Intensity, thresholds and ranges

There is no universal intensity scalar. A dimension records quantity, units, stage, optional value/bounds, conditions and uncertainty. A touch description can include force (N), pressure (Pa), contact area (m²), duration (s), vibration frequency (Hz) and temperature (°C). Acoustic and optical inputs use their own dimensions.

The five stages are stimulus_intensity, sensor_response, encoded_intensity, sensation_intensity, and candidate_perceived_intensity. A count in a camera image is not physical luminance; a temperature reading is not warmth. Mapping/nonlinearity fields describe supported relationships without computing a universal conversion.

Unknown values

Null means unknown or not asserted. Empty bounds are useful when the concept and measurement question are known but a source is not. A zero is a real value and must not be substituted for missing data. Floating-point NaN and infinity are rejected. Lower bounds must not exceed upper bounds, and a provided value must lie within its bounds.

Scoped ranges

Range types include detection/discrimination thresholds, just noticeable difference, operating, comfortable, preferred, strong-stimulus, aversive, saturation, safety-limit and damage-risk ranges. Required scope includes population/device and measurement method. Conditions, species, uncertainty and claim evidence provide further context.

The seed's real-world touch, sound and LiDAR threshold records intentionally have unknown bounds. The only numeric seed range is an explicitly synthetic 0–100 percentage encoding, not a battery operating or safety recommendation. No universal comfort, injury or safety threshold is provided.

Units and uncertainty

Units come from a reviewed controlled vocabulary; invalid spellings are rejected. v0.1 does not perform dimensional analysis or unit conversion, so reviewers must check that a quantity and unit are appropriate together. 1 denotes a dimensionless quantity and requires contextual explanation. Add derived units to the contract when needed instead of overloading a mismatched existing unit.

Uncertainty supports confidence, nonnegative error bounds, method and notes. Numerical confidence needs a method and source support in a claim. Accuracy, precision, noise, latency, sampling rate and other device characteristics can be represented as typed metrics. A confidence number is not a universal truth score.