Excerpted From: Robert Koulish and Kate Evans, Punishing with Impunity: The Legacy of Risk Classification Assessment in Immigration Detention, 36 Georgetown Immigration Law Journal 1 (Fall, 2021) (350 Footnotes) (Full Document)


KoulishEvansThe Biden administration took office amid promises to promote racial justice, decarceration, and a humane approach to immigration enforcement. A flurry of executive orders and policy memoranda announced the administration's efforts toward each goal. However, immigrants as well as their families and advocates have watched a disturbing increase in the number of people detained. Indeed, rather than reversing course from the Trump administration predecessor, these commitments appeared to be sidelined. This increase portends a return to the detention policies of the Obama and Trump administrations: mass incarceration of Black, Latinx, and Asian migrants under the guise of risk mitigation.

Since 2012, the decision to detain or release someone after an arrest for an immigration violation has been guided by the Risk Classification Assessment (“RCA”) system. The RCA is an automated risk tool designed to recommend whether to detain or release migrants pending resolution of removal charges. Modeled on evidence-based criminal justice reforms, the RCA combines database records and interview information into a weighted scoring system that produces public safety and flight risk assessments of “low,” “medium,” or “high” and issues a corresponding custody recommendation. The RCA was programmed to generate one of four recommendations: (1) detain in the custody of the Department of Homeland Security (“DHS”) (no bond); (2) detain, but with an accompanying recommended bond amount; (3) supervisor to determine; and (4) release. An Immigration and Customs Enforcement (“ICE”) supervisor then reviews the RCA recommendation and any accompanying comments and makes a final custody decision including any conditions for detention or release.

In 2014, this research project, called Risk Assessment in Immigration Detention (“RAID”), introduced the immigration RCA for detention to readers of the Georgetown Immigration Law Journal. The article argued that the risk tool would not reduce current levels of over-detention. It contended that competing immigration laws and policies undermined an effective and accurate risk assessment. Author Robert Koulish later introduced data from the tool's pilot in Baltimore, Maryland during 2012. That study remarked on the high detention rates and the inability of risk to predict whether or not an individual would be detained. More recently, in a joint study, the authors analyzed the “black box” algorithm to demonstrate the manipulation of the risk tool over time and linked more restrictive detention outcomes to substantial changes in the risk algorithm.

This Article is the culmination of several years of research on the RAID Project. It breaks down the broad analysis by Mark Noferi and Koulish and goes inside the algorithmic black box previously analyzed by the authors. With the original empirical data presented here, we examine the carceral immigrant state and the mechanisms at play in detaining immigrants despite the best-made rhetoric on taming detention. Our analysis reveals the effects of ICE's manipulation of the RCA algorithm on detention cases and demonstrates how the RCA's recommendations and ICE's detention decisions grew increasingly punitive over time. In sum, we see ICE detaining lower-risk individuals at almost the same rate as higher-risk ones. That is, ICE has used the risk tool as an instrument to legitimize detention, not limit it.

Summary Findings


• About 18 percent of immigrants in ICE custody are mandatorily detained under § 236(c) of the Immigration and Nationality Act (“INA”).

• Almost 20 percent of immigrants in ICE custody are detained despite eligibility for bond under INA § 236(a).

• Almost 55 percent of immigrants in ICE custody are in some expedited process and are not scheduled to see an immigration judge.

• Few detained migrants have committed violent offenses.

• Most cases with “high” flight-risk assessments receive the designation automatically because the individuals are in expedited removal, even if they are asylum seekers with family ties in the United States and strong incentives to pursue their cases in immigration court.

Special Vulnerabilities

• ICE underutilizes the special vulnerabilities designation to protect migrants who face particularly severe harm in detention.

• Only 6 percent of people processed through the RCA were designated as having special vulnerabilities

• People with special vulnerabilities are more likely than those without special vulnerabilities to be released.

• An overwhelming majority of people with special vulnerabilities are nonetheless detained.

Changes to the RCA Algorithm and their Impact:

• Between 2012 and 2019 ICE increasingly detained lower-risk immigrants.

• ICE removed bond eligibility from the RCA recommendations in 2015.

• ICE suspended the release recommendation for all immigrants in 2017.

• By 2019 nearly every migrant was detained without bond regardless of risk level.

[. . .]

Our research tells the story of how the RCA engaged in manipulation, subversion, and bias. It failed to enhance objectivity and transparency in detention decisions, failed to put a dent in mass detention, and instead seemed to rationalize ever more draconian detention outcomes for immigrants. We presented data that show ICE detains low hanging fruit using the risk tool to provide a scientific veneer, beneath which everyone is designated a risk, and almost everyone is detained without bond. Immigration detention policies were hijacked by frontline officers bent on maximizing the detention of immigrants. This occurred because frontline ICE officers were delegated a large amount of discretion in finalizing detention decisions. That discretion was used to manipulate the risk logic and confound efforts to create a rational custody determination process.

ICE's subversion of the risk classification tool provides two teachable lessons for the Biden administration. First, frontline ICE officers have the power to defeat an administration's policy goals through their intransigence and ability to manipulate its implementation. Second, ICE's culture is fundamentally punitive. Rather than political appointees wearing down the intransigence of frontline agents, frontline agents won out over the political appointees. Punitive bias subsumed and subverted efforts to challenge mass detention. Consequently, migrants are unlikely to be treated humanely without subjecting ICE to radical and wholesale change.

Launched on the promise of reducing immigration detention, the RCA ended up permitting and even encouraging the detention of nearly everyone within ICE's grasp. For the Biden administration to make good on its promises of decarceration, racial justice, and a humane immigration system, it must learn from the mistakes in this decade-long, multi-million dollar experiment and heed the calls to dismantle the current enforcement apparatus and put an end to its reliance on immigration incarceration.

Full Research Professor and Director of MLAW Programs at University of Maryland, and Lecturer at Law at Francis King Carey School of Law, Baltimore, MD. B.A. University of Pennsylvania; Ph.D. University of Wisconsin. </p

Clinical Professor of Law and Director, Immigrant Rights Clinic, Duke University School of Law; B.A., Brown University; J.D. New York University School of Law.