The Role of Professional Ethical Practice Standards in Ensuring Scientific Integrity of AI

Advancements in artificial intelligence offer many opportunities for the collection, processing, and dissemination of official statistics. But with these opportunities come some notable challenges, especially regarding quality, transparency, reproducibility, and trust. Fortunately, existing ethical frameworks can help practitioners navigate these challenges. This presentation will explore how ethical frameworks such as the ASA's Ethical Guidelines for Statistical Practice and the Federal Data Strategy's Data Ethics Framework can offer invaluable insight in the context of AI and official statistics. 

Since 1992, Principles and Practices for a Federal Statistical Agency has concisely described the unique responsibilities of federal statistical agencies to provide relevant, timely, accurate, and trustworthy government statistics to inform the public and policymakers. It has guided federal statistical staff as they balance these responsibilities in making decisions--large and small--to carry out their work.
Recently released in its eighth edition, what does P&P ("the purple book") have to say about the ethical use of AI for official statistics? As it turns out, a lot.  

The use of artificial intelligence (AI) is developing at an accelerating pace within National Statistical Organizations (NSO). Just recently, a Centre for AI Research and Excellence (CAIRE) has been created to help achieve Statistics Canada's AI goals such as data processing and automation. However, while AI carries a lot of promising opportunities, it is important to ensure that ethical concerns are addressed. In this presentation, it will be shown how Statistics Canada ensures that scientific and ethical standards are maintained as it expands its use of AI. A case will be made for the importance of contextualizing the use of algorithms within the mandate of a NSO, which, in the case of Statistics Canada, does not allow the use of personal data for administrative purposes. It will be argued that, in many instances, the introduction of AI seems to amplify known challenges for which we already have frameworks. This approach will be contrasted with others that rely more heavily on specific 'AI ethics' frameworks while trying to anticipate the evolution of AI and its impact on society.