People dimension: account for institutional authorship

[aadivar]

Problem

The People dimension (20/100 points) is scored entirely on ORCID coverage. Institutional authors (e.g., OECD, World Bank, WHO) cannot hold ORCIDs — only individuals can. Publishers with significant institutional authorship will always score poorly on this dimension regardless of their metadata quality.

This is structurally similar to #5 (funding penalizing unfunded research) — the scoring assumes a model of individual, funded academic authorship that doesn't hold across all scholarly publishing.

Why this is hard to fix with current data

• No authorship-type signal in the Crossref Member API. Coverage stats report "X% of DOIs have ORCIDs" but don't distinguish individual vs. institutional authors.
• Crossref's author field supports both. At the work level, institutional authors use the name field instead of given/family, but this isn't reflected in aggregate coverage stats.
• ROR could serve as the institutional equivalent of ORCID, but the Organizations dimension already counts ROR IDs separately — double-counting would introduce its own bias.

Possible approaches

1. Work-level sampling — Sample works from publishers flagged as low-ORCID, check the ratio of name (institutional) vs given/family (individual) authors, and adjust the People score accordingly.
2. Content-type heuristics — Certain content types (reports, datasets, standards) are more likely to have institutional authorship. Reduce ORCID weight for these types.
3. Composite People metric — Score People as "ORCID coverage among individual authors + ROR coverage among institutional authors" rather than ORCID alone. Requires work-level analysis.
4. Methodology note — At minimum, document this limitation on the site.

Related

• Data lacks context #3 (broader data context concerns)
• Aggregate scores diluted by non-article content types #5 (funding dimension penalizes unfunded research — same structural issue)

Source

Raised by Toby Green (OECD/Coherent Digital) on LinkedIn