Glossary: Confidence
Confidence is a 0–1 statistic TypeSafe derives from the probabilities map on Choice and Score answers. A peaked distribution is high; a flat one is low. Noul has no confidence field. Use confidence as a gate, not a warranty on the next ticket. Calibration is trained to be group-level under the RLCD vendor claim — not a per-row guarantee.
This glossary entry is for builders who already heard of Jev / System One and need a citable definition before they read answers.foo.confidence on a Noul. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Confirm the confidence page on docs.typesafe.ai. We do not sell or issue API keys.
How it shows up
TypeSafe’s three-option demo uses the approximation (3 × largest probability − 1) / 2. They also say you may compute another statistic from the full map. Official routing examples use floors such as 0.5–0.6 and higher bars (0.85–0.9) for irreversible acts — those numbers are illustrations. Store your confidence threshold in the decision contract.
You cannot replace TypeSafe’s field in the request. You can compute another number from probabilities and log both. Low confidence is usually HTTP 200; abstention happens in route(). Pin jev-1.13.0 once fitted; jev-latest can move.
Choice: unordered, max 255. Score: 2–10 ordered levels; score may be fractional. Noul: threshold noul or |noul − 0.5|. Question IDs are not used in inference. Schema-safe ≠ factually correct: high confidence can still pick the wrong closed key.
Contrast
Confidence vs probability. Probability is mass on each option or level. Confidence collapses the shape of that map into one number.
Confidence vs Noul. Noul’s probability is the signal. There is no second field.
Confidence vs calibration. Confidence is the statistic on the wire. Calibration is whether bins match hit rates on your labels.
Confidence vs confidence threshold. The field vs the cut you apply to it.
Common mistakes
- Reading confidence on a Noul (
undefined). - Treating high confidence as proof this row is true.
- Copying a Noul τ onto Choice confidence.
- Treating TypeSafe’s 0.5 / 0.9 examples as defaults.
- Asking Jev to “only answer if confident” in prose instead of gating in code.
See also
FAQ
Can I read confidence on a Noul?
No. Threshold the noul (or its distance from 0.5).
Is high confidence proof the label is true?
No. TypeSafe trains higher confidence to mean higher accuracy across groups. It is not a per-row guarantee.
Can I send my own confidence algorithm?
You can compute one from probabilities. You cannot replace TypeSafe’s field in the request.
Low confidence HTTP status?
Usually 200. Hold in code.
Limits
This is a jev.pro working definition. Confidence is a TypeSafe answer field on Choice and Score. We do not invent search volume. Schema-safe answers are not automatically factually correct.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.