A peer-reviewed paper calls twelve people a “source of misinformation.” Its author’s own nine-minute conference talk shows how that verdict was reached. In all nine minutes, and in all eleven pages of the paper, the authors never take one statement by any of the twelve and show that it is false. Not one.
That is the whole story. Everything else is how they got around it.
1. The headline number is misquoted. The Center for Countering Digital Hate (CCDH) claimed the twelve produced “up to 65%” of anti-vaccine content, in a six-week sample across Facebook and Twitter. For Twitter, the only platform this study examines, CCDH’s own figure was “up to 17%.” The paper drops “up to,” stretches “anti-vaccine content” into “Covid-19 misinformation,” and backs it with an honors thesis that it lists as a Ph.D. dissertation. Then, on stage, the author says “75%” out loud while his own slide reads 65%. Three different numbers for the claim the whole study stands on.
2. The rebuttal is ignored. Ten months before the paper appeared, Meta said there “isn’t any evidence” for CCDH’s Facebook claim. Meta noted that CCDH had analyzed just 483 pieces of content from 30 groups, and it put the twelve at about 0.05% of views of vaccine content, true or false. The paper never mentions this. I laid out the Meta rebuttal and its timeline here.
3. The “good information” baseline is a follow list. The 260 comparison accounts come from one physicist’s personal web page. His own fine print calls it “only a rough guide,” offered “rather than a verified list of trusted sources.” That page is the study’s benchmark for truth.
4. “Misinformation” was assigned by a website rating. Shared links were scored using Media Bias/Fact Check’s ratings of whole outlets, not by checking what any link actually said.
5. Ordinary things count as guilt. Deleted YouTube videos are counted as evidence. The paper itself admits a video may be removed “by YouTube or its owner” and that removal only “may indicate” a violation. Topics like glyphosate count too, and so do title words like “truth,” “censorship” and “awakening.” Nobody analyzed what a single video claimed.
6. The verdict came first. CCDH chose the twelve because they had big followings. The study then discovers they have big followings and high engagement, and treats that as confirmation. Popularity is measured, and accuracy never is.
7. There is no limitations section. The word “limitation” does not appear anywhere in the paper.
8. It spread. GreenMedInfo’s case study reports that the work was supported by a Swiss National Science Foundation Spark grant, and it tracks later papers that cite the “Dozen” framing. Once a contested claim is published, it gets cited as settled.
This is circular by design. The label goes in as an assumption and comes out as a finding. The study measures labels, not truth. If twelve people really were spreading falsehoods, the simplest proof would be to quote one false statement and show the evidence against it. This study never tries.
Watch the talk. Read the paper. Then ask the only question that matters: where is the false claim?
Sources
Nogara et al., WebSci ‘22, pp. 348-358: https://doi.org/10.1145/3501247.3531573
CCDH, “The Disinformation Dozen,” March 24, 2021: https://counterhate.com/research/the-disinformation-dozen/
Meta, Aug 18, 2021: https://about.fb.com/news/2021/08/taking-action-against-vaccine-misinformation-superspreaders/
Oppenheim list (archived): https://web.archive.org/web/20220122202248/https://www.ucl.ac.uk/oppenheim/Covid-19_tweeps.shtml
Media Bias/Fact Check, About: https://mediabiasfactcheck.com/about/
GreenMedInfo case study: https://greenmedinfo.com/content/disinformation-publication-case-study-academic-failure1
