Why we built CrisisGraph

The web is full of timelines, and it’s full of encyclopedias. What it doesn’t have is a single place that treats history as a causal graph — what led to what, with citations you can click and filters you can trust. CrisisGraph is that place: a century-plus of wars, pandemics, disasters, and terrorism, wired to the laws, agencies, and reforms that answered them — and an AI that can walk the chain for you.

Timelines end at “when.” We wanted “why.”

Try to answer a simple question with the tools we have today: what did the 1918 influenza pandemic actually change?

A timeline gives you a row of dates. An encyclopedia gives you a long, excellent article you have to read end-to-end and then cross-reference by hand. Neither one lets you follow the thread — from the outbreak, to the public-health acts that followed, to the agencies that were stood up, to the next pandemic that tested them. The causation is in there somewhere, spread across a hundred pages. It just isn’t connected.

History is not a list. It’s a graph. Events cause other events; policies are responses to crises; the same failure shows up three disasters in a row until a law finally fixes it. We built CrisisGraph because that structure deserved to be first-class — queryable, citable, and navigable — instead of something you reconstruct in your head every time.

What CrisisGraph actually is

A collaborative causal knowledge graph of global crisis history. Right now it holds 419+ events and 400+ typed causal links — and it grows every night.

Two things make it different from a timeline:

• The links are typed and causal. An edge isn’t just “these two things are related.” It says how — this reform was enacted in response to that disaster; this outbreak escalated into that one. You can walk cause-and-effect, not just chronology.
• Every entity is a real, citable page. Each event, agency, and reform has its own URL, its own sourced citations, and its own structured-data footprint. Nothing is a dead node in someone’s private database — it’s a page you can link to, quote, and check.

How it stays current — and where the AI lives

CrisisGraph isn’t a static dataset we dumped online and walked away from. It grows nightly through an on-premise LLM — our own model, QALA, running on hardware we control — that ingests the day’s Wikipedia Current Events and back-fills history year by year. New crises land within a day; the deep past keeps filling in.

The AI does the reading and the drafting. It does not get the final word. Two places make that concrete:

• Ask the graph. At /ask, you can pose a question in plain English and the AI answers by walking the actual causal links — and shows you the events it traversed to get there, so you can check its work instead of trusting it.
• Compare two events. At /compare, put two crises side by side — their causes, their responses, what each one changed.

And the graph is collaborative. Anyone can explore without an account; contributors sign in with a magic link to add and correct entities. The machine proposes; a human confirms. That’s the same principle behind everything we build — AI in the loop, a person in the chair.

Five ways in

There isn’t one right way to read a graph, so there isn’t one view:

• Timeline — the familiar chronological spine, but every node links out to its causes and effects.
• World map — see where crises clustered, geographically and over time.
• Playboard — explore the graph as a connected board of events and links rather than a flat list.
• Compare & Ask — the two AI-assisted surfaces above.
• Embed — drop any single event onto your own blog or lesson page as an iframe, citations and all.

Who it’s for, plainly

• Students and educators who need the connections, not just a list of dates to memorize.
• Journalists and researchers looking for the precedent — the last three times this happened and what followed.
• Policy and public-health readers tracing how a crisis became a law became an agency.
• Anyone who has ever fallen down a Wikipedia hole at midnight and wished the links told you why, not just what next.

CrisisGraph is free and open at crisisgraph.com — no account required to explore, a magic link to contribute. If you teach with it, research with it, or just want to argue with an edge or two, we’d love to hear about it.