Skip to content

Valuations & Value Semantics

v9 compatibility note: This page documents the retained v6 portable-state semantics. V9 preserves the contract as a frozen substrate and may wrap/project it through Logical Artifacts without redefining it. See Version-Stack-v6-v7-v8-v9.

Kristal v6 replaces the old assumption of one universal certainty_level with typed valuations.

A valuation answers two separate questions:

  1. Which dimension is being measured?
  2. What kind of value represents that dimension?

Supported semantic families

Typical value semantics include:

Semantics Example
boolean document received: true/false
categorical native / introduced / cultivated
set applicable dimensions: {energy, information}
ordinal low < medium < high
scalar wear = 0.72
interval expected pressure = 350–400 kPa
probability P(cause | observations) = 0.72
distribution competing diagnostic hypotheses
vector support + severity + cost + testability
partial_order alternatives incomparable without preferences
state draft → review → approved
temporal instant, interval, duration, recurrence

Value state is separate from value

These are not numbers:

unknown
not_applicable
indeterminate
not_measured

In particular:

unknown ≠ 0
not_applicable ≠ low

Ordinal “grayscale”

The old certainty vocabulary was best understood as an ordinal grayscale: values were ordered, but distances were not necessarily equal.

In v6 an ordinal scale can still be used, but the dimension must be explicit. For example:

dimension = diagnostic_support
value_semantics = ordinal
value = high

is not equivalent to:

dimension = intervention_necessity
value = high

Measurement is not policy threshold

A continuous or ordinal measurement should preserve nuance. Policy can later derive an action boundary:

risk < 0.20       → automatic path allowed
0.20 ≤ risk < .60 → human review
risk ≥ .60        → specialist decision

The threshold belongs to policy/actionability. It does not redefine the underlying measurement.

Avoid fake precision

Do not convert qualitative review into 0.83 just because a scalar field exists. Use a measurement type only when the domain has a defensible interpretation and method.

Scaling and operational boundaries

A rich measurement may later be projected to a simple action boundary, but the measurement and the policy threshold remain separate. See Measurement-Scaling-and-Decision-Boundaries for binary projections, continuous measures, vectors, partial orders and legacy down-projection.