Homeland Security Department file photo

Border Agents Hacked Their 'Risk Assessment' System To Recommend Detention 100% Of The Time

Risk assessment was already problematic. Now immigration authorities are throwing its "scientific veneer" to the wind.

U.S. border agents use a computerized system to help make decisions about immigrants facing deportation. Given information about an immigrant—their criminal history, work status, likelihood of fleeing, and so on—the system suggests whether to “detain” or “release” them.

Now, the Trump administration has modified that system to always give the same answer: “detain.”

“To better align [Immigration and Customs Enforcement] processes to the Executive Order on Border Security and Immigration Enforcement, a modification was made to the RCA that removed the system recommendation of ‘release’,” the Department of Homeland Security said in a statement to Quartz.

The change appears to be focused on Section 6 of that executive order, which calls for “the termination of the practice commonly known as ‘catch and release.'” Under “catch and release,” some immigrants would not be detained while awaiting their deportation proceedings. The “release” recommendation was removed last year, when Trump’s order was announced, according to the statement.

Hacking the system to always recommend “detain” certainly calls into question the purpose of a risk assessment in the first place. But the truth is, the assessment system never really worked in the first place.

Immigrant risk assessment was put into action in 2013 under then-president Barack Obama. Immigrants in the U.S. are detained at vastly higher rates than Americans facing criminal trials, and an automated system, it was believed, might help reduce that gap. It failed miserably; the system almost never suggested “release.” From July 2012 to December 2013, it did so only 0.6% of the time (pdf, p. 30), about 1,500 cases out of over 200,000. And even those suggestions were frequently overridden by ICE officials.

Risk assessment scores are famously problematic. For example, the score used by courts in the U.S. to predict how likely a criminal is to recommit an offense has been shown to be racially biased. Indeed, “the immigration risk assessment may only add a scientific veneer to enforcement that remains institutionally predisposed towards detention and control,” wrote the authors of one paper that analyzed ICE’s methodology.

Those reasons were not cited by ICE to justify interfering with the assessment system. Instead, they suggest that the decision was made largely because the word “release” is not supposed to be in ICE’s vocabulary.

The result is a statistically meaningless model that remains part of the border security system, even as it throws its “scientific veneer” to the wind, forced by executive order to predict that immigrants will be dangerous 100% of the time. According to this logic, if you’re an immigrant facing deportation, you are dangerous; and if you’re not dangerous, well then you must not be an immigrant facing deportation.