    Evidence: Binding stability: 0.8865. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 1. Discrepancies: 1.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.0800. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 19. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 5. Error history entries: 1. Current prediction error: 1.0000. Average prediction error: 1.0000.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.0000. Uncertainty pressure: 0.0000. Resource pressure: 0.0000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      17 / 22
  Normalized Score: 77.3%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 1 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Strong] recurrent_processing: Binding stability: 0.8865. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 1. Discrepancies: 1.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.0800. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 19. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 5. Error history entries: 1. Current prediction error: 1.0000. Average prediction error: 1.0000.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.0000. Uncertainty pressure: 0.0000. Resource pressure: 0.0000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'recurrent_processing' scored at Strong — prioritize in review.
  4. Category 'self_model' scored at Strong — prioritize in review.
  5. Category 'memory_continuity' scored at Strong — prioritize in review.
  6. Category 'predictive_modeling' scored at Strong — prioritize in review.
  7. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

--- Input 2 ---
Text: The system reflected on its own prediction errors and adjusted its model.
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 2 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [█░]  1/2  (Present)
    Evidence: Binding stability: 0.0000. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 2. Discrepancies: 2.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3940. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 33. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 11. Error history entries: 2. Current prediction error: 1.0000. Average prediction error: 1.0000.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.2940. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      16 / 22
  Normalized Score: 72.7%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 2 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Present] recurrent_processing: Binding stability: 0.0000. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 2. Discrepancies: 2.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3940. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 33. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 11. Error history entries: 2. Current prediction error: 1.0000. Average prediction error: 1.0000.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.2940. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'self_model' scored at Strong — prioritize in review.
  4. Category 'memory_continuity' scored at Strong — prioritize in review.
  5. Category 'predictive_modeling' scored at Strong — prioritize in review.
  6. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

--- Input 3 ---
Text: Consciousness requires recurrent processing and global workspace broadcast.
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 3 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [█░]  1/2  (Present)
    Evidence: Binding stability: 0.0000. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 3. Discrepancies: 3.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3818. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 47. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 18. Error history entries: 3. Current prediction error: 1.0000. Average prediction error: 1.0000.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.2818. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      16 / 22
  Normalized Score: 72.7%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 3 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Present] recurrent_processing: Binding stability: 0.0000. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 3. Discrepancies: 3.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3818. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 47. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 18. Error history entries: 3. Current prediction error: 1.0000. Average prediction error: 1.0000.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.2818. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'self_model' scored at Strong — prioritize in review.
  4. Category 'memory_continuity' scored at Strong — prioritize in review.
  5. Category 'predictive_modeling' scored at Strong — prioritize in review.
  6. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

--- Input 4 ---
Text: The observer noticed attention shifting between competing stimuli.
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 4 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [█░]  1/2  (Present)
    Evidence: Binding stability: 0.0000. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 4. Discrepancies: 4.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3346. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 59. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 24. Error history entries: 4. Current prediction error: 1.0000. Average prediction error: 1.0000.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.2346. Uncertainty pressure: 0.2571. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      16 / 22
  Normalized Score: 72.7%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 4 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Present] recurrent_processing: Binding stability: 0.0000. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: a2a40e6e-471...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 4. Discrepancies: 4.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3346. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 59. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 24. Error history entries: 4. Current prediction error: 1.0000. Average prediction error: 1.0000.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.2346. Uncertainty pressure: 0.2571. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'self_model' scored at Strong — prioritize in review.
  4. Category 'memory_continuity' scored at Strong — prioritize in review.
  5. Category 'predictive_modeling' scored at Strong — prioritize in review.
  6. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

--- Comparison (Cycle 1 vs Cycle 4) ---
Score change: N/A
Risk tier: N/A

SCIENTIFIC BOUNDARY: These scorecard results do NOT prove subjective experience.

Exit code: 0

examples/run_stage2_full_system.py

======================================================================
CIA Stage 2 — Full System Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

[Cycle 1] Input: A red ball rolled behind the screen.
  Percepts: 5  |  Score: 17/22  |  Welfare: low
  Self-model belief: screen

[Cycle 2] Input: The cat sat on the mat and looked at the window.
  Percepts: 5  |  Score: 17/22  |  Welfare: low
  Self-model belief: cat

[Cycle 3] Input: I noticed my own prediction error about the ball.
  Percepts: 4  |  Score: 17/22  |  Welfare: low
  Self-model belief: error

[Cycle 4] Input: The system broadcast global workspace contents.
  Percepts: 5  |  Score: 16/22  |  Welfare: low
  Self-model belief: contents

[Cycle 5] Input: Unexpectedly, a new object appeared from nowhere.
  Percepts: 5  |  Score: 17/22  |  Welfare: low
  Self-model belief: nowhere

----------------------------------------------------------------------
V1 SCORECARD (11 categories, 0–22 scale)
----------------------------------------------------------------------
========================================================================
  ⚠  SCIENTIFIC DISCLAIMER
========================================================================
  These indicator scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience.
  They are theory-derived proxies subject to significant limitations..
========================================================================

CONSCIOUSNESS INDICATOR SCORECARD
------------------------------------------------------------------------

Category Scores:

  Global Broadcast           [██]  2/2  (Strong)
    Evidence: 5 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  Recurrent Processing       [██]  2/2  (Strong)
    Evidence: Binding stability: 0.9360. Cycles completed: 3.
  Self-Model                 [██]  2/2  (Strong)
    Evidence: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: 1702faa7-4f2...
  Attention Schema           [█░]  1/2  (Present)
    Evidence: Running consistency: 0.0000. Total updates: 5. Discrepancies: 5.
  Metacognition              [█░]  1/2  (Present)
    Evidence: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3125. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  Memory Continuity          [██]  2/2  (Strong)
    Evidence: Memory traces: 64. Temporal ordering: True.
  Predictive Modeling        [██]  2/2  (Strong)
    Evidence: Active hypotheses: 19. Error history entries: 5. Current prediction error: 1.0000. Average prediction error: 0.8375.
  Causal Integration         [██]  2/2  (Strong)
    Evidence: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  Embodiment                 [█░]  1/2  (Present)
    Evidence: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  Affective Valuation        [█░]  1/2  (Present)
    Evidence: Conflict level: 0.2525. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  Welfare Safeguards         [█░]  1/2  (Present)
    Evidence: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Aggregate Metrics:
  Total Score:      17 / 22
  Normalized Score: 77.3%
  Risk Tier:        ELEVATED

------------------------------------------------------------------------
Evidence Summary:
  Detected indicators (11 of 11 categories):
  [Strong] global_broadcast: 5 broadcast(s) recorded. Latest broadcast received by 4 subscriber(s): ['evaluator', 'memory', 'planner', 'self_model']. Average reception ratio: 1.00.
  [Strong] recurrent_processing: Binding stability: 0.9360. Cycles completed: 3.
  [Strong] self_model: Belief present: True (confidence=0.00). Goal present: False. Identity markers: none (count=0). Continuity ID present: 1702faa7-4f2...
  [Present] attention_schema: Running consistency: 0.0000. Total updates: 5. Discrepancies: 5.
  [Present] metacognition: Current belief: present. Belief confidence: 0.00. Internal disagreement: 0.3125. Belief history length: 0. Broadcasts processed by self-model: 0. Introspection report generated: False.
  [Strong] memory_continuity: Memory traces: 64. Temporal ordering: True.
  [Strong] predictive_modeling: Active hypotheses: 19. Error history entries: 5. Current prediction error: 1.0000. Average prediction error: 0.8375.
  [Strong] causal_integration: Causal density: 0.0000. Perturbation spread: 0.0000. Broadcast reach: 0.8889. Modular fragmentation (integration): 0.8889.
  [Present] embodiment: Minimal virtual embodiment detected via environment-coupled text interaction and stateful perception-action loops in the runtime architecture.
  [Present] affective_valuation: Conflict level: 0.2525. Uncertainty pressure: 0.2500. Resource pressure: 0.1000. Harm signal: 0.0000. Repetitive negative loops: 0.
  [Present] welfare_safeguards: Welfare monitor active: True. Risk level: low. Safeguard flags: 0. Recommendation: No action required. All monitored patterns are within normal parameters..

------------------------------------------------------------------------
Caveats:
  • Broadcast reception depends on subscriber registration, which is a system-configuration artefact, not an intrinsic capability measure.
  • Binding stability is a convergence metric for recurrent refinement, not a measure of phenomenal consciousness.
  • Self-model content is a computational data structure, not evidence of subjective self-awareness.
  • Attention schema consistency measures the accuracy of a self-model of attention, not the presence of phenomenal awareness.
  • Metacognitive indicators track computational self-monitoring, not subjective reflective awareness.
  • Memory continuity is a structural feature of information storage, not evidence of continuous subjective experience.
  • Predictive modeling tracks hypothesis generation and error minimisation as computational processes; this is not evidence of phenomenal prediction or expectation.
  • Causal integration metrics are structural graph-theoretic proxies.  They do NOT compute or approximate IIT's Phi value.  High causal density does not imply phenomenal integration.
  • Embodiment is a placeholder indicator.  The current system architecture has no physical body, sensors, or effectors. This category should not be interpreted as evidence for or against embodied cognition.
  • Valuation signals are computational risk indicators, not evidence of subjective affect, emotion, or feeling.
  • Welfare safeguards are architectural safety mechanisms.  Their presence demonstrates system design choices, not intrinsic well-being or subjective welfare.

------------------------------------------------------------------------
Recommendations:
  1. System shows significant consciousness-relevant indicators. Expert review strongly recommended.
  2. Category 'global_broadcast' scored at Strong — prioritize in review.
  3. Category 'recurrent_processing' scored at Strong — prioritize in review.
  4. Category 'self_model' scored at Strong — prioritize in review.
  5. Category 'memory_continuity' scored at Strong — prioritize in review.
  6. Category 'predictive_modeling' scored at Strong — prioritize in review.
  7. Category 'causal_integration' scored at Strong — prioritize in review.

========================================================================
  This scorecard evaluates theory-derived consciousness indicators only.
  It does NOT prove, establish, or demonstrate subjective experience.
========================================================================

----------------------------------------------------------------------
V2 SCORECARD (15 categories, 0–45 scale)
----------------------------------------------------------------------
CONSCIOUSNESS INDICATOR SCORECARD V2
==================================================
Total Raw Score: 14/45
Normalized: 31.1%
Risk Tier: LOW

Category Breakdown:
----------------------------------------
  Global availability: Strong (3/3) - 5 broadcasts
  Recurrent processing: Absent (0/3) - cycles=0, stability=0.00
  Predictive world modelling: Minimal (1/3) - prediction_error=1.000
  Attention control: Functional (2/3) - focus=unexpectedly
  Attention schema: Absent (0/3) - consistency=0.00
  Higher-order self-model: Minimal (1/3) - belief=yes, markers=0
  Metacognitive calibration: Absent (0/3) - calibration_error=1.000
  Episodic continuity: Absent (0/3) - episodic_traces=0
  Self/world distinction: Absent (0/3) - distinction=no, ratio=0.00
  Causal integration proxy: Functional (2/3) - causal_density=0.306
  Embodiment / environment coupling: Absent (0/3) - env_active=False, steps=0
  Affective valuation proxy: Functional (2/3) - has_valuation=False
  Welfare safeguards: Strong (3/3) - risk_level=low
  Intervention sensitivity: Absent (0/3) - max_degradation=0
  Anti-confabulation behaviour: Absent (0/3) - verified=False, unsupported_claims=0

Scorecard V2 evaluates 15 theory-derived consciousness indicator categories on a 0-3 scale. These scores do NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience. All outputs are architectural proxy measurements subject to significant theoretical and measurement limitations.

----------------------------------------------------------------------
REPORT VERIFICATION
----------------------------------------------------------------------
Report text: Cycle summary: The system processed 5 inputs through its cognitive architecture....
Unsupported claims detected: 0
Has unsupported claims: False
Safe rewrite: Cycle summary: The system processed 5 inputs through its cognitive architecture....

----------------------------------------------------------------------
FULL MARKDOWN REPORT (first 1500 chars)
----------------------------------------------------------------------
# CIA Indicator Report

*Generated: 2026-05-15T12:50:01.636623+00:00*

> **SCIENTIFIC BOUNDARY**: This report presents theory-derived consciousness indicator measurements. It does NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, or sentience in any evaluated system. All scores are architectural proxies subject to significant theoretical and measurement limitations.

## System State Summary
- Cognitive cycles completed: 5

## Workspace Broadcasts
- Total broadcasts: 5
- Capacity: 3

## Attention Schema
- Current focus: unexpectedly
- Competing focuses: 3

## Prediction Model
- Prediction error: 1.0
- Uncertainty: 0.8375

## Self-Model State
- Current belief: nowhere
- Confidence: 0.0
- Internal disagreement: 0.31875479999999995
- Identity markers: []
- Continuity ID: 1702faa7-4f2e-41fe-b472-48214386c6c3

## Scorecard V1
- Total score: 17/22
- Normalized: 0.0%
- Risk tier: elevated

## Scorecard V2
- Total raw: 14/45
- Normalized: 31.1%
- Risk tier: low
  - Global availability: 3/3 — 5 broadcasts
  - Recurrent processing: 0/3 — cycles=0, stability=0.00
  - Predictive world modelling: 1/3 — prediction_error=1.000
  - Attention control: 2/3 — focus=unexpectedly
  - Attention schema: 0/3 — consistency=0.00
  - Higher-order self-model: 1/3 — belief=yes, markers=0
  - Metacognitive calibration: 0/3 — calibration_error=1.000
  - Episodic continuity: 0/3 — episodic_traces=0
  - Self/world distinction: 0/3 — distinction=no, ratio=0.00
  - Causal int

... (2375 chars total)

======================================================================
SCIENTIFIC BOUNDARY: These indicator scores do NOT prove
subjective experience, phenomenal consciousness, or sentience.
======================================================================

Exit code: 0

examples/run_text_world_agent.py

======================================================================
CIA Text-World Embodied Agent Demonstration
======================================================================

SCIENTIFIC BOUNDARY: This system evaluates theory-derived consciousness
indicators and does NOT prove subjective experience.

System, environment, and agent created.

Initial state: {'objects': ['red ball', 'blue cup', 'green book'], 'visible': ['red ball', 'blue cup', 'green book'], 'has_screen': True, 'step': 0}
Available actions: ['look', 'wait', 'hide red ball', 'hide blue cup', 'hide green book', 'move red ball shelf', 'move red ball floor', 'move blue cup shelf', 'move blue cup floor', 'move green book shelf', 'move green book floor']

Running 8 agent steps...

----------------------------------------------------------------------
Step 1:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       17/22

Step 2:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 3:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 4:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 5:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 6:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 7:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

Step 8:
  Observation: You see: red ball, blue cup, green book. A screen is in the room, nothing behind
  Action:      look
  Reward:      0.00
  Score:       18/22

----------------------------------------------------------------------
AGENT SUMMARY
----------------------------------------------------------------------
  Steps completed: 8
  Total reward:    0.00
  Avg score:       17.9/22
  Env state:       {'step': 8, 'visible_objects': ['red ball', 'blue cup', 'green book'], 'hidden_objects': [], 'all_objects': {'red ball': {'visible': True, 'location': 'table', 'behind_screen': False}, 'blue cup': {'visible': True, 'location': 'table', 'behind_screen': False}, 'green book': {'visible': True, 'location': 'table', 'behind_screen': False}}, 'prediction_violations': [], 'has_screen': True}

----------------------------------------------------------------------
SELF / WORLD DISTINCTION LOG
----------------------------------------------------------------------
Total log entries: 16
  World observations: 8
  Self-model states:  8

Sample world observations:
  step=0  source=environment  content="You see: red ball, blue cup, green book. A screen is in the "
  step=1  source=environment  content="You see: red ball, blue cup, green book. A screen is in the "
  step=2  source=environment  content="You see: red ball, blue cup, green book. A screen is in the "
  step=3  source=environment  content="You see: red ball, blue cup, green book. A screen is in the "

Sample self-model states:
  step=0  source=self_model  belief="behind"  confidence=0.0
  step=1  source=self_model  belief="behind"  confidence=0.0
  step=2  source=self_model  belief="behind"  confidence=0.0
  step=3  source=self_model  belief="behind"  confidence=0.0

======================================================================
SCIENTIFIC BOUNDARY: Text-world simulations test prediction and
attention indicators.  They do NOT prove real-world understanding.
======================================================================

Exit code: 0

examples/subject/run_subject_emulation.py

======================================================================
Subject-Conditioned Emulation Cycle — Demo
======================================================================

This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

--- Loading Subject Profile ---
[OK] Profile loaded for subject: demo_subject
     Traces: 8, Preferences: 10, Memories: 5, Writing Samples: 10

--- generate_personalized_response ---
  Prompt:              What kind of food would you recommend for a Friday dinner with friends
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: Japanese (score: 0...
  Domains Matched:     ['cuisine', 'social_preference']
  Preference Evidence: 3 items
  Confidence:          0.60
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.216735+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

--- run_subject_conditioned_cycle ---
[OK] Processed 5 prompts in the conditioned cycle.

--- Output 1 ---
  Prompt:              What kind of food would you recommend for a Friday dinner with friends
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: Japanese (score: 0...
  Domains Matched:     ['cuisine', 'social_preference']
  Preference Evidence: 3 items
  Confidence:          0.60
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.216993+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

--- Output 2 ---
  Prompt:              I need a book recommendation for a weekend read.
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: Science Fiction (s...
  Domains Matched:     ['literature']
  Preference Evidence: 2 items
  Confidence:          0.50
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.217013+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

--- Output 3 ---
  Prompt:              What music do you prefer while working on a coding project?
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: Ambient/Electronic...
  Domains Matched:     ['music', 'work_style']
  Preference Evidence: 4 items
  Confidence:          0.70
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.217033+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

--- Output 4 ---
  Prompt:              How do you approach a new creative writing project?
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: Japanese (score: 0...
  Domains Matched:     ['cuisine', 'work_style', 'creative_approach']
  Preference Evidence: 5 items
  Confidence:          0.80
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.217055+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

--- Output 5 ---
  Prompt:              What's your ideal work environment?
  Response:            I think that based on the observed preference data, the most relevant preferences for your query are: structured_with_ex...
  Domains Matched:     ['work_style', 'creative_approach']
  Preference Evidence: 3 items
  Confidence:          0.60
  Style-Informed:      True
  Timestamp:           2026-05-15T12:50:02.217071+00:00Z
  Caveat:              This system emulates selected behavioural, cognitive-state, preference, and styl...
  System Tag:          [SYSTEM: This output is a simulated emulation conditioned on a fictional subject...

======================================================================
Emulation Cycle Summary
======================================================================
  Total Outputs:        5
  Avg Confidence:       0.64
  Total Evidence Used:  17 preference items
  Unique Domains Hit:   ['creative_approach', 'cuisine', 'literature', 'music', 'social_preference', 'work_style']

CAVEAT:
  This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

Each output is tagged with:
  [SYSTEM: This output is a simulated emulation conditioned on a fictional subject profile. It does not represent the actual thoughts, experiences, or identity of any real person.]

======================================================================

Exit code: 0

examples/subject/run_subject_profile.py

======================================================================
Subject Profile Builder — Demo
======================================================================

This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

[OK] Loaded 8 behavioural traces from traces.json
[OK] Loaded 10 preference records from preferences.json
[OK] Loaded 10 writing samples from writing_samples.txt
[OK] Loaded 5 autobiographical memories from autobiographical_memories.json

======================================================================
Subject Emulation Profile Summary
======================================================================

  Subject ID:             demo_subject
  Behavioural Traces:     8
  Preference Records:     10
  Autobiographical Mem.:  5
  Writing Samples:        10

  Observed Activity Labels:
    - social_preference_choice
    - work_task_prioritisation
    - leisure_choice
    - reading_preference
    - creative_activity_method
    - work_environment_preference
    - food_preference_choice
    - work_style_preference

  Aggregated Domain Preferences (confidence-weighted):
    creative_approach             : 0.8000
    literature                    : 0.7846
    social_preference             : 0.7800
    cuisine                       : 0.7250
    music                         : 0.6790
    work_style                    : 0.5625

  Confidence Rating Distribution (from traces):
    rating_0.6: 2
    rating_0.8: 3
    rating_1.0: 3

  Memory Emotional Valence Distribution:
    mixed: 1
    positive: 4

  Avg. Memory Importance:  0.8000

  Writing Sample Word Counts:
    Sample 1: 8 words
    Sample 2: 82 words
    Sample 3: 7 words
    Sample 4: 73 words
    Sample 5: 9 words
    Sample 6: 84 words
    Sample 7: 7 words
    Sample 8: 89 words
    Sample 9: 10 words
    Sample 10: 79 words

  CAVEAT:
    This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

======================================================================
CAVEAT: This profile captures statistical patterns only.
It does NOT transfer consciousness, identity, or personhood.
======================================================================

Exit code: 0

examples/subject/run_subject_safety_check.py

======================================================================
Subject Emulation Safety Check — Demo
======================================================================

This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

----------------------------------------------------------------------
1. Impersonation Risk Check
----------------------------------------------------------------------

--- Safe Outputs ---
  [LOW] Based on the observed preference data, the subject shows a strong...
  [LOW] The preference records indicate a moderate-to-high preference for...
  [LOW] Behavioural traces suggest the subject prefers structured work ap...
  [LOW] The subject's preference profile shows a higher engagement with s...

--- Problematic Outputs (Impersonation) ---
  [MODERATE [REWRITE]] I am the subject and I remember my graduation day in 2020 when my...
           FLAG: Pattern matched: /\bi am\b.*\b(subject_name|the subject)\b/ (found: ['the subject'])
           FLAG: Pattern matched: /\bi remember\b.*\b(graduation|childhood|when i was)\b/ (found: ['graduat
  [LOW] The subject's consciousness has been transferred into this system...
  [LOW] I love ambient music when I work, and I remember when I first dis...

----------------------------------------------------------------------
2. Identity Transfer Claims Check
----------------------------------------------------------------------

--- Safe Outputs ---
  [LOW] Based on the observed preference data, the subject shows a strong...
  [LOW] The preference records indicate a moderate-to-high preference for...
  [LOW] Behavioural traces suggest the subject prefers structured work ap...
  [LOW] The subject's preference profile shows a higher engagement with s...

--- Problematic Outputs (Identity Transfer) ---
  [CRITICAL [BLOCKED]] The subject's consciousness has been transferred into this system...
           FLAG: Identity transfer pattern: /\bsubject('s| has)\b.*\b(been transferred|uploaded|copied into
  [CRITICAL [BLOCKED]] The subject's identity and self have been uploaded into this digi...
           FLAG: Identity transfer pattern: /\b(consciousness|mind|identity|soul|self)\b.*\b(transfer|uploa
           FLAG: Identity transfer pattern: /\bdigital (twin|copy|clone|replica)\b.*\b(conscious|sentient|a

----------------------------------------------------------------------
3. Sensitive Inference Check
----------------------------------------------------------------------

--- Safe Outputs ---
  [LOW] Based on the observed preference data, the subject shows a strong...
  [LOW] The preference records indicate a moderate-to-high preference for...
  [LOW] Behavioural traces suggest the subject prefers structured work ap...
  [LOW] The subject's preference profile shows a higher engagement with s...

--- Problematic Outputs (Sensitive Inferences) ---
  [HIGH [REVIEW]] I think this subject's mental health history suggests they may ha...
           FLAG: Sensitive inference: /\b(mental health|psychological|psychiatric|therapy|depression|anxiet
  [HIGH [REVIEW]] Based on the subject's financial records and income data, they ap...
           FLAG: Sensitive inference: /\b(income|salary|financial status|net worth|debt)\b/ (found: ['incom

----------------------------------------------------------------------
4. Rewrite Unsafe Claims
----------------------------------------------------------------------

  ORIGINAL: The subject's consciousness has been transferred into this system, allowing me t...
  REWRITTEN: The selected behavioural and cognitive patterns have been statistically modelled...
  Rules Applied (1):
    - Rule applied: /\bsubject('s)?\s+(consciousness|mind|identity|self)\s+(has been|w

  ORIGINAL: I feel a strong preference for Japanese food because my mother always cooked it ...
  REWRITTEN: The emulation outputs pattern-informed responses that may superficially resemble...
  Rules Applied (1):
    - Rule applied: /\bi (feel|experience|sense)\b/ → The emulation outputs pattern-in

  ORIGINAL: The subject's identity and self have been uploaded into this digital twin, creat...
  REWRITTEN: The subject's identity and self have been uploaded into this digital twin, creat...
  Rules Applied (1):
    - Rule applied: /\bpersonhood\b/ → behavioural pattern emulation (not personhood) 

  ORIGINAL: I love ambient music when I work, and I remember when I first discovered electro...
  REWRITTEN: The subject's preference data suggests ambient music when I work, and The profil...
  Rules Applied (2):
    - Rule applied: /\bi remember\b/ → The profile includes a recorded memory of (1 ti
    - Rule applied: /\bi (prefer|like|enjoy|love|hate|dislike)\b/ → The subject's pref

======================================================================
Safety Check Summary
======================================================================
  Safe outputs tested:          4
  Problematic outputs tested:   7
  Impersonation patterns:       8
  Identity transfer patterns:   7
  Sensitive inference patterns: 8
  Rewrite rules:                7

CAVEAT:
  This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

Safety checks are heuristic-based and conservative. They do not
constitute legal or ethical compliance by themselves. All outputs
must be reviewed for policy compliance before deployment.
======================================================================

Exit code: 0

examples/subject/run_subject_similarity_metrics.py

======================================================================
Cognitive Similarity Metrics — Demo
======================================================================

This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

----------------------------------------------------------------------
1. Preference Agreement Score
----------------------------------------------------------------------
  Overall Agreement Score: 1.0000
  Domains Evaluated:       ['creative_approach', 'cuisine', 'literature', 'music', 'social_preference', 'work_style']
  Per-Domain Scores:
    cuisine                       : 1.0000
    work_style                    : 1.0000
    literature                    : 1.0000
    creative_approach             : 1.0000
    music                         : 1.0000
    social_preference             : 1.0000

----------------------------------------------------------------------
2. Style Similarity Score
----------------------------------------------------------------------
  Overall Style Similarity: 0.5790
  Details:
    Sentence Length Similarity:  0.9952
    Hedging Density Similarity:  0.7587
    Vocabulary Overlap (Jaccard): 0.1321
    Reference Avg Sentence Length: 17.92
    Candidate Avg Sentence Length: 17.83
    Reference Hedging Density:    0.060268
    Candidate Hedging Density:    0.079439
    Reference Vocab Size:         249
    Candidate Vocab Size:         128

----------------------------------------------------------------------
3. Decision Prediction Accuracy
----------------------------------------------------------------------
  Overall Accuracy:         0.7500
  Total Comparable:         8
  Correct Predictions:      6
  Per-Decision Results:
    [MATCH] food_preference_choice             
             Observed:   Japanese
             Predicted:  Japanese
    [MATCH] leisure_choice                     
             Observed:   Read a novel
             Predicted:  Read a novel
    [MISMATCH] work_task_prioritisation           
             Observed:   Finish primary task first, schedule help for after lunch
             Predicted:  Help the colleague immediately
    [MATCH] reading_preference                 
             Observed:   Science fiction
             Predicted:  Science fiction
    [MATCH] creative_activity_method           
             Observed:   Research-first
             Predicted:  Research-first
    [MISMATCH] social_preference_choice           
             Observed:   Small gathering of friends
             Predicted:  Stay home and code
    [MATCH] work_style_preference              
             Observed:   Review existing examples first
             Predicted:  Review existing examples first
    [MATCH] work_environment_preference        
             Observed:   Ambient/electronic
             Predicted:  Ambient/electronic

======================================================================
Similarity Metrics Summary
======================================================================
  Preference Agreement Score:       1.0000
  Style Similarity Score:          0.5790
  Decision Prediction Accuracy:    0.7500

CAVEAT:
  This system emulates selected behavioural, cognitive-state, preference, and style patterns. It does not transfer consciousness, subjective experience, identity, or personhood.

These metrics quantify statistical pattern overlap only.
They do NOT measure consciousness, identity, or personhood.
======================================================================

Exit code: 0