What tools for a governance of AI solutions to support Official Statistics?

Abstract

Artificial Intelligence (AI) refers to the ability of computers to perform repetitive tasks that would usually require the presence of human intelligence. AI can become a powerful tool in support of Official Statistics, innovating traditional approaches to statistical information production processes throughout the entire data life cycle. To fully exploit the potential of AI initiatives, it is not enough to adopt advanced technology solutions but it becomes strategic to integrate them into a reliable, transparent and consistent framework.

While the speed at which AI solutions are being developed augurs an increasingly rich future of opportunities, it must also find producers of official statistics ready not only to embrace the innovative stimuli but also to know how to guide and govern them to take into account the ethical demands that such a challenge entails. This translates into the need to adequately increase the skills required in the various areas affected by AI while stimulating a reflection on how to equip oneself with a standardized approach and processes to develop AI-based solutions that are in line with the agency's strategic objectives.

Istat has activated several experimental projects for the use of artificial intelligence to support the production of statistical information.

In parallel, through dedicated workshops, training moments and monitoring of the initiatives, it is pursuing a path to build a specific framework for the introduction of AI in the Institute, with the aim of aligning the ethical priorities of AI, ensuring its inclusiveness and sustainability.

In this paper, we present the Institute's experience in the conception, development, and early AI-based use cases, highlighting its challenges and most significant milestones.

We also illustrate how, through the integration of use cases into a reliable and effective monitoring and governance ecosystem, AI applications can be overseen to mitigate their risks without reducing their innovative scope.