How we work
How a NOOS briefing is made and checked
Who finds the sources, who checks the numbers, who decides what is published, and what we do not do. Only what we already do is listed here.
- 01
We recount published data
NOOS Observatory joins and recounts data that others have already published: peer-reviewed studies, reviews, national surveys and open statistics. We run no new studies on people and collect no personal data for briefings.
- 02
We write down what would prove us wrong
Before checking a claim, we write down what finding would show it wrong, then look for exactly that. What does not fit is shown on the page, not dropped.
- 03
Every claim gets a row in a fact table
Each number and claim on a briefing has a row with its exact source. A separate AI assistant that has not seen the draft checks every row against the original paper or data table.
- 04
The limit stands next to the finding
Every conclusion carries the line “What we will not claim”, right under it, not in a footnote. We ask anyone who quotes us to keep it.
- 05
An outside reader before and after publication
Before a briefing goes live, a reviewer that has not seen the drafts reads the text and the changes. After it goes live, another reader goes through the published page looking for what misleads a reader. Both are AI assistants. What misleads a reader is fixed; when a published claim changes, the change is listed on our corrections page.
- 06
A person answers for it
AI assistants search, draft and check. The founder of NOOS Institute decides whether a briefing is published and answers for it. NOOS is an independent public-interest project started by one person; there is no legal entity yet.
- 07
Expert reviewers: we are looking for them
No outside specialist reviews our briefings yet. AI checks do not replace a specialist reading the conclusion and its limit, so we are inviting researchers and clinicians to do that.
What we do not do
- We do not diagnose, treat or give personal advice.
- We do not rank countries or groups by how ill people are.
- We do not publish a finding without its limit, or a number without its source.
- Employers who pay for diagnostics cannot buy or change research conclusions.