The Ethical Implications of Synthetic Data in Data Publishing: Towards a Coherent Governance Framework

ABSTRACT The proliferation of generative AI has made synthetic data, algorithmically generated datasets modelled on, but not directly recorded from, real‐world observations, an increasingly prevalent resource in scientific research. Two defining properties distinguish synthetic data from empirically collected data: virtuality, denoting the absence of direct observational records tied to identifiable real‐world entities and statistical fidelity, denoting the degree to which a synthetic dataset preserves the distributional structure and variable relationships of its source data. While these properties enable synthetic data to address data scarcity and facilitate privacy‐sensitive research, they do not render it ethically neutral: synthetic data inherits—and may amplify—biases embedded in its training data, and in the case of partially synthetic datasets, re‐identification risks are reduced rather than eliminated. Existing ethical rules for data publishing are established around the stages of data submission, review, storage and dissemination, were designed for empirically observed datasets and are ill‐equipped to govern synthetic data, resulting in systematic deficits in transparency, quality control and long‐term reproducibility. This paper argues that addressing these deficits requires a fundamental reorientation of data publishing governance: from a result‐oriented model, in which final datasets are evaluated as static artefacts, towards a process‐oriented approach, in which the scientific rigour and ethical integrity of the data generation methodology itself are subject to systematic scrutiny. Accordingly, this paper proposes a tripartite governance framework: at the submission stage, mandatory synthetic data type declarations and structured generation methodology summaries should be incorporated into data specifications; at the review stage, evaluation criteria should be extended to encompass the transparency, fidelity and reproducibility of the generative process; and at the storage and sharing stage, metadata structures and licensing agreements should be revised to ensure the legally compliant and epistemically responsible circulation of synthetic data within and across research communities.

Authors

Institutions

Publication Details

Journal
Learned Publishing
Published
2026-11-13
DOI
https://doi.org/10.1002/leap.2117
Primary Topic
Ethics and Social Impacts of AI
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

The Ethical Implications of Synthetic Data in Data Publishing: Towards a Coherent Governance Framework

Junkai Chen
Learned Publishing
Ethics and Social Impacts of AI
article

The Ethical Implications of Synthetic Data in Data Publishing: Towards a Coherent Governance Framework

Junkai Chen
article en

Abstract

ABSTRACT The proliferation of generative AI has made synthetic data, algorithmically generated datasets modelled on, but not directly recorded from, real‐world observations, an increasingly prevalent resource in scientific research. Two defining properties distinguish synthetic data from empirically collected data: virtuality, denoting the absence of direct observational records tied to identifiable real‐world entities and statistical fidelity, denoting the degree to which a synthetic dataset preserves the distributional structure and variable relationships of its source data. While these properties enable synthetic data to address data scarcity and facilitate privacy‐sensitive research, they do not render it ethically neutral: synthetic data inherits—and may amplify—biases embedded in its training data, and in the case of partially synthetic datasets, re‐identification risks are reduced rather than eliminated. Existing ethical rules for data publishing are established around the stages of data submission, review, storage and dissemination, were designed for empirically observed datasets and are ill‐equipped to govern synthetic data, resulting in systematic deficits in transparency, quality control and long‐term reproducibility. This paper argues that addressing these deficits requires a fundamental reorientation of data publishing governance: from a result‐oriented model, in which final datasets are evaluated as static artefacts, towards a process‐oriented approach, in which the scientific rigour and ethical integrity of the data generation methodology itself are subject to systematic scrutiny. Accordingly, this paper proposes a tripartite governance framework: at the submission stage, mandatory synthetic data type declarations and structured generation methodology summaries should be incorporated into data specifications; at the review stage, evaluation criteria should be extended to encompass the transparency, fidelity and reproducibility of the generative process; and at the storage and sharing stage, metadata structures and licensing agreements should be revised to ensure the legally compliant and epistemically responsible circulation of synthetic data within and across research communities.

Learned PublishingVol. 40(1)
University of Chinese Academy of Sciences (CN)
Ethics and Social Impacts of AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon