Tony_tech/veripath
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: reconcile Tier 1/2 GLOSSARY entries against pre-existing Evidence Model rule

Browse Source
This commit is contained in:
Leonard
2026-08-07 05:10:15 +00:00
parent 3f90b569dc
commit 7e5dfad77f
1 changed files with 33 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/GLOSSARY.md
+33 -6
View File
@@ -228,14 +228,41 @@ A prescriptive classification of how urgently an action should be taken. Lives i
## Tier 1 — Verified
> **⚠️ PENDING RECONCILIATION (2026-08-06)** — See `## Evidence Model` above. That pre-existing section states Findings produced by `llm_sample` or any other non-deterministic method are "Indicative by definition" and "Cannot be upgraded to Verified without additional concrete evidence." This entry distinguishes audit-run facts (Tier 1-eligible via stored artifacts) from Findings (Tier 2). Until the side-by-side extraction is done and Tony resolves it, this entry is not fully authoritative; when the two sections appear to disagree, the pre-existing Evidence Model rule governs (ambiguous findings default to Indicative).
RECONCILED (2026-08-06) against the pre-existing Evidence Model section.
That section governs without exception: Findings produced by llm_sample
or any other non-deterministic method are Indicative by definition and
cannot be upgraded without additional concrete evidence. This entry does
not create an exception — it draws a bright line between two different
things that were previously conflated under "audit-run facts."
A fact verified by the appropriate authority. Verification authority differs by fact class:
- Business-entity facts (name, phone, address, hours, services, categories): requires owner confirmation or owner-access verification.
- Audit-run facts (prompt executed, engine captured, timestamp, raw output stored): requires stored filesystem/git/database evidence. Verifies the audit *event* occurred, not that a business fact inside it is true.
- Public claims inside AI outputs: default Tier 2 until owner confirmation upgrades it (see Tier 2).
Two categories, verified by different authorities:
Same terminology as "Verified" in Module A — one scale, no separate axis.
1. Process facts — the mechanics of testing, not its content. Examples:
prompt text as submitted, engine queried, timestamp, raw output
captured and stored (with artifact pointer/hash), screenshot exists.
Tier 1 when supported by stored filesystem/git/database evidence.
These certify only that a data-collection event occurred — they make
no claim about the business, competitors, or anything the AI said.
2. Findings — anything derived from the *content* of an AI output:
mention status, mention strength, description accuracy, competitor
appearance, any claim the AI made about the business. These are
governed entirely by the existing Evidence Model rule above: Tier 2
by definition, promotable to Tier 1 only via independent concrete
evidence (e.g. owner confirmation), never by the audit process itself.
Business-entity facts (name, phone, address, hours, services, categories)
follow the same rule as Findings when the only source is public
observation or an AI output: Tier 2 until owner confirmation or
owner-access verification upgrades them.
Consequence: the Core Visibility Index (Mention rate, strength, accuracy)
is composed entirely of Findings and stays Tier 2/Indicative regardless
of how precisely it's counted. "Quantitative" and "Tier 1" are different
axes — exact counting does not imply evidentiary confidence.
Same terminology as "Verified" in Module A — one scale, no separate axis,
no exception to the pre-existing rule.
## Tier 2 — Indicative