Selective Transmission
The qualifying material sitting next to a quoted figure, in the same document, that does not travel with it.
Reviewed July 16, 2026Stable
When not to use it
- Where the source genuinely does not qualify its finding, which is a different and more serious problem.
- As a claim that the quoted figure is wrong, since it is usually accurate and being used outside its conditions.
- Where the condition was published separately or later, which is ordinary scientific correction rather than selective transmission.
Reach for something else instead
- Quoting the condition with the figure — costs a clause and removes the problem entirely at the point where it starts.
- Stating a minimum quotable unit — authors identifying which results cannot be reported without their conditions, which does not exist as practice and could.
- -
Read more on the blog
- Every chatbot query on earth is 2% of AI's powerTerritory 8 opens on the numbers behind AI's physical footprint, and on the arithmetic in the IEA's own report that almost nobody quotes.
- Inference prices fell 9x a year. Also 900x.The most cited number in AI economics is a single rate. The study behind it reports a range spanning two orders of magnitude, and a caveat that undercuts its own fastest figures.
- 99.98% is a tracking rate, and the field is animationNVIDIA's motion controller is a real advance that routes around the data problem Territory 7 identified. The number attached to it measures something narrow, and it gets weaker the further it travels from the field it was measured in.
- The same PDF says 83% and nobody quotes itTerritory 10 opens on enterprise deployment. The most-quoted statistic in the field is real, measures something much narrower than its use, and is contradicted inside its own source document.
Further reading
- International Energy Agency (2025), Energy and AI — a worked example in the report that undercuts the framing its projections are used for.
- Epoch AI (2025), LLM inference prices have fallen rapidly but unequally across tasks — a hundredfold range and a contamination caveat published alongside the rate that circulated.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Treating a verified citation as a verified claim, when the citation check passes and the conditions were never in it.
- Assuming a missing qualification indicates concealment, when the usual cause is that it did not fit in a summary.
- Reading only the abstract, which is itself a compression of the paper made by the authors under the same pressure.
At a glance
FieldFoundations
Distinct from citation decay, where the figure itself degrades
Distinct from publication bias, where the result is never published
Directionsurprising, quotable and stronger material travels
Defenceread the neighbouring paragraphs
DifficultyBeginner
Where this sits
A starting point. Nothing needs to come before it.
0Levelstarting point
0Needs firstnothing
0Opens upnothing further
1Areastays here
Computed from the prerequisite graph, not assigned. How this works