concurring in part and dissenting in part.
I agree with the Majority’s result with respect to the question of whether Juror 28 was impliedly biased, and I therefore join in holding that we should remand for additional factfinding on Juror 28’s degree of kinship with the prosecutor. But because I conclude that, depending on facts not available on this record, the average person in Juror 97’s position may also pose an inherent risk of bias, and because we must resolve doubts regarding bias by not seating the affected juror, I would remand for additional factfinding concerning the character and frequency of Juror 97’s interactions with her police officer co-workers involved in the case against Mitchell. I therefore respectfully dissent in part.
As the Majority explains, the “implied bias [doctrine] remains available, in appropriate circumstances, to disqualify jurors whose connection with the litigation makes it highly unlikely that they can remain impartial adjudicators.” (Slip Op. at 13 (citing United States v. Calabrese, 942 F.2d 218 (3d Cir. 1991)).) In dicta in Calabrese, we cited with approval Justice O’Connor’s observation in Smith v. Phillips, 455 U.S. 209, 102 S. Ct. 940, 71 L. Ed. 2d 78 (1982), that a “juror [who] is an actual employee of the prosecuting agency” is one example of an “extreme situation^ that would justify a finding of implied bias.”1 Calabrese, 942 F.2d at 226 (quoting Smith, 455 U.S. at 222 (O’Connor, J., concurring)); accord United States v. Polichemi, 201 F.3d 858, 861-64 (7th Cir. 2000) (holding, without citing to Justice O’Connor’s concurrence in Smith, that a juror who was a 15-year secretarial employee in the civil division of the prosecuting agency was impliedly biased).
It is true, as the Majority says, that Juror 97 is not an employee of the prosecuting agency, the United States Attorney’s Office for the District of the Virgin Islands. Rather, Juror 97 is an employee of the investigative agency, the Virgin Islands Police Department. She knows two of the officers who testified against Mitchell because she works with them. The *877Virgin Islands Police Department was responsible for conducting the investigation that led to Mitchell’s prosecution and was cooperating with the prosecuting agency to convict Mitchell. In my view, a law enforcement agency employee with a close working relationship with testifying officers from the same agency has at least the same risk of inherent prejudice as has an employee of the prosecuting agency. In fact, because of the closer proximity to criminal activity and the often dangerous nature of the work done by agencies like the police department here, employees of such agencies may more likely be seen as impliedly biased against criminal defendants than are employees of a prosecuting agency. Compare Coolidge v. New Hampshire, 403 U.S. 443, 481, 91 S. Ct. 2022, 29 L. Ed. 2d 564 (1971), abrogated on other grounds by Horton v. California, 496 U.S. 128, 130, 110 S. Ct. 2301, 110 L. Ed. 2d 112 (1990) (noting that “the warrant requirement... is ... an important working part of our machinery of government, operating as a matter of course to check the well-intentioned but mistakenly over-zealous executive officers who are part of any system of law enforcement” (citation and internal quotation marks omitted)), with Kang v. Att’y Gen., 611 F.3d 157, 167 (3d Cir. 2010) (observing that an attorney for a prosecuting agency “carries a double burden” and, although “he owes an obligation to government... to conduct his case zealously,” he also “must be ever cognizant that he is the representative of a government dedicated to fairness and equal justice to all and ... he owes a heavy obligation to [his adversary]” (alteration in original) (citations and internal quotation marks omitted)).
I agree with the Majority that we should not only be “concerned . . . with the right of the defendant to an impartial jury” but also “with preservation of the appearance of justice in the courts.” (Slip Op. at 18.) Indeed, public perception matters. Like the seating of a close relative, the seating of someone who goes to work with two of the testifying officers could, depending upon the nature of their interaction on the job, “lodge serious doubts in the public’s mind about the neutrality of the proceedings” which would “favor[] legal attribution of bias.” (Id.) In my view, the “[pjublic confidence in the fairness of the proceedings would suffer if a trial court permitted a juror to deliberate and pass judgment in a case in which” certain of her co-workers, with whom she might have had a close working relationship, were an integral part of “procuring] a conviction.” (Id.)
*878The Majority points out that Mitchell, at oral argument, “concede[d] that employment at a police department, standing alone, does not justify an implication of bias.” (Id. at 19-20.) Citing to Dennis v. United States, 339 U.S. 162, 171-72, 70 S. Ct. 519, 94 L. Ed. 734 (1950), and United States v. Polichemi, 219 F.3d 698, 704 (7th Cir. 2000), the Majority suggests that Mitchell’s concession is well-founded since the law mandates that government employment alone is not enough for a finding of implied bias.2 Then, saying that “our precedent disfavors extending implied bias to cover relationships between jurors and Government witnesses” (Slip Op. at 23),3 the Majority rejects Mitchell’s claim that jurors who are co-workers of testifying police officers at a criminal trial can be impliedly biased. The Majority concludes that “[t]he law . . . does not categorically impute bias to coworkers of key Government witnesses.” (Id. at 25-26.)
*879That holding frames Mitchell’s claim too narrowly, by ignoring Juror 97’s employer and the role that her employer played in securing Mitchell’s conviction. The Majority fails to acknowledge that even the government accepts the premise that at least some employees of an investigative agency would be impliedly biased. At oral argument, the government conceded that a field agent employed by an investigative agency could not serve as a juror because that agent would be biased as a matter of law.4 (See Oral Arg. at 25:28, United States v. Mitchell (No. 11-2420), available at http://www.ca3.uscourts.gov/oralargument/audio/ll2420USAv.Mitchell.wma (“[Q:] Would it have been close enough [of a working relationship with the testifying officers for a finding of implied bias] if [Juror 97] was riding in the squad car with [the testifying officer]? [A:] Yes, that would be close enough. [Q:] Would it be close enough if she weren’t riding in the squad car but she was a fellow officer in the field and sometimes saw him in the field? [A:] Yes. [Q:] So would it be close enough if she never saw him in the field but she was still a field agent? [A:] Yes I believe a field agent in this particular matter would be close enough.”).) The government indicated that the test to determine whether a juror who is an employee of an investigative agency is impliedly biased is whether that employee had a “close working relationship” with the testifying officer. In applying its test, because Juror 97 is not an officer in the field but rather is an office worker who issues equipment and uniforms, the government argued that Juror 97 could not have had a “close working relationship” with the officers and therefore could not be impliedly biased.
The District Court likewise seemed to recognize that a juror could be impliedly biased based on her relationship to the agency that collaborated with the prosecuting agency to convict Mitchell. By asking Juror 97 whether she works with the testifying officers on a daily basis, the District Court appeared attuned to the idea that certain employees of an investigative agency can be impliedly biased, depending upon the nature of their interactions with co-workers who would be testifying officers. Juror 97, however, did not indicate how often or how closely she works *880with the testifying officers, and, unfortunately, the District Court made no attempt to obtain that information. All Juror 97 indicated is that her duties include issuing equipment and uniforms to the officers and that she does not work in the field. By allowing her to remain on the jury, the District Court, like the government, seemed to take the position that the only employees of an investigative agency who can be impliedly biased are field agents. Notwithstanding their ultimate position on whether Juror 97 may be impliedly biased, both the government and the District Court recognized that there is an inherent risk of undermining the appearance of justice if certain employees of an investigative agency are allowed to sit on a jury.5
Although I agree with the government’s assertion — and the implication of the District Court’s line of questioning — that we need to examine the parameters of a juror’s working relationship with a testifying officer to determine whether that juror should be seen as impliedly biased, I do not agree that a bright line can rightly be drawn so that only field agents employed by the investigative agency satisfy the requisite working relationship to be deemed biased as a matter of law. Such an approach “fine-tunes matters too far.” Polichemi, 201 F.3d at 864 (rejecting government’s argument that juror who was a secretary in the civil division of the prosecuting agency “should be disqualified [as impliedly biased] only if she was under the actual supervision of the officer signing the indictment”). Rather, we should consider the specifics of a case in deciding whether bias should be imputed to a juror who is an employee of an investigative agency involved in the prosecution at hand. Consideration should be given, for example, to the character and frequency of the co-workers’ interactions. Similar to how degrees of consanguinity are significant in the “close relative” inquiry, the closer the degree of the working relationship between co-workers, the more likely it *881is that an average person in the co-worker’s shoes would be prejudiced, despite any claim to the contrary.6
The Majority posits that a refusal to accept as jurors any “law enforcement employees who share a close working relationship with a law enforcement witness” constitutes the creation of “a new category of implied bias” which is “unrecognized at common law and in decisions from this Court, the Supreme Court, or other Courts of Appeals.” (Slip Op. at 24.) Thus, says the Majority, “[district courts . . . would have no guidance in policing the nebulous boundary between an employee who works closely with a testifying officer and one who does not.”7 (Id.) As an initial matter, I do not believe that I am suggesting anything genuinely new. As already discussed, for purposes of the implied bias question, an investigative officer can be seen as part of the prosecution team. Cf. Messerschmidt v. Millender, 132 S. Ct. 1235, 1249, 182 L. Ed. 2d 47 (2012) (noting that a police officer is not “automatically entitled to qualified immunity for seeking a warrant unsupported by probable cause” on the basis of his supervisor’s review of it “because the officers’ *882superior . . . [was] part of the prosecution team”); Jackson v. Brown, 513 F.3d 1057, 1074 (9th Cir. 2008) (observing that, in the context of nondisclosure of evidence, “investigative officers are part of the prosecution” (citations omitted)); United States v. Antone, 603 F.2d 566, 569 (5th Cir. 1979) (noting that “[i]n considering use of perjured testimony” the United States Court of Appeals for the Fifth Circuit “focus[es] upon the ‘prosecution team,’ which includes both investigative and prosecutorial personnel” (citation omitted)). Moreover, like jurors who are employees of the prosecuting agency, there is a significant risk that jurors who are close colleagues of the investigative agents involved in the prosecution would feel pressured to convict the defendant because, as Mitchell’s trial counsel argued in his motion to remove Juror 97 for cause, “anything other than a guilty verdict... would mean that [Juror 97] went/voted against her various co-workers and their testimony and against the interest of her direct employer the Virgin Islands Police Department.” (App. at 3.)
However, even if recognizing that law enforcement employees may be viewed as biased in favor of their work colleagues were somehow an innovation in the law, that would not undermine the propriety of the innovation. It is hardly dispositive that a “new category” was not recognized at common law. The common law of implied bias, like common law generally, develops over time. See United States v. Haynes, 398 F.2d 980, 984 (2d Cir. 1968) (“Not only have the[] common law grounds for causal challenge [based on implied bias] retained their vitality, but to them have been added others from which prejudice or bias may be implied.” (internal citation omitted)); cf. Leegin Creative Leather Prods, Inc. v. PSKS, Inc., 551 U.S. 877, 899, 127 S. Ct. 2705, 168 L. Ed. 2d 623 (2007) (“[T]he common law adapts to modern understanding and greater experience . . . .”).
The United States Court of Appeals for the Ninth Circuit, sitting en banc in a habeas case, recognized as much when fashioning a new category of implied bias in Dyer v. Calderon, 151 F.3d 970 (9th Cir. 1998) (en banc). Dyer involved a defendant whose trial resulted in a murder conviction and a death sentence. Id. at 972. During voir dire, one of the prospective jurors who ended up sitting on the jury indicated that none of her close relatives or friends had ever been a victim of any type of crime, even though that juror’s brother had been shot and killed approximately six years earlier. Id. The Dyer court concluded that it should “presume *883bias where a juror lies in order to secure a seat on the jury.”8 Id. at 983. It deemed a prior Supreme Court opinion — outside of the implied bias context — “instructive because Justice Cardozo there equate[d] a juror who lies his way onto the jury to a juror who is related to a litigant.” Id. (citing Clark v. United States, 289 U.S. 1, 11, 53 S. Ct. 465, 77 L. Ed. 993 (1933)). In so doing, the Dyer court understood that it did not matter whether or not a prospective juror was part of a recognized category of implied bias that had been established by the common law or other precedents. Rather, it recognized that the focus of the inquiry should be on whether “it is highly unlikely that an individual will remain impartial and objective” even if that “putative juror swears up and down that [the bias at issue] will not affect his judgment.” Id. at 982. To that end, the Dyer court listed a number of representative individuals who — notwithstanding the fact that some of them fell outside the traditional categories of persons thought to be impliedly biased — would have been deemed biased as a matter of law in Dyer’s trial:
No opinion in the two centuries of the Republic — except the dissent in [Dyer] — has suggested that a criminal defendant might lawfully be convicted by a jury tainted by implied bias. Under the dissent’s logic, reasonable jurists could hold that Dyer would have been accorded due process even if he had been convicted by a jury comprised of the following twelve individuals: (1) the mother of... the prosecutor, (2) [the prosecutor’s] former law partner, (3) [the city’s] Chief of Police, (4) the Grand Dragon of the... [a]rea KKK, (5) the sister of [an individual] who died in the shooting, (6) [that victim’s] mother, (7) the victim of Dyer’s prior robbery, (8) Dyer’s ex-wife, (9) the District Attorney, (10) a[] [city] councilman running for re-election on a “tough-on-crime” platform, (11) . . . Dyer’s cellmate, and (12) [Dyer’s cellmate’s] wife... — so long as they had all sworn they would be fair. We, on the other hand, believe that no reasonable jurist would take that position. Rather, jurists of reason would all agree that each of these individuals, *884had they made their way onto the jury, should have been struck without stopping to inquire into their subjective state of mind.
Id. at 985 (first emphasis added) (internal footnote omitted).
Although the Majority in the present case is correct that no decision is exactly on all fours with the fact scenario we are presented with here, see supra note 5, it is also true that “[n]o opinion in the two centuries of the Republic . . . has suggested that a criminal defendant might lawfully be convicted by a jury tainted by implied bias,” Dyer, 151 F.3d at 985. In sum, “[mjore is at stake here than the rights of petitioner; justice must satisfy the appearance of justice. An irregularity in the selection of those who will sit in judgment casts a very long shadow.” Id. at 983 (citations and internal quotation marks omitted). With respect to Juror 97, I have serious doubts whether judicial outsiders would think it fine that an arresting officer’s work colleague sat on the jury.9 And, as the Majority notes, “[djoubts regarding bias must be resolved against the juror.” (Slip Op. at 9 (alteration in original) (citations and internal quotation marks omitted).)
Whether Juror 97 has the kind of working relationship that warrants a conclusion of implied bias is not something I would venture to determine on this record, since “[t]he record is devoid of information about the nature and regularity of her interaction with the officers.” (Id. at 22.) Juror 97 “may well be objective in fact,” but because yet-to-be-developed facts may indicate that “the [working] relationship is so close” that the law requires us to “err[] on the side of caution” and impute bias, I would have the District Court gather more facts. Polichemi, 219 F.3d at 704. In *885particular, I believe it is our duty to direct the District Court to dig deeper into the character and frequency of the interactions that Juror 97 has had with the two testifying officers. Accordingly, I would remand the case for factfinding to determine whether Juror 97 should be considered biased as a matter of law.
At common law, a “servant” of a party in a case was impliedly biased. See 3 William Blackstone, Commentaries 480-81 (W. Hammond ed. 1980) (noting that jurors may face a “principal challenge... where the cause assigned carries with it prima facie evident marks of suspicion, either of malice or favour... that a juror... is the party ’ s... servant...; which, if true, cannot be overruled, for jurors must be omni exceptione majores”).
The Majority seems to equate employment at an investigative agency that is involved in the prosecution with employment by a government agency that is in no way connected to law enforcement. As discussed supra, there is a sound argument that, in the implied bias context, working for a law enforcement agency which is cooperating with the prosecuting agency presents a much different situation than employment at an agency that is not involved in the prosecution.
The precedents that the Majority cites to support that proposition are United States v. Ferri, 778 F.2d 985 (3d Cir. 1985), and Government of Virgin Islands v. Gereau, 502 F.2d 914, 11 V.I. 265 (3d Cir. 1974), abrogated on other grounds by Corley v. United States, 556 U.S. 303, 129 S.Ct. 1558, 173 L. Ed. 2d 443 (2009). In both of those cases, however, the district court during voir dire conducted a meaningful inquiry with respect to the nature of the relationship at issue. In Ferri, “the day after the jury had been empanelled, the defendants discovered that the husband of one juror was a fireman... [and] an issue arose as to whether she knew [a]... government witness” who was the fire department captain and “apparently knew the juror’s husband.” 778 F.2d at 991. “[T]he district court conducted an additional voir dire examination of the juror,” where the juror indicated that she had never heard of the fire department captain’s name prior to the trial. Id. at 992 & n.5. In Gereau, the district court, after learning that a juror was the divorced wife of a police officer who was to testify at trial, questioned that juror to elicit information regarding the nature of their current relationship — that they had been divorced for seven years, had a child together that lived with the juror, saw each other infrequently and only when it related to the care of that child, and that he contributed to child support. 502 F.3d at 934.
The level of detail elicited from the respective jurors in those two cases stands in stark contrast to the lack of inquiry by the District Court here, as discussed infra. The failure of the District Court to conduct additional voir dire with respect to Juror 97 left open material questions regarding the contours of the working relationship between Juror 97 and the two testifying officers that she knew in her employment capacity.
The Majority’s view thus goes well beyond what even the government is willing to claim. By my colleagues’ position, not even a juror who was a testifying officer’s partner in a squad car could be held biased as a matter of law.
That everyone involved in this case except the Majority accepts this proposition is telling, particularly since the Majority makes apoint of observing that I have not “cite[d] a single case that presumes bias in all prospective jurors who work closely with the arresting officer called to testify on behalf of the Government.” (Slip Op. at 24 n.12.) The rejoinder to the “you have no precedent” observation is that there is probably no case on point because no one before us has ever thought to say that a close work colleague of an investigative and testifying agent could fairly sit on a jury. It is that position which seems to be unprecedented.
This type of examination is distinguishable from an actual bias inquiry. In determining actual bias, a court should look into ajuror’s “state of mind that [would] lead[] to an inference that the person will not act with entire impartiality.” United States v. Torres, 128 F.3d 38, 43 (2d Cir. 1997)). In contrast, the aforementioned considerations would not be used to determine Juror 97’s personal feelings about the particular testifying officers. Rather, the considerations would be used to gather information to make the determination of whether Juror 97 is impliedly biased due to the nature of the working relationship. See Skaggs v. Otis Elevator Co., 164 F.3d 511, 516 n.3 (10th Cir. 1998) (noting that implied bias is “a question of law proved by facts which show [the requisite] connection ...” (emphasis added)).
The Majority “contrasts]” this “new category” of implied bias with “the body of case law on the kinship category of implied bias, stemming both from the common law and from centuries of American constitutional interpretation,” which “supplies the district court a basis for discriminating between jurors who are presumptively biased and those who are not.” (Slip Op. at 24-25.) Although the Majority has not said how “close” a “close relative” needs to be for an implied bias finding with respect to Juror 28 — instead “leaving] it for the District Court to consider, in the first instance, the specific contours of the kinship category within the context of this case” (id. at 19) — I do not read its opinion to adopt the common-law rule of presuming bias to “a juror ... of kin to either party within the ninth degree,” 3 William Blackstone, Commentaries 480-81. If a district court were to look to the common law, which would result in a finding that a third cousin once removed would be biased as a matter of law, “[t]hat consequence” surely would “encumber the selection of jurors in less populated areas like the Virgin Islands,” much more so than presuming bias in a select group of employees at an agency like the police department here. (Slip Op. at 25.) In any event, I am confident that federal district judges have the ability to make appropriate distinctions “between an employee who works closely with a testifying officer and one who does not.” (Id. at 24.)
The Ninth Circuit constructed this new category of implied bias less than a decade after its decision in Tinsley v. Borg, which the Majority cites for the principle that “ ‘[pjrudence dictates that courts’ considering an implied bias claim ‘should hesitate before formulating categories of relationships [that] bar jurors from serving in certain types of trials.’ ” (Slip Op. at 23 (alteration in original) (quoting 895 F.2d 520, 527 (9th Cir. 1990)).)
The doctrine of i mplied bias “has a counterpart in the canons of judicial ethics which require judges to disqualify themselves... if they believe that their impartiality might reasonably be questioned.” Dyer, 151 F.3d at 983 n.22. If a judge fails to do so, a party may move to disqualify a judge under 28 U.S.C. § 455(a). When determining whether to grant a motion for disqualification pursuant to § 455(a), we look to “whether a reasonable person, with knowledge of all the facts, would conclude that the judge’s impartiality might reasonably be questioned.” In re Kensington Int’l Ltd., 353 F.3d 211, 220 (3d Cir. 2003). “[T]he hypothetical reasonable person under § 455(a) must be someone outside the judicial system because judicial insiders . . . may regard asserted conflicts to be more innocuous than an outsider would.” In re Kensington Int’l Ltd., 368 F.3d 289, 303 (3d Cir. 2004) (citation and internal quotation marks omitted). In the implied bias context, we should likewise be sensitive to the fact that a judicial outsider may not regard certain juror bias claims as innocuous, even if a judicial insider would.