Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
70 61 80 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
46 33 59 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
53 43 64 -
Guidance
A company's own figures for its results over a period, checked against what it then reported.
44 24 66
All types combined: 58 (likely 52 to 63, high confidence), from 94 claims: 35 factual claims, 21 predictions, 33 promises, 5 guidance claims. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Elon Musk's claims
24 claims
“The way the system works is that people who have historically disagreed must agree that a note is valid for it to be shown.”
Elon Musk, X owner, on X 30 December 2024 Post by @elonmusk on X, 2024-12-30 (X)
Community Notes only shows a note when people who have historically disagreed agree that it is valid.
Mostly true True in substance, wrong in a detail.
What would count as true or false
Judge as of 30 Dec 2024, using X's Community Notes documentation and open-source algorithm. TRUE: notes are shown publicly only after reaching 'helpful' status through a 'bridging' requirement, under which raters who have historically rated differently (contributors with differing perspectives, inferred from their rating history) must both rate the note helpful, and there are no material ways around this. MOSTLY TRUE: bridging is the main rule for showing a note, but there are minor qualifications, e.g. it is done by matrix factorization rather than an explicit 'agreement' check, or there are small exceptions or supplementary scoring paths that are themselves still based on bridging. MIXED: bridging applies, but notes can also be shown through significant other routes that do not require agreement across perspectives. MOSTLY FALSE: diverse-perspective agreement is only a minor factor, and most notes are shown on other grounds. FALSE: notes are shown on simple majority or volume of ratings with no requirement for agreement across differing perspectives.
Evidence
- communitynotes/documentation/index.md at main · twitter/communitynotes GitHub (X's official Community Notes repository) Own record
“Community Notes doesn't work by majority rules. To identify notes that are helpful to a wide range of people, notes require agreement between contributors who have sometimes disagreed in their past ratings. This helps prevent one-sided ratings.”
- communitynotes/documentation/under-the-hood/ranking-notes.md at main · twitter/communitynotes GitHub (X's official Community Notes repository) Own record
“Matrix factorization identifies notes that are liked by people who normally disagree by assigning high intercepts to notes when users with different factors rate the note Helpful.”
- The Making of Community Notes Asterisk · 1 November 2024
“What the algorithm does is find notes where people who've disagreed on their ratings in the past actually agree that a particular note is helpful.”
- Community Notes - Wikipedia Wikipedia
“The threshold of five requests within 24 hours would determine a note being published.”
Reasoning
X's own Community Notes documentation says notes need agreement between contributors who have sometimes disagreed in their past ratings, and that the system does not work by majority rule (E1). An independent account describes the algorithm the same way (E3). The algorithm documentation shows this bridging is done by matrix factorization, which gives high scores to notes rated helpful by users with different factors (E2), rather than by a direct check that both sides agree. That is the minor qualification the criteria list under mostly true. E4 mentions a threshold of five requests, but the quote does not say this route lets a note be shown without bridging, so it does not show a significant other path.
Cite this claim
Elon Musk, “Community Notes only shows a note when people who have historically disagreed agree that it is valid.” (said 30 December 2024). Verdict: mostly true. Reliability API, claim c_beb799901a42, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_beb799901a42
Claims without a verdict
Is a verdict here wrong, or do you speak for Elon Musk? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.