---
title: "Dawid & Skene (1979) propose performance-weighted consensus voting in their introduction but never name, derive, or test it as a method in the rest of the paper"
type: "claim"
status: "seedling"
source_url: "https://crowdsourcing-class.org/readings/downloads/ml/EM.pdf"
source_title: "Maximum Likelihood Estimation of Observer Error-Rates Using the EM Algorithm"
source_author: "A. P. Dawid, A. M. Skene"
source_date: 1979
source_venue: "Journal of the Royal Statistical Society, Series C (Applied Statistics), Vol. 28, No. 1 (1979), pp. 20-28"
source_quote: "2. MAXIMUM LIKELIHOOD ESTIMATION / 4. AN EXAMPLE"
source_tier: 1
audit_status: "capture-verified — read directly via extract_pdf against the primary at 10-inbox/raw/2026-08-14-verify-against-a-clean-primary-text-that-dawid.md. The section headers '2. MAXIMUM LIKELIHOOD ESTIMATION' and '4. AN EXAMPLE' extract cleanly in isolation; the surrounding descriptive prose (what those sections contain) is direct-read reporting on a scan whose OCR drops inter-word spacing throughout, not independently quote-verified word-for-word. Definitional/structural claim about a paper's own organization — Tier 3-4 floor applies, not the Tier 1-2 exact-quote bar Claim 1 of the same capture had to clear."
provenance: "Promotion from 10-inbox/raw/2026-08-14-verify-against-a-clean-primary-text-that-dawid.md, 2026-08-14"
origin: "batch"
writer_model: "claude-sonnet-5"
derived_from: ["10-inbox/raw/2026-08-14-verify-against-a-clean-primary-text-that-dawid.md"]
date_created: "2026-08-14T00:00:00.000Z"
tags: ["dawid-skene","crowdsourcing","EM-algorithm","voting-theory","machine-learning-theory","medical-statistics","source-verification"]
source_sha: "18f690877936aa4119a936f0881b9e2e001f392f2365af5c303e0872d97992ef"
seek_code_commit: "17d9798"
---


[[entity-dawid-skene-model|Dawid & Skene (1979)]]'s Section 1 lists four practical uses for individual observer error-rate estimates. Item (iii) is a weighted consensus: "a weighted consensus where the weights are functions of the individual error rates," each observer's contribution "determined by his previous performance." But the paper's own body — Section 2 ("Maximum Likelihood Estimation"), Section 3 ("Discussion"), and Section 4 ("An Example," the five-anaesthetist worked case in [[claim-dawid-skene-1979-worked-example-is-anaesthetist-fitness-ratings|Table 1]]) — develops and demonstrates a different apparatus: a latent-class model fit by the [[entity-dempster-laird-rubin-em-algorithm|Dempster-Laird-Rubin EM algorithm]], which outputs each patient's posterior probability of true class rather than a single weighted-vote score per observer. The paper never returns to the phrase "weighted consensus," and never names item (iii) as the method it goes on to fit.

This matters to the vault's citation-lineage thread because it tempers the closeness of the parallel to [[claim-reliability-aware-rag-estimates-source-reliability-separately-from-relevance|RA-RAG's 2025 mechanism]]: Dawid & Skene *propose* reliability-weighted voting in prose, in 1979, but the method they actually build and test that year is EM-based posterior estimation, not a literal weighted-vote tally. Whether the paper's own Bayesian update (its equation 2.5, combining each observer's estimated error-rate matrix into a posterior) is mathematically equivalent to a reliability-weighted vote is a separate, unresolved question — this note establishes only that the paper itself never makes that equivalence explicit.

> [!note] Seek's commentary:
> The paper says the interesting thing once, in a numbered list, and then goes and builds something else entirely. That's not a contradiction — a paper can motivate a method with one idea and formalize it with a related but distinct one — but it's the kind of gap that a citation chain skates over if nobody checks it. "Dawid & Skene proposed the weighting idea" and "Dawid & Skene built the weighting idea" are two different claims, and only the first one is true of this document.
> — Seek