FAIR Principles: Foundations & Applications

• FAIR Principles are a framework that makes digital assets findable, accessible, interoperable, and reusable through persistent identifiers and rich metadata.
• The guidelines extend beyond data to include research software, computational models, workflows, and hardware in diverse disciplines.
• Ongoing implementations focus on semantic interoperability, machine-actionable metadata, and automated evaluation tools to ensure reproducibility and compliance.

The FAIR Principles—Findable, Accessible, Interoperable, and Reusable—constitute the dominant paradigm for scientific data management and stewardship across disciplines. Originally formulated for data objects, the principles have been systematically extended to research software, computational models, AI/ML assets, hardware designs, and complex digital objects such as workflows. The essence of the FAIR framework is to maximize the discovery, utility, and reproducibility of digital research assets through persistent identification, machine-actionable metadata, adherence to community standards, and explicit provenance. Across domains from astrophysics to high energy physics, and from bioinformatics to open hardware, the implementation of FAIR is tightly coupled to global identifier infrastructures, open protocol stacks, domain and cross-domain metadata schemas, formal ontologies, and evolving software and service ecosystems.

1. Formal Structure and Scope of the FAIR Principles

The canonical formulation—the Wilkinson et al. 2016 reference—defines four high-level properties, each decomposed into subprinciples:

• Findable: Assign globally unique and persistent identifier (F1); describe data with rich metadata (F2); ensure metadata clearly include the identifier (F3); register data/metadata in searchable resources (F4).
• Accessible: Data and metadata retrievable by identifier using standardized communications protocol (A1); protocol open, free, universally implementable (A1.1); supports authentication and authorization (A1.2); metadata remain accessible even if the data are no longer available (A2).
• Interoperable: Use a formal, accessible, shared, broadly applicable language for knowledge representation (I1); use vocabularies that themselves follow FAIR (I2); include qualified references to other data or metadata (I3).
• Reusable: Describe with a plurality of accurate and relevant attributes (R1); release with a clear and accessible license (R1.1); associate with detailed provenance (R1.2); meet domain-relevant community standards (R1.3).

This structure underpins all rigorous operationalizations, including sector-specific recommendations (astronomy, HEP, life sciences, hardware) and extensions such as FAIR 2.0 for advanced semantic interoperability (Vogt et al., 2024).

2. Domain-Specific Implementations and Community Standards

2.1 Astrophysics: Virtual Observatory and ESCAPE

In the astrophysics community, the International Virtual Observatory Alliance (IVOA) standards are mapped directly onto FAIR principles: IVOA Identifiers (IVOIDs) provide global uniqueness (F1), VOResource records capture rich and protocol-aligned metadata (F2), and the IVOA Registry acts as a machine-actionable, federated discovery system (F4). Data access relies on open protocols—TAP, SCS, SIAP, SSA, DataLink—compliant with A1 and ensuring persistent metadata per A2. Interoperability is guaranteed through community vocabularies (UCDs, VOUnits) and data models (ObsCore, Provenance DM). Curatorial provenance and licensing, while not always mandated by IVOA, are increasingly enforced via best-practice policies (e.g., CoreTrustSeal) (O'Toole et al., 2022, Molinaro et al., 2021, Civera, 2022).

2.2 High-Energy Physics: Datasets and AI Models

In high energy physics, FAIR implementations involve DOI assignment on datasets and models via DataCite or Zenodo (F1), utilization of DataCite schema and community-specific metadata (detector configuration, channels, calibration) (F2), and registration in domain and cross-domain portals (F4). Interoperability is strong at the format layer (ROOT, HDF5, ONNX, TensorRT) and vocabulary level (PDG, HEPMC), while reusability tracks licensing compliance and full data/software provenance. Project templates and open-source assessment tools enable both quantitative and qualitative FAIRness measurement (Roy, 2022, Duarte et al., 2022, Neubauer et al., 2022).

2.3 Research Software and Computational Models

FAIR for software emphasizes persistent identifiers (DOI, SWH-ID), machine-readable metadata (CodeMeta, CFF), modular architecture, and adherence to standard packaging (PyPI, Docker/Singularity). For computational models, the focus is on encoding in open, standard formats (SBML, CellML), versioned identifiers, and full pipeline provenance within public repositories (BioModels, GitHub, Zenodo) (Hasselbring et al., 2019, Mendes, 2023).

2.4 Workflows and Complex Digital Objects

Computational workflows are treated as first-class FAIR objects: each workflow and component gets a persistent identifier, is described by rich, machine-actionable metadata (Bioschemas, RO-Crate), and registered in generic or domain-specific workflow hubs. Interoperability and reusability are enforced by expressing workflows in standard languages (CWL, WDL), containerizing all runtime environments, and capturing execution/log provenance. Provenance subsumes both the design and execution traces, often formalized in PROV-O or RO-Crate structures (Wilkinson et al., 2024, Wilkinson et al., 21 May 2025, Wilkinson et al., 2022).

2.5 Hardware and Non-Software Digital Objects

FAIR for open hardware requires PIDs for each artifact (design files, BOMs, firmware), rich, frequently layered metadata (including explicit licensing for each component), and co-location in repositories supporting long-term preservation and open protocols. The architecture of metadata and the completeness of provenance (including dependency trees across hardware, firmware, and documentation) are central challenges (Miljković et al., 2021).

3. Extensions: Semantic Interoperability and FAIR 2.0

The next phase of FAIR, conceptualized as FAIR 2.0, extends interoperability from basic syntactic conformance to full semantic interoperability. This involves:

• Formalizing machine-actionability: a data object $A$ is machine-actionable if it is machine-interpretable ($A$ references semantic artifacts) and admits a well-defined operation $o \in \mathcal{O}$ such that $o(A)$ is valid.
• Distinguishing terminological (ontological, referential) and propositional (schema, logical) interoperability: requires explicit mappings (e.g., $m: V_1 \to V_2$) with annotated provenance and schema crosswalks $\varphi: S_1 \to S_2$.
• Introducing new sub-principles (F5.x–F7, I4–I6, R1.4): e.g., multilingual labels, explicit schema references, and certainty (confidence) annotations.
• Embedding FAIR Digital Objects (FDOs) with “kernel information profiles” and inter-FDO entity-relationship graphs for automated discovery, operation binding, and provenance chaining (Vogt et al., 2024, Blumenroehr et al., 2024).

4. Metrics, Evaluation Tools, and Automated Assessment

While no universal quantitative formula for “FAIRness” is prescribed, domain-engineered frameworks operationalize compliance via:

• Indicator-based scoring: E.g., FAIR EVA implements each RDA sub-principle as a Python test, assigns per-indicator weights, and aggregates via weighted mean:

$$T_s = \frac{\sum_{i=1}^n P_i \cdot W_i}{\sum_{i=1}^n W_i} \times 100$$

where $P_i$ is the indicator’s percent fulfilment, $W_i$ its weight (Gómez et al., 2023).

• Compliance checklists: Presence/absence of mandatory PIDs, metadata completeness, licensing, provenance, conformity to data standards.
• Validation and continuous integration: Automated workflows test for schema validity, protocol compliance, and metadata presence in both software (GitHub Actions, Travis) and workflow (LifeMonitor, OpenEBench) settings.
• Service- and metadata-driven metrics: For FDOs, the presence of entity-relationship links, unique per-attribute PIDs, and machine-actionable operation bindings serve as compliance proxies (Blumenroehr et al., 2024).

5. Emerging Challenges and Directions

Current and future challenges in FAIR include:

• Persistent identifier management at scale: The completeness and long-term resolvability of PIDs across data, software, models, and hardware.
• Richness and interoperability of metadata: Harmonization of allocation of semantic types, schema crosswalks, and field-level mappings, especially for multidisciplinary or cross-infrastructure assets.
• Machine-actionability and autonomous workflows: Embedding executable semantics, operation registries, and metadata so that digital objects can be processed, validated, and reused without human mediation.
• Balanced FAIRness scoring: Avoiding inflation from infrastructure defaults (e.g., PIDs provisioned at repository level) and improving weighting so richness of I/R indicators is not masked.
• Extending FAIR to new object classes: Addressing unique demands intrinsic to open hardware, AI models, and dynamically composable workflows (Wilkinson et al., 2024, Duarte et al., 2022, Miljković et al., 2021).

Within this landscape, communities continue to evolve governance (open standards consortia, public issue trackers), automate evaluation (FAIR EVA, F-UJI, FAIR-Checker), and promulgate best practices—underscoring that the practical realization of FAIR is an ongoing, globally coordinated project.

6. Reference Architecture: The CEFCA Catalogues Portal Case Study

The CEFCA Catalogues Portal exemplifies full-spectrum FAIR implementation through:

• Findable: Deployment of a VO “harvest” registry (CEFCA Catalogues Publishing Registry), assignment of IVOIDs for each data release, rich VOResource-conforming metadata, and registry-backed searchability for every J-PLUS/J-PAS release.
• Accessible: Data served via IVOA-standard endpoints (SIAP, SCS, TAP, HIPS), combined with a web interface. Each endpoint is validated for protocol compliance, and registry entries are maintained permanently.
• Interoperable: Tabular outputs in VOTable, FITS, CSV; application of UCDs and UTypes for column-level semantics; SAMP integration for seamless tool interoperability; set-linkages for complex release provenance.
• Reusable: Detailed provenance in VOResource records (pipeline version, calibration, reduction), explicit authorship and instructions for publication acknowledgment, support for both re-ingestion and cross-survey integration.

Planned enhancements include DOI assignment, explicit machine-readable licensing, expanded pipeline log exposure, and automated FAIRness dashboards (Civera, 2022).

7. Concluding Synthesis

The FAIR Principles provide a rigorous, yet extensible, meta-framework for digital object management in research, underpinning discovery, automated access, semantic integration, and reproducible reuse. Their impact and technical richness are determined as much by careful technical implementations—identifier infrastructure, layered metadata, open APIs, semantic mapping—as by policy, community engagement, and ongoing evaluation. While interpretations and compliance measures must be adapted to object type and discipline, the universal trajectory is toward programmatic, machine-actionable infrastructures supporting end-to-end research transparency, interoperability, and sustainable reusability.