Guilty After Proven Innocent: Hidden Factfinding in Immigration Decision-Making

The mere fact of arrest can cause an immigrant to lose their status, their freedom, or even their life. Immigration judges and other agency employees regularly use information from the criminal legal system to justify a discretionary denial, even when all criminal charges have been dismissed, and even when an immigrant is able to provide compelling evidence of innocence. Allegations that were laid to rest in a criminal proceeding pop up again in an immigration proceeding like zombies. Also like zombies, they are effectively impossible to kill.   

 Conventional wisdom among scholars and practitioners holds that there are few, if any, limits on how immigration adjudicators can use information from a criminal case in the exercise of their discretionary powers. This belief rests on the mistaken assumption that, when adjudicators use allegations from a criminal court document against an immigrant, they are merely importing facts already found by the criminal legal system—“balancing factors”—rather than engaging in fact finding of their own. This article identifies the problem of “hidden” factfinding in immigration decision-making and proposes a solution: a framework to govern the use of information from a criminal case in the immigration system and beyond. The framework, which classifies information according to the standard to which it has been adjudicated, would transform discretionary immigration decision-making and open up new avenues for judicial review. The possibility of meaningful review is especially important for bond determinations where immigrants are regularly stripped of their liberty based on mere allegations. The framework also reveals the hidden logic behind immigration’s “categorical” approach for analyzing criminal convictions and shows why recent efforts to erode it are misguided. This piece suggests that a simple evidentiary tweak can help bring discretionary immigration decision-making back in line with the “fundamental norms . . . that animate the rest of our legal system.”

Table of Contents Show

Introduction

Your above detailed arrests are concerning to [us] as they suggest a problematic pattern of behavior and a clear disregard for U.S. law . . . . Although your arrests . . . have been dismissed . . . these charges . . . appear to show you are a risk to the property and the public safety of others . . . . These instances of misconduct . . . must be weighed against the positive equities in the record in order to determine whether [we] should use a favorable exercise of discretion.

-        Letter from United States Citizenship and Immigration Services (USCIS) announcing intent to deny a green card to a victim of domestic violence and human trafficking, March 17, 2016.

Many legal practitioners know that a criminal conviction can make a noncitizen deportable, render them ineligible for immigration relief, and lead to their detention for months or years in conditions indistinguishable from criminal incarceration. But few outside the specialized practice of immigration law understand the extent to which a criminal arrest—even where there has been no conviction, and sometimes even after the defendant has been exonerated—can have the exact same effect. In the denial letter quoted above, even though all charges against the immigrant were dismissed, USCIS nevertheless considered the unproven allegations from a criminal court charging document to be true. This example is in no way an outlier. When adjudicating discretionary applications, the immigration system regularly assumes that unproven and untested allegations contained in documents from the criminal legal system are true, and regularly relies on them to justify decisions to deport, detain, and deny relief. [1] This is true even when criminal charges were dismissed or pled down to lower-level offenses. Indeed, documents from the criminal legal system are frequently given so much weight that no evidence, not even a sworn recantation from the original complaining witness, is sufficient to overcome them.[2] This should be rather stunning to most practitioners of criminal law, or anyone with an understanding of how these documents are created and how inherently unreliable they are. Tolerance for this practice, which has the effect of making arrests the functional equivalent of convictions, is premised on a fundamental misunderstanding of the role that these documents are playing in immigration adjudications. This article aims to correct that error.

The immigration system depends heavily on the criminal legal system to do much of its factfinding work for it. Criminal convictions[3] are presumed to be reliable because of the comparatively high level of process that goes into their creation: even when a case resolves in a plea rather than a trial, the criminal conviction has nevertheless been “tested” by an adversarial system, and the guilt of the individual has been established to an acceptably high degree of certainty to justify the imposition of consequences as severe as the deprivation of liberty or even life.[4] As a result, other legal systems—like the immigration system—requiring a lower degree of certainty than “beyond a reasonable doubt” are able to freely take criminal convictions and use them to make decisions within their own systems.[5] Convictions are one form of factfinding produced by the criminal legal system, but the system does not just produce a binary output of guilty or not guilty. A criminal case, from start to finish, produces a wealth of other “outputs,”[6] which are not all equally reliable but which all tend to benefit from a gloss of legitimacy simply for having been created by the (presumptively reliable) criminal legal system.

The term “outputs” roughly corresponds to documents produced by the criminal legal system including arrest reports, charging documents, indictments, hearing and trial transcripts, guilty and not guilty verdicts, plea colloquies, pre-sentence investigation reports, and restitution orders. But, because a single document might contain different types of information, the term more precisely signifies the information contained within a document rather than the document itself. A charging document, for example, may contain both convicted and unconvicted counts. These have very different evidentiary values[7] requiring different treatment in the second system. They should be thought of as distinct “outputs.”

Whenever outputs from the criminal legal system (CLS outputs) are imported into other systems, there is potential for misuse. Nowhere is this misuse so flagrant as in the immigration system, which is uniquely tolerant of the imprisonment and deportation of individuals based on mere allegations. This unique tolerance is rooted in the vast and nearly unreviewable discretion that immigration adjudicators enjoy—a level of discretion that is almost certainly greater than that enjoyed by decision makers in any other civil context.[8] The determination by an immigration judge or a USCIS adjudicator[9] as to whether an immigrant is eligible for relief or a benefit[10] is a nondiscretionary determination, but, with a few exceptions, baseline eligibility is only the first step. For admission to the United States, and for almost every other form of immigration relief, there is a second step in which the adjudicator gets to decide whether to grant the application.[11] The burden at this second step rests squarely with the individual applicant. At the nondiscretionary step, there are rules governing how information from the criminal legal system can be used in aid of decision-making.[12] At the discretionary step, there are none.[13] At least, that is the conventional wisdom.

The belief that the immigration system is unconstrained in how it may use CLS outputs rests on the assumption that adjudicators are not engaged in factfinding of their own but merely using facts already found by the criminal legal system in aid of their discretionary decision-making. This assumption cannot stand up to any degree of scrutiny: immigration adjudicators are necessarily making a determination that unadjudicated allegations are sufficiently likely to be true to justify a discretionary decision. This is factfinding. But this factfinding has been hidden because it is conducted in aid of discretionary decision-making. Adjudicators characterize their work as the “balancing” of “factors” rather than the finding of facts, even as they import wholesale “facts” that were never adjudicated to be true by any system to any standard of proof.

The mischief of hidden factfinding is not so much the use of any particular CLS output but rather the impoverished and unaccountable decision-making process that it engenders. The immigration agency[14] has consistently failed to see its own factfinding as factfinding. As such, it has failed to engage with adjudicators’ factfinding in any meaningful way. Consequently, adjudicators either abdicate responsibility to evaluate CLS outputs, accepting the assertions therein at face value, or engage in a cursory and simplistic analysis that fails to distinguish between different types of outputs created with vastly different levels of process.[15] And reviewing courts, laboring under the same misapprehension, fail to correct them.

While judicial review should be available to correct agency errors of this magnitude, the promise of review has proven illusory because courts have failed to squarely analyze agency factfinding as factfinding. This failure obscures what should be a straightforward due process analysis. In bond determinations, for example, immigration judges are charged with deciding whether to deprive an individual of their liberty for months or even years while their immigration case proceeds. The deprivation of liberty in this context has been found constitutional only when certain facts are found. But to what standard must these facts be found? And what evidence might be sufficient to provide the required certainty? When might it be permissible to require an immigrant to affirmatively establish their innocence of criminal charges, and what evidence might be sufficient to do so? Because they tend to uncritically accept the agency’s erroneous belief that adjudicators are merely exercising their discretion rather than finding facts, reviewing courts have, so far, failed to answer these questions, and due process violations affecting the most fundamental and protected rights remain unaddressed.

Hidden factfinding is a simple error that creates tremendous consequences for individual litigants, affecting hundreds of thousands of immigrants and their families every year.[16] It also undermines the integrity of the entire immigration system. The discretionary decision-making process, enfeebled by hidden factfinding, is even worse than is implied by the famous characterization of immigration law as “death penalty cases in a traffic court setting.”[17] At least in traffic court there are standards of proof and the possibility of review.

This piece attempts to fill a hole in the scholarship by identifying the problem of hidden factfinding and proposing a solution. The scholarship in this area can be roughly divided into two categories: “crim” scholarship focused on the use of CLS outputs in other systems (including “crim-imm” scholarship analyzing the use of these outputs in immigration adjudications) and “discretion” scholarship aimed at explaining, and often decrying, the ways that the discretion unique to the immigration system has affected its decision-making.

In the former category, scholars such as Eisha Jain have examined the use of CLS outputs in other legal systems as proxies for information that the second system does not wish to adjudicate for itself.[18] This includes the use of criminal convictions to trigger consequences in other legal systems and in the immigration system.[19] It also includes scholarship on the ways that arrests themselves can trigger these consequences.[20] Scholars have also examined the use of convictions and other CLS outputs in nondiscretionary immigration adjudications like determinations of deportability and first-step eligibility for relief.[21] Finally, scholars have described and critiqued the use of the criminal legal system as a way to identify and select immigrants for enforcement.[22]

In the latter category, there is no shortage of excellent scholarship grappling with the extent to which discretionary immigration decision-making has become an outlier such that “[p]robably no other area of American law has been so radically insulated and divergent from those fundamental norms of constitutional right, administrative procedure, and judicial role that animate the rest of our legal system.”[23] Scholars, justifiably uncomfortable with this divergence, have proposed various ways that immigration law might be reformed or recontextualized to bring its discretionary decision-making back into the fold. Hiroshi Motomura, for example, explained that procedural due process has been drafted to serve as a “surrogate,” if a flawed one, for substantive rights blocked by the plenary power doctrine, ultimately arguing that the idea of plenary power ought to be abandoned.[24] Gerald Neuman posited that the conception of immigrants as lacking substantive due process rights is not inevitable and proposes that deportation might be recontextualized as an administrative sanction or deprivation of physical liberty.[25] Daniel Kanstroom focused on the supposedly unreviewable discretion of the agency, arguing that discretion has been misunderstood as unitary where really there are different types of discretion, not all of which require the deference they have been given.[26] Shoba Sivaprasad Wadhia, citing the arbitrariness of discretionary decision-making, advocated for the elimination of what she terms “darkside” discretion—meaning discretion to deny a benefit to an otherwise eligible applicant—or, at the very least, that the agency provide clearer standards for how discretion can be exercised.[27] And Shalini Bhargava Ray advocated for reforms that would force immigration officials to articulate a reasoned explanation for exercising discretion for or against an immigrant.[28]

Finally, blending the categories, scholars have argued for the incorporation of norms from the criminal legal system into the immigration system. Scholars including Peter Markowitz and Mary Holper, relying on the Supreme Court’s watershed acknowledgement in Padilla v. Kentucky, 559 U.S. 356 (2010) that deportation is not purely civil and cannot be categorized as a mere “collateral” consequence of a criminal conviction, have argued for incorporating some criminal law protections into immigration proceedings including the confrontation clause of the Sixth Amendment.[29] Others including Kanstroom, Angela Banks, and Jason Cade have advocated for the application of the Eighth Amendment’s proportionality principle to immigration law, a proposal that would require cabining the current broad discretion to deport based on minor criminal infractions or technical immigration law violations.[30] And Fatma Marouf argued that immigration law must incorporate criminal law’s presumption of innocence when dealing with CLS outputs in order to protect the fundamental fairness of the immigration proceedings and the integrity of criminal proceedings, and also to avoid amplifying the disproportionate impact of the criminal legal system on communities of color.[31]

This article exists in the same scholarly vein as the “crim” scholarship in that it critiques the use of CLS outputs in a particular context. It also exists in the vein of the “discretion” scholarship in that it attempts to help discretionary immigration decision-making conform to norms of fair play and justice that advocates, scholars, and many courts sense should apply, except that there is no satisfactory mechanism for applying them.[32] This piece is unlike much of the “discretion” scholarship, however, in that my proposed intervention has a more discrete focus. (Indeed, the problem of a siloed and unaccountable agency is bigger than the problem I identify of CLS outputs being misused in some of its adjudications.) What I am proposing is not a constitutional realignment so much as an evidentiary tweak. At its heart, the use of information from the criminal legal system by any other system presents an evidentiary question: which outputs are reliable enough to use in an adjudication and, if used, to what purpose can they be put? To answer this question, I develop a framework to analyze and guide the use of CLS outputs in other legal systems, including the immigration system.

In Part A, I explain how the immigration system tolerates the misuse of CLS outputs and demonstrate in the process how this acceptance is premised on a fundamental misunderstanding: the failure to recognize “hidden” factfinding. I outline the (meager) existing guidance available to adjudicators making discretionary determinations that require factfinding and explore the role that CLS outputs are actually playing in these determinations, in a way that has not previously been done in the literature.[33] I show that adjudicators relying on CLS outputs are indeed engaged in factfinding, but their factfinding has heretofore been mischaracterized and hidden from scrutiny under a broad cloak of executive discretion.

In Part B, I propose a framework to guide the use of CLS outputs in other systems, including the discretionary immigration system. This framework introduces the concept of “output portability,” which is determined based on whether the output has been adjudicated and to what standard. Only outputs that have been adjudicated to a sufficiently high standard can be imported wholesale into a second system. The remainder must be the subject of further factfinding. The first step is to clearly identify what factfinding the criminal legal system has conducted and the standard to which the relevant facts were found. The next step is to compare that standard with the evidentiary standard required by the importing system. This allows the adjudicator to assess the “portability” of the output. For example, because of the high standard to which the relevant facts are found, a guilty verdict is fully “portable” (usable as evidence without additional factfinding) into most if not all legal systems. Other outputs may be partially portable (usable as evidence, but requiring additional factfinding) or nonportable (unusable). Under this framework, the majority of CLS outputs will be partially portable into the immigration system for use in discretionary adjudications. This means that adjudicators can use them in aid of factfinding but must not treat them as if they have already been proven to be true.

In Part C, I explore the implications of the output portability framework. One important implication is the ability to draw on law that already exists in the nondiscretionary context. Another, perhaps more critical, implication is the change that identifying hidden factfinding works on the due process analysis. Without necessarily assailing assumptions about the breadth and depth of the deference due to the executive in the exercise of its discretionary powers, I nevertheless show how it might not be “fundamentally fair” to use the outputs in the way they are currently being used. This is especially true for bond determinations, which, I argue, are mischaracterized as purely discretionary determinations but are really factual eligibility determinations with a second discretionary step. The factual eligibility determination is not committed to agency discretion but rather serves to limit the agency’s power to detain. Viewed this way, decisions that were previously misunderstood as unreviewable exercises of discretion become amenable to review. Finally, I demonstrate how the output portability framework has implications beyond discretionary immigration decision-making, including helping to clarify certain aspects of immigration’s “categorical approach” for analyzing criminal convictions. I also briefly discuss how the framework might be useful in other legal systems where CLS outputs are employed.

Analysis

A.     Hidden Factfinding and the Cloak of Discretion

It is an oft-repeated maxim that immigration adjudicators are prohibited from relitigating the guilt or innocence of the immigrants before them.[34] As I explain in this section, this maxim is only half true. Immigrants convicted of a crime cannot introduce evidence of their innocence but, for those who were not convicted, immigration decision makers are allowed, and at times encouraged, to consider extraneous evidence of their guilt. This is equally true for immigrants who pled down to a lesser charge to resolve their criminal case and for immigrants whose charges were dismissed and who were never convicted of anything at all.

I represented the applicant who received the denial letter that opens this article. She had been living in the United States since she was a young child and was forced into prostitution beginning at age fourteen when she was in foster care. She “escaped” this horrible situation by running away with a boyfriend who abused her physically and sexually. Her cooperation with police and prosecutors led to her boyfriend’s arrest and conviction and made her eligible for a special humanitarian visa available to crime victims, the “U” visa. After receiving her visa, my client continued to struggle with unstable housing and predatory adults who attempted to coerce her into sexual activity and who, in two instances, involved the police as a means to further exercise control over her. She was an excellent candidate for a permanent green card. Yet, when it came time for her to convert her visa into permanent resident status, the agency announced that it considered the untested and unproven allegations in criminal complaints as indicative of a “pattern of behavior” that justified denial of her green card and, with it, her chance to remain in the United States.

My client’s case is not an aberration. The denial letter was in perfect conformance with agency guidance and is similar to the experiences of many other immigrants and the observations of other advocates. The same pattern plays out across a vast spectrum of discretionary adjudications including bond, cancellation of removal, DACA, asylum, and more. In this section, I explore the legal landscape that allowed for this outcome. In the process, I expose the fundamental misconception about factfinding that makes current agency practice possible.

1.     The Immigration Agency’s Inconsistent Guidance

The Immigration and Nationality Act (INA) is silent on the use of CLS outputs, including criminal convictions, in discretionary decision-making, and the Board of Immigration Appeals (the BIA)[35] has long relied on this silence to justify their use.[36] The clearest expression of a rule for these outputs, for any immigrant who has ever had contact with the criminal legal system no matter how the case resolved, would be: anything can come in. This rule is sometimes, but not always, tempered by the qualification that CLS outputs must be given their proper weight.

a.     Anything Can Come In

The BIA first approved the use of CLS outputs where an immigrant had not been convicted of a crime in the mid-1990s, at a time when the supposed criminality of immigrants was much on the public mind.[37] The case, Matter of Thomas,[38] is unique because it is one of the few cases to acknowledge, though only implicitly, that immigration adjudicators would be engaged in factfinding using these outputs. In Thomas, the BIA dealt with a situation where the conviction was still on direct appeal and could not be considered final. Tellingly, the BIA did not consider an earlier case allowing the use of CLS outputs where a criminal conviction had been voided on a technicality[39] to be controlling since the voiding of the conviction, the result of a mere “peculiarity of state law,”[40] did nothing to undermine confidence in the respondent’s guilt, or, put differently, the criminal court’s factfinding. In contrast to that earlier case, the appeal in Thomas represented a substantive challenge to the conviction. The BIA could not simply import adjudicated facts from the criminal legal system because the system was not finished with its factfinding process. It, therefore, had to consider CLS outputs as weighable evidence:

When an alien’s conduct results in his having had contact with the criminal justice system or being placed in criminal proceedings, the nature of those contacts and the stage to which those proceedings have progressed should be taken into account and weighed accordingly. Hence, the probative value of and corresponding weight, if any, assigned to evidence of criminality will vary according to the facts and circumstances of each case and the nature and strength of the evidence presented.[41]

The BIA characterized CLS outputs without a conviction as “some evidence weighing against discretionary relief.”[42] CLS outputs were “evidence” of criminality, having some probative value, but not necessarily providing definitive proof of the individual’s conduct.[43] Two weeks after Thomas, the BIA issued a rare reversal of an IJ decision using a CLS output as a basis to deny relief. In that case, Matter of Arreguin, the BIA reversed an IJ’s discretionary denial that had relied on facts contained in a police report to justify its decision. Ms. Arreguín had been arrested but had not been convicted of any crime. The BIA found that the IJ had given the report too much weight.[44] This is one of the last times that the BIA seems to have contemplated that immigration adjudicators considering CLS outputs might be engaged in factfinding.

The consistent and rational approach of Thomas and Arreguin was short lived. The following year, the BIA relied on Thomas to approve using allegations in a police report, but jettisoned Thomas’s implicit understanding that the allegations were evidence, not found fact.[45] In that case, the BIA stated that the police report “bears on the issue of the respondent’s conduct.”[46] There was no discussion as to the weight that should be given to the police report and no acknowledgement that the IJ had engaged in factfinding by choosing to credit it. Since then, Thomas has been consistently cited for the proposition that CLS outputs can come in even in the absence of a conviction, but not for the proposition that they can come in to aid factfinding rather than as already-established facts.[47] One minor exception is a case where the BIA acknowledged that an arrest warrant from Mexico should not be held against the respondent. But even there the BIA was ambiguous as to whether it was excluding the evidence or merely giving it little weight.[48]

b.     But Give it its Proper Weight

In addition to its acknowledgement that adjudicators are conducting factfinding on CLS outputs, Thomas is also important because it represents the first attempt by the BIA to create guidance on how to weigh these outputs—how to determine their evidentiary value. It also has the distinction of being one of only a few gestures at such guidance that the BIA has made, albeit guidance that has been largely ignored and contradicted by later decisions.

In Thomas, the BIA instructed that CLS outputs should be evaluated based on the “stage to which [the] proceedings have progressed” with outputs gaining weight as the criminal case moves forward. Two weeks later, in Arreguin, it stated that the facts in CLS outputs should only be credited if “corroborat[ed].”[49] Taking issue with an IJ’s use of an arrest report as a “negative factor,” the BIA said, “[j]ust as we will not go behind a record of conviction to determine the guilt or innocence of an alien, so we are hesitant to give substantial weight to an arrest report, absent a conviction or corroborating evidence of the allegations contained therein. Here, the applicant conceded that the arrest took place but admitted to no wrongdoing. Considering that prosecution was declined and that there is no corroboration, from the applicant or otherwise, we give the apprehension report little weight.”[50]

These two cases can be read together to lay out rudimentary guidance for evaluating the weight of CLS outputs. Outputs gain weight over the course of a criminal case, possibly due to corroborative actions that occur along the way. In this way, the case with the foreign arrest warrant can be read as consistent with Thomas and Arreguin as the BIA, though it did not use the word “corroboration,” may have been saying that it considered the foreign warrant insufficiently corroborated “in the absence of a conviction[.]”[51] Alternatively, the result might be explained by the fact that the warrant came from a foreign jurisdiction, which the BIA may have been unwilling to credit as reliable.

The BIA, however, appears to have abandoned that rudimentary guidance almost as soon as it was issued. A decision from 2006 seems to establish separate and competing guidance on the weight of CLS outputs. In that case, Matter of Guerra, the immigrant respondent was out on bail on a criminal charge and had not been convicted of any crime. The BIA credited the allegations in the criminal complaint because they were “specific and detailed” and the complaint “describe[d] the source of the information that the respondent was involved in the sale of drugs.”[52] The BIA also noted that the complaint had been signed by the DEA agent who prepared it.[53] Finally, the BIA cited the lack of contrary evidence.[54] The BIA seemed not to have considered corroboration at all, as none of the factors it highlighted establishing the supposed reliability of the complaint—its detail and specificity, the fact that it was signed, and the lack of contrary evidence—would constitute corroboration.[55]

Another recent decision[56] illustrates the BIA’s lack of consistency. Puzzlingly, the BIA failed to cite Thomas, Arreguin, or Guerra and ducked any responsibility for overseeing an IJ’s evaluation of the proper weight of a CLS output. The respondent had argued that, even though he pled guilty to the charged crime, the IJ should not automatically credit the allegations in the charging document as true because he only pled guilty to the broadly worded statute and did not admit to the specific allegations in the document.[57] Surprisingly, the BIA did not say anything about a guilty plea to a charge corroborating the allegations. Nor did it say anything about how far the criminal proceedings had progressed (indeed, up to conviction for that very charge). It did not engage with corroboration or weight at all. Instead, it merely reiterated that the IJ had found the criminal complaint “reliable.”[58] If the IJ below provided any justification for why the complaint was reliable, it was not echoed in the opinion, and the BIA did not offer any analysis, positive or negative, of that finding.[59]

In sum, while the BIA in Thomas and Arreguin suggested a way to evaluate CLS outputs, ultimately ad hoc and case-by-case determinations have been the practice. Adjudicators have continued to wing it, and the BIA has failed to enforce fidelity to any guiding framework.

2.     Hidden Factfinding and an Impoverished Decision-Making Process

The agency’s failure to provide comprehensive, or even internally consistent, guidance on the use of CLS outputs in discretionary decision-making is attributable to its failure to recognize that adjudicators are engaging in factfinding on these outputs. The case law is replete with examples of confusion as to the locus of factfinding, conflation of “facts” and “factors,” and general squishy thinking. As a result, as I discuss below, the agency has endorsed practices including reading criminal complaints and police reports as absolute truth, incorrectly assuming complaints and police reports are corroborative of one another, and faulting respondents for testimony that conflicts with these documents.

The agency’s treatment of CLS outputs goes beyond holding respondents to their burden of proof and crosses into what I call burden multiplying: allowing CLS outputs to trigger an additional and impossible burden on respondents to prove their falsity, all the while giving undue weight to the CLS output and refusing to credit any contradictory evidence produced by the respondent. Allegations from the criminal legal system, even those that have been put to rest by dismissal of the charges, return from the dead like zombies. Once reanimated, these “zombie” allegations are impossible, or nearly impossible, to kill. To put it slightly differently, arrests are treated as the functional equivalent of convictions, triggering almost-certain denial of discretionary benefits.

a.     The Problematic Logic of the Agency

Agency decisions are marked by profound confusion as to the locus of factfinding and the difference between weighing factors and finding facts. This confusion dates back to 1977, when the BIA found that a respondent’s conviction was the result of a “particularly despicable criminal scheme[,]” based on information contained in the criminal indictment, including unconvicted counts.[60] Because the BIA appears to have assumed that the necessary facts as to the “nature of the criminal activity” had already been found by the criminal legal system, it did not discuss the weight of the information on which it relied. Several years later, the BIA accepted information contained in a police report indicating that the respondent had been arrested with over five-hundred grams of marijuana.[61] The five-hundred-gram quantity was presumably not part of the charge to which the respondent pled guilty because, if it were, the police report would not have been needed.[62] Even though the police report had not been part of the adjudication conducted by the criminal legal system, the BIA treated the five hundred grams as a proven fact, failing to recognize that it had never been subject to judicial scrutiny. The BIA justified its decision to consider the police report by saying “all relevant factors concerning an arrest and conviction should be considered to determine whether an alien warrants a favorable exercise of discretion.”[63]

Perhaps the clearest expression of BIA confusion as to the locus of factfinding can be found in Guerra:

An Immigration Judge has broad discretion in deciding the factors that he or she may consider in custody redeterminations. The Immigration Judge may choose to give greater weight to one factor over others, as long as the decision is reasonable. In the present matter, the Immigration Judge determined that evidence in the record of serious criminal activity, even if it had not resulted in a conviction, outweighed other factors, such that release on bond was not warranted. In light of the broad discretion afforded under section 236(a) of the Act, we find no error in the Immigration Judge’s consideration of the information regarding the respondent’s alleged involvement in a drug trafficking scheme in determining whether the respondent poses a danger to the community.[64]

This quote epitomizes the problematic logic of the BIA in these cases: the BIA failed to distinguish between balancing factors and finding facts. The IJ had evidently found that the “evidence in the record of serious criminal activity” was strong enough to warrant a decision to detain the respondent. This is factfinding. But, by characterizing such factfinding as the mere balancing of factors, the BIA refused to engage with the IJ’s finding of fact.

USCIS is susceptible to the same faulty logic. USCIS’s reviewing body, the Administrative Appeals Office (AAO), wrote the following opinion denying adjustment of status to a U visa recipient[65] based in part on a ten-year-old arrest that resulted in dismissal of all charges:

[W]e acknowledge . . . that relevant state law provides that, when charges are dismissed and sealed in favor of the accused, the prosecution shall be deemed a nullity . . . . However, the fact that the Applicant has not been convicted of the underlying crime does not equate with a finding that the underlying conduct in question did not occur, and USCIS may take into account all factors in making its discretionary determination.” [As a result,] reliance on an arrest report in adjudicating discretionary relief—even in the absence of a criminal conviction—is permissible provided that the report is inherently reliable and its use is not fundamentally unfair . . . . As such, the police report or other similar documentation related to this arrest would be a relevant consideration regarding the Applicant’s conduct. Ultimately, the decision in this case focused on the Applicant’s conduct or behavior, as opposed to the presence or absence of a criminal conviction[.][66]

Though the opinion contained a footnote asserting that the arrest was “given limited weight,” it later named the arrest as a major reason for the denial.[67] In the end, the applicant’s considerable “positive equities . . . [were] outweighed by the adverse factors present in his case.”[68]

USCIS apparently made a factual finding that the allegations in the police report were true. But the basis for this finding is not explained or examined. Tellingly, in an earlier paragraph asserting an intention to credit the police report, USCIS failed to even mention any of the contrary evidence submitted by the applicant. This evidence included a sworn affidavit from the complaining witness, the applicant’s troubled teenage daughter, stating that she had fabricated the allegations because she was mad at her father, and a sworn statement from the applicant explaining that he had argued with his daughter over a boyfriend but never harmed her physically.[69] A discussion of contrary evidence is an exceptionally odd thing to omit from a discussion of factfinding. It is explicable only with the understanding that USCIS did not think it needed to engage in factfinding at all. It was merely evaluating factors. The allegations in the complaint were a negative factor against the applicant, regardless of any contrary evidence submitted. Indeed, it is not clear how the applicant could have overcome the allegations given that he had submitted the strongest possible evidence of the falsity of any allegations: a sworn recantation from their original source. Had the agency been forced make its factfinding explicit, it is difficult to see how its decision to credit the allegations could be justified.

This is also borne out in the observations of advocates, who report that, even when they provide “a substantial amount of testimony and evidence to refute unsubstantiated allegations” in a CLS output “the information in the [output] is still often construed against their clients . . . . ‘If we present evidence, including declarations and testimony, undermining the veracity of [a] police report, [immigration decisionmakers] should have to actually weigh that against the police report. As it currently stands, if our client testifies to something contrary to [a] police report, it is the client’s credibility that is damaged, not the report’s.’”[70]

b.     The Police Report Problem as a Factfinding Problem

Police reports, as predicted by Holper, are a common theme in agency denials. The “police report problem” that she identifies[71] can also be understood as a factfinding problem. Hidden factfinding obscures what is really going on in adjudications, weakening the adjudicatory process and leading to poorer decision-making.

In a telling example, USCIS denied a victim of domestic violence her green card because she had failed to submit police reports and therefore “precluded USCIS from fully understanding the behavior that led to the arrests.” On appeal to the AAO, the applicant submitted the requested police reports, arguing that they should be given little or no weight because the criminal cases had been dismissed. Instead, the adjudicator read the entirety of the reports for truth:

The record shows significant discrepancies between the Applicant’s explanations for her arrests and the information contained in the police reports . . . . Regarding Applicant’s [] 2001 arrest for prostitution, while the Applicant claimed that she knew nothing about her friend’s involvement in prostitution, the police report indicated that the Applicant agreed to engage in sexual intercourse with an officer in exchange for money . . . . The Applicant’s reluctance to acknowledge or explain these discrepancies, concede any wrongdoing, or express remorse for her actions is a serious negative factor that was not in the record before [USCIS].[72]

This type of analysis belies adjudicators’ lack of understanding of the documents before them. Police reports are probably the least reliable evidence in most immigration adjudications. They are not reliable enough to even be admitted as evidence in criminal court, much less bear the criminal court’s imprimatur of reliability. The types of evidence that a respondent might submit—letters, sworn statements, or testimony—are objectively more reliable. Live testimony is subject to cross-examination and can be evaluated for credibility and basis of knowledge. Statements sworn under penalty of perjury are written by people willing to take responsibility for the truth of every fact therein. Even a simple letter is signed by the preparer and is likely to specify the source of the writer’s knowledge. Meanwhile, police reports contain unsworn allegations and multiple levels of hearsay and may be an amalgamation of information from several unidentified sources.[73]

Further, adjudicators seem to be confused about how different CLS outputs relate to one another. In the adjudication where USCIS believed the criminal complaint over the disavowed statement of the applicant’s teenage daughter, the applicant was faulted for not including a police report to “independently corroborate”[74] his story. Such a statement implies that the allegations in a criminal complaint can only be authoritatively contradicted by the allegations in a police report. This is nonsensical. Because allegations in complaints are often drawn directly from police reports, there is no reason to believe that a police report would ever contradict a criminal complaint.

The real reason USCIS wants police reports is that they often contain detail that is pared down in the drafting of the criminal complaint. As it did with the criminal complaint, USCIS would have considered the allegations in the police report to be true unless proven otherwise. Indeed, at the conclusion of the opinion, the AAO emphasizes the “lack of sufficient evidence in the record to fully assess the specific conduct underlying [the applicant’s] arrest” as a major reason for the initial denial.[75]

c.      Zombie Allegations and Burden Multiplying

When objecting to the use of spurious evidence in immigration court, the stock response I’ve received from DHS is, “It’s your burden, counsel.” Because the respondent bears the burden to prove eligibility for and deservingness of a discretionary benefit, a feather’s worth of derogatory evidence can weigh a ton. Any possible objection to factfinding to undefined standards based on objectively unreliable evidence is waved away with a broad gesture to “discretion” and the respondent’s burden.

Burdens of proof and persuasion[76] are hardly unique to the immigration context. In both the civil and criminal legal systems, one party bears the burden, and the burden sometimes shifts between parties during different phases of litigation. But bearing a burden does not mean that a party must affirmatively disprove every negative factual possibility that might hypothetically exist in the world.[77] Indeed, in discretionary immigration adjudications where CLS outputs are not involved, adjudicators regularly evaluate evidence, both positive and negative, and find facts based on the strength of that evidence while taking the respondent’s burden into account.[78]

When CLS outputs enter a case, however, things look very different. Allegations that were disposed of in the criminal legal system return from the dead. These “zombie” allegations trigger an additional burden on the respondent: prove their untruth or be denied the benefit. As if proving a negative were not difficult enough,[79] zombie allegations in CLS outputs are not evaluated like other evidence: they are treated as true by default, and the respondent’s evidence is deemed incredible to the extent that it contradicts the CLS output.[80] Zombie allegations are effectively unkillable. Thus, rather than merely shifted or heightened, the burden on the respondent is better described as having been multiplied such that it is impossible, or nearly impossible, to overcome.

Burden multiplying is the de facto rule for any immigrant who has ever come into contact with the criminal legal system. Guerra endorses such a scheme: the BIA cited the fact that “the respondent failed to present any evidence or argument that tended to undermine the reliability of the information contained in [a criminal] complaint” as proof of the complaint’s veracity.[81] In another case, the BIA required a respondent to prove the untruthfulness of pending criminal charges in a DUI arrest. DHS, meanwhile, prevented him from appearing in criminal court to answer those charges by refusing to bring him to his court dates.[82] The IJ below explicitly made “adverse inferences” against the respondent because the charging document did not include his blood alcohol level or the reason the officer pulled him over.[83] The respondent, of course, had no control over what the district attorney chose to put in the document. In the case discussed above where the complaining witness—the applicant’s teenage daughter—retracted her allegations,[84] USCIS’s opinion implies that the only way USCIS would have found that the allegations were untrue is if the applicant had produced a police report essentially exonerating him of the charged crime.[85] Such a unicorn does not exist.

The irony here is that adjudicators treat CLS outputs as reliable because of their origins in the presumptively reliable criminal legal system while disregarding the most reliable output that system produces: its factfinding. The guilty/not guilty verdict is the most reliable product of the criminal legal system: it is created with all the processes and protections that that system can muster. Yet, in an immigration proceeding, this factfinding is denied the same presumption of reliability granted to unadjudicated allegations that were created with far less process and without any of the protections. Dismissal of all charges—a systemic finding of a lack of criminal responsibility—is treated as if it were virtually meaningless. A jury verdict of not guilty is treated equally so.[86] After all, a not guilty verdict only shows that guilt was not established beyond a reasonable doubt. It does not prove innocence, and therefore will not be enough to overcome the multiplied burden required in immigration adjudications. To effectively prove his innocence in a way that would be satisfactory to DHS, an immigrant would need a CLS output affirmatively establishing his innocence to a high degree of certainty.[87] This is evidence that the criminal legal system, by its very nature, is not capable of producing.

d.     Guilty After Proven Innocent

As the Supreme Court has repeatedly said, police reports “prove” little more than the fact of arrest and the substance of the allegations, which cannot be fairly used as a proxy for guilt.[88] Yet, as seen in the examples just discussed, this is exactly how CLS outputs are used in discretionary immigration adjudications—as proxies for guilt. Immigrants whose charges were reduced or dismissed, including those found “not guilty” after trial, have already proven their innocence to the extent required by the criminal legal system. In the immigration system, however, they are forced to prove it again to a much higher and more exacting standard.[89] In these adjudications, their evidence is disvalued while inherently unreliable CLS outputs are elevated far beyond their true evidentiary value.

The title of this piece is guilty after proven innocent rather than the less dramatic until. Despite the risk of hyperbole, I believe the choice to use after is warranted. Immigrants who have been arrested are forced to prove their factual innocence to an exceedingly high standard, in front of adjudicators laboring under severe misconceptions about the value of the evidence before them. These adjudicators may not even realize that they are engaged in factfinding at all, but rather may assume that the facts have already been found by a system that they associate with reliability. These adjudicator decisions are not guided by meaningful standards or subject to meaningful review. As a result, arrests become functionally equivalent to convictions, no matter the ultimate disposition of the criminal case.[90]

In this way, the immigration agency has essentially been allowed to amend the eligibility requirements for certain forms of relief: in addition to the other requirements, an applicant must also show that they have never been arrested. Because this rule has been operating under the cover of discretion, the agency has never had to defend against a challenge that it has exceeded its delegated authority.[91]

Of course, arrests do not occur in a vacuum. Given what we know about racial disparities in policing,[92] equating arrests with convictions will inevitably cause disproportionate harm to immigrants of color. The harm is more pronounced than the harm caused by using criminal convictions since it is imposed without any of the protections of our admittedly imperfect criminal legal system. Any arrest, even one that a court later determines to be baseless, will serve to disqualify an immigrant from discretionary relief. Such a practice effectively suborns the laundering of racial bias.

B.     A Better Framework

After establishing hidden factfinding as a problem, I would like to propose a solution: a framework to guide legal systems that wish to use CLS outputs in their adjudications. This framework requires clearly identifying the factfinding that has (and has not) been done in the criminal legal system and the standard to which it has been performed. Only then is it possible to determine how the output can be used in the second system. The framework aims to solve the problem of hidden factfinding specifically in the context of discretionary decision-making in the immigration system, but I believe it to be applicable anywhere CLS outputs are used, or anywhere an output created by one system is used by another.

1.     Output Portability

The framework depends on the concept of “output portability.” The portability of an output refers to its usability in a second forum. Some outputs are completely portable. This means they are usable in the second forum with no additional factfinding. Other outputs are non-portable—because they have little evidentiary value and carry too much potential for prejudice, they must be excluded. Still other outputs are partially portable. They are usable in the second forum, but additional factfinding is required. Whether or not an output is portable, non-portable, or partially portable depends on the standard to which it has been adjudicated when compared with the standard required by the receiving system, and also on receiving system’s tolerance for prejudice.

a.     Complete Portability

For an output to be completely portable, the facts in question must have been found by a neutral decision maker (usually a court or other tribunal) after a process of adversarial testing where the affected party had an opportunity to be heard and to shape the creation of the final verdict, whether by negotiated plea or trial.[93] This process, which is time consuming and costly, need not be repeated in the second forum.

A criminal conviction is completely portable into the immigration system and, indeed, into probably every other legal system. Fact finding has occurred. A court has found the facts necessary to make up the conviction true to beyond a reasonable doubt.[94] This is a higher standard of proof than that required by the immigration system. An immigration adjudicator can take the adjudicated facts as true to the evidentiary standard that they need to meet, whether it be clear and convincing evidence, a preponderance, or another lower standard.

Complete portability is highly efficient. The respondent need not be given a second opportunity to be heard, as he has already had his chance. No additional process is due. The complete portability of criminal convictions into the immigration system explains why immigration judges do not give respondents the opportunity to argue that they were not actually guilty of the things they were convicted of. The issue has already been litigated, and the immigration court is not in a position to re-litigate the case.

b.     Non-Portability

An output is non-portable if it cannot be used in the receiving system and must be excluded. Non-portability is also highly efficient. There is no need for additional fact finding about the output in the second forum because the output is excluded entirely.

An output is non-portable when it is not very probative and very prejudicial, keeping in mind that different systems will have different tolerances for prejudice. Something is considered potentially admissible “evidence” when it tends to prove a fact at issue,[95] making that fact more likely or less likely to be true. Under this standard, a police report has some evidentiary value, but its probative value is still quite low.[96] As such, it is not admissible in a criminal trial because this low probative value is substantially outweighed by the potential for prejudice.[97] The immigration system, when compared with the criminal legal system, is far less concerned with prejudice to the respondent. As explained above, the general rule in immigration proceedings is anything can come in.[98] As such, the immigration system is unlikely to ever consider a relevant output with some evidentiary value to be non-portable.

c.      Partial Portability: Shifting the Locus of Factfinding

Most (or all) CLS outputs will be partially portable in discretionary immigration adjudications. Partially portable outputs have not been adjudicated to a sufficiently high standard to be imported wholesale into the second system without the need for further factfinding. But the rules of the second forum are not such that the output is considered too prejudicial to be considered at all. In the discretionary immigration system, where the rule is anything can come in, partially portable outputs include police reports, pre-sentence investigation reports, transcripts, lab reports, and the unconvicted counts in a criminal complaint (the convicted counts, having already been adjudicated to a sufficiently high standard, are fully portable).

A partially portable output is admissible as evidence and can be considered, but the factfinding that did not occur in the originator system must happen in the second system. Partial portability is, naturally, less efficient than complete portability or non-portability.

2.     The Output Portability Framework in Practice

When making a discretionary decision based on a CLS output, the adjudicator—IJ or USCIS employee—should first determine whether the output is fully portable. To do so, they should look to whether the output was adjudicated in the first forum and to what standard. If adjudicated (as with a guilty plea or trial conviction), the facts in the output can be considered established to that standard (e.g., beyond a reasonable doubt) and can be used in the second forum without further factfinding so long as the adjudication in the second forum did not require a higher standard of proof. Otherwise, the adjudicator should determine whether the output is partially portable or it is non-portable and must be excluded.

If an output is partially portable, the adjudicator can use it in aid of factfinding. The adjudicator should consider the output and any relevant corroborating evidence provided by DHS along with any evidence provided by the respondent. They should weigh all the evidence and determine whether the output has been established and to what level of certainty.

At present there is no defined level of certainty to which the facts in a CLS output must be found to count as a negative “factor.”[99] The agency could fix this by establishing the target level of certainty, which might even be different for different types of applications: for a U adjustment the target level of certainty might be relatively high given the humanitarian considerations involved, the applicant’s ties to the United States, and the help they provided to law enforcement. For other applications where ties are less strong and humanitarian considerations are less important it might be sufficient for the adjudicator to find that it is more likely than not that the output is true. For some other applications, such as an application for a work or student visa, an even smaller chance that the allegations in the CLS output are true might be sufficient for the output to be used as a negative factor against the applicant. If the facts in the CLS output are found to the target level of certainty, they can be considered as negative “factors” in the discretionary decision. If they are not, they should not be.

Or, in the alternative, rather than establishing a target level of certainty and requiring adjudicators to give no weight to allegations not established to that target, the agency might allow more weakly established allegations to be considered, so long as the low level of certainty is reflected in the weight the adjudicator gives to the factor in their discretionary analysis.

To give an example of both approaches, an adjudicator at USCIS is considering an application for U adjustment. The CLS output is an allegation in a criminal complaint that resulted in dismissal of all charges. The respondent was charged with assault after a fight at a drinking establishment. The factual narrative reflects the word of a civilian complainant who the government is unable to locate, and DHS is unable to provide any other corroborating evidence, physical or otherwise, showing that respondent committed the crime. Given the paucity of evidence, the adjudicator is unable to find that the allegations in the output are established to any standard beyond “some possibility” that the crime happened as described.[100]

Under the first approach, because the allegations were not established to the target level of certainty set by USCIS for U adjustment applications (say, a preponderance of the evidence), the adjudicator must give the allegations no weight in the discretionary phase. Under the second approach, the adjudicator might still consider the allegations as a negative factor in the discretionary phase so long as that factor is given a weight proportionate to the low level of certainty (in this case, “some possibility”) to which the allegations are established.

After completing factfinding, the adjudicator moves on to the discretionary phase where the applicant has the burden to establish, by “clear and convincing” evidence, that he deserves his green card. Under the first approach, the adjudicator would not consider the CLS output at all, while, under the second approach, the adjudicator would consider the output, but only for what it is worth. If the applicant had otherwise strong equities, it would be difficult to justify the denial of the green card based on “some possibility” that the applicant once got into a bar fight. If, however, it was a close case, the small possibility that the output might be true could, under the second approach, be enough to tip the balance against the applicant.

The second approach, admittedly, requires less of a clear division between the factfinding phase and the discretionary weighing-of-factors phase and risks the conflation of the two phases. This danger is mitigated, however, because an adjudicator using the output portability framework would be forced to identify the standard to which they were finding the relevant facts. This would allow an agency reviewer to review that factfinding separately from the weight those facts were given in the ultimate discretionary decision. Either approach would represent a marked improvement over the current practice where factfinding is done implicitly and is not subject to meaningful review.

3.     The Framework and Agency Review

Agency-level review of discretionary decision-making would be significantly improved under an output portability framework. Review would look slightly differently depending on whether the case was decided by USCIS or by an IJ, but, in either situation, the review process would allow for readier identification of adjudicator error.

Discretionary determinations by USCIS adjudicators and IJs are reviewed de novo.[101] USCIS adjudicators’ factfinding is reviewed de novo as well,[102] but IJ factfinding can only be reversed if it is “clearly erroneous.”[103] This difference makes sense because USCIS factfinding is done “on the papers” while IJ factfinding often involves live testimony, which is more difficult to evaluate on appeal.

For a USCIS adjudication, the reviewing body (the AAO) would be able to review both the adjudicator’s factfinding and their discretionary analysis de novo. Because the adjudicator would be obligated to identify the level of certainty to which they found the facts, the AAO would be able to easily review that factfinding and, if it disagreed, substitute its own assessment of the facts for that of the USCIS adjudicator. The AAO would then be able to look at how the adjudicator exercised discretion in the same manner.

For an IJ adjudication, the BIA would not be able to substitute its own factfinding for that of the IJ. It would instead review the IJ’s factfinding for clear error. Currently, the BIA, believing it is reviewing IJ discretion, tends to vacillate between deference and de novo consideration.[104] Either way, there is little chance of identifying errors in IJ factfinding. Identifying hidden factfinding means that the BIA will no longer be able to substitute its judgment de novo for that of the IJ on the assumption that the IJ’s determination was a discretionary one which might, at first glance, seem to augur less robust review. But clear error will be more readily apparent than under the existing system because the IJ will be obligated to identify the standard to which the facts were found and there will be no question that the IJ was, indeed, engaging in factfinding. More clarity will prevent the BIA from mistakenly substituting its judgment for the IJ’s where it is not warranted while also making it more likely that the BIA will catch faulty factfinding at the agency level including errors that are escaping notice under the existing system. In the case where the daughter took back the story that led to her father’s arrest, the complaint contained unsworn and unswearable double hearsay allegations, uncorroborated by a conviction of the charge and undermined by the sworn recantation of the original complainant. This is inherently weak evidence. Unthinkingly crediting such flimsy evidence over the sworn statement of the applicant was error. If such a decision was made by USCIS, it should be reversed by the AAO on de novo review. If made by an IJ, it should rise to the level of “clear error” necessary for the BIA to overturn the IJ’s factfinding.[105]

In conclusion, my proposed framework would require adjudicators to consider whether an output has been adjudicated to a sufficiently high level that its findings of fact can be accepted wholesale or else explicitly acknowledge that further factfinding is required and they must be the ones to do it. The framework promotes transparency and avoids the evils of hidden factfinding. These evils including treating allegations from the criminal legal system as true by default and requiring respondents to produce evidence that simply does not exist in order to refute them.

C.     Implications of the Output Portability Framework

In this third and final Part, I examine the implications of the output portability framework. The major intervention achieved by the framework is that it forces decision makers to identify the locus of factfinding, even in discretionary adjudications. Once factfinding is properly identified and isolated, it can be performed and analyzed like any other factfinding. This approach has two major consequences when applied to discretionary adjudications. First, it allows adjudicators to supplement the BIA’s threadbare guidance with law developed in the nondiscretionary context. Second, the identification of hidden factfinding eliminates an artificial barrier that has prevented reviewing courts from identifying due process errors in the way CLS outputs are used. This implication is especially important for bond proceedings where immigrants are regularly stripped of their liberty based on nothing more than allegations.

The output portability framework also has uses beyond discretionary adjudications. It allows for a better understanding—sometimes necessitating rethinking—of the way CLS outputs are used in nondiscretionary “categorical approach” cases. It can also be applied outside the immigration system in any legal system where CLS outputs are used.

1.     Ability to Draw on Existing Law to Evaluate the Probative Value of CLS Outputs

Agency recognition of hidden factfinding as factfinding will necessarily encourage the development of guidance on how it should be conducted. This guidance does not need to be created from scratch. Once factfinding is disaggregated from discretion, it can be treated like any other factfinding. This will allow adjudicators to benefit from a comparatively richer body of law developed in the nondiscretionary context, as well as on the much richer and more developed guidance that can be found in the law of evidence.

In the non-discretionary context, to find an applicant inadmissible as a controlled substance trafficker an adjudicator must find that there is a “reason to believe” that he has engaged in trafficking of a controlled substance. This nondiscretionary question bears a marked resemblance to factual questions that must be answered in discretionary decision-making, specifically the determination of whether a bond applicant represents a danger to the community.[106] In both “reason to believe” and “danger to the community” determinations, an adjudicator must determine if there is enough credible evidence to indicate that the individual has engaged in certain negative behaviors in the past, often relying on CLS outputs to do so. While the danger question is prospective rather than retrospective, the difference is not fatal to the example because the whole point of looking to CLS outputs is to use past behavior as indicative of likely future behavior. The only difference is that “reason to believe” is purely a factual determination without a second discretionary step. As a result, the factfinding has been treated as factfinding, and is subject to scrutiny as such. Without discretion to confuse the issue, the question becomes, more clearly: are we certain enough that this happened?

Because the question has been clearly understood as a factual one, guidance has developed to help answer the question of when unadjudicated allegations in CLS outputs might provide that certainty. Courts and the BIA have held that an arrest report alone is not the type of “reasonable, substantial, and probative”[107] evidence that could provide reason to believe a respondent engaged in drug trafficking.[108] But multiple arrest reports, especially if corroborative of one another, might be enough, as would testimony from law enforcement officers who observed the criminal activity, surveillance reports containing multiple police officers’ first-hand accounts, or the respondent’s own testimony or admission of guilt.[109] Adjudicators making risk determinations in the discretionary context can benefit from this helpful and highly relevant body of law.[110]

Treating factfinding in aid of discretion like any other factfinding would also allow adjudicators to look to another rich source of guidance on the value of CLS outputs: the law of evidence. Neither the Confrontation Clause nor the Federal Rules of Evidence apply in immigration proceedings in the sense that evidence that would be excluded from both criminal and civil trials is not similarly excluded from immigration proceedings. Nevertheless, the principles underlying the Confrontation Clause and the Federal Rules remain relevant to evaluating the probative value of a given CLS output. The first step in evaluating whether a piece of evidence is admissible is to determine its probative value. This probative value must then be balanced against its potential for prejudice.[111] Law from other contexts, including both civil and criminal, that is concerned with evaluating evidence’s probative value in aid of such a determination is potentially an extremely rich source of guidance for immigration adjudicators. Indeed, where factfinding is treated as factfinding, this is already happening.[112]

2.     Facilitation of a Due Process Analysis

The output portability framework allows for judicial review of agency errors that are sufficiently severe as to constitute due process violations. Because jurisdiction to review many discretionary immigration decisions has been stripped from the courts, due process is often the only way to address agency error. While not every factfinding error represents a due process violation, identifying hidden factfinding changes the character of due process review by opening up an avenue by which courts can address egregious errors, particularly if these errors can be traced to a faulty process rather than merely a faulty result.

a.     Judicial Review of Agency Error

In immigration law, judicial review of discretionary immigration decision-making is almost nonexistent. This is unique to immigration: in other areas of law, courts regularly review agency decision-making under the Administrative Procedure Act (APA), and factfinding done to an inexact or improper standard is subject to judicial correction. This includes factfinding done in support of exercises of discretion.[113] Judicial review of immigration decision-making is complicated by an alternative statutory scheme that, in some areas, replaces review under the APA.[114] But, under either scheme, judicial review of discretionary immigration decision-making was the norm until 1996 when Congress passed the numerous jurisdiction-stripping provisions that affirmatively removed the jurisdiction of the courts to review almost all discretionary immigration decisions, including bond.[115]

After jurisdiction stripping, courts can only oversee nondiscretionary determinations. Congress allowed courts to retain normal oversight over only a single form of discretionary relief: asylum.[116] Courts reviewing all other discretionary immigration decisions are allowed to review only “constitutional claims or questions of law.”[117]

b.     Questions of Law and Constitutional Claims

The preservation of jurisdiction over “questions of law” might seem like it would allow for significant review, but courts have generally been reluctant to identify a question of law in how the immigration agency exercises its discretion. This reluctance can be traced back to the idea that courts reviewing discretion would have nothing to review.[118] As a seminal case explains:

The absence of substantive rules means more than just the absence of “liberty or property.” It means the absence of standards for judges to use. . . . When there are no rules or standards there is neither legal right nor legal wrong . . . . [T]he grant of discretionary relief under the immigration laws is a question on which there is “no law to apply,” and when there is no law to apply judicial review is exceedingly constricted. . . .

In contrast, where there are concrete “rules,” courts can and do engage with them:

The Board [errs] when it acts for . . . a reason that a court can determine is erroneous. If, for example, the Board had [found that the respondent] was engaged in employment and therefore ineligible under [the regulation], a court could review the record to determine whether the Board’s belief had substantial support.[119]

Where jurisdiction over discretionary decision-making has not been stripped—as in asylum and motions to reopen—the law has moved away from this view of discretion, and courts have engaged in more robust review including faulting the agency for ignoring evidence in its factual determinations, making unreasoned or nonsensical decisions, or for failing to follow its own internal rules and precedent.[120] And, where jurisdiction has been stripped but the decision is nondiscretionary,[121] courts have found that jurisdiction to review “questions of law” allows them to review and correct blatantly bad factfinding by the agency.[122] In discretionary determinations where jurisdiction has been stripped, however, courts will review agency errors in determining first-step eligibility for discretionary relief as  “questions of law” but will not review how the agency exercises its discretion at the second step.[123] With respect to review of agency factfinding, the decisions seem to turn on whether the court can identify “law to apply.”[124]

This leaves constitutional claims. Due process is the most obvious constitutional claim that might be raised to challenge an agency adjudication. But, in the realm of immigration law, these claims are hampered by the general (though not necessarily inevitable)[125] consensus that immigrants do not have a protected liberty or property interest in obtaining discretionary relief.[126] The lack of a protected interest in discretionary relief is generally understood as originating in the plenary power of Congress—delegated to the executive—to make rules for the admission of “aliens” into the United States, which is taken to also encompass the power to decide on the continuing status of these immigrants after they have been admitted.[127] Without such a protected interest, there can be no substantive due process requirement that any government infringement of that interest be narrowly tailored to serve a compelling state interest.[128] The procedural due process analysis is also curtailed, as the immigrant has no interest that can be held to weigh against the government’s interest in its procedures.[129]

Despite the lack of a protected interest, procedural due process protections do apply in immigration adjudications.[130] The “discretionary nature of certain forms of relief” does not eliminate the constitutional requirement of a fair hearing.[131] Rather than balancing the interests involved, the question is whether the process provided is “fundamentally fair.”[132] Courts, however, tend to take a jaundiced view of claims of procedural due process violations arising out of discretionary immigration adjudications, characterizing them as requests for APA-style judicial review disguised in constitutional clothing.[133]

Consistent with the above, courts considering the treatment of CLS outputs in discretionary decision-making have been mostly unwilling to find constitutional error in the way these outputs are used. The decisions tend to address legal and constitutional error together, usually in a single sentence at the end of the opinion dismissing the claim.[134] Some courts have rejected out-of-hand the idea that a violation of the BIA’s own precedent (e.g., Arreguin’s holding that police reports should be given little weight) would be reviewable as a question of law or constitutional violation,[135] characterizing such claims as mere “quarrel[s] over the correctness of the factual findings or justification for the discretionary choices.”[136] Others have entertained the idea but ultimately found that any problems with agency factfinding did not rise to the level of due process violations[137] or avoided the question by ruling on other grounds.[138] Courts considering Arreguin have found it significant that it is not a blanket prohibition on the use of police reports but merely “implicates matters of degree.”[139] In this view, Arreguin, is not a true “rule,” and the agency is, therefore, not “act[ing] for a reason that [the] court can determine is erroneous”[140] by departing from it.

Only a few courts have found that the BIA’s failure to follow Arreguin was reviewable. The Seventh Circuit found that the BIA committed legal error when it gave significant weight to an uncorroborated arrest report to justify a denial of discretionary relief despite finding that the respondent had testified credibly that he did not commit the crime.[141] The determining factor may have been how the agency laid bare that it was crediting the police report over what it had already determined to be “credible” testimony from the respondent. This fact (which is hard to understand as anything but a tactical error by the decision maker) underscored for the court the burden-multiplying effect that CLS outputs can have, such that they are impossible to disprove even with compelling evidence. In another case, the Northern District of California found a potential due process violation where USCIS gave too much weight to a police report. The respondent had applied for a U visa and, just as it did in the U adjustment applications discussed above, USCIS had treated the report as fully portable and assumed that the respondent had committed the crimes detailed therein.[142] This case too, presented an unusually stark set of facts as the respondent had not been permitted to access the contents of the report to be able to respond to its allegations.[143]

c.      Hidden Factfinding and Judicial Review

The output portability framework has the potential to dramatically change the character of judicial review. It achieves this by supplying courts with the language to identify a previously unrecognized error that may render an adjudication fundamentally unfair.

Courts are (generally) more likely than the agency to understand that they are dealing with factfinding,[144] but when considering agency factfinding in support of an abuse of discretion, they still struggle to identify “law to apply” that would allow them to provide meaningful review. They tend to look for rules that are overly specific to the factual determinations before them, asking, for example, whether it categorically violates due process to give a certain amount of weight to a given CLS output or to value one type of evidence over another. Unsurprisingly, courts typically find no due process violation. The Second Circuit, for example, has stated (in separate cases) that “[d]ue process does not require that the IJ credit [a respondent’s] testimony over the evidence contained in the criminal complaint,”[145] and “the relative weight to be given to uncorroborated arrest reports” is “unreviewable.”[146] Such courts say they cannot identify a rule requiring the agency to give X output Y weight or credit one type of evidence over another. They lack an analytic framework to identify why the agency’s factfinding might have been wrong, not just in result but in methodology.

This lack of “law to apply” is why the output portability framework matters. Factfinding comes with its own set of rules.[147] These rules are not found in internal agency guidance or the INA but in a wider body of law, including the Due Process Clause itself. Reference to this body of law can help courts identify errors in agency factfinding, including errors that are sufficiently egregious to render these decisions fundamentally unfair and, therefore, rise to the level of due process violations.

Decisions interpreting the Due Process Clause lay out the basic rules for fair play in factual adjudications. For factfinding to comport with due process, facts must be found based on evidence to which the respondent has the opportunity to respond. In other words, the respondent must have a chance to influence the decision.[148] For this chance to be more than illusory, two requirements must be met: (1) the facts must actually be found by the decision maker, and (2) the facts must be found to an identifiable standard. These two requirements have not been articulated explicitly in case law or literature, except by implication.[149] This is because, where factfinding is treated as factfinding, their articulation is largely unnecessary: it should be blindingly obvious that adjudicators engaged in factfinding must find facts, and that they must do so to a defined standard lest factfinding collapse into incoherence. The second requirement may be slightly less immediately intuitive than the first but, if individual adjudicators can deny relief using any factual standard they see fit (including where there is any possibility that negative allegations are true), then respondents do not truly have the chance to influence the outcome of the factfinding. Indeed, if any evidence that a respondent might possibly submit will be insufficient to kill a zombie allegation, then the agency is not engaging in factfinding at all but merely importing the allegations and promoting them to found facts without engaging in any “process” along the way.

My framework addresses both of the requirements that are necessary in order for a respondent to have a meaningful opportunity to influence the result of adjudicator factfinding. The concept of “output portability” requires (1) if an output is not fully portable, that factfinding occur in the second forum. It also requires (2) the identification of the standard to which facts must be found. When courts have squarely considered factfinding as factfinding as they do when reviewing nondiscretionary determinations, they have roughly applied the two requirements I set forth above: they have (1) rejected factfinding that treats unproven allegations as if they were adjudicated, and they have (2) implicitly rejected the idea that the agency is absolved of the need to find the necessary facts to an identifiable standard in adjudications where the respondent bears the burden of proof. The Fourth Circuit rejected a government argument that a respondent, who bore the burden of proof to show he was admissible, “failed to meet this burden because he did not provide evidence, such as arrest or police reports, showing that he was innocent of the . . . charges.” The court held that just because a respondent bears the burden of proof “does not mean . . . that [he] must affirmatively prove that he was innocent.”[150] Meanwhile, the Eleventh Circuit held that “[b]urdens of proof notwithstanding, a finding of inadmissibility must be based on something more than the alien’s failure to prove a negative.”[151] Finally, in a case about whether a respondent had been “admitted” to the United States (a factual determination), DHS claimed its internal system showed that he had left the United States on a certain date but refused to provide any information about how that data had been entered into the system. The respondent submitted powerful evidence that he had, in fact, been in the United States on that date, including bank deposit slips (complete with handwriting analysis) and the testimony of multiple witnesses swearing he had been in Rochester, New York. The IJ nevertheless found that the respondent crossed the U.S./Mexico border that same day, and the BIA upheld the determination. The Eighth Circuit reversed, saying “[e]ven assuming that the departure information was presumptively reliable, we are not sure what more [respondent] could have done to rebut the presumption.”[152] While these are statutory review cases rather than due process claims, the courts’ analysis of the fairness of the agency’s factfinding is far from irrelevant to a due process analysis. These cases show courts balking at factfinding based on faulty logic, specifically the assumption that the respondent’s burden of proof means that the agency need not find the facts on which derogatory decisions are based. Not every legal error of this nature will be a due process violation, but legal errors that are severe enough to make the proceedings fundamentally unfair violate due process and are reviewable.

Applying the output portability framework and treating agency factfinding as factfinding (without discretion to cloud the picture), the due process analysis looks very different. For example, in the two Second Circuit cases I discuss above, the court stated that “[d]ue process does not require that the IJ credit [a respondent’s] testimony over the evidence contained in the criminal complaint,”[153] and “the relative weight to be given to uncorroborated arrest reports” is “unreviewable.”[154] While due process may not require an IJ to credit a respondent’s testimony over a given CLS output, it does require the IJ to conduct a fair review of the evidence such that the respondent’s sworn testimony, if found credible, has a chance to overcome the complaint’s allegations. For this to happen, the allegations in police reports must be treated as allegations, not as adjudicated facts.[155] In the example of the father who was arrested after an argument with his teenage daughter,[156] there is no colorable argument that the respondent was given a meaningful say in the determination of the facts relating to his application. The agency’s factual finding, which was not made to any standard, was predetermined from the outset by a burden multiplying scheme that made it impossible for the respondent to rebut the CLS output. There was no evidence that the respondent could have submitted that would have been sufficient to overcome the police report: even a sworn statement from the original complaining witness disavowing her allegations was rejected as insufficient. This was not factfinding at all, but rather wholesale importation of unadjudicated allegations. This is no mere “quarrel”[157] over the weight assigned to any particular piece of evidence.

Courts that consider and reject Arreguin as an enforceable rule miss a bigger picture. They view the mere assignment of too much evidentiary value to a CLS output as, at most, a minor quibble. They fail to see that in many instances factfinding was never conducted at all, with facts never found by any decision maker imported wholesale and assumed to be true. Or they fail to see that the adjudicator’s factfinding unfairly burdened the respondent by forcing him to prove a negative. They fail to understand that, when there was no possible evidence that a respondent could realistically produce that would suffice to overcome the allegations, it cannot fairly be said that factfinding consistent with due process has occurred.

d.     Hidden Factfinding and Bond

The identification of hidden factfinding has the most dramatic implications for bond.[158] It is in bond proceedings, where the fundamental right of liberty is threatened, that hidden factfinding is on the shakiest constitutional ground. Disaggregating factfinding and discretion changes the apparent character of bond determinations, revealing that bond is not the fully discretionary decision that it has been assumed to be but rather a factual determination with a second discretionary step. This insight suggests the government must bear the burden of proof from the outset. It also suggests that an IJ’s factfinding in this area is properly subject to full judicial review (not just the more limited review retained for errors so egregious they rise to the level of due process violations) under the INA.

The Supreme Court has approved the civil detention of immigrants during removal proceedings only under certain conditions. Drawing on historical precedent holding that preventative civil detention is constitutional only when the purpose of that detention is to protect the safety of the community or ensure appearance at future court hearings, the Court held that an immigrant must be found to be a flight risk or danger to the community before they can be detained.[159] Despite not being written into the statute, the due process requirement that the immigrant represent a flight risk or danger nevertheless amends the legal rule defining who is subject to detention. This amendment has been reflected in subsequent BIA guidance.[160] More than 20 years later, the fundamental nature of bond proceedings is still in contention. Unsettled questions include who should bear the burden of proof in establishing danger or flight risk, what evidence might be sufficient to do so in conformance with due process, and how courts can effectively review those determinations.

The BIA has instructed that an individual immigrant bears the burden of establishing “that his release would not pose a danger to property or persons, and that [he] is likely to appear for any future proceeding.”[161] This holding does not rest on statutory authority, and the Supreme Court has never ruled on the constitutionality of placing the burden on the immigrant.[162] What is clear, however, is that the burden cannot remain on the immigrant over a long period of detention. This is because bond decisions squarely implicate the liberty interest. Accordingly, rather than ask merely whether the process is “fundamentally fair,” courts employ the Mathews v. Eldridge[163] balancing test to determine whether the procedures provided by the government are sufficient to safeguard an individual’s interest in his own freedom. Courts have found that, when detention is “prolonged,” the agency’s procedures are not sufficient to satisfy due process, and the burden of proof must, at some point, shift to the government.[164] But when does it shift? The Supreme Court has explicitly rejected the idea that due process requires a bright-line rule whereby the burden of proof would shift to the government after a set period of time.[165] Without such a rule, whether detention has become “prolonged” must be litigated on a case-by-case basis in federal court via petitions for habeas corpus.[166] This results in tremendous uncertainty for detained people: an immigrant who believes that his particular detention has become sufficiently prolonged might file a petition too soon only to have the court deny it as premature while another might wait too long and spend extra unnecessary months behind bars. Additionally, immigrants who do not have the resources or wherewithal to file a habeas petition will be unable to vindicate their liberty interest before the agency and will remain in detention under the wrong burden of proof.

Relatedly, under the current bond regime, there is nothing preventing an IJ from finding that any level of risk whatsoever, no matter how small or attenuated, is sufficient. IJs necessarily find facts along the way to making the required risk determination. For example, an IJ finds, based on a criminal complaint charging a respondent with driving under the influence, that the respondent drove drunk in the past. This determination involves factfinding about past behavior. The IJ then must determine risk of repetition: whether the past behavior is indicative of future behavior. This allows for consideration of other facts that might make it more or less likely that the respondent will repeat the behavior, including: the time that has elapsed since the crime, evidence that the root of the crime is being addressed (e.g., if the respondent is participating in an alcohol abuse program), and other evidence of rehabilitation.[167] But there is currently no set level of confidence to which the IJ must find that past dangerous behavior—predictive of future dangerous behavior—occurred. Nor is there a set risk of future behavior that must be found.[168] In theory, every individual presents a non-zero danger to the community as a potential crime committer. The safest course of action will always be to detain. Such presumptive detention, however, cannot be what the Supreme Court meant when it required that detention be based on a finding of danger or flight risk. There must be some level of risk that is sufficient to justify detention, including a level of certainty about the facts that justify that risk determination.[169]  Finally, there is no satisfactory procedure by which the agency’s risk determinations can receive judicial review.

Disaggregating factfinding from discretion suggests a different approach. The Supreme Court’s holding that detention must be justified by a factual finding of flight risk or danger narrows the agency’s discretion to detain. The question of flight risk/danger, therefore, is a threshold eligibility determination that must be made before discretion can be exercised, not merely factfinding in support of an exercise of discretion. In other words, the categorization of detention as a fully discretionary determination is incorrect.[170]

Taking discretion out of the picture—as it is irrelevant at this stage—we are left with a straightforward factual determination, but one that is of utmost importance to get right, standing as it does between an immigrant’s liberty and his detention. It seems unlikely that the Supreme Court would have intended that the risk of error in such an important factual determination should fall on the individual immigrant.[171] It is therefore difficult to see how the BIA’s placement of the burden of proof on the person being detained can be justified at any point, even at the initial bond hearing.[172] Additionally, it defies belief that the Court would write a factfinding requirement into the detention statute to prevent an unconstitutional deprivation of liberty and then leave the determination of such crucial facts to the discretion of the jailer.[173] When the Court entrusted the immigration agency with making the required factual findings necessary to justify detention, it surely did not intend for the agency to make those findings by any process and to any standard it saw fit, nor did it intend to exempt the agency’s decisions from judicial review. Yet that is exactly what has happened. The factual eligibility component, the only thing keeping bond in line with due process, is currently being made as if it were merely a discretionary determination.

An understanding of bond as a factual eligibility determination with a second discretionary step also suggests that bond is properly subject to judicial review under the INA. A “discretionary judgment” to detain, release, or set bond is not reviewable,[174] but factfinding relating to the non-discretionary threshold eligibility determination should be. Indeed, the Supreme Court has cited the language of the bond statute exempting only discretionary judgments from review as an example of how Congress would write a statute if it intended to leave judicial review of factfinding in support of threshold eligibility determinations intact.[175] Restoring judicial review of the agency’s determination of dangerousness or flight risk would be a tremendous step toward bringing bond decisions back into the fold of administrative law.

The understanding that the agency effectively, through burden multiplying, uses arrests as proxies for convictions is also relevant here. The Supreme Court found that it did not violate due process for Congress to mandate that certain criminal convictions be used as proxies for dangerousness, subjecting immigrants to “mandatory” detention without the need for a bond hearing.[176] Arrests, however, are not the same as convictions, and agency action is not the same as congressional legislation. Allowing the agency to treat arrests as the functional equivalent of convictions means that immigrants whom Congress did not intend to subject to mandatory detention are nonetheless subject to its functional equivalent. As with requests for discretionary relief, an immigrant’s burden of proof does not absolve the agency of the obligation to engage in factfinding consistent with the principles of fundamental fairness. The factfinder must find facts to a defined standard, via a process wherein the immigrant has the opportunity to actually influence the decision. It also means that an immigrant cannot be forced, by mere allegations, to prove her innocence using evidence that cannot and does not exist (like a second police report contradicting the original one stating that she was actually innocent all along, or a judicial determination of innocence beyond a reasonable doubt).[177]

In sum, the Supreme Court has written a factfinding requirement into the detention statute to prevent an unconstitutional deprivation of liberty. It would be one thing for the Court to leave the determination of such crucial facts to the agency, to be made pursuant to established standards and subject to judicial review. It would be something else entirely for the Court to leave the determination of such crucial facts to the discretion of the jailer. This is almost certainly not what the Court intended. Additionally, given what is at stake in this factual determination—the government’s ability to strip an individual of their liberty under a preventative civil detention regime—due process almost certainly requires that the government, not the individual, bear the burden of proof.

3.     Hidden Factfinding and the Categorical Approach

The output portability framework also has important implications beyond discretionary decision-making, including for the traditional “crim-imm” question of whether a state criminal conviction makes out a ground of inadmissibility, deportability, or ineligibility for relief.[178] The framework helps illuminate how the agency and courts have sometimes gotten it right and sometimes gotten it very wrong when using CLS outputs in these cases.

As required by the Supreme Court, adjudicators must apply a “categorical” analysis to determine whether certain state convictions sufficiently match with a ground for deportation defined in the immigration law. Rather than conduct a factual inquiry into the circumstances underlying the crime, adjudicators must look only to the statute under which the respondent was convicted.[179] When the crime of conviction is unclear (as when the statute of conviction is divisible into multiple, discrete crimes), adjudicators are permitted to apply a “modified” categorical approach and look to certain CLS outputs, specifically the documents making up the “record of conviction.” These include the statutory definition of the crime, the charging document, the written plea agreement, and the transcript of the plea colloquy.[180] The purpose of this inquiry is to clarify the crime of conviction rather than to make a factual finding about the underlying behavior that led to the arrest.[181] For still other crimes, adjudicators need not use the categorical approach and are allowed to look to outside evidence to determine additional facts—such as the loss to the victim in a fraud case[182]—that are necessary for the conviction to constitute a ground for deportation.[183] This alternative to the categorical approach is known as the “circumstance-specific” approach.

a.     “Record of Conviction” Documents as Fully Portable Outputs

The law, especially at the agency level, reflects significant slippage between the categorial, modified categorical, and circumstance-specific approaches. The agency, demonstrably hostile to the categorical approach, has attempted to blur the lines between the three approaches,[184] most recently with a decision, Matter of Laguerre, that attempts to turn the categorical approach on its head by nonsensically allowing adjudicators to “peek” at unadjudicated allegations in a criminal complaint as part of their divisibility analysis.[185] The agency also signaled, with an invitation for amici, an intention to blur the lines even further by expanding the “record of conviction” to encompass many more documents and allow these documents to be considered under a (supposed) modified categorical analysis. These documents, which include sentencing transcripts and records held by penal institutions, all incidentally record the fact of a conviction but also contain a good deal of additional information about the factual allegations in the case.[186] Making these documents a part of the “record of conviction” would allow any information that also happens to be contained in the same document to be bootstrapped in as well. It is difficult to view the proposed expansion of the “record of conviction” to include these documents as anything but an invitation to adjudicators to engage in factfinding on unadjudicated allegations, thereby destroying the very foundation of the categorical approach.

The output portability framework helps guard against this type of slippage by providing an alternative way to understand what the record of conviction is and how it can be used. Viewed through an output portability lens, the “record of conviction” should be properly thought of as the CLS outputs that are fully portable because they have been adjudicated true beyond a reasonable doubt.[187] This would include the portion of the charging document listing the elements of the crime for which the respondent was convicted, as well as the facts admitted under oath during the plea colloquy or in the written plea agreement that were necessary to make out the crime of conviction. It would not include extraneous information contained in those documents that was not adjudicated since such information is not helpful in discovering what crime the respondent was actually convicted of.[188] The limiting principle of the categorical and modified categorical approaches, then, is that adjudicators are limited to facts already found by the criminal legal system or, more precisely, those facts that were necessarily found because they were essential elements of the crime of conviction.[189] And the only way to know whether an output is part of the record of conviction is to determine whether it has necessarily been adjudicated.[190] Unadjudicated outputs found in the same document are not properly part of the “record of conviction.”

This understanding helps to cut through the recent confusion emanating from the agency. The BIA has been thinking of the “record of conviction” as a set of documents rather than the adjudicated outputs contained within those documents. Understandably, then, it has mistakenly believed that it can look to unadjudicated information that just happens to be in the same document as these outputs. In Laguerre, the unadjudicated output was the identity of the substance that the respondent was charged with having. Because the identity of the substance was not adjudicated by the state court, it cannot properly be considered part of the record of the respondent’s conviction. Looking at an unadjudicated output to determine whether a statute is divisible makes no sense. Even worse, it effectively does away with the categorical approach altogether by allowing adjudicators to engage in their own factfinding to fill in the gaps left by the state court.

Similarly, the BIA’s gesture at expanding the documents that make up the “record of conviction” to include documents like sentencing transcripts that contain wholly unadjudicated outputs, and to use these outputs to supposedly determine the crime of conviction is equally nonsensical.[191] Such unadjudicated outputs have no bearing on the elements of the crime the respondent was convicted of and cannot properly be considered part of the “record of conviction.” Any use to which the BIA hopes to put such unadjudicated information is, necessarily, in aid of factfinding, which is not permitted under a modified categorical analysis.[192]

b.     Hidden Factfinding and the Circumstance-Specific Inquiry

The output portability framework can also shed light on the circumstance-specific inquiry, in which adjudicators are permitted to conduct their own factfinding on CLS outputs. The framework illuminates how the agency and reviewing courts have sometimes erred by allowing adjudicators to treat outputs as if they had already been adjudicated rather than using them in aid of factfinding in the immigration system.

One area where the circumstance-specific approach is allowed is to determine the amount of loss to the victim in fraud cases. A crime of “fraud” is an aggravated felony under immigration law if the loss to the victim is over $10,000.[193] Many states do not require that the exact loss amount be adjudicated. For example, a state statute might punish a loss over $5,000 or between $5,000 and $100,000. The agency, which would like to deport more people than would be possible under the categorical analysis,[194] has, with Supreme Court approval, allowed immigration adjudicators to abandon the categorical approach and look to other evidence, usually restitution orders, to determine the actual amount of loss to the victim.[195] Because restitution orders are normally determined by a preponderance of the evidence while the immigration system requires a higher standard of “clear and convincing” evidence to establish deportability, it is not acceptable for an IJ to simply assume that the order would have been adjudicated to the higher standard if only the judge had been asked to do so. Instead, they must conduct their own factfinding.

Reviewing courts have, after some false starts,[196] generally arrived at an approach that does not erroneously treat nonportable or partially portable outputs as if they were fully portable. Because restitution amounts can be calculated based on charges for which the respondent was not convicted,[197] meaning that the facts that have been adjudicated by the criminal legal system are not necessarily the same facts that the immigration system requires, courts have correctly required that the loss be “tethered” to the crime of conviction.[198] This, in practice, means that a conclusory restitution order with no underlying information about how the loss amount was calculated will not support a factual finding that the loss from the crime of conviction was greater than or equal to $10,000. An adjudicator with access to the same underlying information relied upon by the criminal court that created the restitution order might conduct its own factfinding and ultimately determine the loss to the victim according to the “clear and convincing” standard required. But what the adjudicator cannot do is simply import the conclusion of the order wholesale while also giving the criminal court’s factfinding a promotion from a lower standard to a higher one.[199]

In contrast, the agency has erred when using CLS outputs in another circumstance-specific inquiry: the question of whether marijuana convictions make a noncitizen deportable. Marijuana possession is a deportable controlled substance offense unless it involves possession of thirty grams or less for personal use.[200] Though the Supreme Court has held that the categorical approach applies to controlled substance offenses,[201] the BIA has said that IJs may use CLS outputs, specifically a pre-sentence investigation report (PSI), to determine whether the thirty-gram exception applies, thereby finding facts that make a non-deportable conviction into a deportable one. Justifying this decision, the BIA stated that the use of the PSI was essentially the same as the use of restitution orders to identify a fraud aggravated felony:

[A]s with the $10,000 victim loss issue . . . the immigration statute requires the DHS to come forward with “clear and convincing evidence” that the respondent possessed more than 30 grams of marijuana (or that he possessed marijuana for some reason other than personal use) . . . . It also contemplates the use of fundamentally fair procedures that give respondents a reasonable opportunity to dispute any DHS claim that the exception is inapplicable.[202]

Setting aside the manifest error in ignoring the Supreme Court’s clear instruction to use the categorical approach in analyzing controlled substance offenses, the BIA is also wrong to equate the way that CLS outputs are used in these two scenarios. Restitution orders in fraud aggravated felony determinations are not treated as fully portable. They are partially portable, to be used in aid of factfinding, and, in this factfinding, the evidence must actually support the necessary facts. In contrast, when using PSIs in marijuana determinations, the agency treats them as if factfinding to a sufficiently high standard has already occurred. Despite the BIA’s assertion that DHS would have to come forward with “clear and convincing” evidence that the marijuana was not for personal use, on remand the IJ relied on the PSI: an output not adjudicated to any standard, let alone clear and convincing evidence.[203] The subsequent BIA appeal affirming the respondent’s removal order made no mention of the sources of information contained in that PSI or whether those sources were even identified, nor did it discuss whether the respondent presented his own evidence or how that was weighed against the PSI. It merely stated that the PSI “reflects the circumstances that resulted in the respondent’s conviction for possession of marijuana.”[204] In other words, the court treated the output as if it were adjudicated to a sufficiently high standard that it was fully portable into the immigration system. Any comparison to the fraud aggravated felony ground, therefore, is facile and does not hold up to scrutiny. Moreover, it is difficult to see how such an adjudication can be considered “fundamentally fair.”[205]

4.     Implications Beyond the Immigration System

The immigration system is far from the only place where CLS outputs are used to determine whether an applicant is eligible for or deserving of some benefit or relief. CLS outputs are used by the child welfare system, in benefits eligibility determinations, in housing court, by professional licensing bodies, and in many other contexts.[206] The output portability framework is applicable to all of these. Granted, many of these determinations are not discretionary, and it would hopefully be clearer that decision makers are conducting fact finding using CLS outputs. But, as can be seen in the discussion of the categorical and circumstance-specific approaches immediately above, even in nondiscretionary adjudications decision makers are sometimes confused about how to conduct fact finding using CLS outputs, and they sometimes assume that an output has been adjudicated when it has not or mistakenly give too much weight to an output simply because it emerged from the criminal legal system. It is my hope that the framework can prove useful in helping decision makers differentiate between outputs that have been adjudicated, and thus rightfully carry the presumption of reliability that, for better or worse, we ascribe to the criminal legal system, and those that have been merely alleged and do not deserve to bear its imprimatur.

5.     Suggestions for Further Scholarship

This piece suggests several areas for further scholarship. First, the possibility that the agency could define the standard to which facts must be established before they can be considered a negative “factor” suggests a way to differentiate among discretionary adjudications depending on the immigrant’s interest in the relief sought. While it might be permissible to deny admission to a would-be first-time entrant to the country based on a low standard of proof (the mere possibility that allegations are true), a higher standard should be required for more established immigrants and a higher standard still should be needed to justify civil detention during immigration proceedings. Identifying the standard to which facts must be found would not require dramatically different procedures but would allow for bond adjudications, and potentially other adjudications where the immigrant has a greater interest in the relief, to be brought more in line with our instinctive ideas of fairness. For bond proceedings, this might prove an alternative way to protect the liberty interest, rather than shifting the burden of proof to the government at an undefined time on a case-by-case basis as detention becomes prolonged.[207]

Second, further discussion of partial portability and non-portability is warranted. The operating assumption that almost nothing is excludable from the immigration system may be due for revision in light of the identification of hidden factfinding and the realization that CLS outputs are being used in aid of (rule-bound) factfinding rather than (rule-free) discretion. Lawmakers or agency rule makers might well decide that certain outputs—those with low evidentiary value combined with a high potential for prejudice, specifically the high likelihood that adjudicators will err by giving them too much weight[208]—should, in fairness, be excluded altogether. While the risk of adjudicator error might be mitigated by rigorous training and strict oversight, the agency could choose to make things easier for itself by excluding certain categories of outputs entirely. It is widely acknowledged that immigration court is not an appropriate place to relitigate the facts of a criminal case. This is due to the lack of competence of immigration adjudicators, the limitations of the forum, the inevitable unfairness to respondents, and the inefficiency of holding mini trials in every case.[209] When “hidden” factfinding is identified, it becomes clear that re-litigating criminal cases is exactly what adjudicators are being asked to do in the discretionary context. The same considerations that compel a categorical approach to analyzing many criminal convictions might similarly compel a stricter exclusionary rule for unadjudicated CLS outputs in discretionary determinations.

Third, a watershed lawsuit has recently opened up the possibility of proving or disproving the assertion that agency adjudicators regularly deny discretionary applications based on the mere fact of arrest. Previously, any attempt to quantify agency practice with respect to CLS outputs was hampered by the agency’s lack of transparency: USCIS adjudications and IJ decisions are not available to the public or to litigants. The same is generally true of AAO opinions, which are unpublished but sometimes show up on legal research sites, and of BIA opinions, of which only a small minority are published each year (twenty to thirty out of 30,000).[210] The BIA, however, was sued under the Freedom of Information Act and entered into a stipulation agreeing to start making its unpublished decisions available to the public.[211] Once a significant number of decisions become publicly available, there will be tremendous potential for scholars seeking to understand agency practices. It will be possible, for example, to study the relationship between arrests and grants of relief. This article posits that there will be a strong negative correlation, even in the absence of a conviction. The data can confirm, or deny, that hypothesis.

Finally, given what we know about racial disparities in arrests, equal protection might be implicated by treating CLS outputs as fully portable, equivalent to criminal convictions, or by creating a nearly unsurmountable burden for respondents to disprove them. Immigrants from communities targeted by police are much more likely to be arrested, making the use of these outputs as stand-ins for criminal convictions highly problematic.[212] Indeed, the selective targeting of minority groups for arrest, especially for crimes like drug possession that are common among the population at large, might counsel against using even convictions for certain types of crimes against an immigrant. Certainly, then, using the mere fact of arrest should be looked at with extreme skepticism as relying on arrests elevates individual police officers to the role of immigration judge, effectively giving them the final say on whether an immigrant will be deported.[213]

Conclusion

My client’s immigration journey did not end with the denial letter quoted at the beginning of this piece. It took years of advocacy from an entire team of attorneys,[214] but USCIS eventually relented and issued her green card. I selected her case to highlight because it was my first encounter, as a young lawyer, with the distortions that can arise when discretion is unchecked. The decision seemed obviously wrong on its face: my client had been adjudicated not guilty by the criminal legal system. But, in the immigration system, all the original allegations against her were back from the dead, considered true unless she could somehow prove their untruth. How she was supposed to do this without access to any of the information relied upon by the drafter of the document and with no chance to cross examine the original source of that information was unclear. This had to be error. Yet every source I consulted said, basically, “It’s a discretionary decision. They can pretty much do what they want.” The case resolved years ago, but I have continued to be obsessed with figuring out why the agency was wrong, not just ethically or prudentially, but also legally. This is the article that grew out of that obsession.

To summarize the conclusions of this piece: discretion is baked into the immigration system, for better or worse. But discretion to decide does not mean discretion to find facts by any procedure the adjudicator sees fit, based on any evidence, to any standard they deem appropriate. Fact finding, even in support of exercises of discretion, must be fair. Today, in discretionary immigration adjudications, it is not. This situation has been allowed to persist because fact finding has been “hidden” under a cloak of discretion so the assumptions that underlie it have not received any meaningful oversight. The result is shoddy decision-making that has tremendous consequences for individual immigrants, for the communities they are part of, and for the nation as a whole. It is my hope that, when hidden fact finding is properly identified, both the immigration agency and the courts will be better able to ensure that it is done fairly and effectively so that no immigrant should be forced to endure detention, denial of relief, and deportation solely because they once had the misfortune to be arrested.

Copyright © 2024 Sarah Vendzules. The author would like to thank Jennifer Chacón, Adam Cox, Alina Das, Eric Fish, Stephen Gillers, Randy Hertz, Mary Holper, Eisha Jain, Nancy Morawetz, Anna Roberts, and Jason Vendzules for their thoughtful feedback.

           [1].     See Nayna Gupta, Prejudicial and Unreliable: The Role of Police Reports in U.S. Immigration Detention and Deportation Decisions, Nat’l Immigrant Just. Ctr., 7–8 (2022) (citing examples of objectively unreliable police reports being used against immigrants and reporting that “every legal service provider [we] interviewed” agreed that immigration decision makers take the information in police reports to be true); see also Erica D. Rosenbaum, Note, Relying on the Unreliable: Challenging USCIS’s Use of Police Reports and Arrest Records in Affirmative Immigration Proceedings, 96 N.Y.U. L. Rev. 256, 260 (2021) (“[I]n an increasing number of cases, USCIS is using . . . ‘facts’ [from police reports and charging documents] to deny applications.”). While anecdotes abound, meaningful data is harder to come by because the vast bulk of the immigration agency’s decisions have not generally been available to the public or to litigants. See Amanda Frost, Deportation Without Disclosure: Immigration Courts Need Transparency, Bloomberg News (Feb. 23, 2021), https://news.bloomberglaw.com/us-law-week/deportation-without-disclosure-immigration-courts-need-transparency [https://perma.cc/ZU84-SZ5U].

           [2].     See infra Part A.2.

           [3].     While it may seem like it is the conviction itself that is the ground upon which the immigration system acts, more accurately it is the facts found by the criminal legal system that are the basis for an action taken. This is evident in the immigration system’s treatment of convictions that have been vacated or otherwise invalidated in the state courts that created them. Where charges are dismissed after completion of a rehabilitative program, the defendant is still considered “convicted” for immigration purposes if the court found facts sufficient to make out the crime. Matter of Punu, 22 I&N Dec. 224 (BIA 1998); Matter of Salazar, 23 I&N Dec. 223 (BIA 2002). A full pardon will not suffice to eliminate the immigration consequences of many crimes. Matter of Suh, 23 I&N Dec. 626 (BIA 2003). And a vacatur that is not based on the “merits” of the conviction (i.e., the ultimate question of guilt or innocence) will be ineffective for immigration purposes. See Matter of Pickering, 23 I&N Dec. 621 (BIA 2003); Matter of Roldan, 22 I&N Dec. 512 (BIA 1999); Matter of Marroquin, 23 I&N Dec. 705 (A.G. 2005); Matter of Thomas and Thompson, 27 I&N Dec. 674 (A.G. 2019).

           [4].     I do not intend to uncritically parrot the idea that the criminal legal system produces reliable results. See generally Anna Roberts, Convictions as Guilt, 88 Fordham L. Rev. 2501 (2020) (despite copious scholarship on the unreliability of criminal convictions, even scholars tend to conflate convictions with crime commission). Given the coercive plea bargaining system and the steep incentives for innocent people to plead guilty, it would be naïve to assume that all, or perhaps even most, of the defendants in the ninety five percent of convictions that are obtained through guilty pleas are factually guilty. For a well-cited summary of the literature on the prevalence of false pleas and the factors that contribute to them, see generally Josh Bowers, Punishing the Innocent, 156 U. Pa. L. Rev. 1117 (2008) (arguing that guilty pleas by innocent defendants are a rational, expected, and beneficial part of the criminal legal system). And, for a compelling critique of the idea that criminal courts engage in factfinding around guilt and innocence, see Issa Kohler-Hausmann, Managerial Justice and Mass Misdemeanors, 66 Stan. L. Rev. 611, 614 (2014) (arguing that misdemeanor courts have largely abandoned an “adjudicative model” aimed at separating defendants by guilt or innocence and embraced a “managerial model” organized around “the supervision and regulation of the population that flows through [the] courts, often with little attention to questions of guilt in individual cases”). This piece concerns the use of criminal legal system outputs created without the process and protections that go into criminal convictions. These outputs are, therefore, even less reliable than the outputs of that notably flawed system.

           [5].     See Anna Roberts, Arrests as Guilt, 70 Ala. L. Rev. 987, 994 (2019) (“Our system for determining legal guilt, which sets up various processes and protections that must be honored in order to permit a valid declaration of legal guilt, is the primary proxy that we have for factual guilt.”); see also David L. Shapiro, Should a Guilty Plea Have Preclusive Effect?, 70 Iowa L. Rev. 27 (1984) (discussing issue preclusion or collateral estoppel and the conceptual foundations of importing criminal convictions into civil proceedings). For a critique of the preclusive value of guilty pleas in subsequent proceedings, see generally Brandon L. Garrett, Why Plea Bargains Are Not Confessions, 57 Wm. & Mary L. Rev. 1415 (2016).

           [6].     This is my term, though other scholars have used it to express a similar concept. See generally Eric Fish, The Paradox of Criminal History, 42 Cardozo L. Rev. 1373 (2021).

           [7].     A convicted count represents an “output” that has been created through a process involving input from both the prosecution and the defense under the oversight of the court. As a result of this process, the relevant facts have been adjudicated to a set degree of certainty (beyond a reasonable doubt). In contrast, an unconvicted count is created exclusively by one side, the prosecution, and has not been subject to confrontation or testing by the adversarial process nor adjudicated to any degree of certainty.

           [8].     See generally Michael G. Heyman, Judicial Review of Discretionary Immigration Decision-making, 31 San Diego L. Rev. 861 (1994).

           [9].     The adjudicator, whether an immigration judge employed by the Department of Justice or a USCIS adjudicator employed of the Department of Homeland Security, is a representative of the executive branch.

         [10].     While immigration practitioners distinguish between “relief” (sought in removal proceedings) and “benefits” (sought before USCIS), there is no meaningful difference in the standards employed or the rights accorded the applicant. For the purposes of this piece, I use the term “relief” when speaking broadly of something that might be sought to allow a person to stay in the country or gain lawful status.

         [11].     Discretionary relief includes adjustment of status, LPR and non-LPR cancellation, 212(c), 212(h), other waivers, U visas, VAWA, and asylum. Nondiscretionary relief includes CAT, withholding of removal, and certain types of adjustment of status.

         [12].     This is the central concern of immigration law’s “categorical approach” line of cases, which lay out rules for when and how CLS outputs from a criminal conviction can be used by immigration decision makers to determine whether that conviction is a ground of deportability or inadmissibility. See Descamps v. United States, 570 U.S. 254, 258 (2013); Mathis v. United States, 579 U.S. 500, 517 (2016); Moncrieffe v. Holder, 569 U.S. 184, 202 (2013).

         [13].     Discretion alone cannot explain how unadjudicated CLS outputs are sometimes misused to justify a finding of, e.g., deportability (a nondiscretionary determination), except that adjudicators might be applying habits of sloppy thinking developed in discretionary determinations. But, as I explain below, the framework that I advance in this piece is useful in curbing the misuse of CLS outputs in both discretionary and nondiscretionary decision-making.

         [14].     The immigration “agency” refers broadly both to the Department of Homeland Security (formerly Immigration and Naturalization Services) (DHS), of which USCIS is a subset, and the Department of Justice (DOJ) which houses the Executive Office for Immigration Review containing the Board of Immigration Appeals (BIA or the Board) and the Immigration Judges (IJs). Both Departments are part of the executive branch. Both DHS and the DOJ promulgate immigration regulations. These regulations are one source of agency rulemaking. The other source is the BIA, which interprets immigration law and regulations. There is also a mechanism by which the Attorney General (AG) can change the law without regulations: because the BIA is not a court but rather an adjudicative body within the DOJ, the AG has the power to overturn any BIA decision, thereby changing the law. Immigration regulations, in turn, are frequently amended to reflect the rulings of the BIA and the AG.

         [15].     I believe this lack of rigor in analysis is also due to the unfamiliarity of immigration adjudicators with the system that produced these outputs. This is based on my own anecdotal observation and the admission of immigration judges that I have practiced before.

         [16].     See U.S. Citizenship and Immigration Services, 2020 USCIS Statistical Annual Report 1, 3 (2020), https://www.uscis.gov/sites/default/files/document/reports/2020-USCIS-Statistical-Annual-Report.pdf [https://perma.cc/4689-LGCV].

         [17].     See Hon. Mark A. Drummond, “Death Penalty Cases in a Traffic Court Setting”: Lessons from the Front Lines of Today’s Immigration Courts, 44 Litig. News 26, 26 (2018-2019) (referencing statement by IJ Dana Leigh during an appearance on Last Week Tonight with John Oliver).

         [18].     Eisha Jain, Arrests as Regulation, 67 Stan. L. Rev. 809, 815 (2015) (“Noncriminal justice actors rely on arrests . . . because they are relatively easy and inexpensive to access and because they regard arrests as proxies for information they value, such as the potential for violence, unreliability, or instability.”); see generally Eric Fish, The Paradox of Criminal History, 42 Cardozo L. Rev. 1373 (2021).

         [19].     Michael Pinard & Anthony C. Thompson, Offender Reentry and the Collateral Consequences of Criminal Convictions: An Introduction, 30 N.Y.U. Rev. L. & Soc. Change 585, 585–87 (2006); McGregor Smyth, From “Collateral” to “Integral”: The Seismic Evolution of Padilla v. Kentucky and Its Impact on Penalties Beyond Deportation, 54 Howard L.J. 795, 825 (2011); Gabriel J. Chin, Race, the War on Drugs, and the Collateral Consequences of Criminal Conviction, 6 J. Gender Race & Just. 253, 255 (2002).

         [20].     Jain, supra note 18, at 810–11 (describing consequences of arrests in decisions about immigration enforcement, public housing, public employment and licensing, foster care, social service, and education: “[i]n each of these contexts, it is the fact of an arrest itself—not only a subsequent conviction—that triggers a regulatory decision, such as deportation, eviction, loss of a professional license, or loss of custody”); Roberts, supra note 5, at 998–99 (identifying a “fusion” of arrests and guilt in the minds of decision makers and listing legal consequences of arrests that arise out of the assumption of guilt or “likelihood of guilt”).

         [21].     See, e.g., Alina Das, The Immigration Penalties of Criminal Convictions: Resurrecting Categorical Analysis in Immigration Law, 86 N.Y.U. L. Rev. 6 (2011) (laying out the historic and textual basis for limiting immigration officials to the statute of conviction when determining whether a given conviction is a ground for removal); see also Stephen Lee, De Facto Immigration Courts, 101 Calif. L. Rev. 553, 555 (2013) (showing how criminal prosecutors making charging and plea decisions are able to effectively determine which immigrants will become deportable and which will be able to remain in the United States).

         [22].     See, e.g., Adam B. Cox & Cristina M. Rodríguez, The President and Immigration Law 87–92, 111–14, 124–130 (2020) (overbroad laws leave the executive with discretion to decide who, amongst a large pool of deportable individuals, to prosecute); Eisha Jain, Jailhouse Immigration Screening, 70 Duke L.J. 1703 (2021). Both might also fairly be categorized as “discretion” scholarship as they deal with prosecutorial discretion—who to prosecute—in contrast with the adjudicative discretion discussed in this piece.

         [23].     Peter H. Schuck, The Transformation of Immigration Law, 84 Colum. L. Rev. 1, 1 (1984) (describing lackluster judicial review even before jurisdiction was stripped from the federal courts to review most discretionary immigration decision-making).

         [24].     Hiroshi Motomura, The Curious Evolution of Immigration Law: Procedural Surrogates for Substantive Constitutional Rights, 92 Colum. L. Rev. 1625, 1628 (1992); see generally Hiroshi Motomura, Immigration Law after a Century of Plenary Power: Phantom Constitutional Norms and Statutory Interpretation, 100 Yale L.J. 545 (1990).

         [25].     Gerald L. Neuman, Discretionary Deportation, 20 Geo. Immigr. L.J. 611, 636 (2006).

         [26].     Daniel Kanstroom, Surrounding the Hole in the Doughnut: Discretion and Deference in U.S. Immigration Law, 71 Tul. L. Rev. 703, 705 (1997).

         [27].     Shoba Sivaprasad Wadhia, Darkside Discretion in Immigration Cases, 72 Admin. L. Rev. 367, 369–70 (2020).

         [28].     Shalini Bhargava Ray, Immigration Law’s Arbitrariness Problem, 121 Colum. L. Rev. 2049, 2095 (2021).

         [29].     Peter L. Markowitz, Deportation is Different, 13 U. Pa. J. Const. L. 1299, 1301 (2011) (discussing how courts should determine which rights from the criminal system should be incorporated into the immigration system); see generally Mary Holper, Confronting Cops in Immigration Court, 23 Wm. & Mary Bill Rts. J. 675 (2015) (arguing for a right of confrontation for police reports).

         [30].     See, e.g., Angela M. Banks, The Normative and Historical Cases for Proportional Deportation, 62 Emory L.J. 1243, 1298 (2013); Jason A. Cade, Enforcing Immigration Equity, 84 Fordham L. Rev. 661, 665 (2015); Daniel Kanstroom, Smart(er) Enforcement: Rethinking Removal, Structuring Proportionality, and Imagining Graduated Sanctions, 30 J. L. & Pol. 465, 466 (2015).

         [31].     Fatma E. Marouf, Immigration Law’s Missing Presumption, Georgetown L. J. (forthcoming).

         [32].     See Motomura, Immigration Law After A Century of Plenary Power, supra note 24, at 563–64 (suggesting that constitutional norms, which do not apply to immigration adjudications because of the plenary power, have nevertheless become part of the legal culture, and continue to influence judges when they interpret the immigration law); see also Motomura, Curious Evolution, supra note 24, at 1629–30 (judges attempt to protect substantive rights through procedural surrogates as a “principled way of discharging their constitutional responsibilities”).

         [33].     See Holper, supra note 29, at 677 (identifying what she terms the “police report problem”: police reports being used in immigration cases as a basis to deny discretionary relief, despite the manifest unreliability of the information therein). Part A expands on that work.

         [34].     Matter of Roberts, Interim Dec. 3148 (BIA 1991) (individual tried to argue entrapment; IJ could not consider the argument because it was about guilt or innocence, which had already been established); Matter of Edwards, 20 I&N Dec. 191, 197 (BIA 1990); USCIS Policy Manual Volume 1, General Policies and Procedures, Part E, Adjudications, Ch. 8, Discretionary Analysis, n.61 (last visited 2023), available at https://www.uscis.gov/policy-manual/volume-1-part-e-chapter-8#footnote-61 [https://perma.cc/KK9E-RZDR]. This was not always the case. Matter of G-, 1 I&N Dec. 8 (BIA 1940) (expressing doubt that the decision-maker does not retain discretion to question a conviction if there is reason to doubt the respondent’s guilt; crediting the testimony of the respondent that he was coerced into the plea).

         [35].     It is worth noting that the Executive Office of Immigration Review, which houses the BIA and the IJs is “neither a judicial body nor an independent agency. It is created by the Attorney General as part of his office, he names its members, and they are responsible only to him. It operates under his supervision and direction, and its every decision is subject to his unlimited review and revision.” Anthony R. Enriquez, Structural Due Process in Immigration Detention, 21 CUNY L. Rev. 35, 48 (2017) (internal citation omitted).

         [36].     See Matter of Gonzalez, 16 I&N Dec. 134, 136 (BIA 1977) ( “[The INA] nowhere states that the criminal activity and the conviction which resulted therefrom cannot be considered in connection with an application for discretionary relief.”).

         [37].     See Patrisia Macías-Rojas, Immigration and the War on Crime: Law and Order Politics and the Illegal Immigration Reform and Immigrant Responsibility Act of 1996, 6 J. Migration & Hum. Security. 1, 1 (Nov. 12, 2018).

         [38].     21 I&N Dec. 20 (BIA 1995).

         [39].     Matter of Seda, 17 I&N Dec. 550 (BIA 1980).

         [40].     Thomas, 21 I&N Dec. at 23.

         [41].     Id. at 24 (emphasis added).

         [42].     Id. (emphasis original).

         [43].     Thomas is not without ambiguity. Even within the same opinion, the BIA speaks broadly of “conduct” in a way that conflates allegations with adjudicated facts. The BIA asked, “whether unfavorable conduct by an alien, even in the absence of a final conviction, may be considered in the exercise of discretion.” Id. at 23 (emphasis added).

         [44].     Matter of Catalina Arreguin De Rodriguez, 21 I&N Dec. 38 (BIA 1995).

         [45].     Matter of Teixeira, 21 I&N Dec. 316, 321 (BIA 1996).

         [46].     Id. at 321.

         [47].     See, e.g., Matter of Guerra, 24 I&N Dec. 37, 40–41 (BIA 2006).

         [48].     Matter of Sotelo-Sotelo, 23 I&N Dec. 201, 205 (BIA 2001) (“[W]e do not agree with the respondent’s contention that we cannot consider this warrant in our adjudication of his . . . claim. However, in the absence of a conviction, we find that the outstanding warrant should not be considered an adverse factor in this case.”).

         [49].     21 I&N Dec. at 42.

         [50].     Id. (emphasis added).

         [51].     Sotelo-Sotelo, 23 I&N Dec. at 205.

         [52].     Guerra, 24 I&N Dec. at 40. According to the complaint, a “reliable” confidential informant had told a DEA agent that the respondent was a drug dealer and, during a police stakeout, was spotted in a car with someone who later got into another car in which narcotics were found and who told the police he was going to sell the drugs with the respondent. Id. at 38.

         [53].     Id. at 41. The signature of the DEA agent in particular has a dubious relationship to corroboration. The signature provides no independent evidence of the truth of the contents therein. The DEA agent who signed the form does not appear to be someone who had direct knowledge of the events in question and may not have even been the same DEA agent who spoke to the confidential informant, making the value of this signature questionable at best.

         [54].     Id. at 39 (“[T]he respondent failed to present any evidence or argument that tended to undermine the reliability of the information contained in the complaint.”).

         [55].     Id. at 41. Guerra can be read consistent with Thomas and Arreguin only with considerable effort. Thomas contemplates that CLS outputs should weigh more as the criminal case progresses. But the criminal case in Guerra had not progressed very far along at all: charges had been filed, and the defendant had been arraigned. The case had, however, progressed farther than in Arreguin where prosecution was declined.

         [56].     Matter of D-A-C-, 27 I&N Dec. 575 (BIA 2019).

         [57].     This argument echoes the animating principle behind the categorical approach in nondiscretionary immigration adjudications: a criminal conviction is necessarily only an admission to or a finding that the respondent engaged in the least-culpable behavior criminalized by a statue.

         [58].     27 I&N Dec. at 578, 579.

         [59].     27 I&N Dec. at 578.

         [60].     See Matter of Gonzalez, 16 I&N Dec. 134, 135–36 (BIA 1977).

         [61].     Matter of Grijalva, 19 I&N Dec. 713, 722 (1988) (emphasis added).

         [62].     Whether the respondent was eligible for the “30 grams or less of marijuana” exception to controlled substance offense inadmissibility would, under today’s law, be characterized as an issue of baseline eligibility for a waiver rather than a discretionary determination.

         [63].     19 I&N Dec. at 722.

         [64].     24 I&N Dec. at 40.

         [65].     To be eligible for a U visa, an individual must have been a victim of a crime in the United States and must have cooperated with law enforcement in the prosecution of that crime. After two years in this status, during which time the U visa recipient has been protected from deportation and eligible for work authorization and a social security card, the U visa recipient may apply to “adjust status” to that of lawful permanent resident. U adjustment applicants must prove, inter alia, that their continued presence in the United States is “justified on humanitarian grounds, to ensure family unity, or is in the public interest.” 8 USC § 1255(m); see also 8 CFR § 245.24(b)(6). The U regulations, following the BIA’s guidance on discretionary adjudications generally, provide that where “adverse factors” are present, the applicant “may offset these by submitting supporting documentation of mitigating equities,” except that some adverse factors, including where the applicant has “committed or been convicted of” a serious violent crime or multiple drug-related crimes, cannot generally be overcome, even with a heightened showing of extreme and extraordinarily unusual hardship. 8 CFR § 245.24(d)(11).

         [66].     Matter of J-V- (AAO Dec. 11, 2018) 2018 WL 6985737, at *6–7 (italics original; underlining added). The AAO’s review was de novo and deference to the original adjudicator played no part in the decision.

         [67].     Id. at 6 n.5, 8.

         [68].     Id. at 8.

         [69].     Id. at 3–4.

         [70].     See Gupta, supra note 1, at 9.

         [71].     Holper, supra note 29, at 677 (decrying the use of unreliable police reports as a basis to deny discretionary relief).

         [72].     Matter of M-A-P- (AAO Oct. 31, 2018) 2018 WL 6075447, at *7 (emphasis added).

         [73].     See generally Gupta, supra note 1; O’Doan v. Sanford, 991 F.3d 1027, 1046 (9th Cir. 2021) (“Police reports can be written quickly, at odd hours, and with other law enforcement matters pressing.”).

         [74].     Matter of J-V-, 2018 WL 6985737, *6. This language is likely drawn from Arreguin.

         [75].     Id. at *8.

         [76].     The burden on a party seeking to prevail on a discretionary determination is sometimes called a burden of “persuasion,” but this tends to obscure the role that factfinding plays in these adjudications. For almost every discretionary decision, the decision-maker makes factual determinations on the way to that decision (e.g., a sentencing judge finding victim impact testimony credible before imposing a discretionary sentence on a defendant). Few applications will ever be purely discretionary such that factfinding plays no role. If an adjudicator must consider various factors in the exercise of their discretion, they must either engage in factfinding themselves or credit already-conducted factfinding as to the presence or absence of those factors before making a discretionary decision.

         [77].     For example, in a criminal prosecution, the prosecutor bears the high burden of establishing guilt beyond a reasonable doubt. This does not mean that they must prove that nobody else could have committed the crime.

         [78].     For example, an adjudicator may choose to discredit a respondent’s statement because it is internally inconsistent or may credit it because it rings true. They may find the respondent’s evidence of good moral character insufficient because his statement is uncorroborated, or they may credit it because he has submitted corroborating letters from reliable sources like clergy or community leaders. See, e.g., Application: Petition For Special Immigrant Battered Spouse, (AAO 2001) 2001 WL 34077958, at *3 (holding that VAWA self-petitioner did not meet burden to show he was a person of good moral character, which requires that he has not willfully failed to support dependents, because he did not corroborate his claim that he had paid off his child support arrears).

         [79].     See Hernandez-Lara v. Lyons, 10 F.4th 19, 31 (1st Cir. 2021) (“[P]roving a negative (especially a lack of danger) can often be more difficult than proving a cause for concern.”).

         [80].     See Gupta, supra note 1, at 9.

         [81].     24 I&N Dec. at 39.

         [82].     Velasco Lopez v. Decker, 978 F.3d 842, 846 (2d Cir. 2020) (affirming district court’s reversal of the BIA).

         [83].     Id.

         [84].     Matter of J-V-, 2018 WL 6985737 at *4.

         [85].     Id. at *7.

         [86].     See Mena-Flores v. Holder, 776 F.3d 1152, 1156 (10th Cir. 2015) (holding it permissible to consider trial testimony where respondent was acquitted at trial).

         [87].     See, e.g., Motion On Administrative Appeals Office Decision Form I-360, Petition For Abused Spouse Or Child of U.S. Citizen, 2023 WL 2402303, at *2 (finding VAWA self-petitioner lacked good moral character where her DUI charge had been dismissed but she did not “not provide additional information, e.g., the transcript of proceedings, to [explain] why”). This assumes, naïvely, that a prosecutor would state on the record that charges had been unfounded, thereby opening the door to a lawsuit for wrongful arrest.

         [88].     Schware v. Bd. of Bar Exam’rs, 353 U.S. 232, 241 (1957) (“The mere fact that a man has been arrested has very little, if any, probative value in showing that he has engaged in any misconduct. An arrest shows nothing more than that someone probably suspected the person apprehended of an offense.”); Michelson v. United States, 335 U.S. 469, 482 (1948) (“Arrest without more does not, in law any more than in reason, impeach the integrity or impair the credibility of a witness. It happens to the innocent as well as the guilty.”).

         [89].     See generally Marouf, supra note 31 (confronting the same problem and arguing for a presumption of innocence in the immigration system, including in discretionary adjudications). The imposition of such a presumption would be an alternative means to address burden multiplying and zombie allegations and, indeed, would be more protective as it would necessarily shift the burden to establish guilt onto the government. Marouf is also implicitly arguing for a separation of the factfinding and discretionary portions of discretionary adjudications. Such a separation is necessary for the government to bear the burden of proof to establish factual guilt while still maintaining adjudicator discretion to determine whether to grant relief. As such, the two proposals might be seen to be complementary rather than mutually exclusive.

         [90].     Such use of arrests as proxies for undesirability echoes the “managerial” model of criminal law identified by Kohler-Hausmann where the criminal legal system is used primarily as a way to mark individuals as potentially problematic. Her research demonstrated that the most important factor in whether a defendant ended up with a conviction was not the strength of the evidence against them but rather whether they had been arrested or convicted in the past. Kohler-Hausmann, supra note 2, at 672–73. Analogously, in the immigration system, the most important factor in whether discretionary relief is granted or denied may well be the existence of a prior arrest.

         [91].     See, e.g., Jafarzadeh v. Nielson, 321 F. Supp. 3d 19, 40 (D.D.C. 2018) (holding that the agency may have overstepped its authority with a policy that effectively changed the statutory requirement for an immigration benefit).

         [92].     See, Race and Policing, National Association of Criminal Defense Lawyers, https://www.nacdl.org/Content/Race-and-Policing (Nov. 29, 2022) https://www.nacdl.org/Content/Race-and-Policing[https://perma.cc/5SV6-Y9YX] (collecting research on racially biased policing).

         [93].     This is conceptually similar to, and draws upon, the doctrine of collateral estoppel, also known as issue preclusion, which deals with judgments rather than fact finding (though the two are often related). See Parklane Hosiery Co. v. Shore, 439 U.S. 322, 336 n.23 (1979) (“[T]he whole premise of [issue preclusion] is that once an issue has been resolved in a prior proceeding, there is no further fact-finding function to be performed.”). The Supreme Court has held that due process requires that the party against whom an earlier decision is asserted must have had a “full and fair opportunity to litigate” the claim or issue, and if not, the issue in question must be redetermined. See Kremer v. Chem. Const. Corp., 456 U.S. 461, 480–81 (1982) (internal quotations marks omitted).

         [94].     But see Roberts, supra note 4, at 2520–21, 31 (estimating that the standard of proof established by guilty pleas better approximates probable cause than reasonable doubt).

         [95].     F. R. Ev. 401 (“Evidence is relevant if it has any tendency to make a fact more or less probable than it would be without the evidence and the fact is of consequence in determining the action.”).

         [96].     See Olivas-Motta v. Holder, 746 F.3d 907, 919 (9th Cir. 2014) (Kleinfeld, J., concurring) (“[P]olice reports are not especially useful instruments for finding out what persons charged actually did. All the defects of hearsay, double hearsay, and triple hearsay apply, since people may speak to the police despite lack of personal knowledge and lack of adequate observation, may be misunderstood, and what they say may be misreported. People sometimes lie or exaggerate when they talk to the police.”).

         [97].     See F.R. Ev. 403.

         [98].     See Thomas, 21 I&N Dec. 20; Teixeira, 21 I&N Dec. at 321; Guerra, 23 I&N Dec. at 205.

         [99].     In current practice, as described above, adjudicators tend to consider any whiff of a possibility that the output is true to be sufficient to force the respondent to conclusively prove it is false.

       [100].     C.f. Thomas, 21 I&N Dec. at 24 (CLS output without a conviction was “some evidence” of criminality). An adjudicator might be tempted to treat the police officer’s determination that there was probable cause for the arrest as fully portable, but this would be error. A one-sided determination by an officer is not an adjudication by a neutral party after an adversarial process. As such, this determination could never be fully portable into another system, no matter the standard of proof required by that system. A probable cause determination made by a court after an adversarial hearing might be fully portable, but adjudicators should take care to determine what was actually litigated. C.f. Roberts, supra note 5, at 991 (“[I]n our legal system crimes generally require, in addition to particular acts (or omissions), other elements such as mental states, and also require the absence of successful defenses. An arrest may speak to law enforcement’s assertion vis-à-vis an alleged act (and allegations about alleged acts may suffice to establish probable cause), but that falls far short of a demonstration of factual guilt [to any standard.]”).

       [101].     8 CFR § 1003.1(d)(3)(ii) (“The Board may review questions of law, discretion, and judgment and all other issues in appeals from decisions of immigration judges de novo.”).

       [102].     8 CFR § 1003.1(d)(3)(iii).

       [103].     8 CFR § 1003.1(d)(3)(i).

       [104].     Compare Arreguin, 21 I&N Dec. at 39 (considering de novo the IJ’s weighing of factors), and Sotelo-Sotelo, 23 I&N Dec. at 205 (same), with Guerra, 24 I&N Dec. at 40 (nominally deferring to IJ discretion but also noting its approval of the IJ’s weighing of the relevant factors).

       [105].     This is not to say that an IJ or USCIS adjudicator must always credit an immigrant’s statement over an unadjudicated CLS output. An adjudicator might find an applicant’s sworn statement or live testimony to be not credible for any number of valid reasons. But assuming a CLS output to be infallible and, therefore, that the applicant must be lying if he contradicts it, is not a valid reason. What I take issue with here is not the ultimate result reached by the adjudicator, but the faulty methodology used to get there. See infra Part C(2)(c). In this case, the adjudicator failed to engage with the value of the evidence because it treated the CLS output as adjudicated.

       [106].     This is the operative question in discretionary bond determinations and is also a question that must be answered in U adjustment applications.

       [107].     Matter of Rico, 16 I&N Dec. 181, 185 (BIA 1977).

       [108].     Garces v. Attorney General, 611 F.3d 1337, 1349 (11th Cir. 2010) (overturning agency for relying on arrest report to establish “reason to believe”); Igwebuike v. Caterisano, 230 Fed. Appx. 278, 283–85 (4th Cir. 2007) (reversing a finding of reason to believe based solely on arrest for trafficking); Olivas-Motta, 746 F.3d at 919 (Kleinfeld, J., concurring) (holding that police reports are not “reasonable, substantial, and probative” evidence of guilt).

       [109].     See Rico, 16 I&N Dec. 181 (discussing testimony of Drug Enforcement Administration special agent, Border Patrol agent, and Customs Inspector); Lopez-Molina v. Ashcroft, 368 F.3d 1206, 1211 (9th Cir. 2004) (finding sufficient “reason to believe” the applicant had committed the acts underlying a previous drug trafficking arrest because the government submitted documents describing the police surveillance of him and the officers’ personal observations of his subsequent attempt to escape with 147 pounds of marijuana); Alarcon-Serrano v. INS, 220 F.3d 1116, 11198 (9th Cir. 2000) (noting that respondent admitted in his testimony that he suspected the car he was driving was carrying drugs); Matter of Favela, 16 I&N Dec. 753, 754 (BIA 1979) (noting that respondent admitted—presumably during his plea colloquy to a more generalized smuggling charge—that he participated in an attempt to smuggle a kilogram of marijuana into the United States); Castano v. INS, 956 F.2d 236, 238 (11th Cir. 1992) (holding that a guilty plea admission to conspiracy to distribute a controlled substance was sufficient where the conviction was later expunged because of the respondent’s youth pursuant to a rehabilitative statute); see also 9 Foreign Affairs Manual 40.23, at note 2(b) (“‘Reason to believe’ might be established by a conviction, an admission, a long record of arrests with an unexplained failure to prosecute by the local government, or several reliable and corroborative reports. The essence of the standard is that the consular officer must have more than a mere suspicion—there must exist a probability, supported by evidence, that the alien is or has been engaged in trafficking.”).

       [110].     This is true even if the agency chooses to establish a different standard of proof than “reason to believe.” While “reason to believe” might be enough to justify a denial of admission, it is almost certainly insufficient to justify a deprivation of liberty. See infra Part (ii)(d).

       [111].     See F. R. Ev. 403 (“The court may exclude relevant evidence if its probative value is substantially outweighed by a danger of one or more of the following: unfair prejudice, confusing the issues, misleading the jury, undue delay, wasting time, or needlessly presenting cumulative evidence.”).

       [112].     See, e.g., Ezeagwuna v. Ashcroft, 325 F.3d 396, 406 (3d Cir. 2003) (drawing on the principles behind the Federal Rules of Evidence to determine the probative value of hearsay evidence).

       [113].     5 U.S.C. § 706(2)(A), (E); see also Motor Vehicle Mfrs. Ass’n v. State Farm Auto Mut. Ins. Co., 463 U.S. 29, 43 (1983) (holding that courts should overturn agency decision that fails to “examine the relevant data and articulate a satisfactory explanation for [the agency’s] action including a ‘rational connection between the facts found and the choice made’”).

       [114].     See Wong Yang Sung v. McGrath, 339 U.S. 33, 53 (1950) (holding that the APA applies to immigration decisions); Supplemental Appropriation Act of 1951, ch. 1052, 64 Stat. 1044 (exempting [removal] proceedings from Sections 5, 7, and 8 of the APA); Shaughnessy v. Pedreiro, 349 U.S. 48, 50–51 (1955) (affirming that APA review applies to immigration decisions); Marcello v. Bonds, 349 U.S. 302, 306 (1955) (noting that the 1952 Immigration Act’s alternative procedures closely tracking provisions of the APA indicates that Congress intended the INA to supplant the APA in these areas).

       [115].     8 USC § 1252(a)(2)(B) (“[N]o court shall have jurisdiction to review any judgment regarding the granting of relief under section 1182(h) [waiver of criminal inadmissibility], 1182(i) [fraud waiver], 1229b [cancellation of removal], 1229c [voluntary departure], or 1255 [adjustment of status], or any other decision or action of the Attorney General or the Secretary of Homeland Security the authority for which is specified under this subchapter to be in the discretion of the Attorney General or the Secretary of Homeland Security, other than the granting of relief under section 1158(a) [asylum] of this title.”); 8 USC § 1226(e) (declaring no jurisdiction to review “the Attorney General’s discretionary judgment” regarding bond determinations).

       [116].     Motions to reopen, not addressed in detail in this piece, are classified as “discretionary” under agency regulations but are not committed to agency discretion by the INA. Accordingly, courts do review them. Kucana v. Holder, 558 U.S. 233, 237 (2010) (holding that the proscription of judicial review applies only to Attorney General determinations made discretionary by statute, not to determinations declared discretionary by the Attorney General himself through regulation). In conducting review, courts apply an “abuse-of-discretion” standard. See Gonzalez-Cantu v. Sessions, 866 F.3d 302, 304 (5th Cir. 2017).

       [117].     8 USC § 1252(a)(2)(D) (“Nothing in subparagraph (B) or (C), or in any other provision of this chapter (other than this section) which limits or eliminates judicial review, shall be construed as precluding review of constitutional claims or questions of law raised upon a petition for review filed with an appropriate court of appeals in accordance with this section.”). This provision was not part of the statute in 1996. It was added in 2005 in response to years of chaos cumulating in the Supreme Court ruling that review must remain for the questions that would normally be available on habeas review. See generally INS v. St. Cyr, 533 US 289 (2001); Nancy Morawetz, Back to Back to the Future?: Lessons Learned from Litigation over the 1996 Restrictions on Judicial Review, 51 N.Y.L. Sch. L. Rev. 113 (2006).

       [118].     See Heyman, supra note 6, explaining the origins of this “positivist” view and its application to judicial review of discretionary decisions.

       [119].     Achacoso-Sanchez v. INS, 779 F.2d 1260, 1264–65 (7th Cir. 1985) (emphasis added); see also Heckler v. Chaney, 470 U.S. 821 (1985) (holding that when a court has “no meaningful standard against which to judge the agency’s exercise of discretion . . . the statute [] can be taken to have ‘committed’ the decisionmaking to the agency’s judgment absolutely”); Heyman, supra note 6, at 864 (“Many . . . immigration statutes[] simply provide no manageable standards for the exercise of discretion. Their interpretation would seem to be confined to the agencies in the first instance, and even then, in the absence of internal standards for interpretation, it is hard to see how they can avoid a fateful brush with Heckler. Untethered to any controlling standards, the notion of discretion reposes virtually uncheckable power in decisionmakers. Judicial review would seem futile, if not impossible.”).

       [120].     See, e.g., Ojo v. Garland, 25 F.4th 152, 157 (2d Cir. 2022) (“[B]road discretion, however, does not allow an agency to apply an improper legal standard, ignore its own precedent, and fail to assess material expert evidence in support of one of the claims”); Zuh v. Mukasey, 547 F. 3d 504, 511 (4th Cir. 2008) (listing factors—drawn from agency regulations, BIA decisions, and court opinions—that an IJ should consider in a discretionary asylum decision, and requiring IJs to demonstrate that they have reviewed the record and balanced the relevant factors); Kalubi v. Ashcroft, 364 F.3d 1134, 1138–39 (9th Cir. 2004) (finding agency erred by crediting respondent’s testimony for purpose of nondiscretionary determination but then turning around and finding the same testimony incredible for the purpose of the discretionary one).

       [121].     Certain CAT claims are nondiscretionary. Courts were stripped of jurisdiction to review them under a separate part of the jurisdiction-stripping statute. 8 U.S.C. § 1252(a)(2)(C). But, just as with discretionary determinations, courts retain jurisdiction to review constitutional claims and questions of law. 8 U.S.C. § 1252(a)(2)(C).

       [122].     See Doe v. Sessions, 886 F.3d 203, 211 (2d Cir. 2018) (“[T]he agency may commit an error of law if important facts have been totally overlooked and others have been seriously mischaracterized.”) (internal citation omitted).

       [123].     Compare Wilkinson v. Garland, 601 U.S. 209, 217 (2024) (IJ’s determination that immigrant was statutorily ineligible for cancellation of removal because he had not established exceptional and extremely unusual hardship as required by the INA was reviewable as a “question of law”) with Patel v. Garland, 596 U.S. 328, 347 (2022) (there is no “question of law” in how the agency conducts fact finding in support of exercises of discretion); see Guyadin v. Gonzales, 449 F.3d 465, 467, 469 (2d Cir. 2006) (IJ’s determination that “a favorable exercise of discretion is not warranted due to the totality of the circumstances” was not reviewable because the court “lack[s] jurisdiction to review any claim that an IJ or the BIA erred in weighing the factors relevant to the grant or denial of [relief]”); but see Avila-Ramirez v. Holder, 764 F.3d 717, 722–23 (7th Cir. 2014) (BIA’s failure to follow its own precedent was “legal error”).

       [124].     The Supreme Court has rejected the idea that the correctness of the agency’s factfinding in support of exercises of discretion could be reviewable as a question of law, Patel, 596 U.S. at 339–40, and has stated in dicta that a pure exercise of factfinding in support of first-step eligibility for relief would not be reviewable either. Wilkinson, 601 U.S. at 225; but see id. at 221–22 (a question of law may require a court to “marshal and weigh evidence”). Where there is law to apply, however, as when facts are applied to a legal standard, the Court has been willing to find the agency’s determinations are subject to review. Id. at 221–23, 225 (there is a question of law to review in the agency’s determination of first-step eligibility for discretionary relief because a “mixed question of law and fact” is a “question of law”). Interestingly, the Court did not distinguish between, on the one hand, blackletter or judicially-created law and, on the other hand, agency-created guidance. Both were law. Id. at 222–24. Whether an agency could violate some other source of law, including its own internal guidance, in how it conducts its factfinding was not a question before the Court. Some lower courts, however, have identified a legal question in how the agency conducted its factfinding. See Avila-Ramirez, supra note 123; see also Hernandez v. USCIS, 643 F. Supp. 3d 1193, 1203 (W.D. Wash. 2022), reconsideration denied, 2023 WL 184220 (Jan. 13, 2023) (finding “question of law” jurisdiction to review USCIS’s denial of U adjustment based on unadjudicated CLS outputs because respondent did not challenge how the agency exercised its discretion but “merely desire[d] the agency to perform its discretionary analysis without considering improper evidence”); Hernandez v. USCIS, No. C22-904, 2023 WL 7386573, at *7–9 (W.D. Wash. Nov. 7, 2023) (treating “the mere fact of arrest” as “probative” of the “underlying conduct” and by placing “substantial weight” on an arrest “without considering the [ultimate] disposition” of the criminal case violated APA’s prohibition on “arbitrary and capricious” decision-making).

       [125].     See Neuman, supra note 23, at 635–36 (discussing an alternative view of deportation as deprivation of liberty).

       [126].     It is possible for Congress or the executive to create a liberty or property interest in a particular benefit by directing the executive in how that benefit should be adjudicated. In doing so, the rule must “substantively limit the exercise of official discretion through specifically defined criteria that guide official decision making.” Aguilera v. Kirkpatrick, 241 F.3d 1286, 1293 (10th Cir. 2001). This would require that the guidance establish an “entitlement to relief turning on objective criteria.” Achacoso-Sanchez, 779 F.2d at 1264. Courts have been loath to find that such an “entitlement to relief” has been created by an immigration statute or regulation. See id.; see also Dave v. Ashcroft, 363 F.3d 649, 652–53 (7th Cir. 2004); c.f. Mamedov v. Garland, 20-cv-01063 (E.D.N.Y. March 27, 2023) (finding a protected property interest in an I-130 relative petition was created by statutory language stating that the agency “shall” approve the benefit if certain conditions are met, making the adjudication non-discretionary).

       [127].     Neuman, supra note 23, at 638–40 (“[T]he liberty of an alien to remain in the United States results from an initial grant of permission to enter, which the Court continues to describe and to treat as a privilege.”); see Cox & Rodríguez, supra note 20 (arguing that broad rules for deportability coupled with the provision for discretion mean that many types of immigration status are essentially probationary).

       [128].     Reno v. Flores, 507 U.S. 292, 302 (1993); Bowers v. Hardwick, 478 U.S. 186, 191 (1986); United States v. Salerno, 481 U.S. 739, 746 (1987).

       [129].     See generally Mathews v. Eldridge, 424 U.S. 319 (1976) (discussing procedural due process).

       [130].     See, e.g., Flores, 507 U.S. at 306 (“It is well established that the Fifth Amendment entitles aliens to due process of law in deportation proceedings.”); Demore v. Kim, 538 U.S. at 523 (same); Landon v. Plasencia, 459 U.S. 21, 41 (1982) (holding that a returning lawful permanent resident is entitled to procedural due process); Yamataya v. Fisher (The Japanese Immigrant Case), 189 U.S. 86, 101 (1903) (holding that deportation procedures must comport with procedural due process); see also INS v. Rios Pineda, 471 U.S. 444, 451–52 (1985) (reviewing whether BIA’s decision was an “unreasoned or arbitrary exercise of discretion” and finding in the negative because it was “grounded in legitimate concerns . . . [and] determined on the basis of the particular conduct of respondents”).

       [131].     Montanez-Gonzalez v. Holder, 780 F.3d 720, 723–24 (6th Cir. 2014); see also Zambrano-Reyes v. Holder, 725 F.3d 744, 750 (7th Cir. 2013) (finding that the procedural due process right to a fair hearing guards against ineffective counsel in immigration proceedings where the 6th Amendment does not apply); Calderon-Rosas v. A.G., 957 F.3d 378, 384 (3d Cir. 2020) (stating that “petitioners must receive a full and fair hearing”).

       [132].     See, e.g., Bustos-Torres v. INS, 898 F.2d 1053, 1055 (5th Cir. 1990); Marlowe v. U.S. Immigration and Naturalization Service, 457 F.2d 1314, 1315 (9th Cir. 1972).

       [133].     See, e.g., Torres-Aguilar v. I.N.S., 246 F.3d 1267, 1271 (9th Cir. 2001) (“[A]lthough we retain jurisdiction to review due process challenges, a petitioner may not create the jurisdiction that Congress chose to remove simply by cloaking an abuse of discretion argument in constitutional garb.”).

       [134].     See, e.g., Marquez v. Garland, 13 F.4th 108, 115 (2d Cir. 2021) (“[Petitioner] raises no constitutional claims or questions of law, so we dismiss his petition for want of jurisdiction insofar as it challenges the agency’s discretionary decision to deny [relief].”).

       [135].     See, e.g., Carcamo v. Department of Justice, 498 F.3d 94, 98 (2d Cir. 2007) (holding that petitioner raised no colorable constitutional claim because “due process does not require that the IJ credit [petitioner’s] testimony over . . . the criminal complaint”).

       [136].     Marquez, 13 F.4th at 114.

       [137].     See, e.g., Schroeck v. Gonzales, 429 F.3d 947, 951 (10th Cir. 2005) (holding that the question of whether the IJ violated due process by admitting evidence from a case where all charges were dismissed with prejudice was reviewable, but there was no due process violation); Esposito v. INS, 936 F.2d 911, 913 (7th Cir. 1991) (holding that it was “fundamentally fair” to consider in absentia convictions in denying discretionary relief even though agency precedent suggested that they were inherently unreliable); Henry v. INS, 74 F.3d 1, 6 (1st Cir. 1996).

       [138].     Padmore v. Holder, 609 F.3d 62, 69 (2d Cir. 2010) (“Padmore argues further that the BIA’s decision contravenes BIA precedent and thereby violated his constitutional right to due process. Though we need not reach this argument here, we note that the Court is concerned by the BIA’s apparent willingness to accept unproven and disputed allegations as true merely because they exist in the record.”); Sierra-Reyes v. INS, 585 F.2d 762, 764 (5th Cir. 1978) (dismissing on grounds unrelated to the equal protection claim advanced).

       [139].     Henry, 74 F.3d at 6; see also Marquez, supra note 134 at 115.

       [140].     Achacoso-Sanchez, supra note 119, at 1264­–65.

       [141].     Avila-Ramirez, 764 F.3d at 722–23.

       [142].     Chuil Chulin v. Zuchowski, No. 21-00016, 2021 WL 3847825, at *6 (N.D. Cal. Aug. 27, 2021).

       [143].     Id. at *9 (holding that the pleadings sufficiently made out a due process violation because, under the facts as alleged, Mr. Chuil Chulín was “prevented from presenting his case”); see Schroeck v. Gonzales, 429 F.3d 947, 952 (10th Cir. 2005).

       [144].     But see Gomez v. McHenry, 19-CV-7373 at *6–7 (S.D.N.Y. Oct. 30, 2020) (IJ’s “finding” that the respondent was not a danger to the community was not a factual finding, but was the “factual underpinning” of a “discretionary determination” that the respondent was not dangerous).

       [145].     Carcamo, supra note 135 at 98 (declining to review respondent’s claim that the IJ had erred in relying on information in an arrest report to reject his oral testimony as not credible).

       [146].     Marquez, supra note 134 at 115.

       [147].     Heyman, discussing discretion and normative and practical reasons for judicial deference to agency decisions, speculates that “[i]t may be that we should defer if and only if the decision is the product of a reasoned inquiry informed by cognizable standards. . . . [T]alk of [deference] only makes sense if the decisional process involves the use of some set of accepted norms.” Heyman, supra note 7, at 886.

       [148].     See, e.g., Greene v. McElroy, 360 U.S. 474, 496–497 (1959) (“Certain principles have remained relatively immutable in our jurisprudence. One of these is that, [where a government decision] depends on fact findings, [the affective individual must] ha[ve] an opportunity to show that [the government’s evidence] is untrue.”); see e.g., Goldberg v. Kelly, 397 U.S. 254, 269 (1970) (“[i]n almost every setting where important decisions turn on questions of fact, due process requires an opportunity to confront and cross-examine adverse witnesses.”); see also Yamataya, supra note 130 at 101 (The government may not “arbitrarily [] cause an alien . . . to be taken into custody and deported without giving him all opportunity to be heard upon the questions involving his right to be and remain in the United States.”); Colmenar v. I.N.S., 210 F.3d 967, 971 (9th Cir. 2000) (due process was violated where IJ indicated, at the start of the hearing, that he had already judged the claim).

       [149].     See, e.g., Shaughnessy v. Pedreiro, 349 U.S. 48, 56 (1955) (reaffirming that court undertaking due process review should determine whether an order of deportation was supported by “substantial evidence”; implying that factfinding must be conducted to a given standard and that facts must actually be found by the agency before an immigrant can be ordered removed).

       [150].     Igwebuike, supra note 108, at 283–85.

       [151].     Garces, supra note 108, at 1346.

       [152].     Rodriguez-Quiroz v. Lynch, 835 F.3d 809, 820 (8th Cir. 2016).

       [153].     Carcamo, supra note 135, at 98.

       [154].     Marquez, supra note 134, at 115.

       [155].     C.f. Marquez, supra note 134, at 115; Carcamo, supra note 135 at 98.

       [156].     Matter of J-V-, 2018 WL 6985737 (see part A, supra); but see Patel, 596 U.S. at 344–45 (questioning in dicta whether the provision of the jurisdiction-stripping statute that restores jurisdiction to review constitutional claims and questions of law would apply to USCIS decisions).

       [157].     Marquez, supra note 134 at 114.

       [158].     I refer here to bond under INA § 236(a), which covers the detention of immigrants during the pendency of their removal proceedings, but which excludes immigrants who have certain criminal convictions that subject them to “mandatory” detention under INA § 236(c).

       [159].     Zadvydas v. Davis, 533 U.S. 678, 691 (2001) (holding that due process requires that preventative detention, including of noncitizens, be based on a finding of flight risk or dangerousness); Demore v. Kim, 538 U.S. 510 (2003) (statute providing for mandatory detention during removal proceedings for immigrants with certain criminal convictions is constitutional because these convictions are an acceptable substitute for an individualized finding of dangerousness or flight risk); see also David Cole, In Aid of Removal: Due Process Limits on Immigration Detention, 51 Emory L.J. 1003, 1010 (2002) (discussing the holding in Zadvydas and the then-pending Demore). For a critique of the validity of the dangerousness rationale, see Francis M. Kreimer, Dangerousness on the Loose: Constitutional Limits to Immigration Detention as Domestic Crime Control, 87 N.Y.U. L. Rev. 1485 (2012) (arguing that, unlike the goal of effecting removal, the “dangerousness” justification for detention is only tenuously related to Congress’s plenary power over immigration; confining deportable noncitizens to “incapacitate” them as is done in civil confinement requires due process safeguards equal to those that would be required in such proceedings, including, at a minimum, that the government must bear the burden of proof).

       [160].     See INA § 236(a) (IJ may detain, or may release on bond or without bond); Matter of Adeniji, 22 I&N Dec. 1102, 1116 (BIA 1999).

       [161].     Adeniji, 22 I&N Dec. at 1116.

       [162].     See Hernandez-Lara v. Lyons, 10 F.3rd 19, 26–27 (1st Cir. 2021) (contrasting the historical “presumption in favor of liberty” with the current regime); Jennings v. Rodriguez, 583 U.S. 281, 312 (2018) (declining to decide whether the Due Process Clause of the Fifth Amendment entitles a noncitizen to a bond hearing at which the government bears the burden to prove flight risk or danger by clear and convincing evidence).

       [163].     424 U.S. 319 (1976).

       [164].     See Hernandez-Lara, 10 F.4th at 39; Velasco- Lopez, 978 F.3d at 854–55; but see Borbot v. Warden Hudson Cnty. Corr. Facility, 906 F.3d 274, 280 (3d Cir. 2018) (finding due process did not require a shift of the burden of proof to the government at any point, but leaving open the possibility that due process might, in some circumstances, require more process than currently provided under the statute).

       [165].     Jennings, 583 U.S. at 311–12 (throwing out a Ninth Circuit rule that, in an effort to avoid the constitutional concerns created by immigration detention, required shifting the burden of proof to the government after six months of detention); c.f. Johnson v. Arteaga-Martinez, 596 U.S. 573 (2022).

       [166].     While the bond statute specifies that the agency’s discretionary determinations with respect to bond are not reviewable, INA § 236(e), habeas corpus remains available to challenge individual deprivations of liberty. See generally Jennings, 583 U.S. 281; see also Singh v. Holder, 638 F.3d 1196, 1202 (9th Cir. 2011) (“claims that the discretionary process itself was constitutionally flawed are cognizable in federal court on habeas”) (internal citation omitted).

       [167].     C.f. Zadvydas, 533 U.S. at 691–92 (stating that preventative civil detention, as in the detention of sexually violent offenders, is only justified based on dangerous past behavior coupled with “special circumstance[s]” indicating continued dangerousness).

       [168].     For example, in Matter of Urena, 25 I&N Dec. 140 (BIA 2009), the IJ below had found that the respondent posed a “potential” danger to the community before setting bond. The BIA did not engage in any discussion of how much danger a “potential” might represent and instead looked at the issue as binary (danger versus no danger). It remanded for the IJ to clarify whether he found that the respondent posed a danger or not. If the respondent did pose (any level of) danger, the IJ was directed to remand him without bond.

       [169].     Under my proposal, an IJ would not necessarily have to find beyond a reasonable doubt that the respondent committed the crime. But they should have to find it to some standard. C.f. Hernandez-Lara, 10 F.4th at 39–41 (finding that the burden should be clear and convincing evidence for dangerousness but only a preponderance of the evidence for flight risk).

       [170].     If this is true, then IJs have less discretion to detain but greater discretion to release than previously thought: if the respondent is not a danger or flight risk, there is no discretion to detain, and a bond must be set or the respondent must be released on his own recognizance. If a respondent does pose a danger or flight risk, the IJ may choose to detain without bond, but may also choose, despite the risk, to set a bond or even release the respondent on his own recognizance. There are many reasons why an individual who poses a risk due to past behavior might nevertheless be released in discretion. These include health concerns, the particular vulnerability of the individual to harm in detention (e.g., transgender immigrants), or family or community need for the individual’s continued presence. Equally validly, the IJ could find that detention for months or years would be an unjust consequence for the crime. The IJ might also, in an exercise in humility, take into account the level of uncertainty that exists in their factual determination that the past behavior occurred and decide that the risk of error is too great to justify depriving a person of their liberty.

       [171].     See Nelson v. Colorado, 581 U.S. 128, 128 (2017) (state law requiring exonerated defendants to affirmatively prove their innocence by clear and convincing evidence before they could recover funds was insufficiently protective of the exonerees’ interest in their money under the Mathews test; the burden of proving innocence by clear and convincing evidence unacceptably risked erroneous deprivation of the funds); see also Colin Brady, Note, Freedom in the Balance: Procedural Due Process Rights and the Burden of Proof in Detention Hearings in Immigration Removal Proceedings, 31 Wm. & Mary Bill Rts. J. 1241, 1271 (2023) (in every other civil detention regime, save the detention of “enemy” aliens, the Supreme Court has required the burden to be borne by the government).

       [172].     The presumption of innocence suggested by Marouf, supra note 31, is also suggested, and is clearly called for in bond determinations.

       [173].     The jailer in this case is the immigration agency. Technically the jailer is DHS while bond hearings are conducted by EOIR in the DOJ, but both agencies are part of the executive branch. This is not the same separation that would exist between judge and jailer were the judge a part of the judicial branch. See generally Anthony R. Enriquez, Structural Due Process in Immigration Detention, 21 CUNY L. Rev. 35 (2018) (arguing that a due process problem is posed by the lack of separation between jailer and judge in immigration detention decisions).

       [174].     INA § 236(e).

       [175].     Patel, 596 U.S. at 341–42 (“Had Congress intended instead to limit the jurisdictional bar to ‘discretionary judgments,’ it could easily have used that language—as it did elsewhere in the immigration code. [For example, INA § 236(e):] ‘The Attorney General’s discretionary judgment regarding the application of this section shall not be subject to review.’”) (emphasis in original); see Wilkinson, 601 U.S. at 217–18, 225 n.4 (drawing a distinction between review of threshold eligibility for discretionary relief and review of decision-making at the second discretionary step).

       [176].     Demore, 538 U.S. at 510.

       [177].     See Igwebuike, 230 Fed. Appx. at 283–85; Garces, 611 F.3d at 1346; Rodriguez-Quiroz, 835 F.3d at 813–14.

       [178].     These are all nondiscretionary determinations.

       [179].     Moncrieffe, 569 U.S. at 202; see also Das, supra note 21.

       [180].     Shepard v. United States, 544 U.S. 13, 16, 20 (2005).

       [181].     Descamps, 570 U.S. 254 (2013); Mathis, 579 U.S. at 517–19.

       [182].     See Nijhawan v. Holder, 557 U.S. 29 (2009) (endorsing abandoning the categorical approach for the fraud aggravated felony ground).

       [183].     Or inadmissibility, or ineligibility for relief.

       [184].     Compare Matter of Silva-Trevino, 24 I&N Dec. 687 (A.G. 2008) with Silva-Trevino v. Holder, 742 F.3d 197, 200 (5th Cir. 2014) (rejecting the approach that requires immigration judges to apply the circumstance-specific approach in instances where the record of conviction is inconclusive) and Matter of Silva-Trevino, 26 I&N Dec. 826 (BIA 2016).

       [185].     28 I&N Dec. 437 (BIA 2022). This case relies on dicta from Mathis suggesting that, when state law is unclear as to whether a statute is divisible into discrete crimes, adjudicators may “peek” at the record of conviction to help determine whether the modified categorial approach should apply. In Laguerre, however, state law was clear: the state’s highest court had held that the identity of the controlled substance was not an element of the crime (and had not been adjudicated). Consequently, the statue could not be considered divisible into discrete crimes, one for each possible substance the respondent might have possessed. It defined, instead, a single crime that could be committed in various ways. There was no call for a modified categorial approach. Stretching the Mathis dicta well beyond its meaning, the BIA allowed the IJ to look at the original charging document and use the fact that it specified cocaine to find that the statute was divisible after all. Id. at 447 (“The indictment in the respondent’s case reflects that he was charged with possessing the controlled dangerous substance of cocaine. Because this charging document ‘referenc[es] one alternative [controlled dangerous substance] to the exclusion of all others,’ the Mathis ‘peek’ supports our view that the identity of the controlled dangerous substance possessed is an “element” of [the statute] as opposed to a ‘means’ of violating the statute.”).

       [186].     Amicus Invitation No. 21-30-09, https://www.justice.gov/eoir/page/file/1445471/download [https://perma.cc/RH6S-3A3C] (inviting amici to brief, inter alia, whether the record of conviction can be expanded to encompass CLS outputs including sentencing transcripts, docket entries, abstracts, hearing transcripts, “any document or record prepared by, or under the direction of, the court”, or “[a]ny document or record . . . maintained by . . . a State or Federal penal institution, which is the basis for that institution’s authority to assume custody of the individual named in the record” and whether information from these outputs can be considered under a modified categorical analysis) (referencing the list of documents considered acceptable evidence of the existence of a criminal conviction at INA § 240(c)(3)(B)). Ultimately, the agency decided it does not currently have the authority to expand the record of conviction due to the existence of controlling Supreme Court precedent. See Matter of K-M- (BIA May 26, 2023) (unpublished) (on file with author).

       [187].     See Mathis, 579 U.S. at 2248 (conviction limited to “what the jury must find beyond a reasonable doubt” or “what the defendant necessarily admits when he pleads guilty”).

       [188].     See id., 579 U.S. at 515–16 (referring to extraneous information contained within record of conviction documents as “statements of ‘non-elemental facts’” which are not necessary to support the conviction, and emphasizing that this extraneous information cannot properly be used to determine the crime of conviction); see also Evanson v. Att’y Gen., 550 F.3d 284, 293 (3d. Cir. 2008) (adjudicator “may not look to factual assertions in the judgment of sentence” because they are not part of the elements of the crime).

       [189].     See Shepard v. United States, 544 U.S. 13, 26 (2005) (holding that reference to the record of conviction is limited to determining what defendant “necessarily admitted” in order to plead guilty under the statute). For this reason, it makes no sense to look to the supposed “record of conviction” to determine whether a statute is divisible because determining whether the statute is divisible is necessary to determine what makes up the record of conviction in the first place. Contra, Laguerre, 28 I&N Dec. 437.

       [190].     Everything else in the Supreme Court’s opinion in Mathis is consistent with this approach, apart from the dicta relied on by the BIA in Laguerre. This dicta suggesting that adjudicators “peek” at the “record of conviction” to help determine divisibility makes no sense in the larger context of the opinion, as the output portability framework helps clarify. If (as the output portability framework suggests) the “record of conviction” reflects the facts that were necessarily adjudicated by the criminal court, then looking to the “record of conviction” to determine whether a statute is divisible would be putting the cart before the horse. The “record of conviction” can, of necessity, only be consulted after determining whether the statute is divisible because the identity of those facts that make up the record will change depending on the answer. In slightly more concrete terms, if the statute is divisible and therefore the identity of the substance (e.g., cocaine) is an element of the crime that must be adjudicated, then the identity of the substance is part of the “record of conviction.” If the statute is not divisible, then the identity of the substance is not an element of the crime and is not part of the “record of conviction,” regardless of what is specified in the charging document or jury instructions. The majority acknowledges as much in Mathis. 579 U.S. at 515 n.6 (rejecting the idea that a jury could be inferred to have found a non-elemental fact). The Mathis “peek” dicta should not be allowed to undermine the entire categorial framework established by the Court.

       [191].     Amicus Invitation, supra note 186.

       [192].     Shepard, 544 U.S. at 26; Descamps, 570 U.S. at 254; Mathis, 579 U.S. at 500.

       [193].     INA § 101(a)(43)(M).

       [194].     This is sometimes referred to by advocates as immigration’s rule of “anti-lenity.” See, e.g., Matter of Soram, 25 I&N Dec. 378 (BIA 2010) (interpreting the crime of child abuse, abandonment, or neglect ground of deportability to sweep broadly because Congress must have intended for the ground to “protect” as many children as possible).

       [195].     Nijhawan, 557 U.S. at 29.

       [196].     See, e.g., Doe v. Sessions, Nos. 15-2026, 16–58 (2d Cir. Sept. 27, 2017) (requiring an immigrant to come forward with “conflicting evidence” that undermines the agency’s reliance on the restitution order; finding lack of conflicting evidence sufficient to convert a preponderance to clear and convincing).

       [197].     Cortney E. Lollar, What Is Criminal Restitution?, 100 Iowa L. Rev. 93 (2014).

       [198].     See, e.g., Chiao Fang Ku v. Att’y Gen. United States of Am., 912 F.3d 133, 136 (3d Cir. 2019) (because of this possibility, restitution order’s loss amount must be “sufficiently tethered” to the crime of conviction); Rampersaud v. Barr, 972 F.3d 55, 60–61 (2d Cir. 2020) (“The BIA and the IJ seem to have assumed that the . . . restitution amount listed on the sentencing form for his two counts of conviction was a reliable indicator of the loss caused by Rampersaud’s insurance fraud. The restitution order, however, was not on its own sufficient evidence that a loss exceeding $10,000 was attributable specifically to the insurance fraud count.”).

       [199].     See Nijhawan, 557 U.S. at 42–43 (“[T]he statute foresees the use of fundamentally fair procedures” in the determination of the loss amount.).

       [200].     INA § 237(a)(2)(B)(i).

       [201].     Mellouli v. Lynch, 575 U.S. 798 (2015).

       [202].     Matter of Dominguez-Rodriguez, 26 I&N Dec. 408, 413 (BIA 2014); see also Bogle v. Garland, 21 F.4th 637, 645 (9th Cir. 2021) (deferring to BIA on the issue).

       [203].     PSIs are reports created by probation after conviction and before sentencing. They contain information taken directly from the charging document and may also contain information from an interview with the defendant, conversations with the victim, and other parties. PSIs do not reflect the results of judicial factfinding. They are an amalgamation of various sources of information presented as an aid to the sentencing judge, who can further explore the facts therein at a hearing if necessary (though, if the sentence has been negotiated in advance, factfinding will likely not be necessary). There is, therefore, a clear mismatch between the function and purpose of PSIs and what they are later called upon to prove in immigration adjudications.

       [204].     In re Jonet Dominguez-Rodriguez, 2019 WL 7859288, *3 (BIA December 16, 2019) (unpublished) (holding, on appeal after remand, that the IJ did not err in using PSI to find that the respondent had more than 30 grams of marijuana in “circumstances . . . consistent with . . . street- level sale”).

       [205].     See Bogle, 21 F.4th at 661 (Pearson, J., dissenting) (“[I]mmigration proceedings . . . [should not] present an opportunity for an IJ[] to adjudicate free of the Constitutional protections, like the right to cross examine and confront adverse evidence or statutory protections like the rules of evidence that traditionally ensure fair resolutions for criminal defendants.”).

       [206].     See Jain, supra note 18, at 810–11; see also Pinard & Thompson, supra note 19; Chin, supra note 19 (discussing benefits, public housing, employment, child welfare, licensing, etc.).

       [207].     See Jennings, 583 U.S. at 311–12; c.f. Johnson, 596 U.S. 573 (2022).

       [208].     See Victor J. Gold, Federal Rule of Evidence 403: Observations on the Nature of Unfairly Prejudicial Evidence, 58 Wash. L. Rev. 497, 503 (1983) (evidence should be considered unfairly prejudicial when it has a tendency to cause the factfinder to commit an inferential error).

       [209].     See United States ex rel. Mylius v. Uhl, 203 F. 152, 153 (S.D.N.Y. 1913) aff’d, 210 F. 860 (2d Cir. 1914) (“[T]he immigration authorities act in an administrative and not in a judicial capacity . . . . Their function is not, as it seems to me, to go behind judgments of conviction and determine with respect to the acts disclosed by the testimony the questions of purpose, motive and knowledge . . . . Besides, the testimony is seldom available and to consider it in one case and not in another is to depart from uniformity of treatment. I do not think the immigration law intends that where two aliens are shown to have been convicted of the same kind of crime, the authorities should inquire into the evidence upon which they were convicted and admit the one and exclude the other. [A prohibition on relitigating the facts of the case is] “necessary for the efficient administration of the immigration laws.”); see also Matter of S---, 2 I&N Dec. 559, at 4 (BIA 1946) (citing Mylius).

       [210].     See Frost, supra note 1.

       [211].     Stipulation, N.Y. Legal Assistance Grp. v. Bd. of Immigr. Appeals, 18-cv-09495 (SDNY, Feb. 9, 2022); see also N.Y. Legal Assistance Grp. v. Bd. of Immigr. Appeals, 987 F.3d 207 (2d Cir. 2021) (vacating district court’s grant of motion to dismiss); see also https://foia.eoir.justice.gov/app/ReadingRoom.aspx [https://perma.cc/PW5X-8QG6].

       [212].     See Jain, supra note 18, at 841 n.169 discussing a lawsuit against the U.S. census bureau for a policy regarding arrests that had the effect of eliminating ninety-three percent of applicants from eligibility and had a significantly adverse impact on Latinos and African Americans.

       [213].     See Jain, supra note 18, at 815, 841 (arguing that “[r]eliance on arrests alone magnifies the significance of a police officer’s decision to arrest” and showing how reliance on arrests sanitizes racial discrimination: “If an employer relies on drug screening . . . it cannot disproportionately target minorities. But when employers . . . rely on arrests, they leave front-end decisions about whom to screen to the police, without similar regard for racially disparate impact”); see also Jain, supra note 22, at 1703 (arguing that focusing immigration enforcement priorities on arrestees “imposes an enormous cost on racial minorities disproportionately subject to low-level arrest” which is exacerbated by the lack of “procedural protections that ought to accompany the criminal process”); c.f. Lee, supra note 21, at 573 (“conviction-based removals [] represent a transfer of power to . . . local prosecutors [who] can control [these] removals through their discretion over whom and what to charge”).

       [214].     Including Onyx Starrett, Michelle Assad, Marie Mark, Talia Peleg, and Glykeria Teji, with guidance from Carmen Maria Rey.