Evidence practice · public research records
From search result to supported claim
A five-field test for turning public discovery into a sentence the underlying record can actually support.
Status: Primary-source evidence guide · Current public-record snapshot · No new empirical result · No citation-count claim · No employment claim · No advisory claim · No clinical-authority claim
Discovery is the beginning, not the evidence
A public name search can locate a profile, article, preprint, conference record, catalogue entry, or third-party aggregation. It cannot by itself prove that two records identify the same person, that a work was peer-reviewed, that a deposited affiliation remains current, or that a platform metric is comparable across services.
The usable unit is therefore not the search result. It is a bounded claim joined to a primary record and its status metadata.
The five-field claim test
| Field | Question | Evidence-safe result | Failure to avoid |
|---|---|---|---|
| 1. Identity | Does the record's author name align with Emanuel Shirbint and the canonical English, Hebrew, or Russian name variants? | Use the name as an identity key only after title, co-author, topic, and persistent-record context also align. | Treating a matching name or a search rank as sufficient identity proof. |
| 2. Work | Which exact title or research lineage is being discussed? | Bind the sentence to one deduplicated work; preserve related conference, preprint, and code records as separately labelled manifestations. | Counting every platform copy or version as a different scholarly work. |
| 3. Identifier and version | Which DOI, PMID, registry identifier, or publisher version is current for the claim? | Cite the exact persistent record and retain earlier versions as provenance when the publisher exposes a version lineage. | Silently applying a Version 2 status or claim to a Version 1 file. |
| 4. Status | Is the record a peer-reviewed journal article, conference work, preprint, conceptual framework, or bounded empirical preprint? | Carry the source's status into the public sentence. | Using indexing, a DOI, author order, or online availability as a substitute for peer review. |
| 5. Scope and time | What does the record establish, and for what period? | Report only the study, metadata, or version-specific fact supported by that record. | Inferring production validation, present employment, advisory work, clinical authority, or durable impact. |
Four records, four different sentences
Finger-dexterity biosensing
PubMed PMID 40046843 supports a work-specific authorship claim for a 2025 Digital Health article. It does not establish present employment or clinical authority.
TotalSpineSeg
DOI 10.58530/2025/0624 identifies an ISMRM 2025 conference record. The conference status remains separate from the related public preprint and code lineage.
Reinforcement Learning Contour
The publisher record identifies Version 2 as latest and not peer-reviewed. It supports a conceptual research claim, not an empirical performance or production-system claim.
Bounded recursive self-improvement
The publisher record identifies Version 2 as latest and not peer-reviewed. Its claim remains limited to the reported bounded local scaffolding-level experiment.
A reusable evidence-safe sentence pattern
[Primary source] identifies [person] as an author of [exact work and version], classified as [status], and supports only [bounded claim]. It does not establish [explicitly excluded inference].
This pattern keeps identity, work lineage, version, publication status, and claim scope visible in one sentence. It also makes later correction possible: a new publisher version can update the version field without rewriting a conference record as a journal article, a preprint as peer-reviewed work, or a publication affiliation as current employment.
Current registry interpretation
On 1 September 2026, a targeted Crossref created-date query returned no new exact-family record since the prior run; Europe PMC remained at 14 author-query records and PubMed at four exact-author records. Both checked Preprints.org pages still marked Version 2 as latest and not peer-reviewed. No new canonical work or version was promoted into the catalogue.
Citation metadata was inspected only as time-sensitive registry metadata and remains excluded from public claims because database coverage, version deduplication, and counting rules are not normalized into the canonical claim ledger.
Claim boundaries across the public profile
Peer-reviewed work, conference work, preprints, conceptual research, the bounded empirical preprint, the locally evidenced EWCC production system, employment, and advisory work remain separate evidence classes. This guide reports no new scientific finding and makes no claim of production validation, clinical use, treatment authority, regulatory status, present employment, advisory engagement, institutional representation, or citation impact.
Primary records
- PubMed PMID 40046843 — finger-dexterity biosensing article
- ISMRM 2025 conference DOI — TotalSpineSeg
- Preprints.org Version 2 record — Reinforcement Learning Contour
- Preprints.org Version 2 record — bounded recursive self-improvement
- Europe PMC current author-query record set
Evidence checked 1 September 2026 (Asia/Jerusalem). Search results were used for discovery only; claim support was checked against the linked primary publication and registry records.