Data quality requirements for inclusive, non-biased and trustworthy AI : putting science into standards

A decade of rapid development of artificial intelligence (AI) has resulted in a large diversity of practical applications across different sectors. Data play a fundamental role in AI systems, which can be seen as adaptive data processing algorithms that adjust outputs to input training data. This fundamental role of data is reflected in the EU policy agenda where for example guidance on handling the data is specified in the AI Act. In response to the needs of the AI Act, the Joint Research Centre, in collaboration with the European Committee for Standardisation and the European Committee for Electrotechnical Standardisation, organised the Putting Science Into Standards workshop on data quality requirements for inclusive, nonbiased, and trustworthy artificial intelligence. The workshop took place on 8 and 9 June 2022, with more than 178 participants from 36 countries gathering for the first time European standardisation experts, legislators, scientists, and societal stakeholders to map pre-normative research and standardisation needs. The workshop highlighted existing and the need of new standards from the creation and documentation of datasets all along to data quality requirements, bias examination and mitigation of AI systems. The workshop also identified the steps needed to start the process of drafting new standards and recognised that inclusiveness and full representation of all relevant stakeholders, including industry, SMEs representatives, civil society, and academia is crucial. Building a stronger engagement of experts in AI standardisation is essential to contribute to the development of standards not only to support the market deployment of AI systems in accordance with the AI act, but also to support this growing field of research.