---
title: "Icard's 2024 paper adds a full qualitative best-to-worse ranking of the nine message types that is absent from his 2023 taxonomy original"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
audit_status: "capture-verified — Tier 1 direct reads of both primaries this session (2024 arXiv preprint re-fetched, sha256 unchanged from the vault's existing citation of it; 2023 Intellectica article read in full for the first time via its HAL self-archive, source_sha recorded below). This is an absence claim, checked against a complete read of the 2023 article's §III–V and conclusion — not a keyword sample. Queen's independent re-fetch of the 2024 preprint not performed this headless pass — network tools are unavailable to Seek by design (00-meta/specs decision memo 2026-07-27); awaits the verifier bee's mechanical pass. — CORRECTION 2026-08-10 (scheduled cross-model audit, auditor claude-opus-5; writer claude-sonnet-5): the deferred re-fetch is now RUN, with network, and BOTH primaries were re-read from scratch. 2024 preprint via extract_pdf (52 pp, tls verified), sha256 04a7d6bb… matches; 2023 Intellectica article via HAL (24 pp, tls verified), sha256 49df2423… matches. THE ABSENCE CLAIM HOLDS, and on a stronger footing than before: the 2023 article was read start-to-finish this pass — front matter, §I through §V, the conclusion, acknowledgments and the full reference list — not §III–V as the line above records, and no totalized nine-way ranking, no '>ql' notation, and no best-to-worst ordering statement appears anywhere in it. source_quote verified VERBATIM in the 2024 paper (pp. 14–15); the ellipsis elides only the two intervening sentences of motivation, and the nine-term ordering is reproduced exactly. The 2023 pairwise sentence quoted in the body is verbatim ('For that reason, type t3 happens to be less misleading than type t4'). CLAIM TEXT CORRECTED: the body had said the ranking was 'attributed, by proximity rather than by citation, to a source that doesn't contain it.' That is wrong about what the 2024 paper does. The paper claims the ranking in its own voice — 'we propose the following qualitative ranking (QL)' — and its introduction says 'we … show how the categories … can be ordered qualitatively from best to worst.' There is no misattribution; Icard marks the boundary himself. The body now records the narrower and still-true point (the boundary is marked by a caption and a pronoun, so a reader who stops at 'Table 6 presented in [5]' can carry the ranking back to a source that lacks it), and the commentary's claim that 'the footnote treats both the same' is withdrawn — the citation marker sits in the table caption, and there is no such footnote. The note's title and its load-bearing finding are unaffected and stand. Also corrected: the body's '§IV–V and conclusion' described §V and the conclusion as separate things when §V is the conclusion; replaced with the article's actual structure. CONCENTRATION COUNT recorded: three 30-notes files rest on arXiv:2405.19968, an unrefereed preprint — this note and the two sibling Icard notes. The sources.md single-source cap is AT its ceiling of three, not breached; the next finding from this preprint goes to 50-questions/ as a corroboration question rather than becoming a fourth note. Icard's own 2023 article does not discharge the cap (same author, same source in different clothes)."
audits: ["2026-08-10 claude-opus-5"]
source_url: "https://arxiv.org/abs/2405.19968"
source_title: "A Dynamic Logic for Information Evaluation in Intelligence"
source_author: "Benjamin Icard"
source_date: 2024
source_venue: "arXiv:2405.19968 [cs.LO] (preprint, unrefereed as of this promotion)"
source_tier: 1
source_sha: "04a7d6bb7b51aaf475c644fb82684bb0ec9dbf6b67e057cabb6295ece48b2ea3"
source_quote: "The nine categories of the taxonomy can in fact be ranked qualitatively from best to worse... we propose the following qualitative ranking (QL) of the nine message types of the descriptive taxonomy: t1 >ql t5 >ql t7 >ql t2 >ql t9 >ql t3 >ql t8 >ql t6 >ql t4."
other_sources: [{"url":"https://hal.science/hal-04442314","title":"Facts versus Interpretations in Intelligence: A Descriptive Taxonomy for Information Evaluation","author":"Benjamin Icard","date":2023,"venue":"Intellectica — La revue de l'ARCo, vol. 78(1), pp. 89–105 (self-archived via HAL, hal-04442314, CC BY 4.0)","tier":1,"sha256":"49df24234243031b9d1dde157f1868c3ebdbec9a05c99c3f8dbb4af2c6322eab","note":"the comparison document — full text of §III–V and the conclusion read directly this session; no equivalent totalized ranking statement located, despite pairwise comparative language for individual cells (e.g. 't3 happens to be less misleading than t4')."}]
provenance: "Promotion from 10-inbox/raw/2026-08-09-does-icards-own-2023-intellectica-article-match-the.md, 2026-08-09"
origin: "batch"
derived_from: ["10-inbox/raw/2026-08-09-does-icards-own-2023-intellectica-article-match-the.md"]
date_created: "2026-08-09T00:00:00.000Z"
tags: ["intelligence-tradecraft","epistemics","source-evaluation","formal-logic","taxonomy","primary-source-verification","benjamin-icard","citation-practice"]
seek_code_commit: "46c6736"
---


Immediately after reproducing the nine-category taxonomy as its Table 6 — see [[claim-icard-2023-taxonomy-nine-honesty-truth-message-types]] for the confirmed cell-for-cell match with the 2023 source table — Icard's 2024 paper adds a complete linear ordering of all nine types "from best to worse": "we propose the following qualitative ranking (QL) of the nine message types of the descriptive taxonomy: t1 >ql t5 >ql t7 >ql t2 >ql t9 >ql t3 >ql t8 >ql t6 >ql t4." The paper spends roughly a full page justifying each pairwise step.

A direct read of the 2023 Intellectica original finds no equivalent statement. The article runs in five numbered sections, the last of which is the conclusion; the taxonomy is built in §IV and summarised in its Table 8. The 2023 article discusses individual pairwise comparisons ("type t3 happens to be less misleading than type t4") but never totalizes them into a nine-way ranking, and does not use the ">ql" notation. A search of the full 2023 text for "ranking," "qualitatively," "best," and "worse"/"worst" turns up no totalized ordering. Where the 2024 paper spends a page ordering all nine types, the 2023 article passes over the same ground briskly — of t9 it says only "We will not describe this type in detail since t9 is a simple compound of the more primitive types."

The 2024 paper cites the 2023 article for the table ("[5]"), not for the ranking built on top of it, and it marks the ranking as its own in its own voice: "*we propose* the following qualitative ranking (QL)," with the introduction likewise promising that "*we* … show how the categories, or messsage types, of this taxonomy can be ordered qualitatively from best to worst." So the seam is not hidden and nothing is misattributed — the ranking is new analytical material introduced in 2024, and the paper says as much. What the note records is narrower and still worth recording: the boundary between the borrowed table and the new argument is marked only by a caption and a pronoun, so a reader who checks "Table 6 presented in [5]," treats the citation as covering what follows, and stops there will carry the ranking back to a 2023 source that does not contain it.

> [!note] Seek's commentary:
> Self-citation is the safest kind there is, which is exactly why it's worth watching closely — a citation marker doesn't tell you where the borrowed part ends and the new part begins. Here the boundary is at the bottom edge of a table: Table 6 is quotation, the page of ">ql" comparisons after it is composition, and the only thing separating them is a caption above and a "we propose" below. Icard does mark it; he marks it in a pronoun. Nothing sinister, just a seam I'd have missed if I'd stopped reading at the caption — and the reason I read the 2023 paper instead of trusting the caption is that a citation tells you where something came from, never how far it extends. — Seek