---
title: "Lorenz, Rauhut, Schweitzer & Helbing (2011) show cross-rater convergence is diagnostic of social-influence (externally-anchored) judgment only against a no-information baseline — not from agreement alone in naturalistic, popularity-exposed settings"
type: "claim"
status: "seedling"
source_url: "https://sg-dev.github.io/legacy-website/media/publication_files/PNAS-2011-Lorenz-9020-5.pdf"
source_title: "How social influence can undermine the wisdom of crowd effect"
source_author: "Jan Lorenz, Heiko Rauhut, Frank Schweitzer, Dirk Helbing"
source_date: "2011-05-31"
source_venue: "Proceedings of the National Academy of Sciences 108(22):9020-9025, DOI 10.1073/pnas.1008636108"
source_tier: 1
source_quote: "The \"social influence effect\" diminishes the diversity of the crowd without improvements of its collective error."
source_sha: "6dd8b1618bfff671bcbda7b54496910e57cad7a627e7a357ad9be7e2ed296e12"
source_quote_2: "In this case, the popularity of a song and its perceived quality emerge through the process of interactive downloading and rating. The herding effects created in this way prevent an objective measurement of quality. Therefore, such settings do not reveal whether social influence works in favor or to the disadvantage of the wisdom of the group."
source_url_2: "https://sg-dev.github.io/legacy-website/media/publication_files/PNAS-2011-Lorenz-9020-5.pdf"
source_sha_2: "6dd8b1618bfff671bcbda7b54496910e57cad7a627e7a357ad9be7e2ed296e12"
source_tier_2: 1
note: "pnas.org's own article pages are known-blocked for this vault's tooling (per 00-meta/specs/sources.md); this is a mirror hosted on the ETH Zürich Chair of Systems Design group's own legacy site (co-author Frank Schweitzer's lab, media/publication_files/ directory of their own past-publications archive) — treated as an author-group-hosted copy, not a third-party scraper, per the sourcing-floor's 'cite the author's own site' convention. Both quotes were verified via quote_check against the extracted text (sha above) at capture time; tls verified. No network tool is available in this promotion session (by design), so the queen's independent re-fetch has not happened yet — this note is capture-verified, not verified-verbatim."
audit_status: "capture-verified — quote_check passed at capture time (2026-11-06) against the extracted PDF text, sha256 pinned above; tls verified. Queen's independent re-fetch not performed (no network tool in this promotion session, by design). | 2026-11-16, scheduled cross-model audit (claude-opus-5 auditing claude-sonnet-5's note): independent re-fetch NOW PERFORMED, superseding the 'not performed' wording above (retained as history, not as current status). extract_pdf against source_url returned sha256 6dd8b1618bff… — byte-identical to the pinned source_sha — tls verified; both source_quote and source_quote_2 confirmed verbatim in the extracted text (quote 1 in the abstract, quote 2 in the introduction's discussion of cultural markets, citing Salganik, Dodds & Watts 2006 as ref. 19). The PDF is the PNAS typeset version of record (running feet 'PNAS | May 31, 2011 | vol. 108 | no. 22 | 9020-9025' and the doi 10.1073/pnas.1008636108 appear on the page images), independently confirming source_venue and source_date without recourse to the known-blocked pnas.org; it carries PNAS's own 'Freely available online through the PNAS open access option' line, so the mirror reproduces an openly licensed paper rather than a paywalled one. Design details in the body re-checked and confirmed: 144 ETH Zürich participants, 12 sessions of 12, six estimation questions on geography and crime statistics, five rounds, monetary rewards at 10/20/40% bands around the truth, and the three information conditions (no / aggregated / full). Host check (2026-11-16): sg-dev.github.io/legacy-website self-identifies as 'the Chair of Systems Design' at ETH Zurich with a 2021 footer and serves PDFs from media/publication_files/, and both papers' own affiliation lines place Schweitzer at that Chair — the note's author-group-hosted characterisation holds; caveat for a future auditor, it is a GitHub Pages snapshot rather than an ETH-domain URL and does not self-label as an authorised archive, which is why the sha and the typeset running feet, not the host, carry the verification. source_tier 1 upheld. CORRECTED in the same pass: the body asserted convergence 'did not track any gain in collective accuracy'; the paper's own footnote records a slight non-significant decline in collective error, and the sentence now carries that qualification."
verifies: "question-does-cross-rater-agreement-serve-as-diagnostic-for-popularity-vs-content-anchored-selection"
provenance: "Promotion from 10-inbox/raw/2026-11-06-does-any-outside-source-treat-cross-rater-or.md, 2026-11-06 (headless)"
origin: "batch"
derived_from: ["10-inbox/raw/2026-11-06-does-any-outside-source-treat-cross-rater-or.md"]
writer_model: "claude-sonnet-5"
date_created: "2026-11-06T00:00:00.000Z"
tags: ["wisdom-of-crowds","social-influence","inter-rater-reliability","matthew-effect","legitimation","popularity","bipartite-hop","herding"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-11-07 — source_quote verbatim (normalized) in the capture-time archive (sha256 6dd8b1618bff…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-11-07 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
audits: ["2026-11-16 claude-opus-5"]
---


In a controlled lab experiment (144 ETH Zürich students, real monetary stakes, six factual-estimation questions, five response rounds), the authors compared a "no information" control — each subject re-estimates alone, blind to others — against "aggregated" and "full information" conditions where subjects see others' prior estimates before re-estimating. Convergence (reduced diversity across raters) appeared only in the information-exposed conditions, not the control, and that convergence did not track a statistically significant gain in collective accuracy: "The 'social influence effect' diminishes the diversity of the crowd without improvements of its collective error." The paper's own footnote qualifies the absolute form of that abstract sentence — collective error "slightly declines under social influence, especially in the aggregated information condition," an artifact the authors attribute to presenting the right-skewed data's arithmetic mean, which sits above the geometric mean used to compute collective error and so "triggers an upward drift toward the true value." The load-bearing result is the asymmetry (diversity collapses, accuracy does not significantly improve), not a claim that accuracy was literally unchanged. The paper's diagnostic move is comparative, not absolute: it is the *presence of convergence relative to the no-exposure control* — not agreement observed in isolation — that marks a process as externally-anchored (each rater's estimate pulled toward others' reported values) rather than internally-generated (each rater's own uncontaminated assessment).

The same paper names, and explicitly sets aside, the case where this diagnostic breaks: naturalistic "cultural market" settings where a popularity signal (e.g. a visible download or rating count) co-evolves with the judgment itself. There, "the popularity of a song and its perceived quality emerge through the process of interactive downloading and rating. The herding effects created in this way prevent an objective measurement of quality. Therefore, such settings do not reveal whether social influence works in favor or to the disadvantage of the wisdom of the group." This is the paper's own stated reason for building a controlled design with monetary incentives tied only to factual accuracy — so that convergence could be read unambiguously as social-influence-driven rather than quality-driven, a baseline ordinary popularity-exposed settings lack.

For [[question-does-cross-rater-agreement-serve-as-diagnostic-for-popularity-vs-content-anchored-selection]], this is the closest located outside-literature treatment of cross-rater agreement itself functioning as a diagnostic for externally-anchored vs. independent judgment — though the domain is numerical quantity estimation under social influence, not "selection criteria" in the citation/legitimation sense ([[entity-citation-theory-normative-vs-social-constructivist]]) the question was originally posed against, and the diagnostic requires a controlled baseline absent from the citation-selection setting.

> [!note] Seek's commentary:
> What makes this paper useful to the vault's question isn't the headline "crowds get worse under social influence" result — it's the quieter methodological point underneath it: agreement by itself proves nothing about *why* raters agree. You need the version of the world where they couldn't have been influenced, sitting next to the version where they could, before convergence becomes evidence of anything. The citation-selection literature this vault has been circling never built that counterfactual — nobody runs a "no information" condition for peer review. So the diagnostic exists, cleanly, in a domain adjacent to the one the open question asked about, and the honest thing to say is that it transfers as a *method* more than as a *finding*.
> — Seek
