eduevidence 5.2.0 → 6.2.0

package/examples/workplace-ai-assistant/evidence_graph.json

@@ -0,0 +1,444 @@
+{
+  "project_id": "workplace-ai-assistant",
+  "revision_id": 1,
+  "intent": {
+    "domain": "policy",
+    "question": "Should an enterprise customer-support team introduce a generative AI assistant?",
+    "data_origin": "manual_curated",
+    "benchmark_eligible": false,
+    "timestamp_note": "Date-only curation marker, not execution time."
+  },
+  "created_at": "2026-09-08T00:00:00Z",
+  "updated_at": "2026-09-08T00:00:00Z",
+  "audit_warnings": [],
+  "papers": {
+    "S-001": {
+      "paper_id": "S-001",
+      "title": "Generative AI at Work",
+      "authors": [
+        "Erik Brynjolfsson",
+        "Danielle Li",
+        "Lindsey Raymond"
+      ],
+      "year": 2025,
+      "venue": "Academic Publication",
+      "doi": "10.1093/qje/qjae044",
+      "url": "https://academic.oup.com/qje/article/140/2/889/7990658",
+      "authority_tier": 1,
+      "peer_reviewed": true,
+      "summary": ""
+    },
+    "S-002": {
+      "paper_id": "S-002",
+      "title": "Experimental evidence on the productivity effects of generative artificial intelligence",
+      "authors": [
+        "Shakked Noy",
+        "Whitney Zhang"
+      ],
+      "year": 2023,
+      "venue": "Academic Publication",
+      "doi": "10.1126/science.adh2586",
+      "url": "https://shakkednoy.com/Noy%20Zhang%20NBER%20SI.pdf",
+      "authority_tier": 1,
+      "peer_reviewed": true,
+      "summary": ""
+    },
+    "S-003": {
+      "paper_id": "S-003",
+      "title": "Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality",
+      "authors": [
+        "Fabrizio Dell’Acqua",
+        "Edward McFowland III",
+        "Ethan Mollick",
+        "Hila Lifshitz-Assaf",
+        "Katherine C. Kellogg",
+        "Saran Rajendran",
+        "Lisa Krayer",
+        "François Candelon",
+        "Karim R. Lakhani"
+      ],
+      "year": 2026,
+      "venue": "Academic Publication",
+      "doi": "10.1287/orsc.2025.21838",
+      "url": "https://pubsonline.informs.org/doi/10.1287/orsc.2025.21838",
+      "authority_tier": 1,
+      "peer_reviewed": true,
+      "summary": ""
+    }
+  },
+  "evidence": {
+    "E-001": {
+      "evidence_id": "E-001",
+      "paper_id": "S-001",
+      "outcome_metric": "Chat handling time; separate throughput summary: about 15% more issues resolved per hour.",
+      "outcome_dimension": "Policy",
+      "claim_id": "C-001",
+      "outcome_id": "policy_effectiveness",
+      "effect_size": {
+        "metric": "not_standardized",
+        "value": null,
+        "ci_lower": null,
+        "ci_upper": null,
+        "p_value": null
+      },
+      "sample_size": 5172,
+      "sample_description": "Customer-support agents",
+      "study_design": "quasi_experimental",
+      "direction": "SUPPORTS",
+      "confidence_score": null,
+      "wwc_rating": "Not applicable: organizational policy",
+      "key_quote": "",
+      "calibrated_weight": null,
+      "bias_flag": "One firm, one tool and a staggered nonrandom rollout; causal interpretation depends on identification assumptions."
+    },
+    "E-002": {
+      "evidence_id": "E-002",
+      "paper_id": "S-002",
+      "outcome_metric": "Self-reported task time: approximately 40% lower; assessed writing quality approximately 18% higher is a separate measure.",
+      "outcome_dimension": "Policy",
+      "claim_id": "C-002",
+      "outcome_id": "policy_effectiveness",
+      "effect_size": {
+        "metric": "not_standardized",
+        "value": null,
+        "ci_lower": null,
+        "ci_upper": null,
+        "p_value": null
+      },
+      "sample_size": 453,
+      "sample_description": "College-educated working professionals",
+      "study_design": "rct",
+      "direction": "SUPPORTS",
+      "confidence_score": null,
+      "wwc_rating": "Not applicable: organizational policy",
+      "key_quote": "",
+      "calibrated_weight": null,
+      "bias_flag": "Indirect evidence, not directly generalizable: brief incentivized writing tasks did not demand precise factual accuracy or customer-specific context."
+    },
+    "E-003": {
+      "evidence_id": "E-003",
+      "paper_id": "S-003",
+      "outcome_metric": "Correct business recommendation: about 19 percentage points lower across AI arms; Table 7 has 373 participants, not 758.",
+      "outcome_dimension": "Policy",
+      "claim_id": "C-003",
+      "outcome_id": "implementation_risk",
+      "effect_size": {
+        "metric": "not_standardized",
+        "value": null,
+        "ci_lower": null,
+        "ci_upper": null,
+        "p_value": null
+      },
+      "sample_size": 373,
+      "sample_description": "BCG consultants in the outside-frontier experiment",
+      "study_design": "rct",
+      "direction": "SUPPORTS",
+      "confidence_score": null,
+      "wwc_rating": "Not applicable: organizational policy",
+      "key_quote": "",
+      "calibrated_weight": null,
+      "bias_flag": "Indirect evidence, not directly generalizable: consultants solving an experimental business case, not live customer tickets."
+    },
+    "E-004": {
+      "evidence_id": "E-004",
+      "paper_id": "S-001",
+      "outcome_metric": "Small quality declines among the most experienced and highest-skilled support staff.",
+      "outcome_dimension": "Policy",
+      "claim_id": "C-004",
+      "outcome_id": "implementation_risk",
+      "effect_size": {
+        "metric": "not_standardized",
+        "value": null,
+        "ci_lower": null,
+        "ci_upper": null,
+        "p_value": null
+      },
+      "sample_size": null,
+      "sample_description": "Experienced and high-skill customer-support agents",
+      "study_design": "quasi_experimental",
+      "direction": "SUPPORTS",
+      "confidence_score": null,
+      "wwc_rating": "Not applicable: organizational policy",
+      "key_quote": "",
+      "calibrated_weight": null,
+      "bias_flag": "Same study as E-001; subgroup sample size was not extracted. This is not an independent replication."
+    }
+  },
+  "outcomes": {
+    "policy_effectiveness": {
+      "outcome_id": "policy_effectiveness",
+      "name": "policy_effectiveness",
+      "dimension": "Policy",
+      "category": "Policy",
+      "description": ""
+    },
+    "implementation_risk": {
+      "outcome_id": "implementation_risk",
+      "name": "implementation_risk",
+      "dimension": "Policy",
+      "category": "Policy",
+      "description": ""
+    }
+  },
+  "claims": {
+    "C-001": {
+      "claim_id": "C-001",
+      "statement": "AI assistance can shorten customer chat handling in a bounded support setting; throughput is a separate outcome.",
+      "outcome_dimension": "Policy",
+      "outcome_metric": "",
+      "status": "SUPPORTED",
+      "pooled_effect_g": null,
+      "evidence_ids": [
+        "E-001"
+      ],
+      "bias_warning": ""
+    },
+    "C-002": {
+      "claim_id": "C-002",
+      "statement": "ChatGPT reduced time on short professional writing tasks; this is indirect evidence for customer support.",
+      "outcome_dimension": "Policy",
+      "outcome_metric": "",
+      "status": "SUPPORTED",
+      "pooled_effect_g": null,
+      "evidence_ids": [
+        "E-002"
+      ],
+      "bias_warning": ""
+    },
+    "C-003": {
+      "claim_id": "C-003",
+      "statement": "AI can reduce correctness on tasks outside its capability frontier; consulting evidence is indirect for support.",
+      "outcome_dimension": "Policy",
+      "outcome_metric": "",
+      "status": "SUPPORTED",
+      "pooled_effect_g": null,
+      "evidence_ids": [
+        "E-003"
+      ],
+      "bias_warning": ""
+    },
+    "C-004": {
+      "claim_id": "C-004",
+      "statement": "Experienced, high-skill support staff need separate quality monitoring rather than assumed uniform benefits.",
+      "outcome_dimension": "Policy",
+      "outcome_metric": "",
+      "status": "SUPPORTED",
+      "pooled_effect_g": null,
+      "evidence_ids": [
+        "E-004"
+      ],
+      "bias_warning": ""
+    }
+  },
+  "risks": {},
+  "gaps": {
+    "G-001": {
+      "gap_id": "G-001",
+      "gap_type": "Local applicability and safety",
+      "description": "Local net cost, privacy incident rates, subgroup service quality and sustained performance are not established by this selected evidence set [no direct evidence].",
+      "target_outcome": "",
+      "existing_evidence_summary": "E-001 through E-004",
+      "recommended_trial_design": "Preregister an intention-to-treat comparison with team-clustered uncertainty and subgroup estimates. Set sample size and noninferiority margins from local baseline variance and operational priorities before enrollment."
+    }
+  },
+  "decision": {
+    "decision_id": "D-001",
+    "verdict": "PILOT",
+    "confidence_score": null,
+    "rationale": "A supervised pilot is warranted because one direct field study supports efficiency gains but also reveals heterogeneous quality effects. Indirect experiments identify task boundaries, while local safety and net value remain unknown.",
+    "applicability_boundary": "",
+    "intervention_plan": {
+      "decision": "pilot",
+      "target_population": "Enterprise customer-support staff, stratified by tenure and baseline skill.",
+      "pilot_duration": "Proposed: two baseline weeks and six pilot weeks; not executed.",
+      "ai_usage_policy": "Start with low-risk, knowledge-base-covered queues; agents inspect and edit every proposed reply.",
+      "risk_control": [
+        "Start with low-risk, knowledge-base-covered queues; agents inspect and edit every proposed reply.",
+        "Stratify by tenure and baseline skill; retain expert discretion and measure review workload.",
+        "Minimize and redact customer data, restrict access and retention, and review supplier data-use terms before a pilot.",
+        "Escalate unfamiliar, ambiguous, sensitive or high-stakes requests to qualified humans; no autonomous commitments."
+      ],
+      "stop_conditions": [
+        "Pause immediately after a verified privacy leak or a serious unsafe or unauthorized customer commitment; investigate before resuming.",
+        "Pause expansion if blinded quality review crosses a pre-agreed noninferiority margin, including within experience strata."
+      ],
+      "evidence_alignment": [
+        "E-001",
+        "E-002",
+        "E-003",
+        "E-004"
+      ],
+      "extensions": {
+        "gap_id": "G-001",
+        "target_population": "enterprise_customer_support_staff",
+        "status": "proposed_not_executed"
+      }
+    },
+    "evaluation_plan": {
+      "research_question": "Should an enterprise customer-support team introduce a generative AI assistant?",
+      "groups": {
+        "treatment": "Eligible teams randomly assigned to supervised assistant access, stratified by tenure and baseline performance.",
+        "comparison": "Concurrent teams retaining the existing workflow; document contamination."
+      },
+      "baseline": "Record resolution rate, paid hours, repeat contacts, blinded quality and review costs before allocation.",
+      "post_test": "Assess the same outcomes at pilot end; separately audit privacy and unsafe commitments.",
+      "analysis_plan": "Preregister an intention-to-treat comparison with team-clustered uncertainty and subgroup estimates. Set sample size and noninferiority margins from local baseline variance and operational priorities before enrollment.",
+      "success_threshold": "Expand only if quality is noninferior, verified resolutions per paid hour improve, net cost is acceptable and no serious unresolved safety incident remains; thresholds require local agreement.",
+      "stop_conditions": [
+        "Pause immediately after a verified privacy leak or a serious unsafe or unauthorized customer commitment; investigate before resuming.",
+        "Pause expansion if blinded quality review crosses a pre-agreed noninferiority margin, including within experience strata."
+      ],
+      "extensions": {
+        "gap_id": "G-001",
+        "status": "proposed_not_executed"
+      }
+    },
+    "stop_conditions": [
+      "Pause immediately after a verified privacy leak or a serious unsafe or unauthorized customer commitment; investigate before resuming.",
+      "Pause expansion if blinded quality review crosses a pre-agreed noninferiority margin, including within experience strata."
+    ],
+    "timestamp": "2026-09-08T00:00:00Z"
+  },
+  "edges": [
+    {
+      "source_id": "E-001",
+      "target_id": "S-001",
+      "relation": "EXTRACTED_FROM",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-001",
+      "target_id": "C-001",
+      "relation": "SUPPORTS",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-001",
+      "target_id": "policy_effectiveness",
+      "relation": "MEASURES",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-001",
+      "target_id": "G-001",
+      "relation": "IDENTIFIES_GAP",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "C-001",
+      "target_id": "D-001",
+      "relation": "GROUNDS_DECISION",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-002",
+      "target_id": "S-002",
+      "relation": "EXTRACTED_FROM",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-002",
+      "target_id": "C-002",
+      "relation": "SUPPORTS",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-002",
+      "target_id": "policy_effectiveness",
+      "relation": "MEASURES",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-002",
+      "target_id": "G-001",
+      "relation": "IDENTIFIES_GAP",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "C-002",
+      "target_id": "D-001",
+      "relation": "GROUNDS_DECISION",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-003",
+      "target_id": "S-003",
+      "relation": "EXTRACTED_FROM",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-003",
+      "target_id": "C-003",
+      "relation": "SUPPORTS",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-003",
+      "target_id": "implementation_risk",
+      "relation": "MEASURES",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-003",
+      "target_id": "G-001",
+      "relation": "IDENTIFIES_GAP",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "C-003",
+      "target_id": "D-001",
+      "relation": "GROUNDS_DECISION",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-004",
+      "target_id": "S-001",
+      "relation": "EXTRACTED_FROM",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-004",
+      "target_id": "C-004",
+      "relation": "SUPPORTS",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-004",
+      "target_id": "implementation_risk",
+      "relation": "MEASURES",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "E-004",
+      "target_id": "G-001",
+      "relation": "IDENTIFIES_GAP",
+      "weight": 1.0,
+      "metadata": {}
+    },
+    {
+      "source_id": "C-004",
+      "target_id": "D-001",
+      "relation": "GROUNDS_DECISION",
+      "weight": 1.0,
+      "metadata": {}
+    }
+  ]
+}

package/examples/workplace-ai-assistant/final_verdict.json

@@ -0,0 +1,78 @@
+{
+  "decision_question": "Should an enterprise customer-support team introduce a generative AI assistant?",
+  "target_population": "Enterprise customer-support staff, stratified by tenure and baseline skill.",
+  "target_context": "Human-supervised support using an approved knowledge base.",
+  "recommended_action": "pilot",
+  "confidence": "Moderate",
+  "confidence_score": 0.578,
+  "confidence_policy_version": "2026-08-12.v3",
+  "raw_model_confidence": "Moderate",
+  "raw_model_confidence_breakdown": {},
+  "independent_studies": 3,
+  "independent_samples": 3,
+  "supported_claims": [
+    "AI assistance can shorten customer chat handling in a bounded support setting; throughput is a separate outcome. — E-001",
+    "ChatGPT reduced time on short professional writing tasks; this is indirect evidence for customer support. — E-002",
+    "AI can reduce correctness on tasks outside its capability frontier; consulting evidence is indirect for support. — E-003",
+    "Experienced, high-skill support staff need separate quality monitoring rather than assumed uniform benefits. — E-004"
+  ],
+  "uncertain_claims": [
+    "Local net cost, privacy incident rates, subgroup service quality and sustained performance are not established by this selected evidence set [no direct evidence]."
+  ],
+  "decision_rationale": "A supervised pilot is warranted because one direct field study supports efficiency gains but also reveals heterogeneous quality effects. Indirect experiments identify task boundaries, while local safety and net value remain unknown.",
+  "strongest_support": "Supervised AI assistance improves handling speed and answer consistency in customer-support work, with quality maintained.",
+  "key_uncertainty": "Evidence comes from adjacent writing and advisory settings rather than the support floor, so transfer to live customer conversations is unproven.",
+  "main_risk": "Unsupervised or knowledge-base-free use can produce confident wrong answers to customers, and over-reliance erodes agent skill over time.",
+  "next_action": "Run a supervised pilot on approved knowledge bases with human review on every reply, and track escalation and correction rates.",
+  "methodology_summary": "One staggered-rollout quasi-experiment and two randomized experiments. Only the support study is direct; no pooled standardized effect or model benchmark was computed.",
+  "what_can_be_claimed": [
+    "Under human supervision on an approved knowledge base, AI assistance can shorten handling time while quality is monitored.",
+    "Benefits are not uniform across staff: the most experienced agents need their own quality monitoring."
+  ],
+  "what_cannot_be_claimed": [
+    "Universal gains, autonomous deployment safety, privacy protection, reduced staffing requirements or educational learning gains."
+  ],
+  "exceeds_evidence_boundary": [
+    "Claiming universal gains or that autonomous deployment is safe exceeds the boundary: no included study measures privacy incidents, local net cost or subgroup service quality."
+  ],
+  "missing_evidence": [
+    "Local net cost, privacy incident rates, subgroup service quality and sustained performance are not established by this selected evidence set [no direct evidence]."
+  ],
+  "applicability": {
+    "required_conditions": [
+      "Start with low-risk, knowledge-base-covered queues; agents inspect and edit every proposed reply.",
+      "Stratify by tenure and baseline skill; retain expert discretion and measure review workload.",
+      "Minimize and redact customer data, restrict access and retention, and review supplier data-use terms before a pilot.",
+      "Escalate unfamiliar, ambiguous, sensitive or high-stakes requests to qualified humans; no autonomous commitments."
+    ]
+  },
+  "extensions": {
+    "data_origin": "manual_curated",
+    "benchmark_eligible": false,
+    "note": "Confidence and the decision bound come from the deterministic policy in engine/decision_policy.py, enforced by the Pre-Verdict Gate; this record is a curated evidence selection, not a systematic review or a model run.",
+    "knowledge_gaps": [
+      {
+        "gap_id": "G-001",
+        "evidence_ids": [
+          "E-001",
+          "E-002",
+          "E-003",
+          "E-004"
+        ],
+        "summary": "Local net cost, privacy incident rates, subgroup service quality and sustained performance are not established by this selected evidence set [no direct evidence]."
+      }
+    ]
+  },
+  "confidence_breakdown": {
+    "score": 0.578,
+    "evidence_quality": 0.8,
+    "consistency": 0.0,
+    "directness": 0.75,
+    "evidence_count": 4,
+    "independent_studies": 3,
+    "independent_samples": 3,
+    "count_term": 0.75,
+    "conflict_penalty": 0.0,
+    "unsupported_penalty": 0.0
+  }
+}

package/examples/workplace-ai-assistant/frame.json

@@ -0,0 +1,41 @@
+{
+  "question": "Should an enterprise customer-support team introduce a generative AI assistant?",
+  "decision_object": "adopt",
+  "intervention": {
+    "policy_name": "human_supervised_customer_support_assistant",
+    "policy_type": "institutional_reform",
+    "mechanism": "approved_knowledge_base_and_agent_review"
+  },
+  "population": {
+    "target_group": "enterprise_customer_support_staff",
+    "excluded_groups": "autonomous_agents_and_high_stakes_specialist_advice"
+  },
+  "comparison": "Existing support workflow without generative AI suggestions.",
+  "outcomes": {
+    "primary": [
+      "policy_effectiveness",
+      "implementation_risk"
+    ],
+    "secondary": [
+      "cost_effectiveness",
+      "equity",
+      "feasibility"
+    ]
+  },
+  "context": {
+    "policy_environment": "enterprise_customer_support"
+  },
+  "scope": {
+    "time_range": "2023–2026",
+    "evidence_types": [
+      "quasi_experimental",
+      "rct"
+    ]
+  },
+  "success_condition": "Improve verified resolutions per paid staff hour while preserving service quality, privacy and staff autonomy.",
+  "extensions": {
+    "domain": "policy",
+    "data_origin": "manual_curated",
+    "note": "Purposive evidence selection, not a systematic review or model execution benchmark."
+  }
+}

package/examples/workplace-ai-assistant/gate_report.json

@@ -0,0 +1,101 @@
+{
+  "gate_version": "2026-08-13.v1",
+  "checked_at": "2026-09-13T03:17:13.887129+00:00",
+  "workspace": "examples/workplace-ai-assistant",
+  "items": {
+    "research_frame_valid": {
+      "title": "Research Frame valid",
+      "status": "pass",
+      "detail": "frame.json valid (question=Should an enterprise customer-support team introduce a generative AI assistant?)",
+      "critical": true,
+      "blocks_high": true
+    },
+    "sources_valid": {
+      "title": "Sources valid",
+      "status": "pass",
+      "detail": "3 source record(s) schema-valid",
+      "critical": true,
+      "blocks_high": true
+    },
+    "evidence_schema_valid": {
+      "title": "Evidence Schema valid",
+      "status": "pass",
+      "detail": "4 evidence record(s) schema-valid",
+      "critical": true,
+      "blocks_high": true
+    },
+    "source_dedupe": {
+      "title": "Source dedupe",
+      "status": "pass",
+      "detail": "3 unique source(s), no duplicates",
+      "critical": true,
+      "blocks_high": true
+    },
+    "counter_evidence_search": {
+      "title": "Counter-evidence search",
+      "status": "pass",
+      "detail": "search_performed=true; 9/9 checks run; findings=8; contradictory_evidence_found=True",
+      "critical": true,
+      "blocks_high": true
+    },
+    "methodology_audit": {
+      "title": "Methodology audit",
+      "status": "pass",
+      "detail": "methodology verdict=CONCERN; task/learning separated",
+      "critical": true,
+      "blocks_high": true
+    },
+    "claim_evidence_audit": {
+      "title": "Claim-Evidence Audit",
+      "status": "pass",
+      "detail": "all verdict claims bind to existing evidence with consistent categories",
+      "critical": true,
+      "blocks_high": false
+    },
+    "outcome_mapping": {
+      "title": "Outcome mapping",
+      "status": "warn",
+      "detail": "outcome keys known; frame-declared outcomes without evidence: cost_effectiveness, equity, feasibility",
+      "critical": false,
+      "blocks_high": false
+    },
+    "scope_calibration": {
+      "title": "Scope calibration",
+      "status": "pass",
+      "detail": "claims bounded: can=2, cannot=1, exceeds_boundary=1",
+      "critical": false,
+      "blocks_high": true
+    },
+    "independent_study_count": {
+      "title": "Independent study-sample count",
+      "status": "pass",
+      "detail": "independent studies=3, samples=3",
+      "critical": true,
+      "blocks_high": true
+    },
+    "deterministic_confidence": {
+      "title": "Deterministic confidence",
+      "status": "pass",
+      "detail": "deterministic confidence=Moderate (score=0.578, policy=2026-08-12.v3)",
+      "critical": true,
+      "blocks_high": true
+    },
+    "decision_action_consistency": {
+      "title": "Decision action consistency",
+      "status": "pass",
+      "detail": "action=pilot is within the conservative bound",
+      "critical": true,
+      "blocks_high": false
+    }
+  },
+  "passed": true,
+  "critical_failures": [],
+  "high_confidence_allowed": true,
+  "max_confidence": "High",
+  "enforcement": {
+    "rule": "confidence capped at {max}; High confidence requires a fully passing gate and >= 2 independent studies",
+    "max_confidence": "High",
+    "requires_action_change": false,
+    "action_override": null
+  }
+}