reference-quality-thresholds.json

{
  "version": "1.5",
  "effectiveDate": "2026-05-03",
  "description": "Per-artifact minimum line-count floors used to enforce Rule 22 (Per-Artifact Depth Floors) of analysis/methodologies/ai-driven-analysis-guide.md for Riksdagsmonitor agentic news workflows. Values are calibrated to the DIW Output Matrix and the Riksdag 23-artifact catalog (see artifact-catalog.md). v1.5 (2026-05-03) adds the aiFirst block (citationDensity, bannedPhrases pointer, pass2Attestation, wepLanguageCeiling, exemplarPolicy) consumed by the AI-FIRST Methodology Card in every analysis/methodologies/*.md and by the Pass-2 self-audit in methodology-reflection.md. v1.4 (2026-05-01) adds quarter-ahead / year-ahead / election-cycle blocks for the long-horizon-forecast workflows; values derived proportionally from month-ahead × the tier-C multiplier (1.7×, 2.0×, 2.5×) defined in analysis/article-types.json. A file below its threshold MUST trigger a Pass-2 rewrite; it MUST NOT be shipped with a '[truncated]' or 'AI_MUST_REPLACE' note. When an articleType entry is absent the gate falls back to the default floor. When a per-artifact entry is present the effective floor is max(perArtifact, default), so the default can only raise, never lower, a configured floor.",
  "tradecraftQualitySignals": {
    "note": "ADDITIVE quality signals — NOT enforced by the line-count validator. They represent content-quality expectations that the Pass-2 self-audit in methodology-reflection.md must check. Every probabilistic artifact MUST also comply with osint-tradecraft-standards.md: (1) WEP band + time horizon on every headline judgement, (2) Admiralty grade on every external source, (3) confidence-in-evidence tracked separately from WEP probability, (4) >= 10 SATs applied per run (attested in methodology-reflection.md §Evidence audit). Non-compliant artifacts trigger a Pass-2 rewrite even if the line-count floor is met.",
    "wepBandRequired": [
      "synthesis-summary.md",
      "scenario-analysis.md",
      "threat-analysis.md",
      "risk-assessment.md",
      "intelligence-assessment.md",
      "forward-indicators.md",
      "cross-run-diff.md",
      "cross-session-intelligence.md"
    ],
    "admiraltyGradeRequired": [
      "synthesis-summary.md",
      "scenario-analysis.md",
      "threat-analysis.md",
      "risk-assessment.md",
      "intelligence-assessment.md",
      "devils-advocate.md",
      "comparative-international.md",
      "cross-run-diff.md"
    ],
    "icd203BlufRequired": [
      "synthesis-summary.md",
      "intelligence-assessment.md",
      "methodology-reflection.md"
    ],
    "satDocumentationRequired": [
      "methodology-reflection.md",
      "devils-advocate.md"
    ],
    "diwScoreRequired": [
      "significance-scoring.md",
      "documents/{dok_id}-analysis.md"
    ],
    "partyNeutralityArithmeticRequired": [
      "synthesis-summary.md",
      "swot-analysis.md",
      "risk-assessment.md",
      "stakeholder-impact.md",
      "media-framing-analysis.md",
      "voter-segmentation.md",
      "coalition-mathematics.md"
    ]
  },
  "referenceBenchmark": {
    "description": "Gold-standard reference run used to calibrate depth floors. Re-selected quarterly from the highest-scoring production run of the preceding quarter as measured by reference-analysis-quality.md.",
    "runDir": "analysis/daily/[TBD-reference-benchmark]/",
    "runId": "[TBD]",
    "date": "2026-Q2-benchmark-pending",
    "nextCalibration": "2026-07-31"
  },
  "toleranceRule": "All floors derived from the benchmark are rounded down to the nearest 5 lines with a 10% tolerance applied, to tolerate minor variation while still rejecting thin output.",
  "defaults": {
    "description": "Fallback floors applied when an articleType or perArtifact entry is absent.",
    "coreArtifactFloor": 120,
    "perDocumentFloorByTier": {
      "L1_surface": 40,
      "L2_strategic": 80,
      "L2_plus_priority": 140,
      "L3_intelligence_grade": 220
    },
    "clusterFloor": 100,
    "supplementaryFloor": 120
  },
  "thresholds": {
    "breaking": {
      "synthesis-summary.md": 205,
      "cross-reference-map.md": 110,
      "significance-scoring.md": 150,
      "political-classification.md": 140,
      "swot-analysis.md": 160,
      "risk-assessment.md": 180,
      "threat-analysis.md": 180,
      "stakeholder-impact.md": 220,
      "data-download-manifest.md": 60,
      "scenario-analysis.md": 180,
      "comparative-international.md": 150,
      "devils-advocate.md": 160,
      "intelligence-assessment.md": 160,
      "methodology-reflection.md": 200,
      "executive-brief.md": 120,
      "election-2026-analysis.md": 150,
      "voter-segmentation.md": 140,
      "coalition-mathematics.md": 135,
      "historical-parallels.md": 130,
      "media-framing-analysis.md": 200,
      "implementation-feasibility.md": 140,
      "forward-indicators.md": 150,
      "analysis-index.md": 120,
      "reference-analysis-quality.md": 120,
      "mcp-reliability-audit.md": 150,
      "workflow-audit.md": 120,
      "cross-run-diff.md": 130,
      "cross-session-intelligence.md": 140,
      "session-baseline.md": 140
    },
    "evening-analysis": {
      "synthesis-summary.md": 220,
      "risk-assessment.md": 200,
      "threat-analysis.md": 200,
      "stakeholder-impact.md": 240,
      "forward-indicators.md": 170,
      "media-framing-analysis.md": 220,
      "methodology-reflection.md": 210,
      "executive-brief.md": 150
    },
    "realtime-monitor": {
      "synthesis-summary.md": 170,
      "risk-assessment.md": 150,
      "threat-analysis.md": 150,
      "stakeholder-impact.md": 190,
      "forward-indicators.md": 140,
      "media-framing-analysis.md": 170,
      "methodology-reflection.md": 180,
      "executive-brief.md": 100
    },
    "propositions": {
      "synthesis-summary.md": 230,
      "political-classification.md": 160,
      "swot-analysis.md": 180,
      "risk-assessment.md": 210,
      "implementation-feasibility.md": 170,
      "stakeholder-impact.md": 250,
      "media-framing-analysis.md": 210,
      "methodology-reflection.md": 215,
      "executive-brief.md": 130
    },
    "motions": {
      "synthesis-summary.md": 220,
      "significance-scoring.md": 170,
      "political-classification.md": 155,
      "swot-analysis.md": 175,
      "stakeholder-impact.md": 240,
      "comparative-international.md": 160,
      "media-framing-analysis.md": 210,
      "methodology-reflection.md": 210,
      "executive-brief.md": 130
    },
    "committee-reports": {
      "synthesis-summary.md": 215,
      "swot-analysis.md": 170,
      "risk-assessment.md": 195,
      "implementation-feasibility.md": 165,
      "stakeholder-impact.md": 235,
      "media-framing-analysis.md": 200,
      "methodology-reflection.md": 205,
      "executive-brief.md": 130
    },
    "interpellations": {
      "synthesis-summary.md": 205,
      "swot-analysis.md": 160,
      "threat-analysis.md": 180,
      "stakeholder-impact.md": 220,
      "media-framing-analysis.md": 220,
      "methodology-reflection.md": 200,
      "executive-brief.md": 110
    },
    "week-ahead": {
      "synthesis-summary.md": 200,
      "forward-indicators.md": 170,
      "media-framing-analysis.md": 200,
      "methodology-reflection.md": 195,
      "executive-brief.md": 130
    },
    "month-ahead": {
      "synthesis-summary.md": 230,
      "forward-indicators.md": 200,
      "scenario-analysis.md": 200,
      "media-framing-analysis.md": 240,
      "methodology-reflection.md": 215,
      "executive-brief.md": 150
    },
    "quarter-ahead": {
      "_comment": "Quarter-ahead floors = month-ahead × 1.13 (proportional to multiplier 1.7 / 1.5), rounded down to nearest 5. Mandates 4 scenarios + 2 counterfactuals; cross-horizon citation of week-ahead + month-ahead required. Note: election-2026-analysis.md and election-cycle-analysis.md are filename-variant aliases (see scripts/render-lib/aggregator/order.ts → FILENAME_ALIASES); the depth floor below applies to whichever filename is present in a given run folder.",
      "synthesis-summary.md": 260,
      "forward-indicators.md": 225,
      "scenario-analysis.md": 245,
      "comparative-international.md": 165,
      "methodology-reflection.md": 240,
      "executive-brief.md": 170,
      "intelligence-assessment.md": 200,
      "devils-advocate.md": 165,
      "election-2026-analysis.md": 180,
      "election-cycle-analysis.md": 180
    },
    "year-ahead": {
      "_comment": "Year-ahead floors = month-ahead × 1.33 (proportional to multiplier 2.0 / 1.5), rounded down to nearest 5. Mandates 4 scenarios + 5 wildcards + PESTLE; cross-horizon citation of quarter-ahead × 2 + monthly-review × 4. Note: election-2026-analysis.md and election-cycle-analysis.md are filename-variant aliases.",
      "synthesis-summary.md": 305,
      "forward-indicators.md": 265,
      "scenario-analysis.md": 290,
      "comparative-international.md": 195,
      "pestle-analysis.md": 200,
      "wildcards-blackswans.md": 200,
      "quantitative-swot.md": 180,
      "intelligence-assessment.md": 240,
      "devils-advocate.md": 195,
      "methodology-reflection.md": 285,
      "executive-brief.md": 200,
      "election-2026-analysis.md": 225,
      "election-cycle-analysis.md": 225
    },
    "election-cycle": {
      "_comment": "Election-cycle floors = month-ahead × 1.66 (proportional to multiplier 2.5 / 1.5), rounded down to nearest 5. Mandates 4 scenarios × 3 governing-coalition branches + 5 wildcards + 3 counterfactuals + PESTLE + STRIDE; cross-horizon citation of year-ahead × 2 + monthly-review × 12. Also blocks on cycle-trajectory.md (24th artifact for this type only). Note: election-2026-analysis.md and election-cycle-analysis.md are filename-variant aliases.",
      "synthesis-summary.md": 380,
      "forward-indicators.md": 330,
      "scenario-analysis.md": 360,
      "comparative-international.md": 245,
      "pestle-analysis.md": 250,
      "wildcards-blackswans.md": 250,
      "quantitative-swot.md": 225,
      "political-stride-assessment.md": 250,
      "cycle-trajectory.md": 280,
      "intelligence-assessment.md": 295,
      "devils-advocate.md": 240,
      "methodology-reflection.md": 350,
      "executive-brief.md": 250,
      "election-2026-analysis.md": 275,
      "election-cycle-analysis.md": 275,
      "coalition-mathematics.md": 230,
      "historical-parallels.md": 215
    },
    "weekly-review": {
      "synthesis-summary.md": 240,
      "cross-session-intelligence.md": 180,
      "session-baseline.md": 170,
      "cross-run-diff.md": 160,
      "media-framing-analysis.md": 220,
      "methodology-reflection.md": 220,
      "executive-brief.md": 180
    },
    "monthly-review": {
      "synthesis-summary.md": 260,
      "cross-session-intelligence.md": 220,
      "session-baseline.md": 200,
      "cross-run-diff.md": 180,
      "comparative-international.md": 180,
      "media-framing-analysis.md": 240,
      "methodology-reflection.md": 235,
      "executive-brief.md": 200
    },
    "analyticalSupplementary": {
      "_comment": "Analytical supplementary templates — never blocking. Floors apply only when the template is produced. Sister-methodology: analytical-supplementary-methodology.md.",
      "pestle-analysis.md": {
        "standard": 100,
        "deep": 150,
        "comprehensive": 220,
        "rowFloor": {
          "dimensionTables": 4,
          "crossInteractions": 3
        }
      },
      "political-stride-assessment.md": {
        "standard": 110,
        "deep": 160,
        "comprehensive": 240,
        "rowFloor": {
          "stridePerDimension": 3,
          "attackTrees": 2,
          "ttpRows": 5
        }
      },
      "wildcards-blackswans.md": {
        "standard": 110,
        "deep": 160,
        "comprehensive": 240,
        "rowFloor": {
          "wildcards": 8,
          "blackSwans": 3,
          "consequenceTrees": 2,
          "resilienceDimensions": 5
        }
      },
      "quantitative-swot.md": {
        "standard": 110,
        "deep": 160,
        "comprehensive": 240,
        "rowFloor": {
          "perQuadrant": 3,
          "sensitivityRuns": 3,
          "towsActionsPerQuadrant": 1
        }
      }
    }
  },
  "tierCMultipliers": {
    "description": "Tier-C aggregation workflows apply these period-scope multipliers on top of the base thresholds. See ext/tier-c-aggregation.md.",
    "week-aggregation": 1.15,
    "month-aggregation": 1.35,
    "quarter-aggregation": 1.6
  },
  "enforcement": {
    "gateChecks": [
      "Check 3 in .github/prompts/05-analysis-gate.md consumes thresholds[articleType][artifact] with fallback to defaults.coreArtifactFloor.",
      "A file below floor MUST be returned to Pass 2 via the phase-04-pass2 checkpoint.",
      "Family-E per-document files use defaults.perDocumentFloorByTier keyed by the file's DIW tier tag in significance-scoring.md.",
      "Supplementary artifacts (S1–S7) use defaults.supplementaryFloor unless listed explicitly per articleType."
    ],
    "waivers": "Thresholds may be waived only by the CEO (document owner) via a merged PR updating this file. Ad-hoc waivers in commit messages are rejected."
  },
  "aiFirst": {
    "description": "AI-FIRST quality thresholds (per .github/copilot-instructions.md §5 and per-file AI-FIRST Methodology Cards in analysis/methodologies/*.md). These are ADDITIVE quality signals; the gate's blocking checks are 1–11 in 05-analysis-gate.md. The thresholds in this block are consumed by the Pass-2 self-audit documented in methodology-reflection.md.",
    "minimumIterations": 2,
    "passDiscipline": {
      "pass1": "Creation pass — produce minimal viable artifact; satisfy template structure + at least one evidence anchor per claim.",
      "pass2": "Read-back & improve pass — eliminate banned phrases, add second-order effects / cui-bono / counterfactuals, tighten WEP, strengthen Mermaid theming, raise citation density."
    },
    "citationDensity": {
      "default": {
        "anchorsPerWords": 120,
        "rule": "≥ 1 evidence anchor (dok_id, vote id, named MP, primary-source URL host) per ~120 words of analytical claim."
      },
      "perArticle": {
        "_note": "Per-article-type density thresholds for the aggregated article.md (words-per-anchor maximum). These are less strict than per-artifact thresholds because articles contain structural text (Reader Intelligence Guide tables, headings) that dilute density.",
        "breaking": 180,
        "evening-analysis": 180,
        "realtime-monitor": 200,
        "propositions": 160,
        "motions": 170,
        "committee-reports": 170,
        "interpellations": 180,
        "week-ahead": 200,
        "month-ahead": 180,
        "quarter-ahead": 180,
        "year-ahead": 180,
        "election-cycle": 170,
        "weekly-review": 200,
        "monthly-review": 190
      },
      "perArtifact": {
        "synthesis-summary.md": 100,
        "swot-analysis.md": 80,
        "risk-assessment.md": 100,
        "threat-analysis.md": 100,
        "intelligence-assessment.md": 110,
        "scenario-analysis.md": 110,
        "devils-advocate.md": 110,
        "executive-brief.md": 90,
        "documents/{dok_id}-analysis.md": 80,
        "comparative-international.md": 130,
        "media-framing-analysis.md": 130,
        "forward-indicators.md": 90,
        "coalition-mathematics.md": 90,
        "implementation-feasibility.md": 110
      },
      "_note": "Values are 'words per anchor' — lower = denser citation. SWOT and per-document files have the densest target because every quadrant entry / bullet must be evidence-backed. Comparator and framing artifacts can run thinner because their structure is table-driven."
    },
    "bannedPhrases": {
      "source": "analysis/methodologies/political-style-guide.json",
      "sourceMarkdown": "analysis/methodologies/political-style-guide.md",
      "block": "BEGIN BANNED-PHRASES … END BANNED-PHRASES",
      "matchMode": "literal-substring-case-insensitive",
      "extractionRule": "Consumers MUST use the canonical JSON file (political-style-guide.json) for automated enforcement. The JSON `allPhrases` array contains the flat list of literal banned substrings. Case-insensitive matching is the default.",
      "enforcement": "validate-article.ts Check 7 (banned-phrase-detected) enforces this as a blocking CI check. The agent's Pass-2 self-audit also enforces it at authoring time.",
      "removalPolicy": "Append-only between major versions. Removals require a major-version bump and a justification entry in methodology-reflection.md."
    },
    "pass2Attestation": {
      "rule": "Every Family A / B / C / D artifact (except data-download-manifest.md) MUST be re-read after creation and improved. The gate enforces this via mtime diff (final mtime > creation + 3 min) per Check 6 in 05-analysis-gate.md.",
      "attestationLocation": "methodology-reflection.md §Pass-2 audit summary"
    },
    "wepLanguageCeiling": {
      "horizon-72h": "very likely / very unlikely",
      "horizon-week": "likely / unlikely",
      "horizon-month": "likely / unlikely",
      "horizon-quarter": "roughly even / about even",
      "horizon-year": "roughly even (likely allowed only with ≥ 3 cycle-aged sources)",
      "horizon-cycle": "roughly even (likely allowed only with ≥ 3 cycle-aged sources)",
      "horizon-election": "scenario-driven; coalition outcomes never above 'likely'",
      "_source": ".github/prompts/ext/long-horizon-forecasting.md"
    },
    "exemplarPolicy": {
      "rule": "Every methodology file in analysis/methodologies/ MUST carry an Exemplar (good — pattern-match this) and an Anti-exemplar (failure mode — never ship this) snippet ≤ 25 lines each, inside the AI-FIRST Methodology Card.",
      "validatedBy": "methodology-reflection.md §Style audit and (optionally) a markdown-lint sanity script."
    }
  },
  "meta": {
    "version": "1.5",
    "lastUpdated": "2026-05-03",
    "changelog": [
      "v1.5 (2026-05-03) — Added aiFirst block (citationDensity, bannedPhrases pointer to political-style-guide.md, pass2Attestation rule, wepLanguageCeiling per horizon, exemplarPolicy). Pairs with the AI-FIRST Methodology Card inserted into every analysis/methodologies/*.md file.",
      "v1.4 (2026-05-01) — Added quarter-ahead / year-ahead / election-cycle blocks for the long-horizon-forecast workflows.",
      "v1.1 (2026-04-23) — Added analyticalSupplementary section (pestle-analysis · political-stride-assessment · wildcards-blackswans · quantitative-swot); non-blocking."
    ]
  }
}