Workflow Claim Grounding

Workflow claim grounding identifies which public scientific sentences are supported by named evidence and which remain bounded, contradicted, or unearned.

bijux-proteomics-knowledge preserves the connection between an exact claim, its evidence revision, source lineage, contradiction state, and support policy. It does not infer support from cautious wording.

What Ships

The grounding surface is built around three workflow-family artifacts:

• a claim citation table for sentences with direct named support
• an unsupported-claim ledger for sentences that still overrun the current evidence
• a contradiction triage report for sentences that are supported and still scientifically narrowed

These surfaces cover claim-bearing prose in flagship trust pages and public review packets. They intentionally do not treat every heading, link label, or metadata field as a scientific claim.

How To Read The Grounding Pack

• start with the claim citation table when the question is whether a sentence has direct support
• move to the unsupported-claim ledger when the sentence sounds smoother than the actual proof
• move to contradiction triage when the sentence is partly supported and still scientifically narrowed by conflict or thin transfer

The result is one explicit state: supported, bounded, contradicted, or not earned.

Ground One Claim

flowchart LR
    sentence["exact claim text"] --> scope["entity, workflow, population, and strength"]
    scope --> support["direct supporting evidence"]
    scope --> context["contextual evidence"]
    scope --> conflict["contradicting evidence"]
    support --> policy["freshness and support policy"]
    context --> policy
    conflict --> policy
    policy --> state["supported, bounded, contradicted, or unearned"]
    state --> decision["decision input"]

Direct support must entail the scoped sentence at the stated strength. Contextual evidence can explain plausibility or mechanism but cannot substitute for direct support. Contradiction remains attached even when supporting evidence exists; support and conflict are not mutually exclusive tallies.

grounding field	required question
exact claim	What proposition would be false if the claim failed?
scope	Which entity, workflow, population, comparator, and confidence are asserted?
support	Which named source directly supports that proposition?
contradiction	Which named source or checked result narrows it?
freshness	Was the evidence valid under the recorded review policy?
state	Which policy rule produced the grounding state?

Workflow Coverage

family	grounding surface role today	likely reason a sentence still narrows
dda	strong scientific support for bounded language	runtime evidence lowers the end-to-end public posture below the strongest grounding request
dia	strong support for bounded outsider language	library incompleteness and downstream consequence still narrow broader wording
lfq	direct support for bounded outsider-auditable language	missingness, normalization, and transfer pressure still block stronger wording
ptm	strong support for localization and bounded trust language	consequence confidence remains weaker than localization evidence
targeted	strong support for bounded outsider language	calibration, interference, and follow-up burden still narrow broader certainty
multiplex	support for internal-support-only language	the current stress packet still defeats outsider trust

Decisions Grounding Does Not Make

• final recommendation posture by itself
• downstream assay worth by itself
• runtime realism by itself

Grounding answers whether the sentence is scientifically supported. The next layers decide whether it still deserves to sound that strong after judgment and consequence pressure are included.

Continue From Grounding

• Open Workflow Recommendation Confidence when the question becomes whether supported wording is still too confident.
• Open Lab Consequence when the question becomes whether the follow-up burden still narrows the supported sentence.
• Open Decision Support when the reader needs the full route from support to judgment to consequence.

Boundary

Grounding settles scientific sentence support. It does not settle recommendation strength, runtime realism, or laboratory meaning.