Manipulating Risk: Immigration Detention Through Automation

67 Pages Posted: 14 Oct 2020

See all articles by Kate Evans

Kate Evans

Duke University School of Law

Robert Koulish

University of Maryland

Date Written: August 25, 2020


The U.S. Department of Homeland Security arrests as many as 500,000 migrants per year and detains more than 350,000 of them through Immigration and Customs Enforcement (ICE). Since 2012, ICE has relied on an automated Risk Classification Assessment (RCA) system to recommend whom to detain and whom to release. The authors are the first to obtain access to its algorithm and this Article is the first to make that system’s methodology public. While purportedly basing these recommendations on indicia of flight risk and risk to public safety, the RCA in fact relies on an algorithm driven by political preferences. By linking detention to enforcement policy rather than risk, the RCA lost its underpinning in the Constitution. In addition, compromises in its logic thwarted the program’s ability to deliver the harm reduction, transparency, and uniformity it promised. Ultimately, our data and analysis reveal that manipulation of the RCA resulted in automated detention recommendations for hundreds of thousands of people in violation of the Constitution. The RCA thus delivers mass incarceration of immigrants with staggering efficiency. In the end, we argue the RCA supplied a veneer of risk to a tool of punishment.

Keywords: immigration, detention, risk, automation

Suggested Citation

Evans, Kate and Koulish, Robert, Manipulating Risk: Immigration Detention Through Automation (August 25, 2020). Lewis & Clark Law Review, Vol. 24, No. 3, 2020, Duke Law School Public Law & Legal Theory Series No. 2020-65, Available at SSRN:

Kate Evans

Duke University School of Law ( email )

210 Science Drive
Box 90362
Durham, NC 27708
United States

Robert Koulish (Contact Author)

University of Maryland ( email )

College Park, MD 20742
United States
301-405-3175 (Phone)

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics