CS009 NISS-FCSM: AI in Federal Government (Pt 2)

Conference: Symposium on Data Science and Statistics (SDSS) 2024
06/05/2024: 3:45 PM - 5:15 PM EDT
Special Event 
Room: James River Salon CD 

Chair

Luca Sartore, National Institute of Statistical Sciences

Discussant

David Matteson, Cornell University

Presentations

Implications of federal policy on AI use in the federal statistical system

Chris Marcum from The Office of the Chief Statistician will present insights on the integration of Artificial Intelligence (AI) with the Federal statistical system in light of President Biden's recent executive order emphasizing secure and trustworthy AI development. This order extends across various sectors, emphasizing responsible AI entry not only within the government but also in broader societal contexts. Key aspects include establishing clear definitions for AI and AI models, aligning with Title 3 of The Evidence Act of 2018 and OMB's regulations for secure access to confidential statistical data. The executive order emphasizes lawful and secure data collection, usage, and retention, addressing privacy and confidentiality concerns within the federal system. Additionally, it guides agencies in advancing privacy-enhancing technologies to safeguard American data against risks, aligning with OMB responsibilities. The order also addresses how federal agencies obtain, quality control, and utilize commercially available information, directing OMB to consult with the Federal Privacy Council and the Interagency Council on Statistical Policy to guide agencies in mitigating privacy and confidentiality risks associated with such information. 

Presenting Author

Christopher Marcum, Executive Office of the President, Office of Science and Technology Policy

First Author

Christopher Marcum, Executive Office of the President, Office of Science and Technology Policy

CoAuthor

Nancy Potok, NAPx Consulting LLC

Artificial Intelligence and Official Statistics: Key Elements of a Successful AI Approach

Implementing an AI project as a federal statistical agency involves navigating a complex landscape with various challenges and considerations. Transparent AI systems enhance accountability, trust, and user acceptance, but complex AI algorithms may be difficult to interpret, and there might be tension between transparency and proprietary considerations. However, when working with official statistics, ensuring the accuracy and reliability of AI systems is crucial, especially in sensitive domains. Even with the best technology and statistical methods, projects need a broader framework to be successfully implemented in a federal statistical agency. Understanding these broader elements, including the statutory and other guidance such as Executive Orders and OMB Memoranda, as well as the scientific integrity framework under which agencies support research and evidence-building is crucial to successful implementation. 

Presenting Author

Nancy Potok, NAPx Consulting LLC

First Author

Nancy Potok, NAPx Consulting LLC