METHOD
How we work
Open Peace Lab publishes short research briefs on humanitarian and peace questions. AI research agents do the first draft; a human editor decides what is published. This page explains each step and what our briefs can and cannot tell you.
1. Choosing the question
Each mission starts from one question set by the editor, for example: what would make food assistance in Gaza safer, more reliable and accountable to the people receiving it? The question, its scope and its safety limits stay fixed while the research develops.
2. Gathering sources
The agents search for and read public reports, mainly from United Nations agencies and other humanitarian and technical bodies. Every day they check the publishers' own report pages for anything newer. They read parts of each report, not every document in full, and every brief says so.
3. Turning sources into evidence
Every statement of fact in a brief is tied to a short quotation from its source. Software checks that each quotation appears word for word in the retrieved copy of that source, and that any figure in a statement also appears in its quotation. Each brief links every claim to its quotation and source.
A matching quotation shows where a statement comes from. It does not prove the statement is true, current or complete.
4. Drafting and challenge
Separate AI roles draft a proposal, look for missing perspectives, challenge the reasoning and define how success or failure would be measured. A critic role must accept the draft before it reaches the editor. We currently use Claude models made by Anthropic. The agents cannot contact anyone, publish anything or take any action outside research.
5. Human review
The editor reads every draft and either returns it with written notes or approves it. Most drafts are returned. Only approved drafts are published, and each published version records what changed.
“Reviewed” means editorial review. It is not peer review, field validation or endorsement by any organisation we cite.
What our briefs are not
- They are desk research, based on published reports.
- They propose ways to find things out. They do not claim that any intervention works.
- They never publish operational detail that could put people at risk, such as aid routes, convoy timings or precise locations.
- Our sources are mostly agency reports. Voices missing from those reports are missing from our evidence, and each brief names them.
Questions or concerns
Write to contact@openpeacelab.org. To report an error, see our corrections policy.