As the use of AI proliferates rapidly, its development and deployment across all stages requires understanding and documentation regarding aspects such as its accuracy, fairness, and transparency—collectively called ‘quality characteristics’. This has led to standardisation at forums such as ISO, and creation of obligations within regulations such as the EU’s AI Act. To support this developing landscape, we assess the sufficiency of the Data Quality Vocabulary (DQV), which is a W3C Working Note, for representing AI quality dimensions. We find that while the DQV can be useful, some concepts require changes to the definition to enable use beyond datasets, for AI models and systems. Based on our findings, we propose updates to DQV concepts to allow a more general quality representation and provide taxonomies based on ISO/IEC DIS 25059 standard on AI quality models and the EU AI Act to represent key AI quality dimensions and examples of their use. We conclude with a discussion on the future of AI quality documentation, the role of Semantic Web standards, and facilitating interoperability across value chains and data spaces.
DQV4AI: Representing AI Quality Dimensions for ISO and EU AI Act Using the Data Quality Vocabulary
Albertoni, Riccardo;
2026
Abstract
As the use of AI proliferates rapidly, its development and deployment across all stages requires understanding and documentation regarding aspects such as its accuracy, fairness, and transparency—collectively called ‘quality characteristics’. This has led to standardisation at forums such as ISO, and creation of obligations within regulations such as the EU’s AI Act. To support this developing landscape, we assess the sufficiency of the Data Quality Vocabulary (DQV), which is a W3C Working Note, for representing AI quality dimensions. We find that while the DQV can be useful, some concepts require changes to the definition to enable use beyond datasets, for AI models and systems. Based on our findings, we propose updates to DQV concepts to allow a more general quality representation and provide taxonomies based on ISO/IEC DIS 25059 standard on AI quality models and the EU AI Act to represent key AI quality dimensions and examples of their use. We conclude with a discussion on the future of AI quality documentation, the role of Semantic Web standards, and facilitating interoperability across value chains and data spaces.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
