Skip to content

Commitment measurement: closure / gray state / delta without conflating resolution with approval #1135

Description

@zachshallbetter

Parent: #1111
Depends on: #1115, #1127

Goal

Test whether the The Last Undo measurement model belongs as a generic measurement layer over CommitmentContract, without turning assurance into an approval score.

For a pinned required-property set R(K) at commitment K:

  • each property is SATISFIED, VIOLATED, or unresolved/GRAY;
  • closure C(K) measures the fraction (or declared weighted profile) whose state has been established either way;
  • gray G(K) is the unresolved remainder;
  • delta compares the same declared yardstick across two measurements.

Load-bearing rule

Closure measures establishment, not desirability. A fully resolved FAIL and a fully resolved PASS can both have C(K)=1.

Research questions

  • Should closure be a scalar only for unweighted required sets, with weighted cases returning a profile instead?
  • How should not-applicable, expired evidence, and escalated human judgment count?
  • What minimum contract identity/version binding is needed before two closure measurements are comparable?
  • Can a downstream commitment inherit established properties from World: portable assurance receipt distinct from snapshot / replay #1116 while recomputing its own closure against a different required set?

Negative controls

  • Same candidate, two commitments with different required sets -> different closure.
  • Seven of seven established with one violation -> closure 1.0, disposition FAIL.
  • Different contract versions/yardsticks -> delta refused as incomparable unless an explicit transform exists.
  • More favorable outcome with less established evidence -> outcome improvement must not inflate closure.

Acceptance

  • C(K)/G(K) can be recomputed from property-level states; no hidden aggregate truth.
  • Delta refuses mismatched yardsticks by default.
  • PASS/FAIL/ESCALATE remain independent from closure.
  • A receipt can carry the property states necessary for a stranger to recompute the measure.
  • No claim that closure is correctness, safety, utility, or business value.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions