Extended definition
The FAIR principles are a set of four principles for research data management, articulated by Wilkinson et al. (2016, Scientific Data): Findable (rich metadata, persistent identifiers such as DOIs, indexed in catalogs), Accessible (open and standardized retrieval protocols, even when the data themselves have restricted access for legitimate reasons), Interoperable (standardized vocabularies, formats, and schemas allowing combination with other datasets), Reusable (clear licenses, rich documentation, sufficient provenance and context for third-party reuse). FAIR does not require data to be open — a critical distinction often missed. Sensitive data (health, industry) can be FAIR without being public: metadata are open and discoverable; data themselves have mediated access. The European Commission adopted FAIR as an operational principle of the European Open Science Cloud (EOSC) in 2018; the NIH adopted it in 2023 (Data Management and Sharing Policy); funders worldwide have aligned guidance.
When it applies
FAIR applies to any project that produces data as substantive output: quantitative, qualitative, observational, experimental, computational research. It is a growing requirement in funding proposals — Data Management Plans (DMP) are a standard deliverable in ERC, NIH, NSF projects, and increasingly in national funder programs. It applies especially in international collaborative projects where reuse by other groups is an explicit goal. It applies in datasets supporting published manuscripts — top-tier journals require deposit in FAIR-compliant repositories (Zenodo, Dryad, Figshare, specific disciplinary repositories). FAIR is also an element of reproducibility: FAIR data are a precondition for third-party computational reproduction.
When it does not apply
FAIR does not replace informed consent — even potentially anonymizable data, if collected with limited consent, have use restrictions that must be respected. It does not replace LGPD/GDPR/HIPAA risk analysis — some data, although technically FAIR-compliant, require protection that limits accessibility. It does not apply directly to industrial data under restrictive IP contracts without renegotiation. In research with Indigenous peoples, CARE principles (Collective benefit, Authority to control, Responsibility, Ethics) complement FAIR — in some cases, they dictate data sovereignty approaches that limit the scope of “Findable” and “Accessible”.
Applications by field
— Health and biomedical sciences: Open Targets, ENA, dbGaP repositories; FAIR is structural in genomic consortia. — Environmental sciences: PANGAEA, GBIF — exemplary FAIR application in climate and biodiversity data. — Social sciences: ICPSR, DataverseNL — FAIR repositories for survey and experimental data. — Digital humanities: transcriptions, text corpora, and digital editions in CLARIN, DARIAH repositories.
Common pitfalls
The first pitfall is confusing FAIR with Open Data — FAIR is compatible with restricted access provided metadata are open and the access protocol is standardized. The second is treating FAIR as an administrative checklist — real implementation requires technical decisions (metadata schemas, controlled vocabularies, persistent identifiers) that require expertise. The third is depositing a dataset in a random repository with low indexing and considering “Findable” achieved — the repository must have reach and discovery infrastructure (federated catalogs, integration with Google Dataset Search). The fourth is neglecting rich metadata: a dataset with poor metadata is technically FAIR but practically unreusable. The fifth is confusing Reusable with data quality — FAIR addresses management and availability, not experimental care or methodological rigor; a FAIR dataset can still be of low scientific quality.