ACUS hosted a panel discussion of Artificial Intelligence and Administrative Law Doctrines: Challenges and Opportunities on Thursday, July 9, in partnership with the Institute for Technology Law & Policy at Georgetown University Law Center. The panel was the second in a four-part symposium on Artificial Intelligence in Federal Agencies. The first panel, on June 25, featured the authors of recent, ACUS-commissioned report Government by Algorithm: Artificial Intelligence in Federal Administrative Agencies.
At the July 9, panel, Penn Law Professor Cary Coglianese, Berkeley School of Information Professor Deirdre Mulligan, and Duke University Law Professor Arti Rai discussed the interplay of federal agencies’ use of AI and core administrative and constitutional law doctrines. Georgetown Law Professor David Vladeck moderated the conversation.
Professor Coglianese, the author of a forthcoming report to ACUS on the subject, emphasized the need to agree on a common definition of AI and consider the current baseline when analyzing the potential benefits and costs of AI in government. “We should not necessarily think about artificial intelligence and machine learning algorithms as a brave new world,” he stressed, “but rather as an opportunity, when used well, to improve on the status quo.”
Commending the recent ACUS-commissioned report Government by Algorithm: Artificial Intelligence in Federal Administrative Agencies as a resource “where we can look for a solution,” Professor Mulligan argued that a centralized body should ensure that agencies across the federal government have access to AI expertise. She also urged agencies to apprise the public of the choices they make when they use AI.
Focusing on the experience of the Patent and Trademark Office, Professor Rai explained how the complexity and opacity of government decision-making algorithms can raise due process concerns. She stressed that agencies must consider the need for public transparency alongside the need to safeguard trade secrets and protect against adversarial learning.
Together, the panelists discussed the need to build and maintain AI expertise within the federal service, for example by recruiting and retaining skilled technologists versed in the the opportunities and challenges of government AI and by ensuring that agencies have access to high-quality resources. Panelists also discussed the tenuous balance between secrecy and transparency in agency algorithms and the costs and benefits of outsourcing the development of government AI systems to the private sector.
ACUS looks forward to the continued exploration of the nexus between AI and administrative procedure, including a forthcoming report from Professor Coglianese and the continued compilation of resources on our website.
The third panel—which will take place Wednesday, July 29, from 2:00 pm to 3:15 pm ET—will address the potential for bias in government AI. ACUS Public Member and former EEOC Commissioner Chai Feldblum will moderate a discussion with Professors Kristin Johnson and David Super and Alex Givens, President and CEO of the Center for Democracy & Technology. Registration for the event is available here. Information about the final panel, Artificial Intelligence “In the Trenches”: A View from Inside the Agencies, is forthcoming.
Visit www.acus.gov/AIsymposium for more information about the symposium and ACUS’s other AI initiatives.