A three-role architecture separates the one who objects from the one who decides. peira's
gates do both: a gate raises the objection and blocks on it.
For the nine enforced gates that is correct and should not change — they are deterministic, so the
objection IS the adjudication, and interposing a decision step would only add somewhere for the
verdict to be softened.
It stops being correct for the eleven catalogued-but-unenforced lenses, where the objection is a
judgement call. There, "this over-claims" and "the objection is sustained" are different records
made by different parties, and collapsing them puts the objector in the adjudicator's seat — the same
defect PEIR-LINT-SELF-GRADED (#10) exists to catch one layer down.
The machinery already exists and is unused for this:
NodeKind::Examination — "the record of a lens having been run against a subject" — is the
objection.- A graded edge, carrying
by=, is the adjudication.
So an unenforced lens's finding should be an Examination node that ATTACKS the claim, and its
disposition a separately-graded edge. What is missing is any documented workflow saying so, and any
check that an Examination recording a finding was ever dispositioned.
Related: #10 (author must not sign off their own finding — the same separation), #3 (ACH, which is
adjudication between competing cases).
A three-role architecture separates the one who objects from the one who decides. peira's
gates do both: a gate raises the objection and blocks on it.
For the nine enforced gates that is correct and should not change — they are deterministic, so the
objection IS the adjudication, and interposing a decision step would only add somewhere for the
verdict to be softened.
It stops being correct for the eleven catalogued-but-unenforced lenses, where the objection is a
judgement call. There, "this over-claims" and "the objection is sustained" are different records
made by different parties, and collapsing them puts the objector in the adjudicator's seat — the same
defect
PEIR-LINT-SELF-GRADED(#10) exists to catch one layer down.The machinery already exists and is unused for this:
NodeKind::Examination— "the record of a lens having been run against a subject" — is theobjection.
by=, is the adjudication.So an unenforced lens's finding should be an
Examinationnode that ATTACKS the claim, and itsdisposition a separately-graded edge. What is missing is any documented workflow saying so, and any
check that an
Examinationrecording a finding was ever dispositioned.Related: #10 (author must not sign off their own finding — the same separation), #3 (ACH, which is
adjudication between competing cases).