Overprotecting: A Comment on Professor Krishnan’s LODE Model for Immigration Hearings

PDF link Download PDF

Tung Yin1

At its heart, the American legal system is an adversarial one in which the lawyers for each side advocate zealously on behalf of their clients. The trial judge, meanwhile, resolves the legal disputes between the lawyers but plays a minimal role in developing the testimony and evidence.

Immigration judges, however, do not fit this paradigm neatly. Scholars have noted that relevant federal statutes, including a key 1996 amendment to the Immigration and Nationality Act, impose upon immigration judges a duty to “engage in what might be thought of as ‘active judging’”—that is, “to fully develop the record in the cases that come before them.”2 This obligation makes sense in a system where the typical noncitizen does not have legal representation, because the assumption of a full and effective adversarial presentation may be unwarranted in such circumstances.

Of course, some noncitizens do have legal representation. Should this active duty still exist? The lawyer-centric position would argue that the duty should not exist, because the noncitizen’s lawyer should be able to control the presentation of the noncitizen’s case, and the lawyer’s presence would obviate the need for the immigration judge to look out for the noncitizen. The judge-centric position would respond that the 1996 amendment to the Immigration and Nationality Act is silent about the presence or absence of counsel, and that the judge’s obligation remains intact. The federal appellate courts have not reached definitive agreement about whether to endorse one or the other position.3

In Overstepping: U.S. Immigration Judges and the Power to Develop the Record, Professor Jayanth Krishnan proposes an innovative approach for harmonizing the statutory duty of the immigration judge with the traditional role of the lawyer as the undisputed advocate for the client. He rejects the judge-centric position in part because of the status of immigration judges as Article II adjudicators who are appointed and supervised by the Attorney General, and therefore may exhibit bias against the noncitizen and in favor of the government.4 Accordingly, the noncitizen’s lawyer should presumptively be able to refuse any record-developing interference by the immigration judge. At the same time, however, Professor Krishnan does not fully embrace the lawyer-centric position, because there may be incompetent lawyers who pressure their clients to opt-out of the potential safeguard of well-intentioned immigration judges.5 To thread the narrow path between this Scylla and Charybdis of potentially overbearing judges and inept lawyers, he argues for what he calls the “Lawyer Opt-Out with Discretionary Evaluation,” or LODE model.6

Under this model, the noncitizen’s lawyer can choose to opt-out of any record development by the immigration judge; if, however, during the course of the hearing, the immigration judge comes to believe that the lawyer is “underperforming,” the immigration judge can seek permission to intervene from the appropriate Court of Appeals.7 This “discretionary evaluation” part of the model is based on discretionary interlocutory appeals under 28 U.S.C. § 1292(b), in which a litigation party can avoid the general restriction against piecemeal appeals and seek appellate review even before the district court has issued a final order.8 As Professor Krishnan argues, “A requirement that an independent Article III court review and approve this matter would give ‘cover’ and legitimacy to the immigration judge if the request were granted. Additionally, having this safeguard in place would provide a check on a hostile judge who wishes to exert unwarranted authority.”9 To be sure, Professor Krishnan acknowledges the limitations of the analogy, but argues that the analogy still applies because “there is a request for the Article III appellate court to help resolve an issue at a lower court level so that that lower court proceeding can move forward in as expeditious and fair a manner as possible.”10

In this short comment, I raise two concerns regarding the implementation of Professor Krishnan’s model. The first is that the LODE model’s analogy to the discretionary interlocutory appeal is imperfect, and the imperfection is not merely technical, but rather involves the very heart of the reason for such interlocutory appeals: it is difficult to see how the LODE model will actually produce appellate rulings that could potentially shorten or terminate the underlying hearing. Instead, the LODE model seems likely only to create delay while the appellate court decides, based on an incomplete record, whether the rest of the immigration hearing will involve record development by the immigration judge.

The second concern is that the LODE model puts the appellate court into the position of having to decide (again, on an incomplete record) whether the noncitizen’s lawyer is inept or incompetent. Such a ruling could have far-reaching implications on any potential ineffective assistance of counsel claim brought by a noncitizen who loses at the hearing.

Mid-hearing appeal versus interlocutory appeals

It is easy to see the appeal of Professor Krishnan’s analogizing of the LODE model’s operational mechanics to interlocutory appeals under 28 U.S.C. § 1292. Both involve judicial process for appellate review of an important ruling in a lower tribunal with the goal of avoiding needless or protracted litigation that would result from requiring an appeal of a final order.11 Section 1292(b) states in relevant part:

When a district judge, in making in a civil action an order not otherwise appealable under this section, shall be of the opinion that such order involves a controlling question of law as to which there is substantial ground for difference of opinion and that an immediate appeal from the order may materially advance the ultimate termination of the litigation, he shall so state in writing in such order. The Court of Appeals which would have jurisdiction of an appeal of such action may thereupon, in its discretion, permit an appeal to be taken from such order. . .12

Implicit in the LODE model is the value judgment that potential inept lawyering on the part of the noncitizen’s attorney should be addressed as soon as possible, rather than waiting for the appeals process to remedy any ineffective assistance of counsel. If this value judgment were not present, then there would be no need to interrupt the immigration hearing to seek appellate permission for the immigration judge to intervene; instead, the hearing would proceed to its conclusion despite the immigration judge’s doubts about the lawyer’s competence, and any ruling against the noncitizen would result in an appeal in which the effectiveness of counsel could be challenged. By allowing a mid-hearing interruption, the LODE model presumably seeks to spare the noncitizen of the burden and stress of protracted litigation and appeal.

This, however, is exactly not the purpose of an interlocutory appeal. One circuit court has noted that it has “consistently declined to allow the prospect of possibly substantial litigation burdens … or even of relitigation itself … to circumvent the policy against piecemeal review.”13 Rather, discretionary interlocutory appeals under section 1292(b) are limited in nature. The Eleventh Circuit, for example, has laid out five criteria that “generally must be met” before it will consider an interlocutory appeal:

(1) the issue is a pure question of law, (2) the issue is “controlling of at least a substantial part of the case,” (3) the issue was specified by the district court in its order, (4) “there are substantial grounds for difference of opinion” on the issue, and (5) “resolution may well substantially reduce the amount of litigation necessary on remand.14

The LODE model diverges significantly from section 1292(b) in at least two ways. First, under section 1292(b), a pure question of law is one that does not require “delv[ing] beyond the surface of the record in order to determine the facts.”15 In contrast, the LODE model requires an appellate court to engage heavily with the factual record to that point in order to determine whether the noncitizen’s lawyer has been ineffective or incompetent. Ineffective assistance of counsel is a mixed question of law and fact,16 and for an appellate court to override a lower court’s determination, it typically does so with a factual record free of conflicts.17 That will not be the case with the LODE model, as the very fact of the litigation in the circuit court means that the immigration judge and the noncitizen’s lawyer are going to be arguing different facts, or at least interpretations of the facts.

The Eleventh Circuit’s opinion in Drummond Co., Inc. v. Conrad & Scherer LLP provides a good illustration of the incongruity between the LODE model and 28 U.S.C. § 1292(b).18 In this case, the law firm C&S represented Colombian plaintiffs against Drummond for allegedly supporting murderous paramilitary groups. After prevailing in the Colombian plaintiff lawsuits, Drummond sued C&S for defamation. During that protracted litigation, Drummond came to believe that C&S—and in particular, its partner Collingsworth—had made numerous false statements to the court. As a result, the district court ruled that Drummond could rely on the crime-fraud exception to pierce C&S’s and Collingsworth’s claims of attorney-client privilege.19 Collingsworth and C&S both sought interlocutory appeals under section 1292(b). Separate motions panels of the Eleventh Circuit denied Collingsworth’s petition but granted C&S’s petition as to whether agency principles could “be used to impute the application of the crime-fraud exception to an agent’s principal where the principal has separately-held privileges as a co-defendant in the suit and there is no finding that the exception applies directly to the principal…” and also “where the agent is operating as an attorney and there is no finding that the client’s behavior triggered the crime-fraud exception or that the exception applies directly to the principal…”20

On full review, a merits panel of the Eleventh Circuit concluded that the motions panel had erred in accepting C&S’s petition.21 Although the court conceded that “whether, in applying the crime-fraud exception, a court may impute a partner’s knowledge and intent to a partnership appears to raise a purely legal question,” C&S did not dispute that such imputation could be appropriate in some circumstances.22 Instead, C&S was arguing that the facts of this case did not warrant imputation of Collingsworth’s intent and actions to the law firm. The Eleventh Circuit held that C&S’s argument would lead the appellate court to engage in a “fact-specific inquiry” that would call for “the court to apply law to the particular facts of the case and thus to take a deep dive into this case’s voluminous record.”23

In the same case, the Eleventh Circuit did resolve another question on interlocutory appeal: “Can the crime-fraud exception be applied to overcome attorney work product protection when the attorney or law firm was engaged in the crime or fraud but the client was not?”24 This was a pure question of law because there was no dispute over whether the clients (the Colombian plaintiffs) were involved in Collingsworth’s fraudulent actions.

The LODE model resembles the first interlocutory appeal question, the one that the Eleventh Circuit declined to reach. It is not a pure legal question, such as whether the immigration should have the power in the abstract to interfere in the lawyer’s presentation of the case, but a factual question of whether in this case, the immigration judge should be permitted to interfere. If anything, the LODE model will most likely require the appellate court to do more than engage with the factual record; the appellate court will almost certainly have to develop the record beyond that amassed in the immigration hearing. This is true because a successful claim of ineffective assistance of counsel requires demonstration of deficient performance on the part of the lawyer as well as prejudice resulting from that deficient performance.25 In determining whether counsel’s performance was deficient, the Court has emphasized that judicial review should be “highly deferential” and be undertaken with a presumption that the lawyer acted reasonably.26 Where that presumption is challenged, though, the noncitizen’s lawyer may feel compelled to articulate their strategic reasoning. In a typical ineffective assistance of counsel case involving a criminal defendant or habeas petitioner, the lawyer could do so through an affidavit or live testimony at an evidentiary hearing—all of which would take place at the trial court level. But the LODE model would force that explanation to take place at the appellate court level, which sets up the oddity of factfinding by an appellate court, and not simply reversal of a lower court’s factual determination, but actual factfinding based on original evidence procured by the appellate court.

The LODE model’s second divergence from discretionary interlocutory appeals is whether the appellate court’s ruling “may well substantially reduce the amount of litigation necessary on remand.”27 There are three ways in which a discretionary interlocutory appeal could achieve this goal: “(1) eliminate the need for trial, (2) eliminate complex issues so as to simplify the trial, or (3) eliminate issues to make discovery easier and less costly.”28 If we think of the immigration hearing as the equivalent of a trial, then we can see that the LODE model is highly unlikely to lead to any reduction on remand of the amount of litigation of the merits of the case. Under the LODE model, the appellate court’s ruling will send the matter back to the immigration court to complete the hearing; the only consequence is whether the immigration judge is permitted to help develop the record against the wishes of the noncitizen’s lawyer. Thus, the need for completing the hearing won’t be eliminated. If there is any reduction of litigation over particular issues, it could only be of arguments between the immigration judge and the noncitizen’s lawyer over whether the immigration judge can develop the record, through questioning, in ways that conflict with the noncitizen’s lawyer’s strategy. What will not be reduced at all is anything having to do with the merits of the noncitizen’s claim, because none of the merits of the claim are addressed by the LODE model’s avenue of appellate review.

There do not appear to be any discretionary interlocutory appeals involving a claim of ineffective assistance of counsel, but there are a number of instances of failed interlocutory appeals of attorney disqualification rulings.29 These cases have consistently concluded that an interlocutory appeal would not “materially advance the ultimate termination of the litigation,” but would actually cause further delay.30 Reversing a trial court’s ruling on attorney disqualification does not shorten any aspect of the litigation; it only changes which lawyer handles the matter. In the same way, the LODE model’s avenue for review by the circuit court does not reduce the litigation in the immigration hearing; it only dictates whether the immigration judge can override the lawyer’s desire for non-interference.

All of this is to say that the LODE model resembles the discretionary interlocutory appeal only at a very high level of generality – and that the differences are not marginal, but rather ones that go to the heart of the limits of interlocutory appeals. The LODE model forces the circuit court to resolve issues that are likely to be heavily fact-based, and potentially with an incomplete record. And when all is said and done, whichever way the circuit court rules will not save any time at the immigration hearing.

Binding determination of ineffective assistance of counsel?

An additional practical question that arises from the LODE model relates to its interaction with the ineffective assistance of counsel doctrine. In the typical LODE scenario that triggers the interlocutory-like appeal to the U.S. Court of Appeals, the noncitizen’s hearing has begun with the lawyer choosing to opt-out of the judge’s duty to develop the record. But as the hearing proceeds, the immigration judge feels the need to intervene out of belief that the lawyer is inadequate or incompetent. At that point, the immigration judge seeks permission from the appropriate circuit court to disregard the lawyer’s opt-out.

Suppose the circuit court concludes that the noncitizen’s lawyer was not incompetent, and thus the immigration judge may not develop the record. Further suppose that when the matter returns to the immigration court, the immigration judge rules against the noncitizen, who appeals to the Board of Immigration Appeals. Does the circuit court’s prior conclusion preclude the noncitizen from raising a claim of ineffective assistance of counsel? It is hard to see how the answer could be anything other than “yes.” After all, the circuit court engaged in de novo review of the record and determined that the lawyer’s performance was not deficient, which would seem to settle the Strickland v. Washington analysis. But this also seems to be a problematic result: how can the circuit court preclude such a claim before the underlying hearing has even ended?

Perhaps the answer is that the circuit court’s determination is binding only up to that point in the immigration hearing, and that the noncitizen would not precluded from claiming ineffective assistance of counsel for any actions or nonactions by the lawyer subsequent to the return of the matter to the immigration court. But if that is the process, then it seems that the determined immigration judge would be able once again to assert that the lawyer is being incompetent or inadequate from that point on, thus triggering another return to the circuit court for permission to develop the record, and so on.31 Rather than save judicial resources, this process actually causes protracted and needless litigation.

Now consider the opposite possibility: the circuit court agrees with the immigration judge that the lawyer is incompetent or inadequate; the immigration judge is permitted to develop the record. Further suppose that upon return to the immigration court, the immigration judge rules against the noncitizen. How does the circuit court’s prior ruling affect a claim of ineffective assistance of counsel on appeal to the Board of Immigration Appeals? Ineffective assistance of counsel requires that the appellant show not just the lawyer’s deficient performance but also prejudice resulting from that deficient performance.32 We have an appellate ruling that the lawyer was incompetent up to the point when the immigration judge sought review by the circuit court. After that point, the immigration judge intervened and hopefully protected the noncitizen from the lawyer’s ineptitude. But what about the part of the hearing that took place before the immigration judge was permitted to act? Would the immigration judge have to redo the previous part of the hearing with the new power to intervene? Or would the Board of Immigration Appeals simply resolve the ineffective assistance of counsel claim by accepting the prior determination of incompetence in the first part of the hearing and balance that against the second part?

It is not hard to diagnose the cause of the problem: the LODE model puts the circuit court in the position of having to issue a ruling about a lawyer’s performance before the performance has been completed. This is another way in which the LODE model’s analogy to interlocutory appeals breaks down. Even the discretionary interlocutory appeal contemplates a completed factual record for the issue for which appeal is sought, thus enabling the appellate court to engage in the binary task of upholding the lower court’s factual findings or reversing them as clearly erroneous.


There is undoubtedly tension in the immigration statutes that call on immigration judges to develop the record with the prevailing conception of the lawyer’s lead role in the adversarial system. Professor Krishnan has articulated a creative model for attempting to resolve that tension while being faithful to both the statute and the role of the lawyer. But in trying to provide an additional layer of protection for the noncitizen, the solution creates a set of new problems that may be thornier than the one that the model is intended to address.

Moreover, the protection that the LODE model offers the noncitizen is not without cost to the noncitizen. By requiring the immigration judge to pause the hearing to seek appellate permission to intervene in the development of the record, the LODE model delays the resolution of the noncitizen’s hearing by weeks to months, depending on scheduling by the Court of Appeals. If the noncitizen is free on bond or other conditions, this delay might not be particularly burdensome. If, however, the noncitizen has been detained by Immigration and Customs Enforcement, the delay adds to the length of the noncitizen’s incarceration, with only speculative benefit. Without the LODE model, under either the judge-centric or lawyer-centric position, the noncitizen would have a more expeditious resolution, and a favorable ruling would result in faster release, and an unfavorable ruling would lead to an earlier appeal.

If the points raised in this comment are persuasive about the practical concerns about the LODE model, it perhaps suggests that the tension may be irresolvable, and that the better approach is either to relieve judges of the obligation to develop the record when noncitizens are represented by counsel (the true adversarial litigation model), or to allow immigration judges to step in as they see fit (more of quasi-inquisitorial litigation model) – but not to have a system where immigration hearings can be interrupted in media res for appellate review on an incomplete record.

  1. ↑ 1 Professor of Law, Lewis & Clark Law School. Thanks to Jay Krishnan for inviting me to comment on his article.
  2. ↑ 2
    Jayanth K. Krishnan, Overstepping: U.S. Immigration Judges and the Power to Develop the Record, 2022 Wisc. L. Rev. 57, 59 (2022) (citations omitted).
  3. ↑ 3
    Id. at 60.
  4. ↑ 4
    Id. at 61–62.
  5. ↑ 5
    Id. at 62.
  6. ↑ 6
  7. ↑ 7
    Id. at 82.
  8. ↑ 8
  9. ↑ 9
    Id. at 83–84.
  10. ↑ 10
    Id. at 83 n.184.
  11. ↑ 11
    Id. at 83.
  12. ↑ 12
    28 U.S.C. 1292(b) (emphasis added).
  13. ↑ 13
    In re Continental Inv. Corp., 637 F.2d 1, 6 (1st Cir. 1980) (internal citations omitted); see also U.S. Rubber Co. v. Wright, 359 F.2d 784, 785 (9th Cir. 1966) (noting that section 1292(b) “was not intended merely to provide review of difficult rulings in hard cases”).
  14. ↑ 14
    Mamani v. Berzain, 825 F.3d 1304, 1312 (11th Cir. 2016).
  15. ↑ 15
  16. ↑ 16
    See, e.g., Strickland v. Washington, 466 U.S. 668, 698 (1984).
  17. ↑ 17
    See id. (“The record makes it possible to [apply the legal standards to the facts of the case]. There are no conflicts between the state and federal courts over findings of fact”).
  18. ↑ 18
    885 F.3d 1324 (11th Cir. 2018).
  19. ↑ 19
    Id. at 1337.
  20. ↑ 20
    Id. at 1334.
  21. ↑ 21
    Id. at 1328.
  22. ↑ 22
    Id. at 1336.
  23. ↑ 23
    Id. at 1337.
  24. ↑ 24
  25. ↑ 25
    Strickland, 466 U.S. at 687.
  26. ↑ 26
    Id. at 689.
  27. ↑ 27
    McFarlin v. Conseco Svcs., 381 F.3d 1251, 1264 (11th Cir. 2004).
  28. ↑ 28
    See, e.g., Coates v. Brazoria County, 919 F. Supp. 2d 863, 867 (S.D. Tex. 2013).
  29. ↑ 29
    See, e.g., Cardona v. General Motors Corp., 939 F. Supp. 351 (D.N.J.1996); Genentech, Inc. v. Novo Nordisk A/S, 907 F. Supp. 97 (S.D.N.Y.1995).
  30. ↑ 30
    Black v. Missouri, 492 F. Supp 848, 876 (W.D. Mo. 1980).
  31. ↑ 31
    I am reminded of Zeno’s Paradox: travel should be impossible, because whatever distance one has to cover, one first covers half of that distance. Next, one covers half of the remaining distance, and so on. Whatever tiny distance that remains can always be sub-divided in half, and this is true for an infinite number of times. One therefore should never be able to cover an infinite number of these sub-divided distances. See, e.g., Ethan Siegel, This Is How Physics, Not Math, Finally Resolves Zeno’s Famous Paradox, Forbes, May 5, 2020, https://www.forbes.com/sites/startswithabang/2020/05/05/this-is-how-physics-not-math-finally-resolves-zenos-famous-paradox/?sh=3e8f315833f8. Fortunately, mathematics provides the solution in that some infinite sequences will nevertheless sum to a finite number. In the same way, we can imagine that in the truly extreme power struggle among immigration judge, noncitizen lawyer, and circuit court, there might be a repeated set of trips to the circuit court even as the remaining part of the immigration hearing dwindles to near completion.
  32. ↑ 32
    Strickland, 466 U.S. at 687.