FAIR Data Principles & AI Ethics: Convergence & Gaps

A systematic analysis of how FAIR, FAIR for Computational Workflows, and FAIR4RS principles align with nine major AI ethics frameworks — and a practical data steward roadmap for bridging the remaining gaps.

Published in FAIR Connect (2025) • DOI: 10.1177/2949799X251403108

The Challenge

Bridging High-Level Ethics and Technical Implementation

AI ethics frameworks published by the EU, OECD, UNESCO, IEEE, and others define principles such as transparency, accountability, fairness, and traceability. However, they remain high-level and policy-focused, lacking concrete technical mechanisms for data governance.

NovaMechanics investigated whether FAIR data principles and their extensions (FAIR for Computational Workflows and FAIR4RS) can provide the missing operational layer — turning abstract ethical requirements into testable, machine-actionable processes.

9
AI Ethics Frameworks Analysed
EU, OECD, UNESCO, IEEE, Singapore, Montreal, Toronto, Beijing, G20
3
FAIR Dimensions Mapped
FAIR, FAIR for Computational Workflows, FAIR4RS principles
5-Level
Alignment Scale
Strong, Moderate, Partial, Implicit, and Weak alignment ratings

Our Approach

A qualitative crosswalk between FAIR principles and global AI ethics frameworks

Survey nine major AI ethics frameworks

Reviewed the EU Guidelines for Trustworthy AI, OECD AI Principles, UNESCO Recommendation on the Ethics of AI, IEEE Ethically Aligned Design, Singapore's Model AI Governance Framework, the Montreal Declaration, the Toronto Declaration, the Beijing AI Principles, and the G20 AI Principles.

Define alignment criteria

Established a five-level evaluation scale (Strong, Moderate, Partial, Implicit, Weak) based on whether frameworks reference specific mechanisms such as GUPRIs, metadata schemas, structured vocabularies, licensing, or provenance tracking.

Map FAIR principles to ethical requirements

Performed a critical text review of each framework against the FAIR, FAIR for Computational Workflows, and FAIR4RS principles. Produced three alignment heatmaps visualising the degree of convergence.

Identify gaps and propose solutions

Outlined where ethical AI guidance remains abstract and proposed seven concrete technical solutions — from GUPRIs for traceability to SPDX licensing for accountability — targeted at data stewards.

Results at a Glance

Strong
Conceptual Alignment
All nine ethics frameworks share terminology with FAIR: traceability, access, documentation, and openness
7
Technical Solutions
GUPRIs, machine-actionable metadata, APIs, metadata persistence, vocabularies, SPDX licences, and provenance records
Heatmap
Visual Crosswalk
Three heatmaps map every FAIR sub-principle against each AI ethics framework
Traceable
AI Governance
GUPRIs and provenance records operationalise traceability and audit trails
Transparent
Decision Systems
Rich metadata and JSON-LD profiles strengthen explainability and reproducibility
Roadmap
For Data Stewards
Practical four-point action plan connecting FAIR solutions to ethical AI pillars

How NovaMechanics Applies This Work

Data & FAIR Infrastructure

NovaMechanics uses FAIR-native architecture, metadata design, and governance mechanisms to support reproducible and accountable scientific AI systems.

Explore FAIR infrastructure

Scientific Databases & Repositories

Deploy governed repositories that preserve provenance, identifiers, and reusable evidence for long-term analytical value.

Explore databases

This study demonstrates that FAIR principles are not just a data stewardship concept — they can serve as the operational backbone for ethical AI by strengthening traceability, accountability, transparency, and reproducibility across the entire AI lifecycle.

Read the full paper