# Dimensions and Scales

## Dimension validity contract

A publishable `CoherenceDimension` MUST:

- measure one named behavioral or operational construct;
- have an ordered or bipolar `CoherenceScale` with unambiguous direction;
- define observable indicators, counter-indicators, plausible alternative explanations, and falsifiers;
- state applicable entity and scope types;
- specify evidence count, context coverage, half-life, observer eligibility, and confidence threshold;
- document cultural/context limitations and calibration status;
- list permitted and prohibited uses;
- avoid clinical, identity, character, moral-worth, or reputation claims unless a separately governed validated instrument and qualified process exists.

Dimensions may be system-library, Realm-governed, or entity-private. Drafting is open; publication follows namespace Authority. IDs use `dimension.<namespace>.<name>` and never change meaning in place.

## Scale contract

A scale defines numeric `minimum` and `maximum`, labels/definitions for both poles, optional midpoint, units, normalization, and ordered landmarks. `-100…+100` is the default because direction and distance are legible; it is not mandatory.

Landmarks are interpretation aids, not hidden thresholds. Response zones live in the profile target configuration; hard absolute regions live in governed policies or explicitly referenced boundary definitions.

## Default library

The portable defaults are directness, initiative, epistemic humility, risk tolerance, formality, experimentation, collaboration, persistence, transparency, and deference. They are starting instruments, not universal requirements. A Realm can publish another valid dimension or omit a default.

For every scale, the UI MUST show pole meanings and behavioral examples. A number without meaning is not a valid configuration surface.

## Authoring workflow

1. State the decision or feedback the dimension will support.
2. Name one construct and define both poles without moralizing either side.
3. Write observable indicators, counter-evidence, and confounds.
4. Choose range and landmarks.
5. Define applicable entities/contexts and prohibited uses.
6. Set evidence, decay, and confidence requirements proportional to consequence.
7. Draft calibration cases including unusual legitimate targets.
8. Review bias, accessibility, privacy, and Source-agency risks.
9. Publish through the namespace's Authority path.

Use `coherence-scale-author` and `coherence-calibrator` for machine-executable authoring and review.