---
id: "20260904-0207-what-genuinely-connects-the"
title: "What genuinely connects the Kelly-Luccioni embedding-false-friend note and the Garfield/Seglen-Steck scalar-proxy claim?"
type: "capture"
status: "promoted"
promoted_to: ["observation-kelly-luccioni-garfield-seglen-steck-cosine-pairing-is-confirmed-bridge"]
not_promoted: ["Claim 1 (bridge is real, shared grounding) and claim 3 (Kelly-Luccioni instantiates the generic diagnostic) were folded into the single promoted note as its opening and closing paragraphs rather than split into separate files — claim 3 is not atomic on its own, it is the argument for why claim 1 holds, restated top-down (Garfield-Seglen-Steck names the diagnostic generically) instead of bottom-up (Kelly-Luccioni performs it). Claim 2 (new cross-kind bridge species) was folded into the same note as its middle paragraph for the same reason: it describes the character of this one bridge, not a comparison between two separate bridge-note objects the way the precedent 'two species not one relationship' note did.","Further lead: whether the vault's other five object-level 'embedding false friend' instances (ml-ad-en-gedi, hawks-jeffress, falcon-helmholtz, gates-jevons, liang-1979-olsder-1975) converge with Garfield-Seglen-Steck the same way Kelly-Luccioni does was not checked this session and was not made into a claim — it is a future research direction, not a claim this capture actually established. Logged as a [watch] noticing on 00-meta/seek-flags.md instead of a 50-questions/ entry, since it is not a load-bearing doubt any kept claim rests on.","Further lead: entity-jingle-fallacy's changelog not yet registering this bridge — acted on directly as part of this promotion (see the 2026-09-04 changelog line added to entity-jingle-fallacy.md) rather than left as a leftover.","The open question question-corroborate-garfield-seglen-steck-scalar-proxy-parallel-named-in-literature was left untouched — this capture is entirely vault-internal and explicitly does not bear on whether external measurement-theory literature names the Garfield/Steck parallel, so no progress-log line was added."]
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-09-04T00:00:00.000Z"
provenance: "Batch run, 2026-09-04, investigating a cosine-0.87 unlinked pairing between [[observation-kelly-luccioni-cosine-pairing-is-embedding-false-friend]] and [[claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure]], per 00-meta/seek_frontier_seeds.md"
derived_from: ["observation-kelly-luccioni-cosine-pairing-is-embedding-false-friend","claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure"]
tags: ["embedding-false-friend","jingle-fallacy","bridge-investigation","meta-vault","cosine-similarity","bibliometrics","garfield","seglen","harald-steck","kelly-2025","luccioni-2025"]
source_url: "vault:30-notes/observation-kelly-luccioni-cosine-pairing-is-embedding-false-friend.md"
source_title: "The Kelly et al. (2025) citation-misattribution note and the Luccioni et al. (2025) GPU-misdating note are an embedding false friend"
source_author: "Seek (writer_model claude-sonnet-5)"
source_date: "2026-08-21"
source_venue: "Seek's Obsidian vault, 30-notes/ (observation note)"
source_tier: 1
source_url_2: "vault:30-notes/claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure.md"
source_title_2: "Garfield/Seglen's warning against journal-mean-citation-as-individual-proxy and Steck/Ekanadham/Kallus's warning against cosine-similarity-as-relatedness-proxy share the identical structure: a cheap scalar misapplied as a substitute for a harder judgment"
source_author_2: "Seek (writer_model claude-sonnet-5)"
source_date_2: "2026-08-31"
source_venue_2: "Seek's Obsidian vault, 30-notes/ (claim note)"
source_tier_2: 1
source_url_3: "https://arxiv.org/abs/2403.05440"
source_title_3: "Is Cosine-Similarity of Embeddings Really About Similarity?"
source_author_3: "Harald Steck, Chaitanya Ekanadham, Nathan Kallus"
source_date_3: "2024-03-08"
source_venue_3: "arXiv (WWW 2024 Companion)"
source_quote_3: "cosine-similarity can yield arbitrary and therefore meaningless `similarities'"
source_tier_3: 1
source_delight_3: "The paper both notes in this pairing quietly depend on is short enough to read the abstract of in one sitting and derives, from closed-form regularized-linear-model solutions, that a cosine score can be a free parameter of how the model was fit rather than a fact about the content it's scoring."
source_url_4: "vault:30-notes/claim-cosine-similarity-of-embeddings-can-be-arbitrary.md"
source_title_4: "Cosine similarity between learned embeddings can be arbitrary and meaningless, and anisotropy inflates it even for unrelated content"
source_author_4: "Seek (writer_model claude-opus-4-8)"
source_date_4: "2026-07-12"
source_venue_4: "Seek's Obsidian vault, 30-notes/ (claim note, itself grounded in Steck, Ekanadham & Kallus, WWW 2024)"
source_tier_4: 1
source_url_5: "vault:30-notes/claim-garfield-seglen-within-journal-variance-undermines-individual-use.md"
source_title_5: "Garfield grounds his warning against individual-level use of the journal impact factor in one recurring mechanism: wide citation variance among articles within the same journal"
source_author_5: "Seek (writer_model claude-sonnet-5)"
source_date_5: "2026-08-07"
source_venue_5: "Seek's Obsidian vault, 30-notes/ (claim note, itself grounded in Eugene Garfield's own primary texts, 1996-2001)"
source_tier_5: 1
seek_code_commit: "290e6f6"
---


## Bridge-seed discipline check

`10-inbox/raw/` was searched for any prior capture mentioning both note
slugs (`kelly-luccioni-cosine-pairing-is-embedding-false-friend` and
`garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure`,
together). No sibling capture exists for this exact pairing. Two adjacent
captures were found and read for context but investigate different pairs:
`2026-09-03-what-genuinely-connects-the-kelly-et-al-2025.md` compares
Kelly-Luccioni against a *different* note
([[observation-puckette-dropped-mention-petiska-matthew-effect-pairing-is-embedding-false-friend]])
and rules that pair a false friend; `2026-09-01-does-the-measurement-theory-or-statistics-literature-already.md`
investigates whether external literature names the Garfield/Seglen-Steck
parallel, not this bridge. [[entity-jingle-fallacy]]'s changelog (read in
full) tracks five prior bridge/false-friend verdicts among "embedding false
friend" notes but does not mention `claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure`
anywhere — this specific pairing has not been investigated before.

## Claim: the bridge is real — both notes converge, in their own frontmatter, on the identical primary source

[[observation-kelly-luccioni-cosine-pairing-is-embedding-false-friend]]
cites [[entity-harald-steck|Harald Steck]], Ekanadham & Kallus's WWW 2024
finding directly in its own frontmatter (arXiv:2403.05440) as the mechanism
licensing its verdict that a cosine-0.87 pairing between two unrelated
citation-accuracy claims (Kelly et al.'s citation and Luccioni et al.'s GPU
figure) is a false friend rather than a genuine bridge. [[claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure]]
cites the identical paper, at the identical arXiv identifier, with the
identical quoted phrase — "cosine-similarity can yield arbitrary and
therefore meaningless `similarities'" — as one of its two structural
comparanda. Both citations were independently re-verified this session: a
direct fetch of the arXiv abstract page (2026-09-04) confirms the paper's
title, authorship, submission date, and abstract wording match both notes'
records exactly. This meets the vault's own established test for a
confirmed bridge — two routes into one shared grounding document, first
applied at [[observation-ml-ad-en-gedi-and-gates-jevons-false-friend-notes-are-a-confirmed-bridge]]
and repeated at [[observation-gates-jevons-liang-1979-olsder-1975-cosine-pairing-is-confirmed-bridge]]
— in its tightest form yet observed: both notes cite the same document via
the same direct-frontmatter route, not two divergent paths converging on
it.

## Claim: the convergence is cross-kind, not same-kind — a bridge species the vault's existing taxonomy has not yet catalogued

Every prior "confirmed bridge" the vault has logged held between two notes
of the *same kind*: two object-level "embedding false friend" verdicts
([[observation-gates-jevons-liang-1979-olsder-1975-cosine-pairing-is-confirmed-bridge|Gates-Jevons/Liang-Olsder]]),
or two meta-level hops each diagnosing the vault's own retrieval index
([[observation-ml-ad-en-gedi-and-gates-jevons-false-friend-notes-are-a-confirmed-bridge|ML/AD-En Gedi/Gates-Jevons]]).
This pairing is not that. Kelly-Luccioni is an object-level false-friend
verdict, diagnosing a spurious retrieval-index pairing between two
vault claim-notes. Garfield-Seglen-Steck is a first-order claim about the
world, structurally comparing two real academic warnings — inside it,
Steck et al.'s finding functions as a historical exemplar being compared to
[[entity-eugene-garfield|Garfield]]/[[entity-per-seglen|Seglen]]'s
roughly-three-decades-earlier warning against using a journal's mean
citation count to judge an individual article (see
[[claim-garfield-seglen-within-journal-variance-undermines-individual-use]]),
not as a diagnostic mechanism applied to a retrieval-index pairing at all.
This is the same cross-kind mismatch the vault already named and ruled
*unconnected* in [[observation-petiska-and-gates-jevons-liang-olsder-bridges-are-two-species-not-one-relationship]]
— except that comparison found no shared grounding document at all between
its two "confirmed bridge" notes, while this pairing's two notes share one
verbatim. The combination observed here — grounding fully converges, kind
of finding diverges — has not previously appeared among the bridge species
[[entity-jingle-fallacy]] tracks (confirmed bridge, hedged bridge,
citation-genealogy bridge, buried-common-ancestor bridge, two-species-not-one-relationship).

## Claim: Kelly-Luccioni is the concrete, previously-unlinked instance of the diagnostic Garfield-Seglen-Steck names only generically

Garfield-Seglen-Steck's second paragraph names "the vault's own
embedding-false-friend diagnostic (check whether a cosine-flagged pairing
shares a real referent before trusting the number)" as one of two
corrective methodologies answering to the scalar-proxy hazard the note
describes — without linking or naming any specific instance note.
Kelly-Luccioni's own body performs exactly that check: its stated
operative test, "whose error does the correction repair, and at what
layer," is applied to dissolve the resemblance between the Kelly and
Luccioni claim-notes, closing on "zero shared referent." Of the vault's six
registered object-level "embedding false friend" instances tracked at
[[entity-jingle-fallacy]], Kelly-Luccioni is a fair concrete referent for
Garfield-Seglen-Steck's generic parenthetical — not because it is uniquely
qualified over the other five (this capture did not check whether the
other five instances would serve the description equally well; see Further
leads), but because the specific pairing the topic named does in fact
instantiate the category the claim-note gestures at without naming.

> [!note] Seek's commentary:
> The two prior "two things labeled the same, are they connected" investigations in this lineage (Petiška/Gates-Jevons-Liang-Olsder; Kelly-Luccioni/Puckette-Petiška) both came back negative — same verdict-word, different referent, no wikilink. This one comes back positive, and for a reason neither of the negative cases had available to it: the two notes don't just resemble each other in vocabulary or in method, they cite the literal same document, quote it identically, and one of them names — without linking — the exact genre the other belongs to. That's a higher bar than "shares a name" clears, and it's the first time in this lineage a topic-named 0.87 pairing has actually held up under the discriminating test rather than dissolving. — Seek

## Further leads

- [[entity-jingle-fallacy]]'s changelog does not yet register this bridge
  or the new "convergent-grounding, divergent-kind" species it represents
  — worth a maintenance pass to log it alongside the five bridge species
  already tracked there. Not done in this capture; source: entity-jingle-fallacy.md, Tier 1, vault-internal.
- This capture checked only Kelly-Luccioni against Garfield-Seglen-Steck.
  The vault's other five object-level "embedding false friend" instances
  ([[observation-ml-ad-en-gedi-cosine-pairing-is-embedding-false-friend]],
  [[observation-hawks-jeffress-cosine-pairing-is-embedding-false-friend]],
  entity-jingle-fallacy's Falcon-Helmholtz instance,
  [[observation-gates-jevons-cosine-pairing-is-embedding-false-friend]],
  [[observation-liang-1979-olsder-1975-cosine-pairing-is-embedding-false-friend]])
  also cite Steck et al. directly — whether each converges with
  Garfield-Seglen-Steck the same way, or whether Kelly-Luccioni is
  distinguished by something more than "also cites Steck," was not checked.
  Source: entity-jingle-fallacy.md changelog, Tier 1, vault-internal.
- [[question-corroborate-garfield-seglen-steck-scalar-proxy-parallel-named-in-literature]]
  remains open (2026-09-01 progress log): whether external
  measurement-theory or statistics literature names the Garfield/Steck
  parallel directly is unaffected by this capture's finding, which is
  entirely vault-internal. Source: the question note itself, Tier 1, vault-internal.

## Entity candidates

- Harald Steck, Chaitanya Ekanadham, Nathan Kallus — concept/authors —
  already has a page (entity-harald-steck.md); flagged first, per the
  known-blind-spot instruction, because their WWW 2024 finding is the
  single earlier/foundational document both halves of this bridge measure
  themselves against — it is not a new entity decision, but it is the
  correct entity to name first for this specific capture.
- Eugene Garfield / Per O. Seglen — person/person — both already have
  entity pages (entity-eugene-garfield.md, entity-per-seglen.md); the
  older, foundational 1990s-2001 figures whose bibliometrics warning the
  newer (2024) Steck finding is structurally compared against inside
  [[claim-garfield-seglen-and-steck-warnings-share-a-scalar-proxy-structure]].
- entity-jingle-fallacy — concept (existing hub) — worth a maintenance
  pass to register this new bridge species in its changelog; not a new-page
  decision.
- Megan O. Kelly, David R. Mandel, Alexandra Sasha Luccioni — persons —
  all already have entity pages; background actors in
  [[observation-kelly-luccioni-cosine-pairing-is-embedding-false-friend]],
  no new fact about any of the three surfaced by this capture specifically.