Paper Number
ECIS2026-2479
Paper Type
CRP
Abstract
Synthetic data is a promising solution to enable privacy-preserving data sharing, yet there is no existing legally-compliant framework to evaluate synthetic data fairness. This paper contributes to closing this gap by providing the first explicit legal analysis of synthetic data fairness in light of the EU Artificial Intelligence Act (AI Act). Complementing this legal perspective, we conduct a systematic literature review that synthesizes state-of-the-art metrics for assessing synthetic data fairness, revealing two separate strands of bias- and diversity-oriented evaluation. The review uncovers three interlinked problem fields: (1) a lack of operationalization and automation, (2) a lack of governance mechanisms, and (3) legal unclarity concerning fairness obligations. We argue that Information Systems (IS) researchers can play a central mediating role between regulatory expectations, technical metrics, and governance requirements by co-developing operationalizable fairness metrics, context-specific thresholds, and domain-sensitive governance structures together with regulators, supervisory authorities, and inter-disciplinary research communities.
Recommended Citation
Kämpf, Nicki Lena; Perez, Sergio; Rößler-von Saß, David; Sivizaca Conde, Daniel Juan; and Kliewer, Natalia, "Between Regulation and Evaluation: An Information Systems Perspective On Fairness For Synthetic Data In the EU" (2026). ECIS 2026 Proceedings. 17.
https://aisel.aisnet.org/ecis2026/litrev/litrev/17
Between Regulation and Evaluation: An Information Systems Perspective On Fairness For Synthetic Data In the EU
Synthetic data is a promising solution to enable privacy-preserving data sharing, yet there is no existing legally-compliant framework to evaluate synthetic data fairness. This paper contributes to closing this gap by providing the first explicit legal analysis of synthetic data fairness in light of the EU Artificial Intelligence Act (AI Act). Complementing this legal perspective, we conduct a systematic literature review that synthesizes state-of-the-art metrics for assessing synthetic data fairness, revealing two separate strands of bias- and diversity-oriented evaluation. The review uncovers three interlinked problem fields: (1) a lack of operationalization and automation, (2) a lack of governance mechanisms, and (3) legal unclarity concerning fairness obligations. We argue that Information Systems (IS) researchers can play a central mediating role between regulatory expectations, technical metrics, and governance requirements by co-developing operationalizable fairness metrics, context-specific thresholds, and domain-sensitive governance structures together with regulators, supervisory authorities, and inter-disciplinary research communities.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.