Methodology
How the instrument is sealed.
Public Perception is how alarmed the public and lawmakers already are, measured from published polling and enacted law. The Good Institute needle is the load, measured from published surveys of how people actually use AI. Every indicator is scored against thresholds we fixed in advance. The difference between the two needles is the miss, and that is our alarm. When a channel doesn't have enough sourced indicators yet, it shows as pending until the sources are there to score it.
5 of 5 channels published. 4 past the line. The widest miss is childhood, +13.
- 01
Panel first
Indicators, cuts, and max ages are fixed in the panel before any number is scored. We do not move the goalposts after a headline.
- 02
Sourced observations only
Each score needs a publisher, URL, tier ≤ 2, verbatim quote, and a published date inside the indicator's freshness window.
- 03
Compute, don't edit
Needles are means of scored indicators on a 0–100 scale. Channels publish when load and perception both clear coverage. The miss is load minus perception.
- 04
Seal the inputs
Every reading carries an inputs_hash over the observations used. Archives keep prior boards. Diff JSON shows what moved.
- 05
Firewall
Good Health family records never enter this instrument or Institute research — by policy and by path.
Hash b14de6870c53bde97477cd6a9bd5b9e187f2cda1fc659b86ad10d2f07983d7b8 · as of 2026-10-20 · computed 2026-10-20T07:09:13.159Z