Research paperWorking paperARR-2026-7S4DYK6VRE8A9RF8 · v1 · 2026-11-19

Weight sensitivity of leadership in the CAIDP AI Index

Lluis Eriksson

Citable version that has not completed AIRR screening. Not peer reviewed. Screening record

Abstract

Canada and Japan share the highest numerical score in both the 2025 and 2026 CAIDP AI Index. We ask how much the twelve item weights must differ before another country can tie them. The exact largest-to-smallest weight ratio falls from 3 to 1.5. The same result holds among the 80 countries present in both editions. Replacing only the leaders' response to the UNESCO implementation item reproduces the fall under either vintage of the other countries' responses. Replacing the other countries' responses alone changes the set of potential entrants but leaves the critical ratio unchanged. The comparison describes the geometry of the published evaluations.

Preview PDF on this page v1

Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.

Reference this work

Cite this paper

v1 · 2026-11-19

Lluis Eriksson (2026). Weight sensitivity of leadership in the CAIDP AI Index. AIRR.SCIENCE, ARR-2026-7S4DYK6VRE8A9RF8, v1. https://airr.science/papers/ARR-2026-7S4DYK6VRE8A9RF8/versions/v1/ Research preprint; ARR-2026-7S4DYK6VRE8A9RF8 v1. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.

Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.

Version history

The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.

  • v1 · source snapshot available · viewing

Verification record

Frontier-model screening
Not assessed
Source integrity
Pass
Bibliographic integrity
Partial
Reproducibility
Partial
Lean 4
Not applicable

Recorded under ARR-VERIFY-1.0. AIRR verification and screening are not peer review.

AI assistance statement

OpenAI Codex participated in analysis, writing and the first review of the exact PDF. Its shared-context assessment is 4.05/10. Report SHA-256 ad78c4aa3b917bbe25c2da6ed2e821ab4e5b8a414051e9bd56852e32b6c50204. The author administers AIRR. Independent peer review has not been documented.

Frontier-model screening

Status: not_assessed. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.

    Longitudinal frontier-model record

    Model assessments and review context

    Read the scale and limits
    4.05 / 10.00Strong
    Distinct models
    1
    Range
    4.05–4.05
    Method
    bounded capability-weighted mean · exact version only

    Limited evidence · 0/1 benchmark matches. This is an evidence description, not a probability of correctness.

    Only one distinct model; One provider; shared blind spots possible; Fresh blinded review context not documented for every model; Mathematical benchmark not applied to this subject; Adverse reports or unresolved objections require editorial disposition

    Score calculation and evidence

    Unweighted mean: 4.05. Policy: AIRR-RATING-1.0. Method and limits.

    ModelOriginal scoreWeightBasis
    OpenAI · OpenAI Codex, GPT-6 family (exact serving revision unavailable)4.051.000Benchmark outside declared mathematical scope

    Download the calculation snapshot

    OpenAI · OpenAI Codex, GPT-6 family (exact serving revision unavailable) · involved in manuscript · review context: shared4.05

    Recommendation: Minor Revision · Material objections: 0

    A concise applied robustness study. The leaders, tie thresholds and boundary entrants are defined precisely; a break in the underlying UNESCO scoring guidance makes the cross-edition comparison substantively informative. I inspected all three pages of the exact PDF and participated in construction and testing in the same context. The analysis supports the deterministic score claim on the specified data. It leaves country-level attribution open.

    Correctness confidence
    The box maximization is attained coordinatewise and the two leader vectors are equal. A half-point gain on each of two entries and three half-point losses give the 2026 Dutch witness. The four crossed cells reproduce.
    Rigor
    Distinguishes the full population, shared cohort, interior partial credits and the tie boundary. The grading paragraphs support a methodological break but do not attribute country decisions.
    Novelty
    Sensitivity to criteria weights is longstanding. The exact comparison for these two policy index editions, with the crossed table and changed Q11 guidance, is the contribution.
    Significance
    Shows that one coding change can halve a headline robustness statistic and that expanding the country set did not cause it. Applicability beyond the index and relevance to actual policy decisions are open.
    Reproducibility
    A pinned map digest, rational calculation and witnesses enable replication, but loss of the exact publisher bundle would impair fresh runs. The public script does not itself reproduce the printed-grid or rubric audit.

    Strengths

    • The constraint set is explicit, and the boundary has an elementary proof plus an attaining witness.
    • Crossing leader and outsider vectors separates effects on the threshold from effects on entrant identity.
    • The duplicate leader vectors justify using one comparison instead of incorrectly maximizing each pair separately.

    Weaknesses

    • The public script reproduces thresholds but omits the extraction and comparison of the source PDF tables and the check of the variable-partial-credit endpoint.
    • The checksum is useful for identifying the data but cannot replace continued availability of the exact map.
    • A short reference to the country-specific discussions and scoring evidence would help readers examine individual attributions.

    Potential errors

    • “Independently confirm” could imply independent evidence. Both grids and maps are publications of the same organization. They provide two representations of its ratings.
    • An overtake of the two leaders beyond the entry boundary need not imply sole overall leadership at every larger ratio. The claim on page 2 is specifically about the pair.

    Strong novelty candidates

    • The exact 3 versus 3/2 sensitivity boundary and the Q11 crossed decomposition on the common 80-country cohort.

    Unresolved material objections

    None reported.

    arr:assessment:c19cd297-6ec4-4b23-abab-a1419857ad08 · prompt ARR-ASSESS-1.0 · response SHA-256 6ffbb681815b32d9511c5ab24cc0796b9c70cea6b981fff6391ddd4d081c8e65 · canonical PDF SHA-256 a4c4116b0775c11175ee319bd519f6b24651c77b3845d798d4957206912ce31a

    A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.

    Editorial disclosure

    Working-paper deposit under exact-hash permission. Collection screening and independent peer review remain outstanding.

    Record timestamps & download statistics
    Deposit recordedFirst repository commit · full Git SHA recorded
    PublishedGitHub release · tag ARR-2026-7S4DYK6VRE8A9RF8-v1
    Canonical PDF downloads2GitHub release downloads only; direct AIRR PDF reads are not included
    Page viewsNot availablePage-view counts are not available in this public snapshot. Optional operator statistics are separate.
    Definitions and rankings →