Sdahrie Howard v. Cook County Sheriff's Office

989 F.3d 587

• Court: Court of Appeals for the Seventh Circuit
• Decided: March 4, 2021
• Docket: 20-1723
• Status: Published

Opinion

In the

United States Court of Appeals For the Seventh Circuit ____________________ No. 20‐1723 SDAHRIE HOWARD, et al., Plaintiffs‐Appellees, v.

COOK COUNTY SHERIFF’S OFFICE and COUNTY OF COOK, Defendants‐Appellants. ____________________

Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 17 C 8146 — Matthew F. Kennelly, Judge. ____________________

ARGUED DECEMBER 3, 2020 — DECIDED MARCH 4, 2021 ____________________

Before SYKES, Chief Judge, and FLAUM and ST. EVE, Circuit Judges. ST. EVE, Circuit Judge. In this case we review the certifica‐ tion of a class‐action lawsuit alleging a horrible “epidemic” of sexual harassment at the Cook County Jail. The named plain‐ tiffs are ten women who work at the jail or an adjoining court‐ house. They sue their employers—the Cook County Sheriff’s Office and Cook County—for failing to prevent male inmates from sexually harassing them. They propose to sue not just 2 No. 20‐1723

for themselves but for thousands of other women who work at the jail or courthouse. The district court certified a class comprising all non‐supervisory female employees who work with male inmates at the jail or courthouse, of whom there are about 2,000. We granted the defendants’ request for an interlocutory appeal, see Fed. R. Civ. P. 23(f), and we now hold that the dis‐ trict court abused its discretion in certifying the class. The court’s primary error was using the peripheral and overbroad concept of “ambient harassment” (i.e., indirect or secondhand harassment) to certify a class of employees who have endured a wide range of direct and indirect harassment. Even without this error, the class cannot stand because it comprises class members with materially different working environments whose claims require separate, individualized analyses. We thus reverse the order certifying the class. I. Background A. The Jail Complex The Cook County Jail is one of the largest single‐site jails in the country. It fills 36 buildings that span eight city blocks. About 100,000 inmates pass through the jail each year. At any given time approximately 6,500 inmates reside at the jail while awaiting trial. On an average day, between 700 and 900 inmates attend court hearings. (These numbers have de‐ creased somewhat during the COVID‐19 pandemic.) The inmates reside in different housing divisions at the jail. Divisions 3, 4, and 5 house female inmates. Divisions 2 and 6 house male inmates with minimum‐ or medium‐secu‐ rity classifications. Division 8 encompasses medical facilities operated by Cermak Health Services and a Residential No. 20‐1723 3

Treatment Unit where inmates receive mental health services. Divisions 9 and 10 house maximum‐ and super‐maximum‐se‐ curity male inmates. The inmates in Divisions 9 and 10 often overlap with the Division 8 population. Other, non‐residen‐ tial parts of the jail, such as the “south campus” and a Divi‐ sion 5 annex, host administrative operations. The George N. Leighton Criminal Courthouse is a sepa‐ rate structure that connects to the jail through a series of tun‐ nels by which officers take inmates to court hearings. The courthouse has dozens of courtrooms spread across eight floors. It also has an administrative wing containing adminis‐ trative offices, jury rooms, judicial chambers, a cafeteria, a law library, and a mental health courtroom. The only place in the administrative wing where inmates go is the mental health courtroom. The Sheriff’s Office controls operations at the jail complex. (We use the term “jail complex” as a shorthand for the entire complex, including the jail, Cermak, and the courthouse.) It employs both sworn staff (i.e., correctional officers and dep‐ uty sheriffs) and civilian staff. Cermak workers are civilian employees of the County. Non‐supervisory employees at the jail have a variety of roles. Many work in the housing divisions. Others provide healthcare to inmates at Cermak or the Residential Treatment Unit. Others work behind the scenes at Cermak as, for exam‐ ple, accountants, programmers, and administrative assis‐ tants. Still others work in non‐residential parts of the jail in administrative roles (e.g., mailroom, visitation, IT, records, HR, drug testing). Employees who work at the courthouse (called court services deputies) also have a variety of roles. Some work as courtroom deputies, for example, while others 4 No. 20‐1723

perform door security or work with juries in the administra‐ tive wing. B. This Lawsuit The named plaintiffs are ten women who work at the jail complex. They include four sworn correctional officers, a ci‐ vilian correctional rehabilitation worker, a civilian para‐ medic, and four sworn deputy sheriffs who work at the court‐ house. One of the correctional officers works in the Strategic Operations and Information Unit reviewing reports and vid‐ eos of inmate misconduct. The other three work in Divisions 2 or 10, as does the civilian rehabilitation worker. Collectively, the correctional officers and the rehabilitation worker have worked in every housing division (including Cermak and the Residential Treatment Unit) and non‐residential parts of the jail. Three of the deputy sheriffs work in the Court Services Department at the courthouse. Their duties include transport‐ ing inmates to hearings. The other deputy sheriff works in “the bridge,” a staging area in the courthouse basement where inmates wait for the court to call their cases. The four deputy sheriffs have worked at various other locations in the courthouse in the past. The paramedic works at Cermak. She also has worked in Divisions 2, 9, and 10. In their suit against the Sheriff’s Office and Cook County, the plaintiffs allege that they have endured frequent and ex‐ treme sexual harassment by male inmates, which the defend‐ ants have failed to take reasonable measures to prevent. The harassment occurs “on a daily or nearly daily basis through‐ out the Jail.” The plaintiffs allege that male inmates expose their genitals to them, masturbate at them, direct sexual re‐ marks and gestures at them, grope and grab them, and threaten and commit sexual violence against them. The No. 20‐1723 5

plaintiffs say they have complained of this horrible harass‐ ment to no avail. On behalf of themselves and other similarly situated women who work at the jail complex, the plaintiffs sue the defendants for permitting a hostile work environment in vio‐ lation of Title VII of the Civil Rights Act of 1964. They also bring claims under 42 U.S.C. § 1983 for gender discrimination in violation of the Fourteenth Amendment’s Equal Protection Clause; gender‐discrimination claims under the Illinois Civil Rights Act, 740 ILCS 23/5; and claims for indemnification against Cook County. A few weeks after the plaintiffs filed suit, the district court entered a preliminary injunction man‐ dating certain preventative measures agreed to by the parties. This appeal arises from the second of two class certifica‐ tion orders entered by the district court. The second order modified the first, so our review touches on both orders. 1. Original Class Certification Order The plaintiffs moved to certify a class under Federal Rule of Civil Procedure 23(b)(3) comprising all female employees of the Sheriff’s Office or the County who work at the jail com‐ plex, other than certain high‐level management employees like the executive director, chief financial officer, and superin‐ tendent. The plaintiffs submitted a range of evidence in support of their motion for class certification, including the defendants’ policies, procedures, and records; expert reports from statisti‐ cal, psychological, and correctional professionals; 144 class member declarations; and deposition testimony from class members and defense witnesses. The defendants submitted 6 No. 20‐1723

their own evidence, similar in kind, to oppose class certifica‐ tion. At a general level, the evidence shows:  Prevalence of sexual harassment. There were at least 1,745 filed reports of male inmates sexually harassing female employees at the jail complex between January 2015 and September 2018. The vast majority involved indecent ex‐ posure or exhibitionist masturbation. Because of underre‐ porting problems, the number of incident reports may sig‐ nificantly understate the actual number of incidents. Ac‐ cording to the plaintiffs’ statistical expert, more than two‐ thirds (1,186) of the reported incidents occurred in Divi‐ sions 8, 9, or 10, which house the highest‐security male in‐ mates. (The defendants’ expert put that number at three‐ quarters.) Only seven percent (121) of the incidents oc‐ curred in Divisions 2 or 6, which house lower‐security male inmates. Nine percent (151) occurred in non‐residen‐ tial parts of the jail. And six percent (110) occurred at the courthouse. Beyond incident reports, the 144 class mem‐ ber declarations recounted extreme episodes and patterns of sexual harassment by male inmates.  Exposure to harassment. A female employee’s exposure to harassment can vary significantly with job assignment, though job assignments can change. Some class members, for example, described seeing little misconduct while working in Divisions 2 through 5 (which house female in‐ mates and lower‐security male inmates), while encounter‐ ing much more upon transferring to Division 9 (a maxi‐ mum‐security male division). Other employees (e.g., ad‐ ministrative courthouse personnel and non‐medical staff at Cermak) do not work in the housing divisions and have less direct contact with male inmates. At the same time, job No. 20‐1723 7

assignment is not the only factor that determines an em‐ ployee’s exposure to harassment. Both employees and in‐ mates traverse the jail, so female employees can encounter male inmates outside of their workspaces. For example, many class members described sexual harassment occur‐ ring in the tunnels that connect the courthouse to the jail.  Policies and Preventative Measures. The Sheriff’s Office controls jail operations through a hierarchical structure. It issues policies on inmate security, conduct, and discipline that apply to all inmates and employees at the jail com‐ plex. These policies prohibit, and impose sanctions for, sexual harassment by inmates. They also establish report‐ ing procedures. They require civilian staff to notify sworn staff or supervisors of inmate misconduct. The sworn staff or supervisor must then speak with the inmate or file a disciplinary report. The Sheriff’s Office has taken other measures to curb the sexual harassment including inmate education, staff training, and physical restrictions on known offenders. Based on this evidence and the parties’ extensive briefing, the district court certified the following class: All women who have been employed by the Cook County Sheriff’s Office at the Jail, or as Court Services deputies at the Leighton Courthouse, or by the County in positions with Cermak Health Services, at any time since April 23, 2015, except women who, during that period, have held the positions identified in Exhibit A to the complaint or who were employed in supervisory roles. 8 No. 20‐1723

Exhibit A listed the high‐level management employees whom the plaintiffs had carved out of the class. The court also ex‐ cluded supervisory employees because it agreed with the de‐ fendants that supervisors’ own wrongdoing might become an issue in the lawsuit. See Fed. R. Civ. P. 23(a)(4) (requiring that “the representative parties will fairly and adequately protect the interests of the class”). In all other respects, the court found that the plaintiffs’ proposed class met the requirements for class certification under Rule 23(b)(3). Relevant here, the court found that the proposed class met the requirements of commonality, typicality, and predominance. See Fed. R. Civ. P. 23(a)(2)–(3), (b)(3). The court’s Rule 23 analysis relied heavily on the concept of “ambient harassment,” which it defined as “the experience of working in an environment highly permeated with sex‐ ually offensive and degrading behavior, that is, a highly sex‐ ualized atmosphere in which crude and offensive sexual be‐ havior is common and employees see that it is normative, whether specifically directed at them or not.” The court bor‐ rowed this definition from the plaintiffs’ expert, Dr. Louise Fitzgerald, who performed a “social‐framework analysis” to opine on the impact and causes of sexual harassment at the jail complex. Although the court excluded Dr. Fitzgerald’s opinions about the extent of sexual harassment at the jail com‐ plex and the adequacy of the defendants’ curative measures, the court deemed her more general “opinions regarding the social‐scientific background of workplace harassment” ad‐ missible. Relying on an ambient‐harassment theory, the court found that the putative class met the commonality requirement be‐ cause the class members shared at least one common No. 20‐1723 9

question: “whether the ambient harassment experienced by female employees at the jail and the courthouse is sufficiently severe and pervasive to support a Title VII hostile work envi‐ ronment claim.” Ambient harassment also factored into the court’s finding of typicality. The court concluded that the named plaintiffs’ claims were typical of the class claims because they relied on the same legal theory as the class members (hostile work en‐ vironment), even if some had experienced direct harassment and others only ambient harassment. The court reasoned that direct harassment differs from ambient harassment only in degree and not in kind. Finally, the court found that three common questions pre‐ dominated over individual questions: (1) “whether all of the putative class members experienced an objectively hostile work environment based on the ambient harassment at the jail and courthouse;” (2) “whether the detainees’ harassment occurred because of sex;” and (3) “whether there is a basis for employer liability based on the defendants’ failure to adopt reasonable policies to combat the harassment.” The court acknowledged that the second question was “unlikely to be a significant point of dispute,” but it considered the other two questions to be substantial issues in the case. The court did not see the individual issues—i.e., the subjective offensive‐ ness of the harassment, damages, and affirmative defenses— as significant aspects of the case. 2. Class Modification and Interlocutory Appeals We granted the defendants’ motion for an interlocutory appeal of the original class certification order. See Fed. R. Civ. P. 23(f). Soon after, however, the plaintiffs moved in the 10 No. 20‐1723

district court to refine the class based on a newly produced staffing spreadsheet that, in their words, “comprehensively describes and summarizes by division and department, the job duties of all correctional officers and all other job catego‐ ries in the Jail.” According to the plaintiffs, the spreadsheet proved that almost every class member had direct contact with male inmates. The plaintiffs identified a few categories of employees (mostly administrative) whose jobs did not bring them into direct contact with male inmates and asked the district court to trim those employees from the class. Given the pending appeal, the plaintiffs requested an indica‐ tive ruling from the district court that it would refine the class based on the spreadsheet. See Fed. R. Civ. P. 62.1. They stressed that modifying the class would “have the advantage of allowing Plaintiffs to establish the common issue of whether sexual harassment was severe or pervasive and ob‐ jectively offensive without reference to ambient or ‘secondhand’ harassment.” As requested, the district court issued an indicative ruling explaining that if we remanded the case it would grant the plaintiffs’ motion to modify the class. The court recognized that the plaintiffs’ motion implicated Rule 23’s commonality and predominance requirements. Given that the spreadsheet showed that some of the class members lacked contact with male inmates, those class members were now “situated differ‐ ently from the others in a material way.” As such, “a narrower class that excludes positions that lack contact with male de‐ tainees is more appropriate than the presently certified class.” Yet the court did not reanalyze the Rule 23 requirements or explain why the modified class met them. It also did not re‐ solve the parties’ dispute about whether ambient harassment would remain an issue for the modified class. It merely No. 20‐1723 11

commented that “narrowing the class is likely to significantly narrow any differences in this regard across the class.” We remanded the case over the defendants’ objections, and the district court modified the class definition as follows: All women who have been employed by the Cook County Sheriff’s Office at the Jail, or as Court Services deputies at the Leighton Courthouse, or by the County in positions with Cermak Health Services, at any time since April 23, 2015, except women who, during that period, have held the positions identified in Exhibit B to the plaintiffs’ Rule 62.1 motion or who were em‐ ployed in supervisory roles. Exhibit B expanded on Exhibit A to the original certification order by excluding those employee categories that (according to the staffing spreadsheet) had no contact with male inmates. The district court found that the modified class met the re‐ quirements of Rule 23 for the reasons stated in its indicative ruling and its initial certification order. The parties do not provide precise numbers, but they agree that, while the original class had more than 2,000 mem‐ bers, the modified class has just under 2,000 members. Ac‐ cording to the defendants, the new class has about 180 fewer members than the original class. The plaintiffs represent that the modified class comprises roughly 1,100 correctional offic‐ ers, 525 medical and mental health personnel, 150 courtroom deputies, 75 social workers or law librarians, and 100 other employees such as assistants or clerks. We granted the defendants’ Rule 23(f) petition for an in‐ terlocutory appeal of the court’s order modifying the class. 12 No. 20‐1723

II. Discussion To achieve certification, a proposed class under Rule 23(b) must meet the requirements of Rule 23(a)—numerosity, typi‐ cality, commonality, and adequacy of representation—and one of the alternatives listed in Rule 23(b). Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 811 (7th Cir. 2012). Where, as here, plaintiffs seek certification under Rule 23(b)(3), a court must find that common questions of law or fact “predominate” over individual ones and that a class ac‐ tion is “superior” to other methods of adjudicating the case. Id.; McFields v. Dart, 982 F.3d 511, 515 (7th Cir. 2020). The plaintiffs bear the burden of proving by a preponderance of the evidence that their proposed class satisfies the require‐ ments of Rule 23. Messner, 669 F.3d at 811. Rule 23 is more than “a mere pleading standard.” Wal‐ Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350 (2011). Thus, in evaluating the Rule 23 factors, a court does not take the plain‐ tiffs’ allegations at face value. Instead, the court “must go be‐ yond the pleadings and, to the extent necessary, take evidence on disputed issues that are material to certification.” Beaton v. SpeedyPC Software, 907 F.3d 1018, 1025 (7th Cir. 2018); see also Szabo v. Bridgeport Machs., Inc., 249 F.3d 672, 675 (7th Cir. 2001). This will often “entail some overlap with the merits of the plaintiff’s underlying claim. That cannot be helped.” Wal‐ Mart, 564 U.S. at 351. Still, a court’s preview of the merits must remain tethered to its Rule 23 analysis. The merits themselves are “not on the table” at this early stage. Beaton, 907 F.3d at 1025. Rule 23 does not permit courts “to engage in free‐rang‐ ing merits inquiries at the certification stage.” Amgen Inc. v. Conn. Ret. Plans & Tr. Funds, 568 U.S. 455, 466 (2013). No. 20‐1723 13

We review an order on class certification for abuse of dis‐ cretion, “leaving considerable room for the exercise of judg‐ ment unless the factual determinations are clearly erroneous or there are errors of law.” Beaton, 907 F.3d at 1025. Review “is deferential, but deferential doesn’t mean abject.” CE Design Ltd. v. King Architectural Metals, Inc., 637 F.3d 721, 723 (7th Cir. 2011) (internal quotations and citation omitted). A district court must “rigorously analyze” the requirements of Rule 23. Beaton, 907 F.3d at 1025. “The decision to certify a class or not can cause a considerable tilt in the playing fields of litigation and therefore is not one to take lightly.” Chi. Teachers Union, Local No. 1 v. Bd. of Educ. of City of Chi., 797 F.3d 426, 433 (7th Cir. 2015); accord Szabo, 249 F.3d at 675 (noting that class cer‐ tification “may induce a substantial settlement even if the [plaintiffs’] position is weak”). The defendants’ appeal centers on three of Rule 23’s re‐ quirements: commonality, typicality, and predominance. These requirements overlap, so they often rise or fall together. See, e.g., McFields, 982 F.3d at 516–19. This case is no exception. The “same basic defects” doom the class on each front. Id. at 518. First, ambient harassment does not unite the modified class because it is no longer a central issue in the case and the plaintiffs have not shown that it manifests in the same way across different parts of the jail complex. Second, and relat‐ edly, the significant variation in harassment levels across dif‐ ferent parts of the jail complex renders certain class members’ work environments materially different from those of others. Some variation among class claims is inevitable, but the vari‐ ation in this case is more than Rule 23 tolerates. We reverse on those grounds. We reject the defendants’ contention that ade‐ quacy of representation is an additional barrier to class certi‐ fication. 14 No. 20‐1723

A. Commonality We begin with commonality. Rule 23(a)(2) requires “ques‐ tions of law or fact common to the class.” One common ques‐ tion is enough, but not just any question will do. Wal‐Mart Stores, 564 U.S. at 359; Jamie S. v. Milwaukee Pub. Sch., 668 F.3d 481, 497 (7th Cir. 2012). As Wal‐Mart explains, the commonal‐ ity inquiry is easy to misinterpret, as “[a]ny competently crafted class complaint literally raises common ‘questions.’” 564 U.S. at 349 (quoting Richard A. Nagareda, Class Certifica‐ tion in the Age of Aggregate Proof, 84 N.Y.U. L. Rev. 97, 132 (2009)). Superficial common questions like whether each class member “suffered a violation of the same provision of law” do not suffice. Id. at 350; Jamie S., 668 F.3d at 497. Rather, the class claims “must depend on a common contention” that is “capable of classwide resolution—which means that determi‐ nation of its truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.” Wal‐ Mart, 564 U.S. at 350; accord 1 William B. Rubenstein, Newberg on Class Actions § 3:18 (5th ed. 2012) (the common question must “not be peripheral but important to most of the individ‐ ual class member’s claims”). “Dissimilarities within the pro‐ posed class are what have the potential to impede the gener‐ ation of common answers.” Wal‐Mart, 564 U.S. at 350 (quoting Nagareda, supra, at 132). The district court’s initial class certification order found a common question to be “whether the ambient harassment ex‐ perienced by female employees at the jail and the courthouse is sufficiently severe and pervasive to support a Title VII hos‐ tile work environment claim.” In reaching this conclusion, the court relied on Dr. Fitzgerald’s explanation of the ambient‐ harassment theory. In her expert report, Dr. Fitzgerald No. 20‐1723 15

explained that the ambient‐harassment theory focuses on the general environment of harassment in a workplace, which af‐ fects both direct victims of harassment as well as “bystanders and coworkers who observe or hear about other women being directly targeted.” Because ambient harassment “reaches be‐ yond the focal individual to affect the entire workgroup,” the ambient‐harassment theory allowed the district court to reject the defendants’ argument that there was too much variation among the individual experiences of the class members, some of whom lacked direct contact with male inmates. After the plaintiffs submitted the staffing spreadsheet, however, the court modified the class to exclude nearly 200 employees who, according to the spreadsheet, lacked direct contact with male inmates. As the court saw it, the class mem‐ bers who lacked contact with male inmates were “situated differently from the others in a material way,” and thus did not belong in the same class. Although the court recognized that its modification of the class implicated the commonality requirement, it did not reassess commonality or identify a new common question for the modified class. It also did not resolve the parties’ dispute as to the continuing significance of ambient harassment in the case, noting only that “narrow‐ ing the class is likely to significantly narrow any differences in this regard across the class.” The court’s failure to reassess commonality when modify‐ ing the class was error, given that the ambient‐harassment theory was the focal point of the commonality ruling in the original order. A district court must assure itself at all stages of the litigation that a certified class meets the requirements of Rule 23. Newberg on Class Actions § 7:37; see Phillips Petro‐ leum Co. v. Shutts, 472 U.S. 797, 812 (1985). The court properly 16 No. 20‐1723

exercised its discretion to modify the class in light of new ev‐ idence, see Fed. R. Civ. P. 23(c)(1)(C); Fonder v. Sheriff of Kanka‐ kee Cnty., 823 F.3d 1144, 1147 (7th Cir. 2016), but it was wrong to “presume[]” that the modified class complied with Rule 23, Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 160 (1982). To be sure, not every adjustment to a class requires a fresh Rule 23 evaluation. In many cases, a modified class may sat‐ isfy Rule 23 for the reasons the original class did, and a court may simply say so. Not so here. Ambient harassment was cen‐ tral to the court’s original commonality finding. Yet, in modi‐ fying the class, the court excluded nearly 200 class members, apparently because their claims depended solely on ambient harassment. The remaining class members—all of whom had direct exposure to male inmates—could (at least in theory) rely on direct harassment to prove their claims. Although the district court recognized that the new class materially differed from the previous class, it failed to identify a new common question for the class. This complicates our review: on one hand, the court seemed to acknowledge that ambient harass‐ ment was no longer central to the case. At the same time, the court did not identify a new common question for the modi‐ fied class, choosing instead to incorporate the reasoning of its earlier order—which had hinged entirely on ambient harass‐ ment. The court’s failure to reassess commonality leaves us in the dark as to the legal justification for the modified class. Is ambient harassment still the common question? Or is ambient harassment now on the sidelines? We do not know. Neither do the parties, for that matter. They posit completely different readings of the court’s orders. According to the plaintiffs, the court’s indicative ruling disavowed ambient harassment. No. 20‐1723 17

What unites the class now, they say, is direct harassment by male inmates. They maintain that ambient harassment is a “severable and independent” basis for class certification that we need not address. The defendants see things much differ‐ ently. They point out that the court’s order modifying the class incorporated the reasoning from its initial certification order, so, they say, ambient harassment is still the “glue” holding this class together. Wal‐Mart, 564 U.S. at 352. Ordinarily, we would vacate and remand for the court to supplement its reasoning. See, e.g., Red Barn Motors, Inc. v. NextGear Capital, Inc., 915 F.3d 1098, 1101 (7th Cir. 2019). But we have already remanded the case once, and we worry that remanding the case again will prompt a third appeal present‐ ing the same issues. More importantly, we believe that re‐ manding for the court to clarify the common question would ultimately be pointless because, under either party’s interpre‐ tation of the court’s rulings, commonality is lacking. We ad‐ dress the parties’ interpretations in turn. 1. Ambient Harassment We start by assuming, as the defendants do, that common‐ ality still depends on ambient harassment—which is to say, the common question remains “whether the ambient harass‐ ment experienced by female employees at the jail and the courthouse is sufficiently severe and pervasive to support a Title VII hostile work environment claim.” We hold that this is not a common question for the class. Although the plaintiffs bring a few claims, their bid for class certification focuses almost entirely on their hostile work environment claim. We accept their invitation to focus on that claim, with one caveat that we address later in the opinion. 18 No. 20‐1723

To succeed on a hostile work environment claim, “a plain‐ tiff must show that she was (1) subjected to unwelcome sexual conduct, advances, or requests; (2) because of her sex; (3) that were severe or pervasive enough to create a hostile work en‐ vironment; and (4) that there is a basis for employer liability.” EEOC v. Costco Wholesale Corp., 903 F.3d 618, 625 (7th Cir. 2018) (internal quotations and citation omitted). The third fac‐ tor “requires the unwelcome conduct to be severe or perva‐ sive from both a subjective and an objective point of view.” Id. (citing Faragher v. City of Boca Raton, 524 U.S. 775, 787 (1998)). “To be severe or pervasive enough to create a hostile work environment, conduct must be ‘extreme.’” Id. (quoting Fara‐ gher, 524 U.S. at 788). “This is not, and by its nature cannot be, a mathematically precise test.” Harris v. Forklift Sys., Inc., 510 U.S. 17, 22 (1993). Whether a work environment is hostile de‐ pends on “all the circumstances,” including “the frequency of the discriminatory conduct; its severity; whether it is physi‐ cally threatening or humiliating, or a mere offensive utter‐ ance; and whether it unreasonably interferes with an em‐ ployee’s work performance.” Id. at 23. The district court’s commonality analysis focused on the third element of a hostile work environment claim, which asks whether the harassment was severe or pervasive enough to create a hostile work environment. Costco Wholesale Corp., 903 F.3d at 625. The court recognized that not every class member had experienced the same level of harassment, but it thought they shared at least one thing in common: they all worked in an environment plagued by “ambient harass‐ ment.” Relying on Dr. Fitzgerald’s expert report, the court de‐ fined ambient harassment as “the experience of working in an environment highly permeated with sexually offensive and degrading behavior, that is, a highly sexualized atmosphere No. 20‐1723 19

in which crude and offensive sexual behavior is common and employees see that it is normative, whether specifically di‐ rected at them or not.” As an aside, the district court’s reliance on Dr. Fitzgerald’s expert opinions was improper. The court had previously ex‐ cluded her case‐specific opinions as unreliable. Yet, when an‐ alyzing commonality, the court concluded without explana‐ tion that Dr. Fitzgerald’s “opinions regarding the social‐scien‐ tific background of workplace harassment” were admissible. We do not follow that reasoning. We have held that “[w]hen an expert’s report or testimony is ‘critical to class certifica‐ tion,’” the district court “must make a conclusive ruling on any challenge to that expert’s qualifications or submissions before it may rule on a motion for class certification.” Messner, 669 F.3d at 812 (quoting Am. Honda Motor Co. v. Allen, 600 F.3d 813, 815 (7th Cir. 2010) (per curiam)); see also Wal‐Mart, 564 U.S. at 354 (“doubt[ing]” that “Daubert did not apply to expert testimony at the certification stage of class‐action proceed‐ ings”). We have also held that Daubert and Rule 702 apply “to social science experts,” just as they apply “to experts in the hard sciences.” Tyus v. Urban Search Mgmt., 102 F.3d 256, 263 (7th Cir. 1996); see also Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579 (1993); Fed. R. Evid. 702. The district court’s Rule 23 analysis relied heavily on Dr. Fitzgerald’s general opinions about the ambient‐harassment theory. Before relying on those opinions—which the defendants had moved to exclude, along with the rest of Dr. Fitzgerald’s report—the court should have ensured that they lived up to the standards of Daubert and Rule 702. Messner, 669 F.3d at 812. The only authority that the district court cited for its deci‐ sion to admit Dr. Fitzgerald’s background opinions was Van 20 No. 20‐1723

v. Ford Motor Co., No. 14‐CV‐8708, 2018 WL 4635649 (N.D. Ill. Sept. 27, 2018), which denied Ford’s motion to exclude Dr. Fitzgerald’s expert opinions as premature. Id. at *13. But the cited portion of Van expressed doubts about the admissibility of Dr. Fitzgerald’s opinions and disclaimed any reliance on them. Id. at *12–13. Moreover, a court does not fulfill its obli‐ gation “to act as a vigorous gatekeeper to ensure the reliabil‐ ity of expert testimony” simply by citing another case that considered similar testimony. Robinson v. Davol Inc., 913 F.3d 690, 696 (7th Cir. 2019). The “rigorous analysis” requirement “applies to expert testimony critical to proving class certifica‐ tion requirements.” In re Blood Reagents Antitrust Litig., 783 F.3d 183, 187 (3d Cir. 2015); see Comcast Corp. v. Behrend, 569 U.S. 27, 35 (2013). Expert issues aside, we readily acknowledge that “ambi‐ ent harassment,” to the extent it refers to indirect or secondhand harassment, is a familiar concept in employment‐ discrimination law. See generally Kristin H. Berger Parker, Comment, Ambient Harassment Under Title VII: Reconsidering the Workplace Environment, 102 Nw. U. L. Rev. 945 (2008). We have held that harassing comments directed at others, while they “carry less weight” than remarks directed at the plaintiff, may still be relevant to a hostile work environment claim. Johnson v. Advocate Health & Hosps. Corp., 892 F.3d 887, 902 (7th Cir. 2018). We have also made clear that not all indirect har‐ assment is created equal. For example, a discriminatory com‐ ment made in the plaintiff’s presence might be actionable if it applies equally to the plaintiff, while a mere “rumor” might not be. Yuknis v. First Student, Inc., 481 F.3d 552, 554–55 (7th Cir. 2007); accord Johnson, 892 F.3d at 902–03. We have even suggested that, on extreme facts, indirect harassment stand‐ ing alone could create a hostile work environment. Dandy v. No. 20‐1723 21

United Parcel Serv., Inc., 388 F.3d 263, 272 (7th Cir. 2004) (opin‐ ing that the “[r]epeated use of [racial epithets] in the work en‐ vironment … may create an objectively hostile work environ‐ ment, even if they are heard secondhand.”). We have never held, however, that ambient harassment can serve as a standalone basis for a hostile work environment claim in the jail or prison context, and the defendants urge us to hold that it cannot. They acknowledge, as they must, that Title VII covers employees who suffer harassment from in‐ mates. See, e.g., Erickson v. Wis. Dep’t of Corr., 469 F.3d 600 (7th Cir. 2006); Beckford v. Dep’t of Corr., 605 F.3d 951, 958 (11th Cir. 2010) (collecting cases). But, they maintain, hostile work envi‐ ronment claims depend on context, and jails are not “typical work environments.” Vajdl v. Mesabi Acad. of KidsPeace, Inc., 484 F.3d 546, 550 (8th Cir. 2007). Jails “by definition” house people who have been detained for “breach[ing] prevailing societal norms.” Slayton v. Ohio Dep’t of Youth Servs., 206 F.3d 669, 677 (6th Cir. 2000). Not only that, but jail administrators have limited tools at their disposal to control inmate behavior. Beckford, 605 F.3d at 959. Many of the incentive structures that allow employers to control employee behavior in the typical workplace are not available to jail administrators. See id. There are both “practical and constitutional limits on what [jails] can do to protect staff.” Id. Unlike employees, inmates cannot be fired. And while other sanctions are available, including dis‐ cipline and criminal charges, these sanctions (the defendants tell us) do not necessarily deter individuals who are already confined and often facing prolonged imprisonment on exist‐ ing charges. What is more, some of the most effective sanc‐ tions—e.g., separate housing—could raise Eighth Amend‐ ment concerns. Id.; see Hutto v. Finney, 437 U.S. 678, 685 (1978). 22 No. 20‐1723

In short, “some harassment by inmates cannot be reason‐ ably avoided.” Beckford, 605 F.3d at 959. For that reason, the defendants contend that ambient harassment cannot support a hostile work environment claim in the penal context. Like the district court, we do not entertain that argument because it goes straight to the merits. See Schleicher v. Wendt, 618 F.3d 679, 687 (7th Cir. 2010) (“The chance, even the certainty, that a class will lose on the merits does not prevent its certifica‐ tion.”). We do hold, however, that the district court erred in‐ sofar as it found that ambient harassment supplied a common question for the modified class. The plaintiffs’ theory for refining the class was that every member of the new class had likely experienced—or, at the very least, had been positioned to experience—direct harass‐ ment, given that their job descriptions involved direct contact with male inmates. If that is so, then the class members, by and large, have limited use for ambient harassment. Recall that ambient harassment, while relevant to a hostile work en‐ vironment claim, “carr[ies] less weight” than direct harass‐ ment. Johnson, 892 F.3d at 902; accord Ellis v. CCA of Tenn. LLC, 650 F.3d 640, 647 n.2 (7th Cir. 2011) (“In the context of a hostile work environment claim, secondhand harassment is less se‐ vere than firsthand harassment.”). Indeed, we have repeat‐ edly rejected hostile work environment claims that rest pri‐ marily on secondhand harassment. See, e.g., Yuknis, 481 F.3d at 556; Ezell v. Potter, 400 F.3d 1041, 1048 (7th Cir. 2005); Dandy, 388 F.3d at 272; Patt v. Family Health Sys., Inc., 280 F.3d 749, 754 (7th Cir. 2002); Johnson v. City of Fort Wayne, Ind., 91 F.3d 922, 938 (7th Cir. 1996). Our precedent makes clear, then, that generally a hostile work environment claim predicated on ambient harassment No. 20‐1723 23

alone would be very difficult to mount. For that reason, the named plaintiffs and the members of the modified class have a clear incentive to prioritize evidence of direct harassment over evidence of ambient harassment. Evidence of ambient harassment may still be relevant to the third element of the class claims, but it is not central. In other words, the answer to the district court’s common question—was the ambient harassment sufficiently severe or pervasive to create an objec‐ tively hostile work environment?—will not “drive the resolu‐ tion of the litigation” because it will not resolve the issue that the third element of the class claims now depends on: whether the direct harassment that the class members experienced was sufficiently severe or pervasive to create an objectively hostile work environment. Wal‐Mart, 564 U.S. at 350 (quoting Na‐ gareda, supra, at 132). The district court’s continued focus on ambient harass‐ ment when modifying the class might be understandable if the plaintiffs had told the district court that ambient harass‐ ment was still the centerpiece of the case. But they said essen‐ tially the opposite. When asking the court to refine the class, the plaintiffs argued that the court could certify the modified class “without reference to ambient, or ‘secondhand’ harass‐ ment,” signaling that it had faded into the background and was no longer the “glue” holding the class together. Wal‐Mart, 564 U.S. at 352. All of this is to say that ambient harassment is now, at most, a “peripheral” issue for most class members. Newberg on Class Actions § 3:18. Resolving it one way or another will thus do little to advance the class claims. As such, it cannot be the common question for the modified class. See Wal‐Mart, 564 U.S. at 350; McFields, 982 F.3d at 517 (holding that a question 24 No. 20‐1723

was not common because it was “relevant to just one small part of the analysis” and “leaves us far from resolving the lit‐ igation on a classwide basis”). Even apart from its diminished importance to the modi‐ fied class, the ambient‐harassment theory is a problematic ba‐ sis for commonality because it overlooks meaningful distinc‐ tions among the class members’ individual experiences. As discussed in more detail below, class members’ experiences can vary dramatically depending on where they work. The district court, however, treated ambient harassment at the jail complex as if it were one homogenous phenomenon that af‐ fects every class member in the same way. This conception of ambient harassment—which derives from Dr. Fitzgerald’s untested opinions about the general nature of ambient harass‐ ment—cannot be reconciled with our case law. We have made clear that ambient harassment, like direct harassment, comes in materially different forms and affects different workers dif‐ ferently. Yuknis, 481 F.3d at 554–55; see Johnson, 892 F.3d at 902–03. Even if ambient harassment pervades the jail com‐ plex, the plaintiffs have not demonstrated that it manifests in the same way across all parts of the jail, such that it could be a common question for the class. Cf. Wal‐Mart, 564 U.S. at 353– 54 (holding expert’s “social framework” analysis of Wal‐ Mart’s “corporate culture” could not “bridg[e] the gap” for commonality because expert could not say with any specific‐ ity how often stereotypes impacted employment decisions). On this basis too, ambient harassment fails as a common ques‐ tion. 2. Direct Harassment As mentioned, the plaintiffs dispute the premise that the district court relied on ambient harassment when modifying No. 20‐1723 25

the class. According to the plaintiffs, the district court’s indic‐ ative ruling disavowed ambient harassment and replaced it with direct harassment. The plaintiffs articulate the new com‐ mon question as: “Whether the severity or pervasiveness of detainee sexual harassment has rendered the working envi‐ ronment objectively hostile.” In its indicative ruling, the district court did not hold that it was no longer relying on ambient harassment. But it did modify the class to exclude class members who lacked direct contact with male inmates. And we acknowledge that the court’s reasoning for modifying the class is arguably suscep‐ tible to the plaintiffs’ interpretation. We therefore address the plaintiffs’ argument. Although the plaintiffs distance themselves from ambient harassment, their proposed common question has problems of its own. Namely, it begs the question—what “sexual har‐ assment” are they referring to? It would be one thing if all class members had experienced the same harassment. But that is not what the evidence shows. Rather, the evidence re‐ veals that class members’ experiences can vary significantly depending on where they work. Some class members work (or have worked) in Divisions 8, 9, and 10, where the vast ma‐ jority of sexual harassment has been reported. Others work in Divisions 2 or 6, where reports of sexual harassment have been far less common. Still others work in non‐residential parts of the jail or at the courthouse where, again, reports have been far less common. Even if every class member has en‐ dured sexual harassment, that does not mean that they have endured the same harassment, such that they could rely on the same evidence to prove their claims. See Tyson Foods, Inc. v. Bouaphakeo, 136 S. Ct. 1036, 1045 (2016). 26 No. 20‐1723

Hostile work environment claims are fact intensive. They turn on the frequency, severity, character, and effect of the harassment. Harris, 510 U.S. at 23. Here, these are “worker‐ specific” inquiries because they depend on a class member’s unique experience—which correlates to where she works. Bol‐ den v. Walsh Const. Co., 688 F.3d 893, 896 (7th Cir. 2012). To be sure, some—maybe many—class members will have had comparable experiences. But the plaintiffs have not proven that for the entire class. The jail complex is massive. It fills dozens of buildings that sprawl across eight city blocks. There are many different work assignments, both within the jail and at the connected court‐ house. Unfortunately, sexual harassment occurs throughout the jail, but the evidence shows that it is heavily concentrated within a few residential divisions where a fraction of class members and named plaintiffs work. The plaintiffs’ own evi‐ dence suggests that employees who have worked in Divisions 8, 9, or 10 likely have endured far more frequent harassment than employees working elsewhere in the jail or at the court‐ house. This does not mean that one class member (e.g., a Di‐ vision 10 employee) has experienced a hostile work environ‐ ment while another (e.g., a law librarian) has not. Rather, the questions of whether these two employees have endured ob‐ jectively severe or pervasive harassment “must be answered separately” because they depend on “individualized ques‐ tions of fact and law,” whose “answers are unique to each [class member’s] particular situation.” Jamie S., 668 F.3d at 498. The plaintiffs acknowledge that the frequency of sexual harassment varies by location. In response, they point to evi‐ dence showing that employees often change jobs, and both No. 20‐1723 27

employees and inmates traverse the jail. The evidence sup‐ ports those assertions, and indeed the district court found based on this evidence that “class members frequently occupy the same work environments, even if only briefly.” Still, it is a leap too far to conclude from this evidence that all class mem‐ bers share essentially the same work environment. The plain‐ tiffs have shown that inmates and class members often pass through the tunnels connecting the jail to the courthouse. The plaintiffs have not shown, however, that employees who work in the courthouse or other parts of the jail frequently pass through—much less work in—Divisions 8, 9, and 10, where most incidents occur. Nor have they shown that most class members worked in Divisions 8, 9, or 10 during the rel‐ evant period. Indeed, if the named plaintiffs are representa‐ tive of the class, it appears that employees who work in the courthouse generally do not transfer to positions at the jail. Without a stronger evidentiary showing, the plaintiffs cannot demonstrate that class members working in disparate parts of the massive jail complex have experienced the same work en‐ vironment. For these reasons, the class does not share a common ques‐ tion as to whether the severity or pervasiveness of sexual har‐ assment by male inmates has created an objectively hostile work environment. Importantly, we do not suggest that none of the class members here could band together to form a smaller class or classes. Conceivably, a smaller class comprising a subset of class members who have had comparable experiences could form a coherent class. Along those lines, the defendants’ counsel conceded at oral argument that a class of employees who work in Divisions 8, 9, and 10 might satisfy the 28 No. 20‐1723

requirements of Rule 23. We will leave that issue to the district court’s discretion. But the current class is overbroad, and the plaintiffs have failed to establish commonality. B. Typicality Beyond commonality, Rule 23 requires that “the claims or defenses of the representative parties are typical of the claims or defenses of the class.” Fed. R. Civ. P. 23(a)(3). As a general matter, “[a] plaintiff’s claim is typical if it arises from the same event or practice or course of conduct that gives rise to the claims of other class members and his or her claims are based on the same legal theory.” Keele v. Wexler, 149 F.3d 589, 595 (7th Cir. 1998) (internal quotations and citation omitted). Typ‐ icality “is meant to ensure that the named representative’s claims have the same essential characteristics as the claims of the class at large.” Lacy v. Cook Cnty., 897 F.3d 847, 866 (7th Cir. 2018) (internal quotations and citation omitted). The logic behind the typicality requirement “is that a class representa‐ tive will adequately pursue her own claims, and if those claims are ‘typical’ of those of the rest of the class, then her pursuit of her own interest will necessarily benefit the class as well.” Newberg on Class Actions § 3:28. Typicality differs from commonality, but the two require‐ ments are “closely related.” Rosario v. Livaditis, 963 F.2d 1013, 1018 (7th Cir. 1992). “Both serve as guideposts for determin‐ ing whether under the particular circumstances maintenance of a class action is economical and whether the named plain‐ tiff’s claim and the class claims are so interrelated that the in‐ terests of the class members will be fairly and adequately pro‐ tected in their absence.” Wal‐Mart, 564 U.S. at 349 n.5 (quoting Falcon, 457 U.S. at 157 n.13). The distinction between common‐ ality and typicality is that “the commonality inquiry focuses No. 20‐1723 29

on what characteristics are shared among the whole class while the typicality inquiry focuses on the desired attributes of the class representative.” Newberg on Class Actions § 3:31; accord Wal‐Mart, 564 U.S. at 349 n.5. As with commonality, the district court failed to reassess typicality when modifying the class. Indeed, the court’s indic‐ ative ruling did not mention typicality. We thus look to the reasoning of the court’s original class certification order, which the court adopted in modifying the class. The court found that the named plaintiffs’ claims were typical of the original class claims because the named plaintiffs and the class members relied on the same legal theory—even if some of the class members would have to resort to ambient harass‐ ment to prove their claims. We cannot accept this reasoning because, again, it elides material differences between direct harassment and ambient harassment. According to the complaint, the named plaintiffs are each victims of frequent direct harassment. If that is so, then, for reasons we have explained, the named plaintiffs have limited use for ambient harassment. Typicality ensures that class representatives have an “incentive to litigate vigor‐ ously” the claims of the absent class members. Muro v. Target Corp., 580 F.3d 485, 493 (7th Cir. 2009). Even if the named plaintiffs were to rely on ambient harassment to buttress their evidence of direct harassment, they would have no reason to place it front and center. As such, the named plaintiffs are poor proxies for any class members whose claims rise or fall on ambient harassment. See id. The same would be true if the situation were reversed and the named plaintiffs had materi‐ ally weaker evidence than some of the class members. See 30 No. 20‐1723

Fonder, 823 F.3d at 1146; see also Van v. Ford Motor Co., 332 F.R.D. 249, 282–83 (N.D. Ill. 2019). The district court apparently thought that the named plaintiffs’ claims were typical of the class because ambient harassment differs from direct harassment only in degree, and not in kind. Cf. Yuknis, 481 F.3d at 554 (opining that the term “second‐hand harassment” performs “no analytic func‐ tion and is better avoided” because it “tends to obscure” im‐ material differences between certain types of harassing con‐ duct). Even if that is true, we fail to see how it compels the conclusion that the named plaintiffs’ claims are typical of the class claims. It is not enough for typicality that the named plaintiffs and the class members rely on the same legal theory. See Keele, 149 F.3d at 595. The named plaintiffs’ claims must also share the same essential characteristics as the class claims. Lacy, 897 F.3d at 866. Whether ambient harassment differs from direct harassment in degree or in kind, the differ‐ ence is essential. The reasoning of the district court’s indicative ruling called into question whether typicality still hinged on ambi‐ ent harassment. But the district court did not reanalyze typi‐ cality for the modified class, or otherwise suggest that its ear‐ lier analysis did not fully apply. We will not speculate as to how the court might have analyzed typicality with ambient harassment out of the picture. Spano v. The Boeing Co., 633 F.3d 574, 577–78 (7th Cir. 2011) (“[O]ur task is to review only the class certification orders issued by the district court in these two cases. We are not here to review any or all hypothetical orders that the court might have crafted.”). We note, however, that much of what we have said about commonality would No. 20‐1723 31

apply to typicality, even without regard to ambient harass‐ ment. C. Predominance We turn next to predominance. Rule 23(b)(3) requires that “questions of law or fact common to class members predom‐ inate over any questions affecting only individual members.” This requirement builds on commonality; whereas Rule 23(a)(2) requires the existence of a common question, Rule 23(b)(3) requires the common question(s) to “predominate” over the individual ones. As a result, predominance “is far more demanding” than commonality. Amchem Prod., Inc. v. Windsor, 521 U.S. 591, 624 (1997). “There is no mathematical or mechanical test for evaluating predominance.” Messner, 669 F.3d at 814. Efficiency is the animating principle. Chi. Teachers Union, 797 F.3d at 444. To gauge whether a class ac‐ tion would be more efficient than individual suits, “[t]he pre‐ dominance inquiry ‘asks whether the common, aggregation‐ enabling, issues in the case are more prevalent or important than the non‐common, aggregation‐defeating, individual is‐ sues.’” Tyson Foods, 136 S. Ct. at 1045 (quoting Newberg on Class Actions § 4:49). The district court’s original class certification order identi‐ fied three common questions that supposedly predominated: (1) “whether all of the putative class members experienced an objectively hostile work environment based on the ambient harassment at the jail and courthouse;” (2) “whether the de‐ tainees’ harassment occurred because of sex;” and (3) “whether there is a basis for employer liability based on the defendants’ failure to adopt reasonable policies to combat the harassment.” As discussed above, the first is not a common question. And the district court acknowledged that the second 32 No. 20‐1723

question was minor. So we focus on employer liability, asking first whether it qualifies as a common question, and if so, whether it predominates. As with typicality, the court’s indic‐ ative ruling did not revisit its earlier determination on pre‐ dominance. As always, an evaluation of predominance begins with the elements of the underlying claim. Messner, 669 F.3d at 815. An employer is liable in a hostile work environment case if (1) a supervisor participated in the harassment (giving rise to strict liability) or (2) the employer was negligent in discovering or remedying the harassment by a coworker or a third party. Montgomery v. Am. Airlines, Inc., 626 F.3d 382, 390 (7th Cir. 2010). The harassers in this case are inmates, not supervisors, so the plaintiffs proceed on a theory of negligence. See Erick‐ son, 469 F.3d at 609; see also Dunn v. Washington Cnty. Hosp., 429 F.3d 689, 691 (7th Cir. 2005). To prove negligence, the plaintiffs must show that (1) the defendants knew or should have known about the harass‐ ment and (2) failed to take reasonable steps to prevent it. Berry v. Delta Airlines, Inc., 260 F.3d 803, 811 (7th Cir. 2001). Preven‐ tative measures “can involve proactive steps such as con‐ structing a reporting system for instances of sexual harass‐ ment, training employees about sexual harassment risks and what can be done to ameliorate them …, and taking reasona‐ ble steps to prevent harassment once informed of a reasonable probability that it will occur.” Erickson, 469 F.3d at 606. The reasonableness of an employer’s response to harass‐ ment is a fact‐bound inquiry. The response “must be reason‐ ably calculated to prevent further harassment under the par‐ ticular facts and circumstances of the case at the time the alle‐ gations are made.” Berry, 260 F.3d at 811 (quoting McKenzie v. No. 20‐1723 33

Ill. Dep’t of Transp., 92 F.3d 473, 480 (7th Cir. 1996)). “We are not to focus solely upon whether the remedial activity ulti‐ mately succeeded, but instead should determine whether the employer’s total response was reasonable under the circum‐ stances as then existed.” Id. (quoting McKenzie, 92 F.3d at 480). The relevant circumstances include “the gravity of the harass‐ ment.” Baskerville v. Culligan Int’l Co., 50 F.3d 428, 432 (7th Cir. 1995); accord Erickson, 469 F.3d at 604 (“The greater the poten‐ tial injury to the employee, the greater care the employer must take.”). In opposing class certification, the defendants submitted evidence of various preventative measures that they have taken to curb sexual harassment by inmates. These measures include: policies prohibiting sexual harassment and establish‐ ing reporting procedures; a disciplinary system with a range of sanctions; educating inmates and staff about policies and the consequences of violations; physical restrictions such as handcuffing, waist restraints, and unique jumpsuits for known offenders; a special housing unit for known offenders; and modifications to reduce visibility in certain parts of the jail. In addition, the Sheriff’s Office sometimes refers sexual harassment complaints to the State’s Attorney Office for po‐ tential criminal prosecution. The adequacy of these preventa‐ tive measures is not at issue in this appeal, but they illustrate some of the tools available to curb harassment. The plaintiffs tell us that the reasonableness of these pre‐ ventative measures is a common question because the Sher‐ iff’s Officer has one set of policies to control sexual miscon‐ duct at the jail complex and it enforces those policies through a centralized, hierarchical management structure. As the plaintiffs see it, the defendants’ policies are either reasonable 34 No. 20‐1723

or they are not, and their reasonableness is a common ques‐ tion. The district court apparently agreed with that reasoning. It found that the defendants’ common policies distinguished this case from cases in which the delegation of discretionary authority to local managers precluded commonality. See Bol‐ den, 688 F.3d at 896; Wal‐Mart, 564 U.S. at 352–54. For their part, the defendants submit that their liability for failing to prevent inmate harassment is not a common ques‐ tion because the same preventative measures might qualify as reasonable for some class members but not others, given the class members’ differing job assignments and exposure to harassment. (The plaintiffs say the defendants have forfeited this argument. We disagree; the defendants point out that they disputed employer liability as a common question in their opposition to class certification.) We grant that, insofar as control over the jail complex is centralized and its policies are uniformly applied, this case differs from Bolden and Wal‐Mart, where the discretion of local managers derailed commonality. Even so, we do not see how this difference translates to a finding that employer liability is a common question that predominates. The defendants’ poli‐ cies may be uniform throughout the jail, but the reasonable‐ ness of those policies (and any other preventative measures) still depends on the specific circumstances of the plaintiff(s) or class member(s) challenging the policies. See McFields, 982 F.3d at 517. The policies could be reasonable for one class member (e.g., a law librarian), but unreasonable for another (e.g., a Division 10 employee). Or, they could be unreasonable for both; but if they are, they are unreasonable for different legal and factual reasons. Jamie S., 668 F.3d at 498. To reiterate, the jail complex is not a homogenous workplace. In some No. 20‐1723 35

parts, the harassment is worse—much worse—than in others. The adequacy of a particular response is evaluated by refer‐ ence to the underlying problem. Whether preventative measures are reasonable for a given employee depends on the “gravity” of harassment that she endures, see Baskerville, 50 F.3d at 432, and the gravity of harassment, in turn, depends on where she works. For these reasons, employer liability is not a common question for the current class. As before, we do not foreclose the possibility that smaller subsets of the class could share a common question as to employer liability. Because employer liability is not a common question, we have no occasion to ask whether it predominates over individual questions. Perhaps sensing that the certified class stands on shaky ground, the plaintiffs propose a new common question on ap‐ peal: whether the defendants’ tolerance of the inmates’ sexual harassment amounts to a pattern or practice of discrimina‐ tion. See Int’l Bhd. of Teamsters v. United States, 431 U.S. 324, 334–36 (1977). The plaintiffs did not raise this common ques‐ tion below, so they have waived appellate review of it. Puffer v. Allstate Ins. Co., 675 F.3d 709, 718 (7th Cir. 2012). We close out our discussion of employer liability by flag‐ ging an issue that the plaintiffs have largely bypassed: the dif‐ ference between their hostile work environment claims, which we have thus far focused on, and their state‐law and equal protection claims, which we have not yet addressed. The district court accepted the plaintiffs’ invitation to treat all three claims as parallel, so it did not separately analyze the state and constitutional claims. See Alamo v. Bliss, 864 F.3d 541, 550 n.16 (7th Cir. 2017). That is perfectly acceptable for the state‐law claims, given that Illinois courts “look to cases 36 No. 20‐1723

concerning alleged violations of federal civil rights statutes” when interpreting the Illinois Civil Rights Act. Cent. Austin Neighborhood Ass’n v. City of Chi., 1 N.E.3d 976, 980 (Ill. App. Ct. 2013). But the plaintiffs’ equal protection claim is not par‐ allel to their hostile work environment claim: the equal pro‐ tection claim requires proof that the defendants “were moti‐ vated by a discriminatory purpose,” Chavez v. Ill. State Police, 251 F.3d 612, 635–36 (7th Cir. 2001), whereas negligence is enough for the hostile work environment claim, Montgomery, 626 F.3d at 390; see also Nabozny v. Podlesny, 92 F.3d 446, 453– 54 (7th Cir. 1996); Washington v. Davis, 426 U.S. 229, 239–42 (1976). The plaintiffs may choose to rely on the same evidence to prove both claims, but the claims’ legal elements do not perfectly overlap. D. Adequacy of Representation The final Rule 23 requirement that the defendants chal‐ lenge is adequacy of representation. Rule 23(a)(4) requires that “the representative parties will fairly and adequately pro‐ tect the interests of the class.” This inquiry “serves to uncover conflicts of interest between named parties and the class they seek to represent.” Amchem Prod., 521 U.S. at 625. It also screens for conflicts of interest among class members because the same representative parties cannot adequately represent class members with divergent interests. Johnson v. Meriter Health Servs. Employee Ret. Plan, 702 F.3d 364, 372 (7th Cir. 2012). We have made clear, however, that “the mere possibil‐ ity that a trivial level of intra‐class conflict may materialize as the litigation progresses” does not prevent class certification. Abbott v. Lockheed Martin Corp., 725 F.3d 803, 813 (7th Cir. 2013). “If and when [potential conflicts] become real, the dis‐ trict court can certify subclasses with separate representation No. 20‐1723 37

of each.” Kohen v. Pac. Inv. Mgmt. Co. LLC, 571 F.3d 672, 680 (7th Cir. 2009) (citing Fed. R. Civ. P. 23(c)(5)). The defendants maintain that the modified class fails the adequacy‐of‐representation requirement because it includes both civilian employees, who have a duty to report miscon‐ duct, and sworn officers, who have a duty to act on reported misconduct. They posit that sworn officers should be ex‐ cluded from the class for the same reason the district court excluded supervisors: because their own misconduct might become an issue in the case (if, for example, the evidence showed that one of them failed to follow through on a report of misconduct). We reject the plaintiffs’ preliminary accusation that the de‐ fendants have forfeited this argument. In opposing class cer‐ tification, the defendants argued that “conflicts of interest amongst putative class members who are supervisory and nonsupervisory staff, and between sworn and civilian class mem‐ bers, render the class representatives inadequate.” (Emphasis added.) The plaintiffs also responded to the argument in their reply brief below. The district court did not address the argu‐ ment, but the defendants made it. Regardless, the defendants’ argument is unconvincing. As the plaintiffs point out, the defendants offer almost no evi‐ dence that the class includes sworn female officers who failed to report misconduct. They supplied no such evidence in the district court. (Perhaps that is why the district court passed over the argument.) On appeal, they muster only two exam‐ ples. On remand, if the defendants present evidence of a gen‐ uine conflict, the district court can explore the need for sub‐ classes. Kohen, 571 F.3d at 680. For now, the defendants’ argu‐ ment is premature. 38 No. 20‐1723

III. Conclusion For these reasons, we REVERSE the district court’s order certifying the modified class and REMAND for further pro‐ ceedings.