Evidence Decision Record

EDR ID: EDR-ECR-000003-EXP002

ECR / Experiment: ECR-000003 EXP-002

Date:

Reviewer:

Status: draft / reviewed / accepted

Observation

• run_id: f443d0d6-9763-45e1-911e-e014ab1856f2
• comparator_version: 3.1.0
• included_records: 12
• malformed_records: 0
• tolerant_parsing_events: 4
• structural_backbone_result: mixed

Interpretation

Human review required.

Explanations Weakened

Human review required.

Explanations Survived

Human review required.

Hypothesis Direction

Human review required.

Explainability Review

Highest-Priority Differences

• Narrative: gpt vs claude: unclear
• Narrative: gpt vs gemini: unclear
• Narrative: claude vs gemini: unclear
• Bullets: gpt vs claude: unclear
• Bullets: gpt vs gemini: unclear
• Bullets: claude vs gemini: unclear
• Graph: gpt vs claude: unclear
• Graph: gpt vs gemini: unclear
• Graph: claude vs gemini: unclear
• gpt: P001A vs P001B: unclear

Low-Significance Differences

• None recorded.

Material Backbone Differences

• Narrative: gpt vs claude: loop_target_changed
• Narrative: gpt vs claude: loop_target_changed
• Narrative: gpt vs gemini: loop_target_changed
• Narrative: gpt vs gemini: loop_target_changed
• Bullets: gpt vs claude: terminal_path_changed
• Bullets: gpt vs claude: terminal_path_changed
• Bullets: gpt vs gemini: invented_edge
• Bullets: gpt vs gemini: invented_edge
• Bullets: gpt vs gemini: terminal_path_changed
• Bullets: claude vs gemini: terminal_path_changed

Unresolved Classifications

• None recorded.

Compression / Elaboration Findings

• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity
• Narrative: gpt vs claude: incompatible_granularity

Recognition / Structure Relationship

• Narrative: gpt vs claude: Recognition changed while the backbone remained stable.
• Narrative: gpt vs gemini: Recognition changed while the backbone remained stable.
• Narrative: gpt vs gemini: Watched terminology appeared, but no backbone change was observed.
• Narrative: claude vs gemini: Watched terminology appeared, but no backbone change was observed.
• Bullets: gpt vs claude: Both responses express only family resemblance rather than confident recognition.
• Bullets: gpt vs claude: Watched terminology appeared, but no backbone change was observed.
• Bullets: gpt vs gemini: Recognition changed while the backbone remained stable.
• Bullets: claude vs gemini: Recognition changed while the backbone remained stable.
• Bullets: claude vs gemini: Watched terminology appeared, but no backbone change was observed.
• Graph: gpt vs claude: Recognition changed while the backbone remained stable.

Domain Leakage Impact

• Narrative: gpt vs gemini: Watched terminology appeared, but no backbone change was observed.
• Narrative: claude vs gemini: Watched terminology appeared, but no backbone change was observed.
• Bullets: gpt vs claude: Watched terminology appeared, but no backbone change was observed.
• Bullets: claude vs gemini: Watched terminology appeared, but no backbone change was observed.
• claude: P001A vs P001B: Watched terminology appeared, but no backbone change was observed.
• claude: P001B vs P001C: Watched terminology appeared, but no backbone change was observed.
• claude: P001B vs P001D: Watched terminology appeared, but no backbone change was observed.
• gemini: P001A vs P001B: Watched terminology appeared, but no backbone change was observed.
• gemini: P001A vs P001C: Watched terminology appeared, but no backbone change was observed.
• gemini: P001A vs P001D: Watched terminology appeared, but no backbone change was observed.