Editorial policy

Last reviewed: 2026-08-01

CrisisGraph is a knowledge graph of the events, laws, agencies, and reforms that shaped modern crisis management. This page describes how we decide what belongs, how contributions are reviewed, and what we do when we get something wrong.

What qualifies as an entity

An entity is a discrete, name-able node in the causal record: a specific incident, a named statute, an agency, an executive directive, a formal after-action report, a legal case, a preparedness framework, or a documented reform.

• Verifiable. Every entity must be describable by an authoritative source — a Wikipedia article, a .gov page, an academic paper, an established news report, or a court filing. Speculation and rumor do not qualify.
• Non-trivial. Events must have measurable consequences (casualties, damage, displacement) OR have generated a documented policy response (a law, an agency change, a formal report).
• Named. "The 2005 hurricane season" is not an entity. "Hurricane Katrina" is.
• Neutral naming. Entity names follow common historical usage; loaded framings ("the tragic X", "the heroic Y") are not used.

How contributions are reviewed

Every contribution — whether from a signed-in user via the Contribute portal, from our nightly historical scanners, or from our news-monitoring pipeline — enters a queue. Before publishing, each proposed entity is checked for:

• Duplication against existing entities (fuzzy-matched by name, aliases, and date).
• Source quality — the reference must be authoritative and reachable.
• Sufficient description — a summary that explains the event, not a headline.
• Reasonable dates, locations, and (where applicable) casualty estimates.

Contributions that meet the bar are approved and appear on the site within an hour. Contributions that don't are either rejected (with a reason returned to the contributor) or sent back for revision.

Confidence tiers

Every relationship and every impact number carries a confidence tier. This is not marketing — it tells readers how strongly the claim is anchored.

high Widely-agreed authoritative figure. Example: WWI deaths, Katrina damage estimates, PATRIOT Act signing date.

medium Reasonable historical consensus but with published caveats or ranges. Example: Bhopal death toll (35,000–40,000, sources vary).

low Extrapolated from context; single-source or disputed. Example: cyber attack impact where formal reporting hasn't happened yet.

Numbers below "medium" confidence are surfaced with a badge on the entity page so readers can weight them accordingly.

Disputed facts

When authoritative sources disagree — death tolls that vary across studies, causal links historians contest — we prefer the widely-published range over a single point estimate, and we say so in the entity's notes field. We do not adjudicate contested history; we report the range and cite both anchors.

Neutrality

CrisisGraph does not take positions on contested policy questions ("was the invasion of Iraq justified", "was the Great Reset a success"). We describe what happened, what caused what, and what the record shows. Editorial judgments are limited to selecting entities that meet the qualification bar.

Attribution and licensing

All CrisisGraph data is licensed CC BY 4.0 — reuse freely, with attribution to “CrisisGraph — crisisgraph.com”. Every citation on an entity page links back to its original source; if you republish, please carry those citations forward.

Corrections

Everyone gets things wrong sometimes. When we do, we correct the record and log it publicly at /about/corrections. If you spot an error, email crisisgraph@zafronix.com or use the "Suggest an event" link in the footer to propose a fix.

What we do not publish

• Personal information about private individuals (crime victims by name, contested identities in ongoing investigations).
• Unverified security claims (undisclosed intelligence, uncorroborated leak-based reporting).
• Op-eds, opinion pieces, or advocacy content of any kind.