When Judges Become the Jury

McClarin v. City of New York, the “Reasonable Jury” Problem, and the Judicial Factfinding That Quietly Reshapes Civil-Rights Cases

I. The Problem Is Larger Than McClarin

The importance of McClarin v. City of New York, 181 F.4th 260 (2d Cir. 2026), extends well beyond the particular warrantless entry that produced the litigation. At its core, the decision exposes a persistent problem in civil-rights litigation: courts sometimes cross the line between determining whether a factual dispute exists and deciding that dispute themselves. The distinction is fundamental. Judges decide questions of law. Juries resolve genuinely disputed questions of historical fact, determine credibility, weigh competing evidence, and draw reasonable inferences from that evidence. Yet doctrines such as summary judgment, qualified immunity, probable cause, and exigent circumstances repeatedly place courts in the position of predicting what a “reasonable jury” could conclude. That procedural responsibility can become something very different when a court stops asking what a jury could reasonably find and instead decides what the judge believes actually happened.

The Second Circuit returned to first principles in McClarin. In reviewing the officers’ request for judgment as a matter of law, the court emphasized that a judge must consider the record as a whole while drawing reasonable inferences in favor of the nonmoving party. More importantly, the court reiterated that credibility determinations, the weighing of evidence, and the drawing of legitimate factual inferences are functions of the jury, not the judge. That proposition should be unremarkable. It comes directly from the Supreme Court’s summary-judgment jurisprudence. Yet its repeated invocation by appellate courts reflects an uncomfortable reality: the boundary continues to require enforcement.

Civil-rights cases are particularly susceptible to the problem because they frequently turn on competing accounts of the same event. Police officers and civilians may describe the same encounter in materially different ways. An officer may testify that a person consented to entry; the person may testify that police forced their way inside. Officers may describe an immediate threat; other evidence may show delay, hesitation, inconsistent conduct, or circumstances inconsistent with urgency. A government defendant may characterize information as undisputed even though the only source of that information is the defendant’s own contested testimony. In discrimination and retaliation cases, the dispute may concern motive, intent, comparative treatment, credibility, and competing explanations for an employment decision. These cases frequently cannot be reduced to a single uncontested document or mechanical fact.

That makes the allocation of decision-making authority critical. A judge may believe one account is more plausible. A judge may believe an officer’s explanation makes more sense. A judge may consider the plaintiff’s evidence weak or regard an inference as unlikely. But that is not the relevant inquiry when the governing procedural standard asks whether a reasonable jury could accept that evidence or draw that inference. The difference between “I do not believe this evidence” and “no reasonable jury could believe this evidence” is not semantic. It is the difference between judicial review and judicial factfinding.

McClarin demonstrates the distinction because the officers repeatedly attempted to transform contested facts into supposedly undisputed ones. In resisting the jury’s verdict, they argued that the evidence established exigent circumstances as a matter of law and, alternatively, entitled them to qualified immunity. The Second Circuit rejected both approaches. It observed that the officers themselves acknowledged “competing accounts” concerning virtually every material aspect of the entry, including the circumstances surrounding the officers’ arrival and the reliability of the statement that supposedly supported their actions. The officers testified that the apartment door had been opened for them. McClarin and Samantha Miranda testified that officers kicked the door down and entered with guns raised. The officers characterized information obtained from Miranda’s aunt as effectively undisputed, but the aunt did not testify, and the Second Circuit expressly explained that the officers’ testimony about what she allegedly told them was not thereby dispositive. The jury remained free to accept or reject the officers’ testimony in whole or in part.

That is the larger significance of the case. McClarin is not merely another decision applying the Fourth Amendment’s warrant requirement. It illustrates how constitutional rights can be lost not only through the underlying governmental conduct but through the procedural handling of the later civil-rights case. If a court resolves contested factual predicates before the jury ever hears them, then constitutional doctrine can become a mechanism for avoiding factual adjudication rather than conducting it. The danger is particularly acute when the dispositive legal question depends upon what actually occurred.

The problem therefore deserves examination beyond one case and beyond one doctrinal category. Federal courts have spent decades articulating the prohibition against weighing evidence at summary judgment while simultaneously deciding thousands of cases under standards requiring them to determine whether any “reasonable jury” could rule for the nonmoving party. That structure necessarily creates tension. The question is whether courts are consistently maintaining the line between assessing evidentiary sufficiency and substituting their own judgment for the judgment of a jury. Increasing empirical evidence suggests that the answer deserves far greater scrutiny.

II. An Emergency That Could Wait

The facts of McClarin make the problem unusually visible because the officers’ claim of exigency was directly at odds with evidence concerning their own conduct.

According to the record, New York City Police Department officers received information that Samantha Miranda was being held against her will, beaten, and forced to use drugs inside a basement apartment at 393 Warwick Street in Brooklyn. Those allegations were unquestionably serious. Miranda’s aunt, Marisol Lopez, had contacted police and identified Justin McClarin as the person allegedly holding and abusing her niece. Lopez reportedly directed officers to the location where Miranda was supposedly being confined. If officers reasonably believed a person inside that apartment faced an immediate threat of serious physical harm, the law recognizes circumstances in which police need not stand outside waiting for a warrant while the danger continues.

But that is not what the officers did.

Before reaching the apartment, officers encountered two men near 399 Warwick Street who attracted their attention because they appeared to be drinking in public. The officers stopped their progress toward the supposed hostage situation and investigated those men. They searched, located contraband, made arrests, transported the men to the precinct, and participated in processing them. Only after completing that unrelated enforcement activity did they return to the apartment where Miranda was allegedly being held and beaten. The Second Circuit noted that the diversion consumed what the officers themselves described as “a couple of hours.” It further emphasized that the time spent investigating, arresting, transporting, and processing the unrelated men would have provided ample time to obtain a warrant.

The officers’ own testimony made the contradiction even more significant. Officer Grieco testified that he would not describe the situation as “urgent.” Sergeant Martinez testified that he did not consider it an “emergency.” Officer Ardolino described the investigation as ongoing and the underlying information as not yet “verified.” The jury was entitled to consider those statements together with what the officers actually did. The Second Circuit concluded that the officers’ decision to spend substantial time on an unrelated investigation constituted circumstantial evidence supporting the jury’s inference that they did not believe Miranda faced an immediate emergency.

That chronology matters because exigent circumstances cannot be determined simply by attaching an alarming label to the original complaint. The constitutional question is not whether allegations of kidnapping, assault, or unlawful restraint sound serious when later summarized in a litigation brief. Of course they do. The question is whether the facts known to the officers at the time created a reasonable belief that immediate entry was necessary and that obtaining a warrant was impracticable.

The district court correctly recognized that the answer depended upon facts: what information the officers actually received, what they observed or heard before entering, and the credibility and weight of the officers’ testimony concerning when, if ever, they perceived an emergency. Those are precisely the kinds of questions that cannot be resolved merely by accepting the government’s description of the incident. The sequence of events itself was evidence from which a jury could infer whether the supposed exigency genuinely existed.

The scene encountered at the apartment also mattered. The Second Circuit noted that there was no officer testimony that they observed lights, heard screaming, heard a struggle, or detected any other contemporaneous indication of an emergency before demanding entry. Grieco testified that the basement area was “dark” and “quiet.” That did not necessarily prove that no danger existed. But neither could it be ignored when evaluating whether the officers reasonably believed immediate warrantless entry was necessary.

This is where courts can improperly alter the constitutional analysis. If a judge begins with the proposition that officers had received a report of a woman being held and beaten, the judge may understandably conclude that police intervention was warranted. But the Fourth Amendment asks a more demanding question when the intervention involves entry into a home without a warrant. It requires examination of the circumstances that supposedly made obtaining judicial authorization impracticable. A serious allegation is not itself an exigency. Otherwise, virtually any sufficiently grave accusation could become a warrant exception.

The distinction between seriousness and immediacy is essential. Police frequently investigate serious crimes without entering homes without warrants. What transforms an ordinary investigation into an exigent circumstance is the reasonable need for immediate action. The officers’ own decisions are therefore highly probative. If officers claiming an emergency have time to stop, investigate unrelated conduct, conduct searches, make arrests, travel to a precinct, and process other prisoners, a jury may reasonably question whether the asserted emergency actually required bypassing the warrant process.

That was the factual determination the officers sought to undo after trial. The jury heard the competing versions and found for McClarin on the unlawful-entry claim. On appeal, the officers attempted to convert their preferred factual account into a legal conclusion by characterizing the evidence as effectively undisputed. The Second Circuit refused to do so. It held that the evidence permitted the jury’s inference and that the existence of exigent circumstances therefore could not be resolved in the officers’ favor as a matter of law.

The lesson extends far beyond McClarin. Courts evaluating claimed emergencies should examine conduct, not merely labels. What did the officers actually do after receiving the alleged emergency information? How much time passed? What did they observe? What did they hear? Did they call supervisors? Did they seek a warrant? Did they undertake unrelated activity? Did their radio transmissions, body-worn-camera recordings, dispatch records, or testimony reflect urgency at the time, or did urgency first become central after litigation began?

An emergency should be demonstrated by the contemporaneous facts, not reconstructed from the seriousness of an accusation after the constitutional challenge is filed.

III. The “No Reasonable Jury” Standard Has a Built-In Institutional Risk

The problem exposed by McClarin is embedded in one of the most familiar phrases in federal civil procedure: “no reasonable jury.”

Summary judgment and judgment as a matter of law necessarily require judges to evaluate evidentiary sufficiency. Courts cannot send every allegation to trial merely because the parties disagree. Rule 56 exists because some disputes are not genuinely factual and some evidentiary records are incapable of supporting a verdict for the nonmoving party. The same principle governs judgment as a matter of law under Rule 50. A federal judge must therefore determine whether the evidence is legally sufficient for a reasonable jury to find for the party bearing the relevant burden.

The institutional danger lies in what happens next.

The proper inquiry is whether a reasonable jury could reach the nonmoving party’s conclusion. It is not whether the judge would reach that conclusion. It is not whether the judge finds one witness more credible than another. It is not whether the judge believes the defendant’s explanation is more persuasive. And it is not whether the judge considers the plaintiff’s inference weak, improbable, or contrary to what the judge believes probably occurred.

Yet those questions can quietly merge.

The Second Circuit’s formulation in McClarin reflects why the distinction matters. It emphasized that courts reviewing the evidentiary record may not make credibility determinations or weigh evidence and must disregard evidence favorable to the moving party that the jury is not required to believe. The rule exists because the moment a judge determines that one witness is more believable, one inference is more sensible, or one version of events is more probable, the court has begun performing the jury’s function.

For years, critics of aggressive summary-judgment practice have argued that this danger is particularly pronounced in civil-rights cases. What has often been missing, however, is empirical evidence directly comparing judicial predictions about reasonable juries with how potential jurors actually evaluate the same facts.

That evidence is beginning to emerge.

In 2026, Professors Elizabeth C. Tippett and Jamillah Bowman Williams published Misjudging a Reasonable Jury: Evidence That Courts Dismiss Meritorious Harassment Claims, a mixed-methods empirical study examining 80 federal Title VII harassment cases. In each case, the employer had argued that no reasonable jury could find the alleged harassment sufficiently severe or pervasive. The researchers presented the underlying facts and relevant jury instructions to study participants and asked them to assess the same legal question the courts had confronted. The divergence was substantial. Judges had granted summary judgment or dismissal in 65 percent of the sampled cases. The simulated juries would have dismissed fewer than 20 percent.

The significance of that result extends beyond harassment doctrine. The researchers were testing the institutional prediction at the center of summary judgment: what would a reasonable jury do with this evidence? Their results suggest that judges and potential jurors may evaluate the same record in materially different ways. The study also found a qualitative difference in how those assessments occurred. Lay participants tended to examine the factual patterns more holistically, while judicial analysis was more likely to separate and discount individual pieces of evidence. The authors concluded that courts should exercise substantially greater caution when resolving harassment claims at summary judgment.

That finding should concern anyone interested in civil-rights adjudication. The summary-judgment standard presumes that judges can distinguish cases in which a jury could reasonably find liability from those in which no reasonable jury could do so. But if judges systematically process evidence differently from jurors, then the phrase “no reasonable jury” risks becoming less a neutral procedural standard than a judicial prediction that may itself be unreliable.

Earlier research points in the same direction. In What Would a Reasonable Jury Do? Jury Verdicts Following Summary Judgment Reversals, 115 Columbia Law Review 1773 (2015), Michael Pfautz examined cases in which district courts granted summary judgment, appellate courts reversed, and the cases later proceeded to jury verdicts. His study identified cases in which actual juries ultimately found for the very party whom the district judge had previously concluded could not persuade any reasonable jury. Pfautz also found a disproportionately high concentration of civil-rights cases among cases in which summary judgment was reversed and the nonmoving party later prevailed before a jury. The implication is difficult to dismiss: in some cases, an actual jury ultimately does precisely what a federal judge previously determined no reasonable jury could do.

None of this means summary judgment should disappear. Nor does it establish that every judicial disagreement with a jury reflects improper factfinding. The standard performs an essential function in federal litigation. But the empirical evidence makes it increasingly difficult to treat judicial assessments of hypothetical juries as infallible.

The problem is especially acute where liability depends upon context, credibility, motive, intent, perception, or competing factual narratives. Those characteristics describe a substantial portion of civil-rights litigation.

That is why McClarin matters beyond its particular facts. The Second Circuit did not decide that the officers’ account was false. It did not decide that McClarin’s account was true. It recognized that the record permitted competing factual conclusions and that the jury had authority to choose among them. The officers’ inability to persuade the jury did not entitle the court to reconstruct the facts afterward under the label of exigent circumstances or qualified immunity.

The persistent danger is not that judges consciously decide to usurp juries. The more subtle problem is structural. Federal procedure repeatedly asks judges to determine what a hypothetical “reasonable jury” could find. Unless courts rigorously maintain the distinction between evidentiary sufficiency and evidentiary persuasion, that inquiry can gradually become a substitute for the jury itself.

The question is therefore larger than whether an individual judge reached the correct result in an individual case. The question is how frequently the federal courts, across thousands of civil-rights cases, are crossing that line—and whether modern empirical tools, including artificial intelligence capable of examining decisions at unprecedented scale, can finally allow us to measure the problem comprehensively.

IV. How the “Reasonable Jury” Inquiry Becomes Judicial Factfinding

The problem is not difficult to identify doctrinally. Federal courts repeatedly state that judges may not weigh evidence, determine credibility, or choose between competing reasonable inferences when deciding summary judgment or judgment as a matter of law. The difficulty lies in application. A court rarely announces that it is deciding which witness to believe. Judicial substitution usually occurs more subtly, through the characterization of evidence, the fragmentation of the factual record, and the conversion of disputed inferences into supposedly legal conclusions.

One recurring method is to describe contested evidence as “undisputed.” A government defendant provides an account of what occurred, often through testimony, an affidavit, an incident report, or other records created by government personnel. If there is no documentary record directly disproving that account, a court may begin treating the government’s version as an established fact rather than one piece of evidence. But the absence of independent contradiction does not automatically transform testimony into an undisputed fact. A jury may reject testimony based upon inconsistencies, surrounding circumstances, contradictory conduct, other circumstantial evidence, or simply its assessment of credibility.

McClarin demonstrates that distinction. The officers characterized significant portions of their account as undisputed, yet the Second Circuit observed that “very little was undisputed.” The parties offered competing accounts concerning the circumstances of entry, whether the door was opened or forced, and the reliability of the information supposedly supporting the officers’ actions. The fact that police officers were the only witnesses to certain conversations did not make their account conclusive. The jury remained entitled to determine whether to credit their testimony.

A second method is factual fragmentation. Courts may examine each item of circumstantial evidence separately, determine that no single fact proves liability, and then dismiss the combined record as insufficient. But juries do not necessarily reason that way. A jury may infer intent, knowledge, pretext, fabrication, retaliation, or unreasonable conduct from the cumulative interaction of multiple facts, none of which would independently establish the claim.

That difference is critical. A suspicious chronology may mean little standing alone. A procedural irregularity may mean little standing alone. An inconsistent explanation, disparate treatment, missing documentation, unusual delay, selective enforcement, or contradiction between testimony and conduct may each be susceptible to an innocent explanation. But the relevant question is not whether each fact independently establishes liability. It is whether the evidence, considered collectively and with reasonable inferences drawn for the nonmoving party, would permit a jury to find liability.

A third problem arises when courts select among competing inferences. Suppose the evidence supports two explanations. One favors the defendant; the other favors the plaintiff. A judge may find the defendant’s explanation more logical, more probable, or more consistent with ordinary experience. That may ultimately prove correct. But at the dispositive-motion stage, the existence of a competing reasonable inference is ordinarily the reason the issue belongs to the jury, not the reason the judge should select the inference the judge considers superior.

This distinction can easily disappear in written opinions. A court may state that evidence “shows” an officer reasonably perceived danger when the evidence actually permits competing conclusions about what the officer perceived. It may say that an employer “acted because of” a legitimate business reason when the plaintiff has produced evidence from which a jury could find the explanation pretextual. It may conclude that an inconsistency is “insufficient” when what the court has actually done is decide how much weight the inconsistency deserves.

Those formulations sound like legal conclusions. Frequently, however, they depend upon antecedent factual judgments.

A fourth mechanism is the treatment of credibility as something other than credibility. Courts know they cannot ordinarily write that they simply do not believe the nonmoving party. Instead, credibility decisions may appear through language describing testimony as “self-serving,” “conclusory,” “implausible,” “uncorroborated,” or contradicted by the “record as a whole.” Some of those characterizations can be legally appropriate. Conclusory allegations unsupported by admissible evidence do not create genuine factual disputes. But sworn testimony based on personal knowledge is evidence, even when it benefits the person giving it. And where testimony conflicts with other testimony or circumstantial evidence, deciding which account is more believable ordinarily remains a jury function.

The Supreme Court’s admonitions in Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 255 (1986), and Reeves v. Sanderson Plumbing Products, Inc., 530 U.S. 133, 150–51 (2000), address more than formal credibility findings. The prohibition against weighing evidence requires courts to resist these indirect forms of factual resolution as well. As the Second Circuit reiterated in McClarin, a court must draw reasonable inferences in favor of the nonmoving party and may not make credibility determinations, weigh competing evidence, or displace legitimate factual inferences that belong to the jury.

A fifth problem occurs when courts convert fact-dependent legal standards into pure questions of law. Concepts such as reasonableness, probable cause, exigency, intent, pretext, and causation contain legal components, but their application frequently depends upon disputed historical facts. Calling the ultimate issue “legal” does not eliminate those predicates.

That is particularly evident in McClarin. Whether the warrantless entry could be justified depended upon what the officers knew, what they observed, what they heard, how they acted, and whether they genuinely confronted circumstances requiring immediate entry. The district court correctly recognized that the exigency question depended in part upon the credibility and weight of the officers’ testimony about when they perceived an emergency. Once those facts became disputed, the legal doctrine could not legitimately be applied by simply selecting the defendants’ preferred version of events.

This is the institutional problem hidden inside the phrase “no reasonable jury.” Judicial substitution does not ordinarily occur because a court openly declares that it is assuming the jury’s role. It occurs when evidentiary characterization gradually becomes evidentiary weighing; when circumstantial evidence is broken into pieces and discounted; when one inference is characterized as reasonable and another as speculation; when contested testimony becomes “undisputed”; and when fact-dependent doctrines are treated as though the underlying facts had already been resolved.

The result is the same. The court moves from deciding whether a jury could find for the nonmoving party to deciding whether the court believes the nonmoving party should win.

Those are fundamentally different inquiries.

V. Civil-Rights Cases Are Especially Vulnerable to This Judicial Drift

The mechanisms described above do not operate uniformly across all litigation. They are particularly consequential in civil-rights cases because those cases so often depend upon context, credibility, motive, perception, and competing factual inferences—the very matters courts are prohibited from resolving at summary judgment.

A commercial dispute may sometimes turn on an executed agreement, an accounting record, or an objectively verifiable transaction. Civil-rights cases are frequently different. A police-misconduct case may turn on what an officer saw before using force, what was said before an arrest, whether consent was given, whether a suspect resisted, whether an officer actually perceived a threat, whether an emergency existed, or whether the officer’s later account is consistent with contemporaneous conduct. Employment-discrimination and retaliation cases may depend upon motive, comparative treatment, timing, shifting explanations, departures from established procedures, and whether seemingly separate events form a recognizable pattern.

Those questions rarely arrive in neat documentary form. They require factual judgment.

That is what makes civil-rights cases particularly susceptible to the judicial drift described in Section IV. When a court characterizes one party’s account as “undisputed,” breaks circumstantial evidence into isolated pieces, selects among competing reasonable inferences, or discounts testimony because it is supposedly “self-serving” or “uncorroborated,” the effect can be especially severe in cases where the plaintiff necessarily relies upon circumstantial evidence and conflicting testimony.

Police-misconduct litigation presents an additional structural problem: the government frequently creates and controls much of the official evidentiary record. Officers prepare arrest reports, aided reports, memo books, command entries, use-of-force documents, body-worn-camera records, investigative reports, and statements summarizing what witnesses supposedly said. That documentation is evidence, but it is not self-authenticating truth. An official report does not become fact merely because the government created it.

McClarin illustrates the problem. The officers sought to rely on what Miranda’s aunt allegedly told them as support for the warrantless entry. But Lopez did not testify. The only trial testimony concerning portions of that conversation came from the officers themselves. The Second Circuit rejected the suggestion that this circumstance rendered their account dispositive. The jury remained entitled to evaluate their credibility and to accept or reject their testimony in whole or in part.

That principle is particularly important in civil-rights cases because otherwise governmental control over documentation can quietly become governmental control over the factual narrative. If an officer’s written report is treated as establishing what occurred unless a plaintiff can produce independent documentary evidence disproving it, the evidentiary burden has effectively been transformed. The plaintiff is no longer required merely to produce sufficient evidence from which a jury could find in his or her favor; the plaintiff is effectively required to disprove the government’s version before being permitted to reach a jury.

That is not the summary-judgment standard.

The same problem arises in employment-discrimination litigation, although the evidence looks different. Employers rarely announce discriminatory or retaliatory motives directly. Plaintiffs therefore rely upon circumstantial evidence: comparators, timing, inconsistent explanations, selective discipline, deviations from policy, unusual scrutiny, procedural irregularities, historical treatment, and the sequence of events surrounding the challenged action. No single item may conclusively establish discriminatory intent. That is precisely why juries are permitted to draw inferences from the total record.

When courts instead analyze each fact separately and determine that each has an innocent explanation, they risk performing the very weighing process reserved for the jury. A comparator difference may be explainable. A shifting explanation may be explainable. A suspicious timeline may be explainable. A departure from established procedure may be explainable. But the legal question is not whether the defendant can articulate an innocent explanation for every individual fact. The question is whether a reasonable jury, examining those facts collectively, could infer unlawful motive.

That distinction matters because civil-rights liability is frequently established through accumulation rather than admission.

The danger is therefore not simply that civil-rights plaintiffs lose dispositive motions at high rates. The deeper institutional concern is that the nature of these claims places courts in repeated contact with precisely the types of factual questions that summary judgment forbids judges from resolving. Motive must be inferred. Credibility must be assessed. Context matters. Conduct must be evaluated cumulatively. Competing explanations must be compared.

Those are ordinary functions of a jury.

The more a claim depends upon those functions, the more vigilant courts must be about maintaining the distinction identified in Section IV: determining whether evidence is legally sufficient is not the same thing as determining which evidence is true.

A court does not preserve the jury’s role merely by reciting that all reasonable inferences must favor the nonmoving party. The protection exists only if the court actually applies that rule to the entire factual record.

VI. Qualified Immunity Can Turn Judicial Drift Into a Dispositive Doctrine

Qualified immunity presents an even greater danger because it combines an issue ultimately decided by the court with factual predicates that may belong to the jury.

The legal component of qualified immunity is familiar. Courts determine whether the conduct alleged violated a constitutional right and whether that right was clearly established at the relevant time. But the legal analysis frequently cannot be performed in the abstract. Whether an officer acted reasonably may depend upon what the officer knew, saw, heard, perceived, or confronted before acting.

Those are historical facts.

The distinction becomes critical when the parties dispute them.

Suppose an officer claims that a person suddenly reached toward his waistband, while the plaintiff denies making any such movement. Suppose officers claim that they heard screams from inside a residence before entering without a warrant, while other witnesses deny that any screams occurred. Suppose an officer claims that a person actively resisted arrest, while video evidence or testimony supports a competing account. The ultimate immunity question may be one for the court, but the court cannot legitimately answer it by first deciding which version of the disputed event actually occurred.

That is precisely the boundary McClarin reinforces.

The Second Circuit explained that qualified immunity is an affirmative defense on which defendants bear the burden. More importantly, where entitlement to immunity depends upon a particular factual finding, the defendant must obtain that finding from the jury. If the defendant fails to ask the jury the necessary factual question, the court may not supply the finding afterward in order to establish immunity.

That principle deserves more attention than it ordinarily receives. It means that qualified immunity cannot function as a procedural device for converting contested historical facts into judicial findings.

In McClarin, the officers argued that they were entitled to immunity because a reasonable officer could have believed the warrantless entry lawful. But the answer depended upon what information the officers possessed and what they perceived before entering. The Second Circuit emphasized that the relevant inquiry was fact specific and required examination of the information possessed by the officers at the time.

The factual record was sharply contested. The officers testified that circumstances justified their entry. McClarin and Miranda disputed central portions of that account. The officers contended that the door had been opened; McClarin and Miranda testified that officers forced their way inside. The officers claimed exigency; their own conduct included spending substantial time investigating unrelated individuals before returning to the purported emergency. The area outside the apartment was described as dark and quiet, with no testimony that officers heard screams or sounds of an ongoing assault before demanding entry.

A court could not legitimately resolve qualified immunity by choosing which of those competing accounts was more believable.

Yet this is precisely how judicial drift can occur. A court begins with a legal doctrine—qualified immunity—but must decide what factual circumstances existed before applying it. If the court accepts the officer’s account of those circumstances, discounts contrary evidence, or selects the inference most favorable to the officer, the court has already performed the factual work necessary to determine the legal result.

The label “qualified immunity” does not make that antecedent factfinding judicial.

This distinction also exposes a tension in the traditional justification for qualified immunity. The doctrine is often described as protecting public officials not merely from liability but from the burdens of litigation and trial. That objective encourages courts to resolve immunity early. But early resolution is impossible where the legal analysis depends upon genuinely disputed facts unless the court either permits factual development or chooses among the competing versions itself.

That creates institutional pressure in exactly the wrong direction.

Professor Joanna C. Schwartz documented another aspect of that tension in How Qualified Immunity Fails, 127 Yale Law Journal 2 (2017). Her empirical study examined 1,183 actions brought under 42 U.S.C. § 1983 against state and local law-enforcement defendants across five federal districts. She found that qualified immunity disposed of far fewer cases at the earliest stages of litigation than the doctrine’s conventional justification would suggest. Among the law-enforcement cases studied, qualified immunity resulted in dismissal of only a small percentage of cases at the pleading and summary-judgment stages.

That research does not establish that judges improperly resolve factual disputes when applying qualified immunity. It demonstrates something different but important: the doctrine’s real-world operation does not neatly correspond with the assumption that qualified immunity principally functions as an efficient mechanism for eliminating insubstantial constitutional claims before significant litigation occurs.

Professor Alexander A. Reinert identified the related jury problem in Qualified Immunity at Trial, 93 Notre Dame Law Review 2065 (2018). His work examined the uncertain division of responsibility between judge and jury when immunity depends upon disputed facts. The problem is structural. Courts are charged with resolving the ultimate immunity issue, but factual disputes relevant to that decision may remain within the jury’s province. When those roles are not carefully separated, the immunity analysis risks absorbing the factfinding function.

That is why the sequence matters.

First, determine whether material historical facts are genuinely disputed.

If they are, identify the factual questions necessary to the immunity determination.

Allow the jury to resolve those factual disputes where required.

Then apply the legal qualified-immunity doctrine to the facts the jury has actually found.

Reversing that order invites the very problem this article examines. A court may begin with the desire to decide immunity as a matter of law and then implicitly resolve the factual record in whatever manner makes that legal determination possible.

That is not simply an error in qualified-immunity doctrine. It is the same institutional problem running throughout modern dispositive-motion practice: the transformation of a question about what a reasonable jury could find into a judicial determination of what actually happened.

McClarin provides the corrective. The Second Circuit did not hold that a jury decides qualified immunity. It held something more precise and more important. When immunity depends upon disputed historical facts, a court cannot bypass the jury by making those factual findings itself.

The distinction preserves both functions.

The jury decides the disputed facts.

The court applies the law to those facts.

When courts collapse those two stages, qualified immunity stops being merely a defense to liability. It can become a mechanism for removing factual disputes from the jury altogether.

VII. McClarin Shows That Protecting the Jury’s Role Runs Both Ways

The principle at issue is not that courts should defer to civil-rights plaintiffs. It is that courts should preserve the jury’s function regardless of which party benefits from it.

That point is essential because McClarin did not end with the Second Circuit simply affirming the jury’s verdict against the officers. The court rejected the officers’ attempt to obtain judgment as a matter of law on the unlawful-search claim, but it nevertheless vacated the judgment and ordered a new trial because the district court improperly excluded evidence that the jury should have been permitted to consider.

The excluded evidence consisted of recorded conversations between McClarin and Samantha Miranda. The defense sought to use those recordings to impeach McClarin’s testimony that he had not offered or promised Miranda money in exchange for favorable testimony. The recordings were also relevant to Miranda’s credibility because she had repudiated the statement she originally gave police accusing McClarin of holding and abusing her. The district court excluded the recordings on the ground that they had not been disclosed before trial. The Second Circuit concluded that the exclusion was erroneous and sufficiently important to require a new trial.

That ruling carried substantial consequences. The jury had awarded McClarin $40,000 in compensatory damages and $275,000 in punitive damages on the unlawful-search claim, together with $75,000 in compensatory damages and $500,000 in punitive damages on the malicious-prosecution claims. The Second Circuit nevertheless vacated the judgment because the jury had not been permitted to hear evidence bearing directly on the credibility of two central witnesses.

The significance is not that the recordings proved the officers’ version of events. The Second Circuit did not make that finding. The significance is that the jury should have been allowed to hear the evidence and decide what weight, if any, to give it.

That distinction reinforces the larger thesis of the case.

A court improperly substitutes itself for the jury when it resolves disputed facts in favor of a government defendant. But a court can also interfere with the jury’s function by preventing jurors from hearing admissible evidence necessary to evaluate credibility. In both circumstances, the institutional error is the same: the factfinding process has been altered by judicial decision-making that exceeds the court’s proper role.

The appellate outcome therefore strengthens, rather than weakens, the broader argument. McClarin is not a decision announcing that juries should favor civil-rights plaintiffs. It is a decision insisting that juries be permitted to perform the task assigned to them.

That principle is especially important in cases built around competing narratives. If credibility is central, the jury must generally be permitted to hear the evidence bearing upon credibility. If material facts are disputed, the court cannot decide which version is true merely because resolving the dispute would make a legal issue easier to decide.

The role of the court is not to engineer the factual record toward the outcome it considers most justified. It is to apply the governing evidentiary and procedural rules while preserving the jury’s authority over factual disputes.

That is what makes McClarin more consequential than a conventional Fourth Amendment decision. On one side of the case, the Second Circuit refused to allow the officers to replace the jury’s factual determinations with their preferred version of events. On the other, it refused to allow the district court’s evidentiary ruling to prevent the jury from hearing material impeachment evidence.

Both rulings protect the same institution.

The jury.

VIII. Artificial Intelligence May Finally Allow Us to Measure the Problem at Scale

The existing empirical literature establishes enough to justify concern, but it does not yet tell us how extensive the problem is across federal civil-rights litigation.

That limitation is understandable. Until recently, studying judicial substitution at scale would have required extraordinary amounts of human review. Researchers would need to identify thousands of dispositive decisions, reconstruct the relevant factual record, determine which facts were genuinely contested, analyze how courts characterized those disputes, identify whether competing inferences existed, and then distinguish permissible determinations of evidentiary insufficiency from improper credibility assessments or factual weighing.

Traditional empirical legal research can perform that work, but generally only across limited datasets, selected jurisdictions, particular doctrinal categories, or manageable samples of decisions.

Artificial intelligence changes what is technically possible.

Modern natural-language systems can process large bodies of judicial opinions and identify recurring linguistic and analytical patterns at a scale that previously would have required teams of researchers working for years. That does not mean artificial intelligence can simply declare that a judge improperly invaded the province of the jury. The legal judgment remains too contextual for that kind of automated conclusion to be reliable. But artificial intelligence can identify the decisions most likely to present the problem and make comprehensive human review substantially more feasible.

A serious national study could begin with federal decisions resolving motions under Federal Rules of Civil Procedure 12, 50, and 56 in civil-rights cases. The dataset could include actions under 42 U.S.C. § 1983, Title VII of the Civil Rights Act of 1964, the Americans with Disabilities Act, the Age Discrimination in Employment Act, and other federal civil-rights statutes. Researchers could then use artificial-intelligence-assisted analysis to identify opinions containing the doctrinal language associated with factual substitution: references to “no reasonable jury,” “self-serving testimony,” “uncorroborated” accounts, “implausible” testimony, “undisputed” facts, competing inferences, credibility disputes, and determinations that circumstantial evidence is insufficient as a matter of law.

The study could then ask much more precise questions.

How often do courts acknowledge a factual dispute but nonetheless resolve the case as a matter of law? How often do courts characterize one party’s account as “undisputed” even though that characterization depends upon accepting testimony from an interested witness? How frequently are plaintiffs’ sworn accounts discounted as self-serving while defendants’ sworn accounts are treated as affirmative evidence of what occurred? How often do courts consider pieces of circumstantial evidence separately rather than evaluating their cumulative effect? How frequently do appellate courts reverse because the district court improperly resolved factual disputes or failed to draw reasonable inferences for the nonmoving party?

The analysis could also distinguish among substantive areas. Police excessive-force claims may produce one pattern. Unlawful-search and false-arrest cases may produce another. Employment-discrimination and retaliation cases may reveal different mechanisms of judicial substitution. Qualified-immunity decisions could be studied independently because of the doctrine’s distinctive interaction between legal and factual questions.

The research could also examine institutional variables that have historically been difficult to study comprehensively: circuit, district, case type, procedural posture, presence of video evidence, whether the plaintiff was represented, whether factual disputes were expressly acknowledged, whether the court relied on official government records, and whether the decision was later reversed.

Most importantly, artificial intelligence could permit researchers to examine the language courts use immediately before resolving factual disputes. That may prove more revealing than raw disposition rates. Courts rarely announce that they are weighing credibility. The problem, as discussed earlier, often appears through subtler moves: describing testimony as uncorroborated, dismissing an inference as speculative, fragmenting circumstantial evidence, or treating one account as objective fact while characterizing the competing account as allegation.

Those patterns are exactly the kind of large-scale textual phenomena that computational analysis is well suited to identify.

But artificial intelligence should be used as an investigative instrument, not as the final decisionmaker. Machines can locate patterns, cluster opinions, compare language, trace citations, and flag cases for examination. Lawyers and scholars must still determine whether a particular ruling actually crossed the line between evaluating evidentiary sufficiency and deciding disputed facts.

The value of artificial intelligence is therefore not that it replaces legal judgment.

It is that it may finally allow legal judgment to be applied comprehensively enough to determine whether the problem is isolated, episodic, doctrinally concentrated, or systemic.

For decades, lawyers have encountered cases in which courts appeared to do more than determine whether a jury could reasonably find for the plaintiff. Appellate courts have repeatedly reminded trial courts not to weigh evidence or resolve credibility. Empirical studies now demonstrate meaningful differences between judicial predictions and juror assessments in at least some civil-rights contexts.

The next step should be to measure the phenomenon nationally.

We now have tools capable of making that project realistic.

IX. The Question Is No Longer Whether the Problem Exists, but How Large It Is

The law already supplies the rule.

Courts may determine whether evidence is legally sufficient. They may decide questions of law. They may dismiss claims where the record cannot support a verdict. But they may not weigh competing evidence, determine credibility, or choose among reasonable factual inferences merely because one version of events appears more persuasive.

The Supreme Court has said so repeatedly.

The courts of appeals repeat the rule constantly.

And McClarin repeats it again.

The persistence of those reminders should itself prompt a question. If the boundary between judicial decision-making and jury factfinding were consistently respected, why must appellate courts continue returning to the same prohibition?

The available empirical evidence makes that question more difficult to dismiss. Studies comparing judicial determinations with juror assessments demonstrate that judges and juries do not necessarily evaluate factual records in the same way. Research involving summary-judgment reversals shows that actual juries sometimes reach precisely the result a district judge previously concluded no reasonable jury could reach. Civil-rights cases appear repeatedly in that literature because they so often depend upon context, inference, credibility, and cumulative evidence.

McClarin adds another current example of why the distinction matters. The Second Circuit refused to permit the officers to transform disputed historical facts into a judgment as a matter of law. It reiterated that the jury was entitled to assess credibility and draw reasonable factual inferences from the officers’ conduct. At the same time, it vacated the verdict because the district court had improperly prevented that same jury from hearing important impeachment evidence.

The lesson is not complicated.

Judges decide law.

Juries decide disputed facts.

The difficult part is enforcing that division when modern civil procedure repeatedly asks judges to determine what a hypothetical “reasonable jury” could conclude.

That is where the danger lies. A court can begin by asking whether sufficient evidence exists and end by deciding which evidence deserves to be believed. It can begin by identifying whether an inference is legally permissible and end by deciding which inference is more convincing. It can begin by applying qualified immunity and end by resolving disputed historical facts necessary to reach the immunity question.

Each individual step may appear modest. Collectively, they can eliminate the jury from a case without ever expressly acknowledging that a factual determination has been made.

That is why this issue deserves comprehensive empirical study.

The central question is no longer whether examples can be found. They can. The Supreme Court’s repeated warnings, appellate reversals, McClarin, and existing empirical research establish that the risk is real.

What remains unknown is its scale.

How many civil-rights cases are terminated each year after courts resolve factual disputes that a jury could reasonably have decided differently? Are particular doctrines especially susceptible? Are some categories of cases affected more than others? Does the problem vary by circuit or district? Are particular forms of judicial language reliable indicators that the court has moved from assessing evidentiary sufficiency into weighing the evidence itself?

Those questions are now capable of being studied in a way they were not even a decade ago.

Artificial intelligence will not answer whether every individual judicial ruling was correct. Nor should it.

But it can help us finally assemble and examine the universe of decisions necessary to ask a more consequential question:

When federal courts say that “no reasonable jury” could find for a civil-rights plaintiff, how often are they accurately applying a procedural standard—and how often are they simply becoming the jury themselves?

About the Author