Abstract

The field of artificial intelligence (AI) and machine learning (ML) has seen such a rapid and disruptive progress in recent years that it has become ubiquitous. Private companies that produce and disseminate statistics have already explored the potential of AI/ML and can now provide their services in a timely and accessible manner, drawing the attention of various stakeholders and policymakers. In response to the growing demand for more relevant, timely, detailed, and accessible statistical information, national and international statistical organizations have started investigating AI/ML algorithms to enhance their statistical production and better serve the public good. To keep up with the private sector and remain competitive, statistical organizations must move quickly and take advantage of AI/ML models. However, in doing so, they must not compromise on their responsibility to ensure ethical, safe, and confidential use of AI/ML models. While striving to produce results faster and more efficiently, statistical organizations must ensure that their products are compliant with ethical standards and are safe and aligned with existing confidentiality and security rules. they must be prompted towards quality, ethics, sound methods and algorithmic accountability. Improperly deployed AI/ML models can violate privacy, reinforce existing biases, threaten security, and support questionable decisions that can ultimately harm society. Hence, it is crucial that statistical organizations adhere to solid principles while developing AI/ML models and deploying them in production. In practice, this means there is a need for governance frameworks that ensure the responsible design, development, and use of AI/ML algorithms and systems. To address this need, the "Applying Data Science and Modern Methods" Group of the UNECE High Level Group for the Modernization of Official Statistics is working on an international framework that will support the community of official statistics in developing AI systems and ML algorithms that respect human rights and privacy, are fair, transparent, explainable, robust, secure, and safe.