The Standard · Part A normative textAnchors are stable across patch versions · #A17.3
The Standard  /  Part A — Normative  /  §17 Uncertainty

§17 — Evidence classes and the uncertainty factor

Part A is normative. Part B (calibration) is issued separately and is referenced, not reproduced, here.
v0.6.0  current
Compare versions
You are reading  v0.6.0, published 12 August 2026. This clause changed in this version — see the diff in the right rail.

17.1  Purpose

#A17.1 🔗

The uncertainty factor exists so that a determination distinguishes between a cost that is known and large and a cost that is assumed and large. Assurance sized without that distinction charges the diligent applicant and the undiligent applicant the same amount, and therefore purchases no diligence.

17.2  Evidence classes

#A17.2 🔗

Each input to a determination shall be assigned exactly one evidence class, reflecting the strength of the artifact the input is drawn from and the independence of its author.

ClassBasis of the inputAuthor independence
E1Executed instrument, tariff or statuteBinding on a third party
E2Independent professional study, reviewedIndependent of the applicant
E3Professional study, unreviewedRetained by the applicant
E4Applicant attestation or internal estimateNone
E5Assumption, analogue, or an input whose bound artifact has been amended or withdrawnNone
Class is a property of the artifact and its author, not of the number. A precise figure in an unreviewed estimate is E3; a rounded figure in an executed agreement is E1.

17.3  Assignment rules

#A17.3 🔗  ·  changed in v0.6.0
Normative

An input shall be bound to the artifact it is drawn from by cryptographic digest, together with the located span relied upon. An input that cites no stored artifact shall not be assigned a class above E5.

Normative

Class shall be assignable from the document type. An assessor shall not assign E1, E2 or E3 on the basis of an automated reading alone; such a class may be proposed but shall be confirmed by the accredited assessor of record before it enters the determination.

The second rule is the boundary between what a machine settles and what a person answers for. Digest comparison, recomputation and completeness are machine-settleable and are settled that way. The judgment that a particular study was genuinely independent is not, and is therefore reserved.

Rationale — the assignment step is the only point at which an unbound judgment can move the requirement. It is therefore the only step the standard removes from automation entirely, rather than merely constraining.

17.4  Uncertainty factor

#A17.4 🔗

The uncertainty factor for a bucket shall be computed as a linear function of the evidence credit of the class assigned:

UAFb  =  1 + k · ( 1 − cb )
  c  evidence credit of the assigned class — Part B
  k  uncertainty coefficient — Part B

The factor is bounded below at unity: better-than-required evidence removes uncertainty loading but shall not create a credit against the attributed cost. Credit for protection is granted under §13 and shall not be granted a second time here.

17.5  Re-verification

#A17.5 🔗

Where a bound artifact is amended, withdrawn, or fails digest comparison at any subsequent verification, every input bound to it shall be demoted to E5 and the determination of record shall be marked stale until recomputed.

Cite this clause
DICA Standard, Part A §17.3 (v0.6.0, 2026‑08‑12). Permanent anchor: /standard/v0.6.0#A17.3
What changed in v0.6.0

+ added  the reservation of E1–E3 to the assessor of record

+ added  digest binding as a precondition of any class above E5

− removed  the discretionary “equivalent evidence” allowance in v0.5

Version history
  • Currentv0.6.0
  • Supersededv0.5.0
  • Supersededv0.4.1
  • Withdrawnv0.3.0
View full changelog
Conformance
An assessment conforms to v0.6.0 only if every input carries a class, a bound artifact and a located span, and the determination names the parameter set and engine build under which it was computed.
DICA Standard, Part A — version 0.6.0·Patent pending·Version history·Limitations register·Contact