How Evidence Gets Gathered
Before a jury ever scores an organization, something has to go find out what's actually known about it. Here's that process.
The rest of this Research System section documents how the scoring methodology itself evolved — V4.0 through V6.1. This page is about something else entirely: the pipeline that gathers real-world evidence about a specific candidate organization, before that organization is ever scored. The two share a name, not a mechanism.
When a new organization is proposed for the dataset, the system works through a fixed order of sources — cheapest and most structured first, most expensive last — and stops as soon as it has enough:
- Existing structured datasets on terrorism, political violence, and extremism
- Established institutional watchdog research and analysis
- General encyclopedic sources
- Historical web archives, for organizations with a thin present-day footprint
- Academic literature search
- Public government records — court filings, regulatory filings, spending and disclosure databases, dozens of agency sources
- Only as a last resort: paid, general-purpose web search, when every free avenue above comes up short
Whatever is found gets synthesized into a structured evidence brief, organized around the same ten scoring criteria the jury evaluates against (see the system overview). This brief — not raw search results — is what the jury actually reads when it scores an organization.
Some organizations warrant more than a brief — a full narrative dossier tracing history, key figures, and documented practices in depth. Those go through an additional fact-checking pass before anything from them is used: every citation is checked against its source.
That check isn't a formality. In practice it has caught meaningful fabrication — one dossier's automated first draft invented roughly a quarter of its citations outright, all of which were removed before the material was used for anything.