R0052 — Methodology Claims¶
Mode: Claim · Status: Active · Tags: methodology, intelligence-community, scientific-method
Input¶
- ICD 203 defines nine tradecraft standards that govern how intelligence analysts produce assessments.
- No prior work in published, accessible literature has systematically combined intelligence community analytical standards with scientific methodology frameworks into a single unified research methodology.
- GRADE separates the quality of evidence from the strength of conclusions drawn from it — these are independent axes that must be scored separately.
- The IPCC uses a two-axis confidence model: evidence quality (Limited/Medium/Robust) and source agreement (Low/Medium/High).
- As early as 1987, Mulrow documented that none of the 50 reviews she examined met all eight basic scientific reporting criteria.
- CONSORT 2010 was a 25-item checklist; CONSORT 2025 expanded to 30 items.
- Chamberlin first published "The Method of Multiple Working Hypotheses" in 1890 (revised 1897). Platt published "Strong Inference" in 1964, explicitly citing Chamberlin's work.
- Platt deliberately numbered his final step "1'" (one-prime, not four) to signal that it's a loop, not a sequence.
- ICD 203's probability scale defines seven points with dual terminology and explicit numeric ranges, capping at "Almost Certain" (95-99%) — never reaching 100%.
- The NAS published 21 standards with 82 elements of performance organized across four stages of review.
- The Wardle and Derakhshan Information Disorder Taxonomy classifies information failure along two dimensions — falseness of content and intent to harm — producing three categories: misinformation, disinformation, and malinformation.
- Journalism is principles-based, not methodology-based — no journalistic framework has a hierarchical evidence quality scale, calibrated uncertainty language, structured bias assessment domains, or source reliability tiering.
- Different domains use different terms for the same phenomenon, and single-term searches create systematic blind spots when searching across disciplines.
- The process self-audit (ROBIS) catches process errors but not interpretation errors — an agent can follow every step correctly and still mischaracterize what a source says.
Runs¶
2026-03-31 — Initial run
Mode: Claim · Claims: 14 · Prompt: ai-research-methodology v1 research.md · Model: unknown
First research run for methodology claims verification.
2026-03-31-02 — Rerun 02
Mode: Claim · Claims: 14 · Prompt: ai-research-methodology v1 research.md · Model: unknown
Second research run for reproducibility comparison.
2026-03-31-03 — Rerun 03
Mode: Claim · Claims: 14 · Prompt: ai-research-methodology v1 research.md · Model: claude-opus-4-6
Third research run. 10 claims almost certain, 4 very likely. 344 files produced.