devils-advocate.md

May 15, 2026 · View on GitHub

😈 Devil's Advocate Template

📊 Systematic Red-Team Challenge to the Day's Analytical Consensus
🎯 ACH · Counter-Evidence · Alternative Hypotheses · Assumption Audit

📋 Document Owner: CEO | 📄 Version: 1.2 | 📅 Last Updated: 2026-04-25 (UTC) 🏢 Owner: Hack23 AB (Org.nr 5595347807) | 🏷️ Classification: Public

📌 Template instructions: Produce on every run. Save as analysis/daily/${ARTICLE_DATE}/${DOC_TYPE}/devils-advocate.md. Scope depth by DIW: for lower-DIW runs, provide a concise but explicit challenge to the main assessment; for higher-DIW runs, provide full ACH, multiple alternative hypotheses, assumption audit, counter-evidence, and falsifiable indicators. Pairs with political-risk-methodology.md and intelligence-analysis-techniques.

✨ What to produce: An honest, structured challenge to the main assessment. Apply Analysis of Competing Hypotheses (ACH), surface at least three alternative explanations, audit the assumptions, list falsifiable predictions, and state what evidence would change the judgement.

📐 Template Contract — every fill of this template MUST satisfy this row.

Slot Value
Owning methodology per-artifact-methodologies.md
Owning gate check Check 7 (Family C — ≥ 3 ACH hypotheses) — see 05-analysis-gate.md
Required inputs synthesis-summary.md, sibling Family A
Horizon band per-run (per scripts/horizon-context.ts)
Output family Family C — Strategic Extensions
Aggregation order 25 of 30 in canonical order (see scripts/render-lib/aggregator/order.ts)
Reader Intelligence Guide row generated from devils-advocate.md (see scripts/render-lib/aggregator/reader-guide.ts)
Canonical evidence anchor | claim | evidence (dok_id / vote / MP intressent_id / primary-source URL) | retrieved_at | confidence | — every analytical claim row uses this schema.

Cross-reference: README.md §Template ↔ Methodology ↔ Gate-Check Matrix.

Slot	Value
Owning methodology	`per-artifact-methodologies.md`
Owning gate check	Check 7 (Family C — ≥ 3 ACH hypotheses) — see `05-analysis-gate.md`
Required inputs	`synthesis-summary.md`, sibling Family A
Horizon band	per-run (per `scripts/horizon-context.ts`)
Output family	Family C — Strategic Extensions
Aggregation order	25 of 30 in canonical order (see `scripts/render-lib/aggregator/order.ts`)
Reader Intelligence Guide	row generated from `devils-advocate.md` (see `scripts/render-lib/aggregator/reader-guide.ts`)
Canonical evidence anchor	`\| claim \| evidence (dok_id / vote / MP intressent_id / primary-source URL) \| retrieved_at \| confidence \|` — every analytical claim row uses this schema.

🔄 Tradecraft Context

Element	Value
F3EAD Stage	ANALYZE — challenge prevailing view through structured contrarian analysis
PIRs Served	Serves same PIRs as target assessment; applies adversarial rigor to prevent groupthink
Admiralty Floor	[B2] for alternative-hypothesis evidence; challenge logic even if evidence is same as main assessment
WEP + ODNI	Alternative hypotheses state probability vs. main hypothesis; use WEP comparison ("Main: likely 70%; Alternative A: unlikely 25%")
Source Diversity Floor	P1 (alternative hypotheses): ≥2 independent evidence sources per hypothesis; re-interpret existing evidence from different angle
SAT(s) Applied	ACH (Analysis of Competing Hypotheses), Devil's Advocacy, Red Team Analysis, Key Assumptions Check
ICD 203 Standards	3 (judgments vs assumptions), 4 (alternative analysis — PRIMARY), 7 (explain consistency / change)

📋 Challenge Context

Field	Value
Devil's-Advocate ID	`DEV-YYYY-MM-DD-NNN`
Generated	`YYYY-MM-DD HH:MM UTC`
Target assessment	`e.g., "FiU48 signals coalition discipline before September 2026 election"`
Source file	`synthesis-summary.md §Finding 1`
Confidence of main claim	`🟩 HIGH`
Post-challenge confidence	`🟧 MEDIUM (downgraded) / 🟩 HIGH (confirmed) / 🟦 VERY HIGH (strengthened)`

🎯 Key Judgment Coverage Matrix (Required)

[REQUIRED] Before writing hypotheses, list every KJ from intelligence-assessment.md and map it to at least one challenge row. Gate expectation: 100% KJ coverage — every row must end with ✅ (no ❌). See analysis/methodologies/admiralty-rubric.md.

KJ ID	KJ summary	Challenged by hypothesis ID(s)	Challenge present?
KJ-1	`[REQUIRED: summary from intelligence-assessment.md]`	`H1 / H2 / H3`	✅ / ❌
KJ-2	`[REQUIRED]`	`H#`	✅ / ❌
KJ-3	`[REQUIRED]`	`H#`	✅ / ❌
...	...	...	...
Coverage	—	—	N / N (must be 100%)

Gate enforcement (Check 7b): the analysis gate extracts the unique set of KJ-N IDs from intelligence-assessment.md and verifies each appears on a non-❌ row inside this ## Key Judgment Coverage Matrix section. Missing KJs or any ❌ row blocks the commit.

🧪 Analysis of Competing Hypotheses (ACH)

Mandatory requirement (ICD 203 §4; gate Check 7b): ACH MUST challenge every Key Judgment (KJ) declared in the sibling intelligence-assessment.md. Coverage is documented above in the ## 🎯 Key Judgment Coverage Matrix section — gate enforcement reads that section and verifies each KJ from intelligence-assessment.md has a non-❌ row there.

List each hypothesis, then score each piece of evidence as C (consistent), I (inconsistent), or N (neutral). The hypothesis with the fewest inconsistencies survives. Each hypothesis MUST correspond to a KJ challenge (see Coverage Matrix above). WEP probability and base-rate prior MUST be stated for each hypothesis; citing only "analyst judgement" without a base-rate source is a banned pattern (flagged by banned-phrase scanner; see analysis/methodologies/base-rates/ for calibration datasets).

graph LR
    E1["📎 Evidence E1<br/>Unanimous FiU vote"] --> H1["🅐 Coalition discipline<br/>(main)"]
    E1 --> H2["🅑 Pre-election ritual<br/>(alternative)"]
    E1 --> H3["🅒 SD pivotal — not discipline<br/>(alternative)"]
    E2["📎 Evidence E2<br/>SD fuel-tax statement"] --> H3
    E3["📎 Evidence E3<br/>EU Commission silence"] --> H1
    E3 --> H2

    style H1 fill:#4CAF50,color:#FFFFFF
    style H2 fill:#FFC107,color:#000000
    style H3 fill:#FF9800,color:#FFFFFF

Evidence	H1: Coalition discipline	H2: Pre-election ritual	H3: SD-pivotal, not discipline
E1 — Unanimous FiU vote 2026-04-21	C	C	C
E2 — SD lead-spokesperson public support	N	N	C
E3 — EU-Commission silence to date	C	C	N
E4 — Internal L reservation on proportionality (leaked)	I	N	N
E5 — Prior coalition discipline metric 88.5 %	C	N	N
Inconsistent count	1	0	0
Consistent count	3	2	2

ACH verdict: H2 and H3 both survive with zero inconsistencies. H1 is plausible but carries one inconsistency (L reservation). The assessment should therefore qualify "discipline" with "conditional on SD posture" (H3) and "pre-election timing" (H2).

🔄 Alternative Hypotheses (minimum 3)

Alternative 1 — Pre-election Ritual

Claim: Coalition unity reflects electoral timing, not durable agreement.
Evidence for: Fuel-tax cut expires before Q4 2026 budget; electoral pressure peaks July–Aug 2026.
Evidence against: Coalition also united on NATO and justice packages this riksmöte.
Implication if true: Unity may fracture post-election regardless of result.

Alternative 2 — SD-Pivotal Compromise

Claim: Coalition cohesion is purchased through SD policy concessions; discipline is external.
Evidence for: Fuel-tax cut + justice-package composition match SD priorities.
Evidence against: Wind-power revenue law is inconsistent with SD's historic climate position.
Implication if true: Coalition-Mathematics risk of SD withdrawal rises after September 2026.

Alternative 3 — Signalling Under Duress

Claim: Unity signals internal weakness the government fears will leak.
Evidence for: Unusually low dissent rate in a pre-election quarter.
Evidence against: Budget-continuity pattern consistent with historical incumbency behaviour.
Implication if true: Expect defensive posture on scandals; reduced appetite for fresh initiatives.

🧰 Assumption Audit

#	Assumption	Source	Status	Vulnerability
A1	"Voting discipline = coalition durability"	Conventional coalition theory	🟡 Contestable	Discipline can be purchased or coerced
A2	"Pre-election polling gap closes by September"	Historical Swedish election cycles	🟢 Supported	Depends on economy and crises
A3	"EU Commission silence = tacit approval"	Diplomatic pattern	🟡 Contestable	Silence often precedes formal probe
A4	"SD maintains confidence & supply"	2024–2026 record	🟡 Contestable	Policy conflicts on climate can trigger withdrawal

🧮 Base-Rate Check

Question	Base rate	Source	Implication
How often do coalition governments survive a pre-election quarter without dissent?	~55 % since 2000	Swedish coalition record	Current unity is above average but not unprecedented
How often does the EU Commission open a fuel-tax review within 6 months?	~30 %	EU Commission state-aid archive	Non-trivial downside risk
How often do incumbent governments win the following election from a 4-pt polling deficit?	~20 %	Swedish electoral history	Retention probability is lower than current polling alone suggests

🎯 Falsifiable Predictions

#	Prediction	By when	What would falsify it
P1	Coalition holds unified on FiU48 chamber vote	2026-04-24	Any coalition-party `Avstår` or `Nej`
P2	EU Commission issues no state-aid letter within 60 days	2026-06-21	Any formal notification to Sweden
P3	Government approval gap closes ≤ 3 pt by August 2026 SIFO	2026-08-31	Gap widens or stays ≥ 5 pt
P4	SD maintains confidence-and-supply posture through election	2026-09-13	SD withdrawal / defection signal

🧭 What Would Change the Assessment

Trigger	Resulting change
Any one of P1–P4 falsifies	Downgrade main claim from 🟩 HIGH to 🟧 MEDIUM
Any two falsify	Downgrade to 🟥 LOW and rewrite `synthesis-summary.md §Finding 1`
All four hold through election	Upgrade to 🟦 VERY HIGH

🚨 Cognitive-Bias Checklist

Bias	Exposure	Mitigation applied
Confirmation bias (toward government narrative)	⚠️	Alt 1 and Alt 2 explicitly considered
Recency bias (over-weighting today's vote)	⚠️	Base-rate check included
Availability bias (media framing)	✅	Cross-checked with `media-framing-analysis.md`
Anchoring (to prior confidence level)	⚠️	Re-scored from evidence this run
Groupthink	⚠️	ACH forces comparison of hypotheses

📎 Links

Link	Path
Main assessment being challenged	`synthesis-summary.md`
Risk register	`risk-assessment.md`
Media framing (for availability check)	`media-framing-analysis.md`
Methodology	intelligence-analysis-techniques SKILL

Document Control

Template path: /analysis/templates/devils-advocate.md
Referenced by: ai-driven-analysis-guide.md § Step 5
Classification: Public
Next Review: 2026-07-21

✅ Pass-2 Self-Audit Checklist (v4.4 — required)

Purpose: AI-FIRST principle requires a Pass-2 read-back-and-improve. After producing this artifact in Pass 1, re-read it end-to-end and verify each item below. Document any remediation in methodology-reflection.md §"Pass-2 audit log". Any unchecked ❌ box at the end of Pass 2 forces a Pass-3 rewrite of the affected section.