FAIR Data Principles

The FAIR data principles — Findable, Accessible, Interoperable, and Reusable — are a set of four guiding goals for how research data and its accompanying metadata should be described, deposited, and structured so that both humans and machines can locate, retrieve, and reuse it with minimal manual intervention. They were designed as guidelines for good data stewardship, not as a data-sharing mandate, a single technical standard, or a certification with one pass/fail test — a dataset satisfies FAIR by degree, across fifteen underlying sub-principles, and different disciplines, repositories, and data types satisfy them through different concrete means.

Origin: the 2016 Wilkinson et al. paper

The FAIR principles were first published as Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J. et al. (2016), “The FAIR Guiding Principles for scientific data management and stewardship,” Scientific Data 3, 160018. The paper grew out of a 2014 workshop (the “Jointly Designing a Data FAIRport” meeting in Leiden) that brought together researchers, funders, publishers, and infrastructure providers who were independently converging on similar concerns about data reuse, and consolidated their thinking into a single, deliberately technology-agnostic set of principles. The paper’s authors and the wider community that has since maintained and elaborated the principles operate as GO FAIR, an international support and coordination network that publishes the canonical principle text and related implementation guidance. A key design choice, stated explicitly by the authors, is that FAIR describes characteristics data should have to support reuse (by both humans and machines), not a specific implementation, technology, or piece of software — the same four goals can be satisfied by a genomics-specific repository and a general-purpose one using very different underlying infrastructure.

The four pillars and their sub-principles

Each of the four letters expands into a small number of more specific sub-principles. The wording below follows the original publication and GO FAIR’s current reference text.

Findable

• F1 — (Meta)data are assigned a globally unique and persistent identifier.
• F2 — Data are described with rich metadata.
• F3 — Metadata clearly and explicitly include the identifier of the data they describe.
• F4 — (Meta)data are registered or indexed in a searchable resource.

In practice, F1 is most often satisfied with a persistent identifier such as a DOI minted through a repository (e.g. via DataCite), and F4 by depositing in a repository that is itself indexed by a discovery layer (a domain repository, a generalist repository, or a registry such as re3data).

Accessible

• A1 — (Meta)data are retrievable by their identifier using a standardized communications protocol.
• A1.1 — The protocol is open, free, and universally implementable.
• A1.2 — The protocol allows for an authentication and authorization procedure, where necessary.
• A2 — Metadata are accessible, even when the data are no longer available.

This is the pillar most often misread as a synonym for open. It is not: A1.2 explicitly anticipates that some data will sit behind an authentication or authorization gate — FAIR is compatible with controlled or restricted-access data, including data that must remain closed for legal, ethical, cultural, or commercial reasons. What Accessible actually requires is that the retrieval mechanism be standardized and, per A2, that the metadata record stays discoverable and readable even where the underlying data cannot be. “As open as possible, as closed as necessary” — the phrasing used across EU funder guidance — is a FAIR-compatible position, not a departure from it.

Interoperable

• I1 — (Meta)data use a formal, accessible, shared, and broadly applicable language for knowledge representation.
• I2 — (Meta)data use vocabularies that follow FAIR principles.
• I3 — (Meta)data include qualified references to other (meta)data.

Interoperability is what lets metadata from different datasets, repositories, and disciplines be combined and queried together — it depends on shared, controlled vocabularies and ontologies (domain-specific ones where they exist, general ones such as Dublin Core where they don’t) rather than free-text description alone.

Reusable

• R1 — (Meta)data are richly described with a plurality of accurate and relevant attributes.
• R1.1 — (Meta)data are released with a clear and accessible data usage license.
• R1.2 — (Meta)data are associated with detailed provenance.
• R1.3 — (Meta)data meet domain-relevant community standards.

R1.1 is the sub-principle most directly relevant to rights and licensing work: GO FAIR frames the absence of a clear, machine-readable usage license as a legal-interoperability failure in its own right, independent of whether the data is technically retrievable. R1.2 (provenance) and R1.3 (community standards) are why FAIR implementation looks different by field — what counts as an adequate, domain-relevant metadata standard for a genomics dataset differs from what counts as adequate for a social-science survey or a clinical dataset.

What FAIR is not

Three confusions come up often enough to be worth stating directly. First, FAIR does not mean open — see Accessible above; a fully restricted dataset with a public, standards-compliant metadata record can be FAIR, while an openly posted spreadsheet with no persistent identifier, license, or structured metadata typically is not. Second, FAIR is not a certification with a single pass/fail test or an ISO standard — no central body issues a “FAIR-compliant” stamp; maturity is usually assessed against the sub-principles qualitatively or via community-built self-assessment tools, and compliance is a matter of degree rather than a binary. Third, FAIR is not specific to research software — a related but distinct set of principles, the FAIR4RS Principles (RDA Working Group, approved 2022), adapts the original four goals specifically for research software, which has different reuse mechanics (versioning, dependencies, executability) than a static dataset. FAIR is also frequently discussed alongside, but is conceptually distinct from, the CARE principles for Indigenous data governance — FAIR addresses machine-actionability and technical stewardship, CARE addresses collective benefit and Indigenous authority to control data; the two are complementary rather than competing, and increasingly cited together in funder and institutional policy.

Relationship to funder data management plan requirements

FAIR has moved from an academic proposal to an explicit expectation in funder policy over the decade since 2016. Major funders now reference FAIR directly in their Data Management Plan requirements rather than leaving data-sharing expectations implicit:

• The European Commission’s Horizon Europe programme requires that data management plans address making data “as FAIR as possible,” applying the FAIR principles explicitly as a framing for the DMP’s data-sharing and metadata sections (see CASRAI’s Horizon Europe FAIR Data Management Plan guide for the specific template and requirements).
• Tri-Agency (Canada) policy on research data management explicitly directs that data deposited to a digital repository be “aligned with FAIR principles,” subject to the caveat “where ethical, cultural, legal and commercial requirements allow” — the same open-vs-accessible distinction discussed above, made explicit in funder text.
• US funder policy (e.g. the NIH Data Management and Sharing Policy) does not always use the word “FAIR” itself, but the substantive requirements — persistent identifiers, standardized metadata, deposit in an established repository, clear reuse terms — map onto the same four goals in substance even where the funder’s own language differs.

In practice, a DMP’s job is to state how a specific project will satisfy each FAIR category — which repository, which identifier scheme, which metadata standard, which license — rather than to assert FAIR compliance as a general claim. CASRAI’s step-by-step FAIR checklist guide walks through that operationalization in more detail, and the Research Data Governance guide covers where FAIR implementation sits within an institution’s broader roles, policies, and lifecycle framework for managing data beyond any single project’s DMP.