Can Artificial Intelligence (AI) help or hinder important decision making by the government? What is the growing role that AI and related technologies play in the federal administrative state? “Administering by Algorithm: Artificial Intelligence in the Regulatory State,” a policy practicum offered in spring 2019 at SLS, tackled these questions.
“Our policy lab was a gathering of 25 students—15 were law students and 10 came from the engineering quad,” says Professor of Law David Freeman Engstrom, who co-taught the policy practicum with with Professor Daniel Ho and Justice Mariano-Florentino Cuéllar. “We advised a federal agency called the Administrative Conference of the United States [on AI use]. This is an agency that has some oversight authority over the rest of the administrative state.”
Freeman Engstrom’s award-winning scholarship has appeared in Stanford Law Review, Yale Law Journal, and Columbia Law Review, among others, and has been cited in scores of federal and state court decisions and litigation briefs. His current work includes a study of the effect of advances in “legal tech” on civil procedure and the civil justice system.
Law student Cristina Ceballos enrolled in the practicum. Her group worked on a project involving border security and facial recognition.
“[We] specifically focused on customs and border protection which is using facial recognition just to identify people at the border,” says Ceballos. “A passport is easily faked. A VISA is easily faked. But your face is very hard to fake.”
Working with artificial intelligence and other technologies raise privacy and civility questions, which participants of the practicum highlight.
“There are a couple of recent studies that show that facial recognition tends to be better on lighter skinned faces and on male faces,” says Ceballos. “Companies are waking up and realizing that there’s ethical problems [with their software].”
The practicum also offered recommendations to companies who are struggling to find the ethical and legal balance in technology—one recommendation involves creating a benchmark comparing an algorithmic tool to the human analogue way.
“Administrative law, as currently structured, does not do a good job of reaching these tools and providing real, legal or political accountability for their use,” says Freeman Engstrom, who is also Associate Dean for Strategic Initiatives at Stanford Law. “And, so, we are developing a series of prescriptions.”
To learn more about this, join Stanford Legal co-hosts Pam Karlan and Joe Bankman for a discussion on AI and the administrative state with Professor David Freeman Engstrom and student Cristina Ceballos. You can listen to Stanford Legal on SiriusXM Insight Channel 121, iTunes, SoundCloud, and YouTube.
This episode originally aired on SiriusXM on April 27, 2019.