Editorial standards

Human First publishes short evidence summaries about what helps people live well, and about how technology, including AI, should serve people. These are the rules every page follows.

What we believe

Human dignity does not depend on productivity, wealth, ability, nationality or status. Technology should be judged by what it does for people's health, freedom, relationships and security. We say this openly, and we keep it separate from the evidence.

How a page is made

1. Research. A research agent searches scholarly databases for systematic reviews, meta-analyses and major studies on one question. A person chooses which sources go into the registry.
2. Drafting. A page may be drafted with AI assistance. The draft may only cite sources already in the registry, and only make claims those sources support.
3. Verification. A verification agent checks that every claim has a source, every source exists, every DOI resolves to the right paper, and every link works. Pages with errors cannot be approved.
4. Human approval. A human editor reads the page against its sources and approves it. Nothing is published without that approval, and every page shows who approved it and when.

How to read a page

• Findings are numbered claims, each linked to its sources.
• Interpretations are boxed and labelled. They are our reading of the evidence, not part of it.
• Principles and law are labelled as such. A principle is a moral or legal commitment, not a scientific result.
• What this does not show lists the main limits of the evidence on every page.
• Evidence level describes the strongest type of source behind the page: strong (meta-analyses and systematic reviews), moderate (large individual studies), limited, expert guideline, or principle.

What we will not do

• Invent citations, quotes or data.
• Present opinion as scientific fact, or overstate how certain the evidence is.
• Hide credible disagreement or manufacture a consensus.
• Use fake accounts, fake reviews, automated comments or spam to spread our pages.
• Target vulnerable people with persuasive messaging.
• Collect personal data from readers. This site has no accounts, no cookies and no paywall. We count daily page views per page, the referring website and the country, and which search engines and AI crawlers visit. We never store IP addresses or follow individual visitors.

Corrections

If a page is wrong, we correct it, add a dated note to the page, and list it on the corrections page. Retracted pages stay online with a notice so links and citations do not break.

Use and reuse

All text is licensed CC BY 4.0. Search engines, researchers and AI crawlers are welcome. Structured data is on the data page.