Berkeley Technology Law Journal Podcast: Bias in Algorithms with Professor Robert Koulish

SoundCloudSpotifyGoogle PodcastsApple PodcastsPocketCasts

[Nikki Seichepine] 0:13

You’re listening to the Berkeley Technology Law Journal Podcast. I’m Nikki Seichepine. In today’s episode we will be discussing the use of risk-assessment to inform criminal sentencing and how it’s used in ICE 1 detention centers.

 

Risk-assessments are used to help quantify the likelihood of an individual’s noncompliance with the law. 2 Variables such as age, race, and gender are used when making these tools. 3These assessments are used in decisions surrounding pre-trials, probation, supervision, bond, and re-entry plans. Individuals with higher scores are assigned less favorable conditions, such as stricter release conditions, and often are under more restrictive supervision.4 Initially, policy makers thought risk-assessments would foster consistency and objectivity while reducing bias. However, in recent years, they have been ridiculed by researchers, media, and formerly incarcerated people for promoting common stereotypes surrounding the most vulnerable populations. 5

 

The U.S. Department of Homeland Security (DHS) operates the world’s largest immigration detention center. 6 In 2012, they adopted their own risk classification assessment (RCA) to assist with decisions related to migrants in custody of ICE, such as their bond amount upon release or whether or not to detain them. 7The RCA’s purpose was to limit detention rates by only detaining migrants that posed a danger or flight risk. The assessment was programmed to generate one of the following four recommendations: 1) detain the individual with no bond, 2) detain with the recommended bond amount, 3) give the supervisor discretion to make the decision, and 4) release the individual. However, in 2018, ICE had removed any “release” option from the RCA, allowing supervisors the opportunity to exercise more discretion. 8

 

Today, we sit down with Robert Koulish. Professor Koulish is here to assist us with learning more about how risk assessments are used in ICE and the flaws inherent in the RCA.

 

Okay, Professor Koulish. Thank you again for joining us. I will give you a brief introduction. Professor Koulish is a political scientist at the University of Maryland and serves as the Director of MLaw, which is the University’s law programs. Much of his current research has focused on risk assessment within immigration detention centers. Professor Koulish is the author of Immigration and American Democracy: Subverting the Rule of Law and co-editor of Immigration, Detention, Risk, and Human Rights: Studies on Immigration and Crime.9

 

Okay, to go ahead and get started, we wanted to begin with a brief overview of risk assessments in general to give our audience a little bit better understanding and background. So, in a nutshell, what are risk assessment tools and how are they used in criminal sentencing?

 

[Professor Koulish] 2:46

Well, thank you for having me. I look forward to the conversation. Risk tools are kind of a form of AI, artificial intelligence. They have taken the place of human decision makers, or the alternative they work with human decision makers. So, we don’t want to leave the human element out of the out of the conversation. But a risk tool is supposed to lend objectivity, fairness, consistency, sense of objectivity, rationality, to what otherwise are kind of classified as discretionary, sometimes ad hoc and arbitrary, subjective decisions.

 

So, a risk tool predicts future behavior. It’s used in the criminal justice system in a lot of ways. I can give you a brief overview of how I’m going to be talking about it in the in the immigration context. It takes us over 100 different factors in the risk assessment tool. The risk assessment tool was first introduced in 2009 by the new Obama administration. It was piloted in 2011–2012. It was unveiled nationally in 2013. It’s been used nationally since 2013. And what it does is it takes biographic information­–the age, national origin, sex, country of birth of immigrants that are coming through, immigration history, criminal history, supervision history, the immigration case status, any removal charges that have been filed against the immigrant, the charging document–and with all that information, going through an algorithm, the algorithm then feeds out recommendations. And it feeds out recommendations to release or detain. If they’re recommending to detain with bond, where somebody is eligible for bond, or without bond, and in some circumstances, the risk algorithm plays no part whatsoever. And the ERO or the enforcement removal officer within ICE is the one that makes the entire decision.

 

So, the risk tool is supposed to kind of temper what had been a highly subjective immigration detention process, it was supposed to tailor detention to dangerousness and, beyond that, people that had committed low grade crimes, or just immigration offenses, really were not supposed to be detained. They were supposed to be released on bond or through other forms of community supervision. What we found with our research is quite the opposite happened that the immigration tool failed. It didn’t provide any transparency. It didn’t provide consistency. It didn’t provide objectivity. But rather, we saw it as a tool of punitive bias against immigrants that kind of gave a veneer of objectivity covering a process that was some of the most punitive in our country’s history.

 

[Nikki Seichepine] 6:10

Thank you so much for that. You touched on this a little bit. But my next question was, how do these methodologies used in tools manifest certain biases such as gender, racial or economic?

 

[Professor Koulish] 6:20

So, the tool doesn’t cover economic status. And it doesn’t cover race, but we’re dealing with the questions of race and race bias. In our research, almost everybody that is brought into the immigration system, really the detention system and the use of the risk tool is part of a larger system of immigration enforcement. It’s part of a larger immigration enforcement regime. That is part, there’s a field of study called Crimmigration Studies that kind of studies the increasingly punitive and criminalization of immigrants going back to the 1980s, and the 1990s. And I could talk a little bit more about that, if you will.

 

Overwhelmingly, the enforcement system and as part of that the risk tool has been used to bring Latinx immigrants into the system. And so overwhelmingly, the people that are brought into the system through the enforcement net that has been widened, over a period of time since the 1990s, and much more in recent years, brings in people from Mexico and from countries in the Northern Triangle of Central America, Honduras, Guatemala, and El Salvador, into the system. And the racism is kind of baked into the algorithm and the use of the risk process by being punitive, in that it applies to overwhelmingly to people of color. So, in our study, really, there weren’t any Caucasians in this in our study. So, we were not able to kind of do a comparison of any of any significant sort, due to prove that people of color are treated differently than white people, because the only people that were brought into the system are people of color. So, what you’re looking at is a punitive bias against people of color. And we see the system of racism that’s kind of baked into not only the algorithm, but also the ICE responses to the algorithm in their final decisions, to detain or to release immigrants.

 

[Nikki Seichepine] 8:34

So as far as trying to remedy that what other factors or characteristics should we consider should be considered when creating these tools?

 

[Professor Koulish] 8:43

Well, there’s a lot that can be done. So, the risk tool itself, as it’s currently been used, as I said, is kind of a veneer over a system that is systemically biased against immigrants of color. And so, the risk tool as it’s currently made up, it can be tweaked around the edges, but it’s really not going to make any significant changes and creating a fairer system. I think that there is, I’m not opposed to using the technology and a risk tool in and of itself, if it measured different factors, and if it weighted different factors differently.

 

For example, the risk tool really does not measure or assess too many positive factors, certainly not in the public safety component, it does a little bit in the in the flight component of the risk tool. But if there was a shift in paradigm, and there was a way of kind of measuring the benefits the merits of an immigrant’s case, for example, for relief from removal, the risk tool currently does not measure that at all. If somebody has a credible claim for political asylum, for example, that should count in their favor, because quite frankly, these are victims of torture and persecution and abuse. These are not offenders. These are individuals if they have a credible claim for relief from removal, they have every reason to show up in immigration court for their hearings, these are the things that a risk tool, ostensibly measures. But the risks will simply doesn’t get at these factors at all.

 

So, if you’re going to use a risk tool, the purpose of the risk tool is to lessen detention. And if you’re going to do that, you kind of need a paradigm that is welcoming people into the country that are seeking safe haven and refugee status and other forms of relief from removal. And you need a system that’s not constructed towards the removal of virtually everybody that that comes across the border or that enters ports of entry and seeks relief from removal.

 

[Nikki Seichepine] 10:57

Again, you kind of touched on this a little bit. But we’re still interested in learning a bit more about how risk assessments address individuals with special vulnerabilities, such as people who are elderly, maybe pregnant, gender identity, or sexual orientation. So, could you tell us a little bit more about that?

 

[Professor Koulish] 11:13

Yeah, that’s a fascinating question. So special vulnerabilities was really the most important part of the risk tool when it was being constructed. When ICE and DHS went before UNHCR to kind of present the risk tool. They led their presentation was really focusing on the fact that they said that they were going to be planning on making sure that people with special vulnerabilities were not going to be detained. And so, you also look at the risk tool itself, how it’s constructed, and special vulnerabilities make up the first module within the risk assessment tool itself. That goes and advances the argument that that special vulnerability should be important.

 

As you suggested, there are several special vulnerabilities that that are listed. There’s a total of eleven of them. Seven are kind of listed as priorities. Individuals with special vulnerabilities are, again, not supposed to be detained unless they fit within a part of the law that mandatorily detains immigrants. But, for discretionary decisions, people with special vulnerabilities simply aren’t supposed to be detained. We found in our study that a majority… well, a couple of things.

 

The first thing is that there was a severe undercount of people with special vulnerabilities we found only about 6% of all the people that were assessed for risk were listed as having special vulnerabilities. We believe that the numbers are far higher, especially since a lot of these people are applicants for political asylum. And if you’re an applicant for political asylum, you’re claiming that you’re even into endured, you’re the victim of persecution or well-founded fear of persecution. So those numbers just by virtue of the case status of the individual, the type of case that they have, those numbers should be much higher. So one, the risk tool doesn’t cover a lot of people with special vulnerabilities.

 

Two, of the people with special vulnerabilities we found, well over half of them were detained and many of them were detained without bond. And so, even though the rhetoric was to kind of focus attention on people with special vulnerabilities, our empirical findings were that really special vulnerability, people with special vulnerabilities were a little less likely to be detained than the general population of people that were in custody. But that said, most people with special vulnerabilities were indeed detained. And by the latter stages of the Obama administration and the Trump administration, virtually everyone with special vulnerability was detained without bond. And then under the Trump administration, they were simply not released at all. They weren’t given that option.

 

[Nikki Seichepine] 14:10

Great, thank you so much. I think that covered my general questions. I can go ahead and move along to my partner who will get a little bit more specific and discuss some of your research on ICE, unless you have anything to add so our listeners have a bit more background information or anything you’d like to add, in general, for our audience.

 

[Professor Koulish] 14:27

Well, I mean, I think it’s very, very important to talk about the kind of the framing of the risk assessment tool in the fact that it really was built to be subverted. If you look at the architecture of the risk tool, it was designed in such a way that it made it very easy to be used, as I said before, as a veneer to cover a punitive system in an increasingly punitive system at that. Part of that has to do with the fact that the algorithm simply made a recommendation, and the final decision on whether or not to detain an individual had to do with the discretionary decision of the ICE detention officer. And we have found in parts of our research that increasingly, and at different times, that immigration detention officers when they increase their rates of dissenter overriding the risk recommendation. So, a risk recommendation, let’s say somebody had a low, low, low flight risk, low public safety risk, and they should not have been detained. What you saw is that in the detention of the detention officer, they would change that gradation, and they would decide to detain these individuals anyway.

 

We pushed our analysis a little bit deeper and we saw that that was due to a combination of endogenous and exogenous factors. And the exogenous factors are really, I think, the interesting ones to talk about. And it shows about, really how front level bureaucrats responding to a punitive culture within ICE and more specifically, kind of dictates from the law enforcement union of ICE that I think three fourths of the ERO officers are part of this particular union, the National ICE Council. The National ICE Council is a strongly anti-immigrant organization that tracks back to anti-immigrant organizations around the country. The President of the ICE Council is somebody who has nativist ties and ties to these right wing organizations, and they have fought every step of the way efforts to kind of have the ICE tool actually work which in the beginning, I think it was designed with the best of intentions, and I think it was systematically subverted largely by the kind of frontline bureaucrats and behind that, organizations such as the ICE Council, and we show that statistically, with the filing of publicly release of filing of different lawsuits against the political appointees, and, and prosecutorial discretion, related to DACA and other efforts that were going to make it a little easier on immigrants.

 

In that filing the public statements and of these lawsuits, you saw, we found an increase in the dissents of Enforcement and Removal officers, which had the result of keeping more people detained. And there’s something I think very important about that. And those issues going to the punitive culture of ICE kind of leads us to conclude that any kind of risk tool that or any type of shift in priorities, even in the new Biden administration, to make things easier on immigration detention, are likely to be undermined by rank and file and by the law enforcement unions that are over overseeing the rank and file and kind of making public statements that then lead to changes in the discretionary behavior of these detention officers.

[Nikki Seichepine] 18:36

Great, thank you so much. I think that was a really good overview. I learned so much.  I’m sure my audience did as well. So if you’re okay, I will go ahead and pass this along with Kermit. And we can learn a little bit more about the specifics and how ICE uses risk assessment tools within their own administration.

[Kermit Rodriguez] 18:52

Thanks, professor, and thanks, Nikki, for the opening there. So to dive deeper here, can you explain like what data the algorithm uses to determine the weighted scores that produce assessments on individuals considered low, medium, and high risk to public safety and flight?

 

[Professor Koulish] 19:13

Yeah, so that goes back to the kind of the biographic information that’s taken. The information is taken in two ways during the intake, and one is static, and one is dynamic. The static is mostly the criminal history of the immigrants. And this information is taken from criminal databases that ICE has access to, such as the NCIC criminal database. Alright, so that information goes right into the immigration ident database and a lot of that information then is automatically uploaded onto the intake form and that that information is there for the intake interview.

 

The information on flight risk, is more dynamic. And these questions and the information recorded here has more to do with, on the one hand community ties, so how long the immigrant has been in the country, whether or not they own any property or assets, whether or not they have a US citizen spouse or child in the United States, how long they’ve been living at their current address, how long they’ve been living in the US. These are seen as positive factors that would lower one’s flight risk score.

 

On the other hand, more attention on the flight risk side is paid to kind of more punitive pieces of information that go to whether or not their disciplinary infractions while the individual is being detained needs to be said that there’s no similar question or information taken about good behavior within detention. And there’s every reason to believe that most immigrants are engaging good behavior. That doesn’t count to their benefit. Efforts to abscond in the past that where they didn’t show up for a hearing, efforts to escape, these are factors that are put together to create the flight risk. In terms of the public safety risk, you’re looking at the you’re supposed to be looking at the dangerousness of the of the individual.

 

So, here you’re looking at a combination of previous criminal history. And the scores are weighted depending upon the seriousness or this ostensible seriousness of the offense. And that’s combined with there… so you have a public let me let me step back you have a public safety risk score and a flight risk score. Both risk scores are low, medium, and high. So you do an assessment of the public safety risk score: low medium high. Then, that’s added with the flight risk of low, medium, high and that’s combined with the algorithm to render the recommendation. The recommendation is either to release, to detain bond eligible, to detain without bond, or to defer to the discretion of the ERO officer, in those cases, bypass the risk algorithm altogether. What we’ve seen over time, is that there’s been a shift.

 

Let me go back. One more to add a little bit more detail to the public safety score is that the public safety score measures these offenses within a gradation of severity. And then also looks at it in terms of how long ago these offenses were, were committed. And so all these things come together within the algorithm, before the algorithm kind of spits out its recommendation.

 

What we see over time is that the risk levels, and the risk recommendations have become more stringent between the time that that risk tool was first started in 2012, and 2013. We see that the risk tool was changed, it shifted 19 times, which goes against pronouncements of stability and consistency. Then within those 19 times, there are three qualitatively different periods that we’ve identified, one leading up to December 2013. And then the change in January 2014. That gets us into the second period. And then the third period followed, the Obama “felons not families” policy change, that led to a change in the detention risk tool in February 2015. 10

 

So you see these qualitative gradations over time. As far as the immigrant is concerned, and as far as our studies are concerned, we see that the risk tool has gotten increasingly more stringent and strict and punitive over time from when it was it started in 2012, 2013 through February 2015. By the time we got to February 2015, the Obama administration stopped recommending bond for nearly every one that was in immigration detention. And then during the Trump administration, they got rid of the release option within the risk tool altogether. So pretty much everybody that came into custody as a result of the enforcement efforts were detained. Pretty much everybody was detained without bond. So that’s, that’s kind of a long way to answer your brief question. But we could go into more detail if you there’s more to be said there.

 

[Kermit Rodriguez] 25:08

So, I would actually love to dive deeper into the changes you highlighted. So, in an article you co-authored titled, Punishing with Impunity: The Legacy of Risk Classification Assessment and Immigration Detention, 11 you highlighted four changes to the risk assessment algorithm and their impact. So, as you said, between 2012 and 2019, ICE increasingly detained lower risk immigrants and second, ICE removed bond eligibility from the RCA recommendations in 2015. Third, OCE suspended the release recommendation for all immigrants in 2017. And by 2019, nearly every immigrant was detained without bond, regardless of the risk level. So, can you explain these drastic changes to the risk assessment algorithm?

 

[Professor Koulish] 25:58

Yeah, so the question is, why. And I think it’s a combination of factors in one is politics and policy questions. Also, very, very important are so in the Obama administration, you saw the Obama administration getting increasingly more strict with the changes within the ICE RCA tool. And this had to do with, I think, political concessions that the Obama administration was making the Congress to try to get larger immigration comprehensive immigration reform through and so he kind of conceded a lot of important issues dealing with immigration detention. That’s one.

 

Two, his DACA pronouncements were responded to or translated as far as within the RCA tool to kind of create a hard line that any immigrant that entered the country after January 1, 2014, would be detained automatically without bond. And so, it’s this shift within the risk tool, and within the criteria within the flight risk of flight considerations that made it impossible or virtually impossible for any new arrivals into the country to be released. So the detention rates skyrocketed during that time where almost everybody was detained, and very few people were released.

 

When we get into the Trump administration, there you see a combination of politics and ideology in the Trump administration was by all accounts just egregiously racist towards immigrants. And so, Trump ran on an agenda, the anti-immigrant agenda, and referring to them as murderers and rapists, and it was in incredibly racist, and that translated into shift of policies where nobody was being released.

 

Let me add the other point to this, though, goes to the role of the ICE council, and the impact that the ICE council had on rank and file that made the final decisions on whether or not to detain. So one of the interesting things about the risk tool is that this is kind of in contrast to the criminal justice risk tool. The criminal justice risk tool is kind of measured against a predicate of recidivism. The risk tool is seen to be working as long as recidivism rates go down. There is no correlate to that within the immigration context. And so, a risk tool was really assessed internally as being effective or not, as measured against the rate of dissents or the rate of overrides by the discretionary final decisions of the ICE detention officers.

 

So, what you see is that rather than kind of using the logic of risk that would detain people on the count of degrees of dangerousness, we saw the risk tool kind of manipulated in such a way that it was measured against the descent rate, which was the discretionary unchecked discretion of detention officers that we found were strongly influenced by the ICE council and the ICE council was making strong statements against the enforcement priorities and priorities to only detain higher risk individuals. Their argument was that this forced ICE enforcement and detention officers to violate the statutes that that they say, told them, that they had to enforce the law against all immigration infractions equally.

 

And so what this led to was an enforcement effort really allowing, rather than giving this prosecutorial discretion to the political appointees to come up with enforcement priorities, gave that discretion to frontline enforcement officers, and then frontline detention officers to make decisions. Basically, to go after the low hanging fruit. And so, the example of that is that you had really the lowest league violators, undocumented entries–people that cross the border without papers–being brought into custody, and that these individuals, maybe with one traffic infraction, ending up being detained for weeks, months, even years, without an opportunity for bond or for release. So, you see these two different factors, one are kind of political factors that kind of explain these qualitative changes, then you see within the system, these increase of the dissents, and that these dissents, kind of it was a feedback loop into the algorithm, and led to changes of the algorithm where the algorithm itself became more punitive.

 

 

 

[Kermit Rodriguez] 31:20

Right. And, you know, it’s very clear from your research and a plethora of other articles about this topic that the risk assessment tool has been manipulated. And part of that has to do with the discretionary culture within ICE. And my question for you is why is it difficult to find clear and convincing evidence showing that ICE did manipulate the tool?

 

[Professor Koulish] 31:43

I believe we did.

 

[Kermit Rodriguez] 31:47

But for a long period, would you agree that it was quite difficult because many organizations were filing Freedom of Information Acts? And it took a long time to get to get that evidence.

 

[Professor Koulish] 31:57

Yeah. And that goes to look, ICE and DHS have violated every, almost every aspect of the risk logic. And one of the big aspects is transparency that went that they’re supposed to lend transparency to the system. The risk tool, the use of the risk tool, the business rules, defining how the risk tool is to be used all this was secret, and private, and was not released. I have been trying since 2013, to get this information. And it was not until I was joined by my co-authors and co-colleagues and their brilliant efforts to file not just one, but several Freedom of Information Act requests. And it was only through really protracted litigation, that we were able to reach some kind of a deal with ICE where we never got all of the information that we were looking for. But we got enough in order to begin to do this research.

 

So, I think we began. I think that there’s a lot more information and data to be received by other scholars that have entered the fray and but ICE and DHS, they are secretive organizations. And they do not release information and data on their internal workings unless they are forced to and unless people are incredibly persistent and using legal means to try to get this information. And so, I consider ourselves fortunate.

 

Just as a brief aside, that if we had taken even a couple of weeks longer, we probably would not have received any of this data. We got a judge to order the release of the data that we did get literally 24 hours before Trump’s 2017 inauguration. And from the information, ad hoc information that I have and stories I’ve been told is that a freeze was put on a lot of FOIA requests after Trump took office. So, chances are, we’d still be waiting for that initial information had the judge not acted when it did. So, it was a combination of people, and a combination of luck and persistence and hard work and brilliant lawyers kind of going after DHS and ICE to force the release of this information.

 

[Kermit Rodriguez] 34:37

I agree they’re very secretive. And it’s quite lucky that you did receive the information 24 hours before the inauguration. But what is very confusing to me, though, is that if ICE is blatantly ignoring the risk assessment tools’ recommendations in this choosing to detain immigrants who recommended for release, why not just have a policy of no release? Why still rely on the algorithm in the first place?

 

[Professor Koulish] 35:06

That’s a really good question. And I think that under the Trump administration, had there been a second term, they would have gotten rid of the ICE tool altogether. They totally rendered it useless through no release. I believe, though, that it serves a purpose of legitimacy and it serves a purpose of, again, creating a patina of due process in constitutionality. That does serve as some defense within different courts circuits that can point to that and say that there is this tool and so there is this objectivity and fairness. And so, to that extent, I think it works in on behalf of DHS and ICE to have a tool even if they don’t take it seriously. So, it lends a covering of legitimacy. And I think it’s it serves that purpose more than more than anything else.

 

[Kermit Rodriguez] 36:08

Understood. So if, if it’s possible, I would like to shift now towards discussing immigration populations that are categorized as special vulnerability. We touched on this subject moments ago with Nikki. And so, the DHS has categorized LGBT immigrants under special vulnerability. But according to a report to the Center of American progress, ICE under the Obama administration, were likely to detain LGBT immigrants 88% of the time when they were clearly eligible for release. In the article, you’re right, that people with special vulnerabilities are more likely than those without special vulnerabilities to be released. 12
And yet an overwhelming majority of people with special vulnerabilities are detained nevertheless. So is the risk assessment tool, in your opinion, specifically biased against the LGBT community? Or would you argue that it’s the discrepancy? Or what would you argue that the discrepancy in your conclusion is attributable to the discretion of ICE supervisors?

 

[Professor Koulish] 37:10

Well, it’s combination of the algorithm and ICE supervisors. I think that look, I don’t think that ICE is a friend of the LGBTQ community. And you see that enforcement and you see that in, in story after story of mistreatment and abuse. The LGBTQ community has a very, very difficult time within detention. And that’s been documented by human rights studies for decades at this point. In terms of the risk tool, I think that the risk tool I would attribute I mean, so we found with the research itself, that there’s not one group with special vulnerabilities that is singled out for worse treatment than others when it comes to decisions of whether or not detain, to detain or not are detained with bond or not. So, we don’t see evidence of the LGBTQ community being singled out for even worse treatment. What we do see is kind of an undercount. So we found very, very few people listed as having special vulnerabilities for being LGBTQ. And so I think that the undercount is an important component there. And then the other thing that we want to look at is, as you said, most people with special vulnerabilities generally are subject to being detained.

 

[Kermit Rodriguez] 38:51

And just to follow up with that, like, why should the public be concerned about ICE’s policy to detain immigrants who are considered special especially vulnerable? Like how do they how do their experiences and attention differ from the rest of the immigrant potential population if we can get more to the specifics?

 

[Professor Koulish] 39:12

Well, gosh, this goes to the question of whether or not there should be immigration detention. And I think there’s a strong argument against it. There’s a movement to abolish immigration detention. I think that under the terms of the risk tool, only the most dangerous individuals should be detained. That’s questionable in terms of how that’s measured. And what we have at the Constitution applied, or it would be kind of questions akin to double jeopardy, where you have individuals that already paid, you know, who are being detained for criminal offenses had already been detained for criminal offenses, and then they are handed over to ICE where they’re detained again for the same offense. And so again, it goes to questions of fairness. When you’re talking about individuals with special vulnerabilities. You want to ask, “What purpose is being served by their detention?” And the answer that all of my research has shown is absolutely none. And so, individuals that are victims of persecution, individuals that are vulnerable to being abused in detention, simply should not be detained. If ICE and again, many of these individuals have credible or meritorious claims for relief, and that’s simply not considered as part of the risk tool. If the risk tool is to be revamped. These factors need to be given weight. So special vulnerabilities are given are given no weight here to count against somebody’s likelihood of being detained. Detention should be a last resort; if there’s detention at all. And what we see really since 2014 2015, is that for people in custody, detention has been the default. And so, the system is backwards. And if you’re looking at it that way, individuals that have special vulnerabilities, there’s no argument to be made that they should be detained. If you’re going to find that there are high flight risks that individuals are high flight risks, there are less punitive ways of making sure that they show up. The Biden administration is experimenting with more of these. Some of these were introduced during the Obama administration alternatives to detention were you’re using–and I’m not arguing in favor of them–there’s a long argument here to be made that this enhances the net of social control against immigrant populations, but they can be monitored without being detained. And so, that’s cost efficient. It’s effective. It’s been shown to be effective by the Vera Institute going back decades. And so, there’s simply no reason why these alternatives shouldn’t be used.

 

[Kermit Rodriguez] 42:26

Touching on that, Professor, would you say that there is an economic incentive to keep immigrants detained as opposed to releasing them to the public with these procedures in place to track them?

 

[Professor Koulish] 42:40

Yeah, I think that’s a great question. So it depends. Financial incentives to whom? So, immigration detention is increasingly a big part of the private, the private prisons industrial complex. Quite frankly, private prisons started with immigration back in the mid 1980s, where the first private jails were used to detain immigrants and the state of Tennessee, and then it was largely taken over by the criminal justice system. As immigration detention grew, particularly after the 1996 Immigration Acts that created mandatory detention requirements, and that mandatory detention then created a need for more detention centers to be built. And a majority of these detention facilities were contracted from the federal government to private prison corporations. It’s in their interest to build as many detention centers and to detain as many people as possible.

 

So, and to bring that around politically, is that the lobbying efforts from private prison complex on members of Congress senators and congressmen, particularly on immigration subcommittees, and national security subcommittees, has been grotesque in my opinion. And so they have, in a sense, help to buy a policy that is good for their shareholders. Now in terms of the taxpayer, it’s in the interest of the taxpayer to move from detention-based system to alternative to detention-based system that immigration detention costs between $154 and $165 per night per individual. And an individual could be monitored through an ATD (alternative to detention) for between $7 and $11 per day. So, it’s in our interest to shift away from immigration detention, it’s in the interest of the private prison facilities to have the opposite. And therein lies one of the important political debates that needs to continue, again, in my opinion.

 

[Kermit Rodriguez] 45:06

Thanks, professor. And I know we’re soon at time and I just wanted to tie go back to the constitutional and due process provisions that we had previously discussed earlier. And so there have been many public interest groups that have alleged and lawsuit that the risk assessment tool violates the constitutional guarantee to due process and federal immigration law that calls for individualized determinations about release. And as you know, ICE’s decision is administratively final. So, there’s no process within the agency for people to challenge either the tool’s recommendation or ICE custody determination. So, can you spell out for our listeners, like what are the most convincing arguments as to why the tool violates either the constitutional guarantee to due process or federal immigration law?

 

[Professor Koulish] 45:58

Yeah, I think the case law is pretty clear on that you go back to the Zadvydas decision of 2001. 13And it tells us that detention in almost all instances is a violation of due process. I’m going to just read a quote, “unless the detention is ordered in a criminal proceeding with adequate procedural protections or in a civil case.”14 Immigration is civil. In certain special and narrow non-punitive circumstances, we have shown in our research, and we are not alone, most studies of immigration detention finds it to be punitive and, and unnecessary punishment. Punishment is reserved for the criminal justice system. There’s no place for punishment within the civil law. So, if you’re looking at immigration detention, and if you can show–which I believe we have–that it is punitive. It’s a violation of due process of law, and it’s clearly unconstitutional.

 

I think that you look at that case, and you look at the Demore case, Demore v. Kim, which came in 2003. 15 And these two cases together, really make a strong argument. And many recent cases that have come up to the court and have the ACLU is fighting, among other organizations are kind of leading the battle here. And we’ve shared research with them and findings and signed on to briefs that, to me, it’s a clear-cut case that the detention of immigrants cuts against questions of liberty and fairness, all part of due process. The racist biased within it brings in other aspects of equal protection under the law. It’s a clear issue in terms of whether or not this stuff is constitutional or not.

 

One of the big issues here goes back to our larger question dealing with immigration law. And that the case is that, such as the Demore case, which kind of allowed was a pro detention case, 16 they go back to kind of these plenary power cases, which go back to the earliest Chinese exclusion cases, which is the kind of the history and the origins of federal immigration detention, which kind of suggests that the role of the Constitution is minimalized by a larger role of sovereignty and kind of protecting national security. So the problem is, is that every time the federal government rhetorically uses the word national security, you see judicial deference kind of jump in, and then the courts kind of allow, kind of allow the DHS and an ICE to continue. And so that goes to a different question. But I think that kind of the foundations of immigration law needs to shift and that might help the situation a little bit.

 

[Kermit Rodriguez] 49:11

Great. And so, in light of everything we’ve discussed, right–changes in political administration, questions of liberty, and the constitutional guarantee to due process–where do you see the future of risk assessment tools heading? And where would you like to see it go?

 

[Professor Koulish] 49:31

Under the current incarnation, they’re not serving any positive purpose whatsoever. I think, as I started out, saying, technology has its place, technology is becoming more and more prevalent in all aspects of law. And so I think that it needs to come together with more professional within the immigration context. There needs to be a professionalization: (1) of immigration decision makers  (2) the algorithms have to be… there needs to be oversight. There needs to be accountability. There needs to be a change in paradigm where it’s non-punitive, but it’s welcoming and it’s providing incentives and inducements for people to show up rather than a punishment mechanism that’s looking to just deport remove and separate people from their families and communities for negative factors. And so, I think that it needs to be restructured. And then there needs to be the discussion of whether or not a risk tool should be part of or whether or not there should be immigration detention at all, which is discussion we can have but we haven’t touched on yet. But in the meantime, I think it needs to be restructured and look more in terms of inducements for people to be welcomed into society, rather than as a punitive tool of punishment to help get rid of individuals.

 

[Kermit Rodriguez] 51:01

Thank you so much, Professor. Well, that concludes our interview. And so, thank you very much for your time professor, and Nikki and I are very grateful.

 

[Professor Koulish] 51:10

Thank you. It’s been great. Thank you very much.

 

Closeout 51:12

Thank you all for listening to the Berkeley Technology Law Journal podcast! This episode was brought to you by Kermit Rodriguez, Nikki Seichepine, and Apoorva Singh.

 

The BTLJ Podcast is brought to you by Podcast Editors Seth Bertolucci and Isabel Jones.

Our Executive Producers are BTLJ Senior Online Content Editors Karnik Hajjar and Thomas Horn. BTLJ’s Editors-in-Chief are Loc Ho and Natalie Crawford. If you enjoyed our podcast, please support us by subscribing and rating us on Apple Podcasts, Spotify, or wherever you listen to your podcasts. If you have any questions, comments, or suggestions, write to us at btljpodcast@gmail.com. The information presented here does not constitute legal advice. This podcast is intended for academic and entertainment purposes only.