Methodology
The project's only real claim is that it is an auditable instrument — one whose provenance you can check. This page is that check.
What it measures
It measures attention: how much a topic is written about and read. It does not measure sentiment — whether coverage is favourable, hostile, credulous or mocking. A congressional hearing and a wave of ridicule produce the same spike in these numbers. Tone is a separate signal we may add later; until then, read every rise as "more attention", never "more belief".
The two sides
- Supply — who publishes. Scientific preprints from arXiv today; broad news coverage (GDELT) and US-government activity (GovInfo) join as those sources are enabled.
- Demand — who then reads. Wikipedia pageviews, the one signal here where absolute magnitude is real, so "was this spike bigger than 2023?" is answerable.
Supply and demand are never merged into one number and never share a chart axis — pageviews and paper counts are different things measured differently.
The three strands
| Strand | What it covers | Sources |
|---|---|---|
| Disclosure & government | Hearings, agencies, official statements | GovInfo (US), Wikipedia |
| Phenomenon & folklore | Sightings, incidents, cultural coverage | GDELT news, Wikipedia |
| Science & SETI | Biosignatures, technosignatures, the search | arXiv, Wikipedia |
What "unusual" means
Each series is compared to its own recent history, not to other places or topics. That is deliberate: a raw count glows brightest wherever there are simply more newsrooms or more readers, which tells you about the sensor, not the signal. Comparing a series to itself cancels that out — a rise means attention is unusual for that series. (The charts here currently show a plain 14-day average; the full deviation detector is being built.)
What you may and may not conclude
| Supported | Not supported |
|---|---|
| "Coverage matching our terms rose against its baseline" | "Global interest in aliens rose" |
| "Public readership peaked N days after coverage" | "Coverage caused the interest" |
| "Preprint output on technosignatures increased" | "A discovery was made" |
Known gaps — stated, not hidden
- No social-media signal. Volume there is a poor proxy for attention and mostly unavailable on fair terms; excluding it is a deliberate quality choice.
- The government signal is US-only (GovInfo). It is labelled as such, never as "worldwide".
- Western and English-language media are over-represented in news sources.
- Every figure is traceable to the exact query that produced it, published alongside the data.
The underlying series will be downloadable and openly licensed, with a public changelog for any change to how something is measured.