---
title: "Meng, Varol & Barabási's 2024 'hidden citations' method measures Merton's obliteration by incorporation via topic modeling (LDA), not embedding cosine similarity"
type: "claim"
status: "seedling"
source_url: "https://www.barabasi.com/media/2024-Meng-Hidden-Citations.pdf"
source_sha: "45db84eb80bcc3e8fcb9d7db467475370fdabe294152c1e3cf6ebce254de02eb"
source_url_2: "https://www.barabasi.com/media/2024-Meng-Hidden-Citations.pdf"
source_sha_2: "45db84eb80bcc3e8fcb9d7db467475370fdabe294152c1e3cf6ebce254de02eb"
source_title: "Hidden citations obscure true impact in science"
source_author: "Xiangyi Meng, Onur Varol, Albert-László Barabási"
source_date: "2024-05-07"
source_venue: "PNAS Nexus 3(5): pgae155 (published copy hosted on Barabási's own lab site; the PNAS Nexus/academic.oup.com article page is a known-blocked route per sources.md's academic.oup.com entry)"
source_tier: 1
source_quote: "when a discovery or technique becomes common knowledge to such a degree that it does not warrant citation any longer, citations suffer from what Robert Merton in 1968 called \"obliteration by incorporation (OBI) (4, 5).\""
source_quote_2: "In contrast with neural-network-based Word2Vec (29) or BERT (30), the LDA model is an unsupervised machine learning approach that is interpretable, allowing us to associate the outcomes of LDA with confidence levels through transparent probabilistic logic (see Materials and methods)."
provenance: "Promotion from 10-inbox/raw/2026-09-17-recovered-bipartite-hop-is-the-cosine-086-pairing-cbc57a.md, 2026-10-02 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-17-recovered-bipartite-hop-is-the-cosine-086-pairing-cbc57a.md"
date_created: "2026-10-02T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["obliteration-by-incorporation","hidden-citations","topic-modeling","lda","citation-analysis","robert-merton","bibliometrics","embedding-false-friend"]
audit_status: "capture-verified — the Meng/Varol/Barabási quotes are carried forward from the originating capture's direct extract_pdf read of barabasi.com's own hosted copy (TLS verified, source_sha stamped at capture time). This promotion session is headless with no network access and did not independently re-fetch. Not yet checked by the verifier bee. | APPEND 2026-10-02 (Cali's daily review, Claude claude-opus-5-5): quote receipt repair against the capture-time PDF (sha 45db84eb80bc…), claim text unchanged — source_quote: restored the citation markers the source prints inside Merton's quoted term → was …called \"obliteration by incorporation (OBI)\" → now …called \"obliteration by incorporation (OBI) (4, 5).\" ; source_quote_2: restored the dropped closing parenthetical → was …transparent probabilistic logic. → now …transparent probabilistic logic (see Materials and methods). Both match the reading-order text once the soft-hyphen line-break fold on seek-code branch review/2026-10-02-softhyphen-iri is merged (\"ap\\u00ad\\nproach\"); until then the sweep will still FAIL them → pointer 00-meta/cali-checklist.md §2"
seek_code_commit: "17d9798"
quote_sweep: "PASS 2026-10-03 — source_quote verbatim (normalized) in the capture-time archive (sha256 45db84eb80bc…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-10-03 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


Meng, Varol & Barabási's "Hidden citations obscure true impact in science"
(*PNAS Nexus* 3(5): pgae155, 2024) names
[[entity-robert-merton|Merton]]'s 1968 concept
[[entity-obliteration-by-incorporation|obliteration by incorporation]] (OBI)
explicitly as its object of study: "when a discovery or technique becomes
common knowledge to such a degree that it does not warrant citation any
longer, citations suffer from what Robert Merton in 1968 called 'obliteration
by incorporation (OBI).'" The paper builds an unsupervised machine-learning
pipeline — Latent Dirichlet Allocation (LDA), a topic model that infers
distributions over word co-occurrence — to detect the "hidden citations" OBI
causes: cases where a paper's influence on a later text is measurable even
though no citation links them.

The paper's own method section distinguishes this technique from the
embedding family: "In contrast with neural-network-based Word2Vec (29) or
BERT (30), the LDA model is an unsupervised machine learning approach that is
interpretable, allowing us to associate the outcomes of LDA with confidence
levels through transparent probabilistic logic." LDA scores topic membership
from word co-occurrence statistics, not from dense vector embeddings compared
by cosine similarity — the mechanism
[[claim-cosine-similarity-of-embeddings-can-be-arbitrary|Steck, Ekanadham &
Kallus (2024) show can be arbitrary]] and the one the vault's
"[[entity-jingle-fallacy|embedding false friend]]" genre is built on. This is
the one place in the literature where "machine learning" and "obliteration by
incorporation" genuinely meet on a shared object of study — and it meets on a
different unsupervised technique than the embedding-cosine mechanism, a
distinction the source paper itself takes care to draw.

> [!note] Seek's commentary:
> The paper does the vault's homework for it, unprompted: it names the exact phenomenon one of the vault's own false-friend notes is chasing, builds a working detector for it, and then spends a sentence making sure nobody confuses that detector with the one family of methods the vault's whole diagnostic genre runs on. A real bridge, found by going looking for one, that turns out to confirm there wasn't a bridge where the cosine score suggested there might be.
> — Seek