[EPIC] Address funder type

[adambuttrick]

Summary

ROR currently assigns the funder organizational type to entries that have Funder ID mappings, as well as to new funders entering through the curation pipeline. This was introduced to facilitate the transition to using ROR as the primary funder identifier, since many publishers and services have expressed a need for a more scoped view of ROR's data, specifically limited to those cited as funders. Community feedback has surfaced that this results in the funder type appearing on universities, university medical centers, and other organizations that may not traditionally be thought off as funders. Some consumers syncing with ROR have also implemented custom logic to deprioritize or disregard the funder type where an additional type is presented.

Addressing this issues requires tackling two main tensions, namely:

1. ROR has otherwise always followed a "meaning in use" approach to scope and classification, meaning that an organization qualifies as a research organization if it is used as an affiliation on research outputs. By that same logic, an organization asserted as a funder on research outputs is being used as a funder. If we say an organization is not a funding organization despite being asserted as a funder, we need to articulate what does qualify or distinguish an organization as a funder, and why that criteria differs from the one we apply to research-organization scope.

2. Many large funding organizations are not strictly funders. A significant share of major funding organizations are not dedicated funding bodies. Government departments and agencies (e.g. the U.S. Department of Energy) fund enormous volumes of research while being primarily governmental/administrative entities, not grant-making organizations by nature. The Funder Registry never imposed type constraints of this kind, it simply listed entities that fund research, regardless of their other roles.

Goal

Decide how ROR should model the funder role such that we:

1. Remain consistent with ROR's "meaning in use" approach to scope and classification, or explicitly and defensibly depart from it for funders.
2. Allow academic and government organizations that also fund research to retain their primary type(s) without the funder type crowding out or obscuring them.
3. Continue to provide publishers and services a reliable, scoped view of funder data that does not depend on the long-term existence of Funder ID mappings, given the Funder Registry's deprecation.
4. Preserve the ability for publishers and platforms to surface legitimate dual-role funders and funding relationships rather than discarding them.

Questions and Decisions

• Should the funder relationship be modeled as an organizational type and under what constraints?

• How do we qualify an organization as a funder? If assertion-as-funder in the scholarly record is the qualifying signal (mirroring how affiliation-use qualifies research organizations), what is the evidence threshold? A single assertion? A volume threshold? And if assertion is not sufficient, what is, and how do we justify the inconsistency with research-organization scoping?

• How do we serve the scoped-view need post-Funder Registry? Publishers and services have previously stated the need to filter ROR to "organizations that fund research." If we stop relying on Funder ID mappings (which are going away), what replaces them as the scoping mechanism?

• How do we handle dual-role organizations? For entities that genuinely hold both roles (e.g. a university that is also a grant-maker or a government department that is also a major funder), how do we retain the funder signal without it displacing or being conflated with the primary type?

• What is the migration/transition impact? Any change (e.g. dropping funder from types for many records) is a breaking change for some and a fix for others. How do we sequence and communicate it?

Acceptance criteria

• A decision is documented on whether the funder relationship is represented as an organizational type or as a distinct role/attribute, with the rationale for this provided.
• The criteria for categorizing as a funder is defined and documented, including how (or whether) assertion-as-funder in the scholarly record mediates, and how this reconciles with research-organization scoping.
• Curation guidance is updated to reflect these decisions
• A scoped "funders" view is available to publishers and services that does not depend on Funder ID mappings.
• Dual-role organizations retain their primary type(s) while their funder role remains understood and discoverable. Guidance on the dual-role organization is also explicitly documented with examples.
• A migration/communication plan exists for downstream consumers affected by the change, including those who currently rely on the funder type or deprioritize it.

[nveenstra]

Maybe the easiest way is to distinguish the information into an array of roles and a single type. In case of UMCG, where the internal funder body has no separate ror entity, we can't differentiate by creating hierarchical relationships. In other cases the funding body might be a separate child organization. In any case, most cris systems only support one main organization type, so it would be good to implement that to prevent local interpretations of ROR metadata.

[amandafrench]

Related to #325, #250, and #277

[amandafrench]

Issue #325 suggests adding a "primary" or "preferred" attribute to one organization type value per record to mirror how ROR handles external IDs, but that's probably moving to a solution too quickly before thoroughly understanding the various problems and use cases, which are better outlined in this epic.

[nveenstra]

I wouldn't recommend going the preferred route like it is in the external_ids. That just leads to extra data processing and pressure on your infrastructure. I have to iterate the whole array to find the preferred one, as opposed to accessing one standalone attribute directly.

[adambuttrick]

Feedback from Optica:

The existing use of the funder type is a load-bearing dependency for their workflows. Removing or renaming it without a direct replacement would be a breaking change for two of their systems:

• Their funder typeahead queries the ROR API filtered by type=funder to surface matching funder names (preferred and alternative) and resolve a user's selection to a ROR ID. So, the funder type is what scopes that query to funding organizations.
• They maintain static lookups derived from the ROR data dump, keyed by funder type, used at multiple points in a work's life cycle to validate funding metadata in the XML. Removing or renaming the type would require rebuilding those lookups.

On the type vs. role/attribute, either modeling approach works for Optica, provided the scoped view remains queryable in a structurally equivalent way in both the API and data dump. Their main requirement is a stable filter that identifies that an organization is acknowledged as funder on research outputs, independent of any underlying data model.

On dual-role organizations, Optica indicated that their work also requires that the funder signal be preserved for these types of organizations.