Opinion
23-md-03084-CRB (LJC)
07-09-2024
IN RE UBER TECHNOLOGIES, INC., PASSENGER SEXUAL ASSAULT LITIGATION This Relates To ALL ACTIONS
ORDER GRANTING IN PART AND DENYING IN PART JOINT DISCOVERY LETTER BRIEF REGARDING DISCOVERY RELATED TO SAFETY DATA AND STATISTICS
REDACTED
On June 13, 2024, Plaintiffs filed an administrative motion for the removal and replacement of the Joint Discovery Letter Brief at Dkt. No. 592 because “under certain conditions, a portion of the content that was intended to be redacted can be viewed.” Dkt. No. 624. Uber consented to the requested relief. Id. The Court granted the motion, locked Dkt. No. 592 from public view, and on June 21, 2024, Plaintiffs filed a corrected Joint Discovery Letter Brief for the record. Dkt. No. 651. The Court will cite to the corrected Joint Discovery Letter Brief in the rest of this Order. However, the Order will cite to the original exhibits attached to the Joint Discovery Letter Brief, which Plaintiffs did not request to remove and refile.
LISA J. CISNEROS UNITED STATES MAGISTRATE JUDGE
Pending before the Court is a Joint Discovery Letter Brief where Plaintiffs request documents and data related to Defendants Uber Technologies, Inc., Raiser, LLC, and Raiser-CA, LLC's (collectively, Uber) 2017-2018 and 2019-2020 U.S. Safety Reports (Safety Reports). Dkt. No. 651. The Court held a hearing on June 13, 2024. Dkt. No. 626. For the reasons detailed below, the Court GRANTS IN PART and DENIES IN PART Plaintiffs' request for an order to compel discovery from Uber as to data and statistics underlying the Safety Reports as well as related documents.
I. BACKGROUND
A. Uber's U.S. Safety Reports
During the period covered by Uber's Safety Reports, the company facilitated, on average, approximately 3 million trips a day. Dkt. Nos. 592-1 at 11, 592-2 at 15. In 2019, Uber issued its first Safety Report to “share[] details on Uber's safety progress, processes, and data related to reports of the most critical safety incidents on our platform.” Dkt. No. 592-1 at 11. This report drew on data from January 1, 2017 through December 31, 2018. Id. at 34. Uber published its second, similar Safety Report in 2022. See Dkt. No. 592-2 at 10. This report covered data from January 1, 2019 through December 31, 2020. Id. at 39. The Safety Reports focused on three categories of critical safety incidents: (1) motor vehicle fatalities; (2) fatal physical assault; and (3) sexual assault. Dkt. Nos. 592-1 at 15-16, 592-2 at 15. Both reports describe their methodologies and Uber's processes to collect and respond to safety incident reports.
Unless specified otherwise, the Court refers to the page number located in the header generated by the Court's e-filing system.
As to the third category of critical safety incidents, Uber adopted very specific definitions of sexual assault and sexual misconduct in both Safety Reports. Uber defines “sexual assault” as “any physical or attempted physical contact that is reported to be sexual in nature and without the consent of the user,” whereas “sexual misconduct” is “non-physical conduct . . . of a sexual nature that happens without consent or has the effect of threatening or intimidating a user against whom such conduct is directed.” Dkt. Nos. 592-1 at 5, 592-2 at 5. Sexual misconduct, as defined in the reports, can range from staring or leering to a verbal threat of sexual assault. Id.
To prepare the Safety Reports, Uber reviewed more than 800,000 user reports over the 2017-2020 time period. Dkt. Nos. 592-1 at 47 (“The [audit] team reviewed approximately 500,000 user reports, representing a range of safety- and non-safety related consumer issues to ensure that all necessary information was documented and all incident reports were categorized accurately and comprehensively”), 592-2 at 46 (“[T]he [audit] team reviewed over 350,000 user reports to ensure that all necessary information was documented and that the incident reports were categorized accurately and comprehensively.”). Uber collects historical safety incident reports through various channels, including from post-trip in-app support, on-trip in-app reporting, social media mentions (Twitter, Facebook, etc.), and law enforcement. Dkt. Nos. 592-1 at 35, 592-2 at 30. Using key words and phrases, in addition to an advanced natural language processing technology, Uber sorts through user feedback to identify reports that may indicate safety concerns. Dkt. Nos. 591-1 at 35, 592-2 at 40.
Frontline support agents for Uber are responsible for handling and responding to reported safety issues and taking action with respect to user accounts as necessary. Dkt. Nos. 592-1 at 5, 592-2 at 5. As explained in the first Safety Report, safety support agents, who sit on multiple teams within Uber's U.S. Incident Response Teams (IRT), receive training to review, document, and recommend appropriate action. Dkt. No. 592-1 at 29. Support agents are responsible for providing an initial classification of the incident using Uber's Safety Taxonomy, which is simply a set of categories that Uber relies on to classify and prioritize safety incidents, apply action on individual reports, and inform Uber's efforts to prevent future incidents. Id. at 35; see also Dkt. No. 592-2 at 30-31 (IRT safety support agents receive training on Incident Classification and use of the Sexual Misconduct and Violence Taxonomy, discussed in greater detail below). In addition, Uber created a specialized team within IRT in 2017 to provide further support in connection with reports related to the most serious and urgent of incidents, including sexual assault. Dkt. No. 592-1 at 29. This specialized team gathers data pertaining to an incident report, including GPS information, timestamps, photos/videos submitted, and in-app communications, and they may speak with all parties involved, including reporting parties, potential victims, and accused parties. Id. Uber relies on “hundreds of people” to advance its safety efforts, including these support agents who receive special training. Id. at 21.
In its Safety Reports, Uber did not report data trends based on the incident reports as categorized by the frontline support agents. Dkt. Nos. 592-1 at 46, 592-2 at 17, 47. Instead, Uber “created a specialized [audit team] dedicated to re-classifying safety incident reports.” Dkt. Nos. 592-1 at 46; see also Dkt. 592-2 at 46. The audit process entailed three phases with the following objectives: (1) “Ensure all relevant safety incident reports were audited with the necessary data documented”; (2) “Audit to a high standard of quality”; and (3) “Update [Uber's] historical data with the most accurate classification, addressing any discrepancies in auditor opinion.”. Dkt. No. 592-1 at 46; see also Dkt. No. 592-2 at 46. Uber reasoned that frontline support agents' primary responsibility involves providing support to reporting parties and collecting user statements and information related to potential safety incidents, not precise data classification. Dkt. No. 592-1 at 46. The Safety Reports explained the need for quality data, and the challenges associated with fragmented data and categorizing unwanted sexual experiences. Dkt. Nos. 592-1 at 33, 42-43, 48, 592-2 at 16. Uber noted that reports of sexual assault can be interpreted subjectively, and the lack of shared definitions and inconsistent tracking can affect the accuracy, reliability, and consistency of data. Dkt. No. 592-1 at 42. In one example, Uber explained that some incident reports, including contacts from law enforcement, “may simply state that a user was sexually harassed or sexually assaulted,” with no clarifying details contained in the report or uncovered in any follow up communications. Id. at 48-49. Reports that are “unable to be sufficiently classified within Uber's Safety Taxonomy and [] therefore classified as ‘Insufficient Information'” are excluded from the Safety Reports. Dkt. Nos. 592-1 at 49, 592-2 at 48.
To refine the sexual assault and sexual misconduct portions of the first Safety Report, Uber partnered with experts from the National Sexual Violence Resource Center (NSVRC) to develop a Sexual Misconduct and Violence Taxonomy. Dkt. No. 592-1 at 42. According to Uber, prior to this effort, a standardized tool for corporations to consistently classify reports of sexual violence from their consumers did not exist. Id. The Sexual Misconduct and Violence Taxonomy classifies unwanted sexual experiences into two overarching categories-sexual assault and sexual misconduct-which are further divided into a total of twenty-one secondary categories. Id. The categories are listed below, ordered from least severe to most severe:
SEXUAL MISCONDUCT
1. Staring or Leering
2. Comments or Gestures > Asking Personal Questions
3. Comments or Gestures > Comments About Appearance
4. Comments or Gestures > Flirting
5. Comments or Gestures > Explicit Gestures
6. Comments or Gestures > Explicit Comments
7. Displaying Indecent Material
8. Indecent Photography/Video Without Consent
9. Soliciting a Sexual Act
10. Masturbation/Indecent Exposure
11. Verbal Threat of Assault
SEXUAL ASSAULT
12. Attempted Touching of a Non-Sexual Body Part
13. Attempted Kissing of a Non-Sexual Body Part
14. Attempted Touching of a Sexual Body Part
15. Attempted Kissing of a Sexual Body Part
16. Non-Consensual Touching of a Non-Sexual Body Part
17. Non-Consensual Kissing of a Non-Sexual Body Part
18. Attempted Non-Consensual Sexual Penetration
19. Non-Consensual Touching of a Sexual Body Part
20. Non-Consensual Kissing of a Sexual Body Part
21. Non-Consensual Sexual PenetrationId. at 84-85. Uber utilized the Sexual Misconduct and Violence Taxonomy again for the 20182019 Safety Report. Dkt. No. 592-2 at 44. However, both Safety Reports only analyzed data related to what Uber deemed to be the five most serious categories: Non-consensual kissing of a non-sexual body part; Attempted non-consensual sexual penetration; Non-consensual touching of a sexual body part; Non-consensual kissing of a sexual body part; and Non-consensual sexual penetration. Dkt. Nos. 592-1 at 15-16, 592-2 at 15. These five categories combined comprised only about 10,000 incidents over the 2017-2020 time period. Dkt. Nos. 592-1 at 60, 592-2 at 57. Uber stated in the first Safety Report, “In determining which categories of sexual assault were appropriate to include in this report, we prioritized: 1. Including the most serious categories of sexual assault outlined in the taxonomy[;] 2. Maintaining a high degree of confidence and consistency in the quality of the overall dataset[; and] 3. Remaining as consistent as possible with the types of sexual assault that are already published in external research and national estimates.” Dkt. No. 592-1 at 17, 43.
Both Safety Reports also included an external third-party validation of Uber's use of the Sexual Misconduct and Violence Taxonomy. Dkt. Nos. 592-1 at 78-80, 592-2 at 69-71. The assessment for the 2019-2020 Safety Report, prepared by sexual violence prevention experts, reviewed a random and non-random sample of incidents reported in 2019, representative of reports across the entire Taxonomy. Dkt. No. 592-2 at 70. The non-random sample, which focused on the most serious incidents, included instances of attempted or actual contact, i.e., sexual assault, and incidents of sexual misconduct that involved solicitation, masturbation/incident exposure, and verbal threats of sexual assault. Id. Both reports concluded that Uber staff are effectively using the Sexual Misconduct and Violence Taxonomy and coding the identified incident data with a high degree of adherence relative to coding from experts working on issues of sexual misconduct and assault. Dkt. Nos. 592-1 at 80, 592-2 at 70-71.
Based on its analysis of the data, Uber emphasized in both the 2017-2018 and 2018-2019 Safety Reports that 99.9% of trips on the Uber platform ended without any safety-related issues. Dkt. Nos. 592-1 at 72, 592-2 at 68. Uber also claimed in 2017-2018 Safety Report that rates of sexual assault incidents on the Uber rideshare platform in the U.S. declined year-over-year, and in the 2018-2019 Safety Report, it noted that the rate of sexual assault had decreased by over 30% since the last report. Dkt. Nos. 592-1 at 72, 592-2 at 19.
B. Parties' Arguments
hr the Joint Discovery Letter Brief, Plaintiffs request an order compelling Uber to produce discovery responsive to Plaintiffs' Requests for Productions Nos. 83 and 84. Dkt. No. 651 at 1. Specifically, Plaintiffs seek “statistics and data regarding safety-related complaints Uber received in the United States from 2017 to the present” and “the documents and ESI underlying Uber's U.S. Safety Reports.”. Id. Plaintiffs want documents and data related to all twenty-one Uberaudited Sexual Misconduct and Violence Taxonomy categories (not just the five most serious categories of sexual assault or misconduct included by Uber in the Safety Reports) as well as the underlying trip data for each incident (e.g., GPS information, trip date and time, etc.). Id. at 3. hi the alternative, Plaintiffs request that the Court order Uber to produce a corporate witness to testify on the Safety Reports, including the data collection, auditing, categorization, and storage process. Id. at 4. n.10. In support of then request. Plaintiffs point in part to (Redacted)
(Redacted)
At the hearing, Uber asserted that Plaintiffs' request for the data underlying the Safety Reports is, in effect, a request for data related to 800,000 safety and non-safety incident tickets that were assessed as part of the reporting process. Dkt. No. 640, Hearing Transcript (Tr.) at 10. Uber emphasized that there was a rigorous review of the incidents that were placed within the five most serious categories of sexual assault, and that the same extensive review was not applied to the incident reports in the other sixteen categories. Id. at 28. Uber also indicated that all incidents were “associated with a category,” though they did not reside in a review database, separate from the existing systems in which incident tickets reside for use in the ordinary course of business. Id. at 19. Uber argued that to the extent information could be gathered from these systems concerning incidents that fall in categories of sexual assault and misconduct apart from the five most serious categories, the information would be “unverified, incorrect and wouldn't be accurate” because the same rigorous review that was applied to the five most serious categories was not also applied to those other incidents. Id. at 20.
II. LEGAL STANDARD
Under Rule 26 of the Federal Rules of Civil Procedure, “[p]arties may obtain discovery regarding any nonprivileged matter that is relevant to any party's claim or defense and proportional to the needs of the case, considering the importance of the issues at stake in the action, the amount in controversy, the parties' relative access to relevant information, the parties' resources, the importance of the discovery in resolving the issues, and whether the burden or expense of the proposed discovery outweighs its likely benefit.” Fed.R.Civ.P. 26(b)(1). The Court is required to limit discovery that is unreasonably cumulative or duplicative, that the party seeking discovery has had ample opportunity to obtain, or that is outside the scope of permissible discovery described in Rule 26(b)(1). Fed.R.Civ.P. 26(b)(2). A party may serve requests for documents on any other party so long as the request is within the scope of permissible discovery as defined in Rule 26(b)(1). Fed.R.Civ.P. 34(a). The requests “must describe with reasonable particularity each item or category of items” to be produced. Fed.R.Civ.P. 34(b)(1)(A).
“The party seeking discovery has the initial burden of establishing that its request satisfies Rule 26(b)(1)'s relevancy requirement.” Planned Parenthood Fed'n of Am., Inc. v. Ctr. for Med. Progress, No. 16CV00236WHODMR, 2019 WL 311622, at *3 (N.D. Cal. Jan. 24, 2019). “The test for relevance is not overly exacting: evidence is relevant if it has ‘any tendency to make . . . more or less probable . . . [a] fact [that] is of consequence in determining the action.'” In re Volkswagen “Clean Diesel” Mktg., Sales Practices, & Prods. Liabl. Litig., 2017 WL 4680242, at *1 (N.D. Ca. Oct 18, 2017) (quoting Fed.R.Evid. 401). On the other hand, the party opposing discovery “has the burden of showing that discovery should not be allowed, and also has the burden of clarifying, explaining and supporting [his] objections with competent evidence.” Sayta v. Martin, No. 16-CV-03775-LB, 2019 WL 666722, at *1 (N.D. Cal. Feb. 19, 2019) (quoting La. Pac. Corp. v. Money Mkt. 1 Inst. Inv. Dealer, 285 F.R.D. 481, 485 (N.D. Cal. 2012)).
III. DISCUSSION
As to the threshold relevancy requirement, Plaintiffs argue that the documents related to, and the data underlying, the Safety Reports bear on what Uber knew, when it had notice of widespread sexual assault and harassment committed by its drivers against its passengers, and what response the company undertook. Dkt. No. 651 at 1. Uber, however, contends that the data are unreliable, and thus, do not offer proof of such notice, and on that account, Plaintiffs' discovery request should be denied. Id. at 4-5. Uber fails to recognize that evidence is relevant if “it has any tendency to make a fact more or less probable than it would be without the evidence.” Fed. R. of E. 401. If certain incident reports are unverified or have less robust corroboration, that may affect the strength of the evidence as to notice, but it does not preclude discovery.
Plaintiffs have established relevancy not just in relation to the incident reports themselves, but also the underlying trip data for each incident (e.g., GPS information, trip date and time, etc.). These data are important for Plaintiffs to detect patterns associated with claims of sexual assault and sexual misconduct. The existence of patterns based on the incidents reported may tend to show whether Uber disregarded the risk of sexual assault and misconduct on a systematic basis in particular circumstances where passengers may be especially vulnerable to victimization.
Nevertheless, to the extent that Plaintiffs' request is for documents and data related to all 800,000 user reports reviewed by Uber during the course of preparing both Safety Reports, this request is patently overbroad. The 800,000 reports involve non-safety related consumer issues and two other categories of critical safety incidents analyzed in the Safety Reports: motor vehicle fatalities and fatal physical assault. Dkt. Nos. 592-1 at 47, 592-2 at 46. Hence, many of the user reports reviewed by the auditors are not related to the key issues in this case. Uber, therefore, is not obligated to produce documents and data regarding reports that are not safety-related, or those that concern fatalities and do not involve sexual assault or sexual misconduct.
As to documents concerning sexual assault and sexual misconduct, Uber appears to advance an alternative argument that the data Plaintiffs seek is not important enough that its production is proportional to the needs of this case. Uber asserts that custodial files, such as the records from its Director of Women's Safety, Director of Data Science, and others, will have relevant data and statistics concerning the incidents related to Uber's Safety Reports, and communications discussing these data and what should be included or not within the Reports. Tr. at 17; see also Dkt. No. 651 at 6. This argument is unpersuasive because the nature of the incidents reported to Uber and the information associated with the incident tickets may lend significance to what is or is not discussed in each custodian's communications or reflected in their files. For example, if certain types of incidents are not discussed or scrutinized at all in the key custodians' records and communications, though such incidents are repeatedly reported or captured through various channels, then that may tend to show a lack of oversight when Uber should have known of a risk of misdeeds. In addition, if certain incident trends exist and are discussed by key custodians, then that may reflect on the quality of Uber's oversight efforts.
Uber also urges that some categories of the Sexual Misconduct and Violence Taxonomy have “inherent data quality problems,” including those categories that concern “staring” or “flirting” or “comments about appearance.” Dkt. No. 651 at 5. On the one hand, these categories may be especially unreliable, and the type of conduct at issue less severe, such that requiring Uber to produce the underlying data concerning these alleged incidents of flirting, staring and foul remarks is not proportional to the needs of the case. On the other hand, Plaintiffs have alleged that at least until 2018, Uber lacked a formal taxonomy for tracking sexual misconduct through its platform, leading to mislabeling of sexual assault as flirtation, inappropriate comments, or inappropriate communications. ECF No. 269, Master Long Form Complaint ¶ 254. Plaintiffs further allege that Uber's information gathering and investigations continue to limit its ability to identify and prevent sexual misconduct by Uber drivers. Id. ¶ 257. In light these allegations, even incidents reviewed as part of the preparation of the Safety Reports that fall within the less severe categories of misconduct may be important to Plaintiffs' case, if they seek to show a breach of a duty of care by demonstrating that Uber mishandled incident reports on an ongoing, systematic basis. These allegations may ultimately lack merit, but Plaintiffs are entitled to this discovery under Rule 26.
Next, Uber contends that it is unduly burdensome for it to produce new statistical analyses, and that it should not be required to compile nationwide data that are not readily accessible and not generally kept in the ordinary course of business. Dkt. No. 651 at 4. Plaintiffs, however, are not asking Uber to generate new statistical analyses or other reports. Instead, they request data regarding safety-related complaints that Uber received in the United States from 2017 to the present, which Uber indisputably has. Uber's Director of Data Science, Katherine McDonald, does not attest to any burden based on the export of existing data related to these incident tickets.
Uber contorts Plaintiffs' request by arguing that such data are not discoverable because they do not exist, and they are not readily available in the form that Uber prefers. Id. Uber has represented that the incidents which Plaintiffs request have been associated with sexual assault and sexual misconduct categories from the Sexual Misconduct and Violence Taxonomy, even if on a preliminary basis. Tr. at 19. The Safety Reports also state that Uber's auditors reviewed over 800,000 user reports to ensure that incidents were categorized accurately and comprehensively. Dkt. Nos. 592-1 at 47, 592-2 at 46. Uber's contention that the data Plaintiffs request would be wholly unverified, incorrect, and inaccurate is belied by Uber's established process to collect information about such incidents, hi addition, Uber's third-party experts assessed Uber's safety support agents and auditors for both Safety Report and concluded that they were effectively using the Sexual Misconduct and Violence Taxonomy and coding incident data with a high degree of adherence. Uber asserts that an additional review process was applied to the five most serious categories included in the Safety Reports and not extended to the other sixteen categories of sexual misconduct and assault, but it has not demonstrated that this additional review process is the only methodology that yields relevant information, or that the auditors' review methods for the other categories of sexual assault and misconduct are unduly burdensome to rely upon for discovery.
See Dkt. No. 592-1 at 78-80; Dkt. No. 592-2 at 71, App. I, n. 2 (citing a website with the publication, RALIANCE Business, Examining Uber's Use of the Sexual Assault Misconduct and Violence Taxonomy and the Development of Uber's United States Safety Report 2 (Nov. 21, 2019) (describing verification analysis of how Uber agents and auditors categorize user-reported incidents into the Taxonomy)).
IV. CONCLUSION
For the reasons explained above, Plaintiffs' request is GRANTED IN PART and DENIED IN PART. The Court will not compel Uber to produce all 800,000 plus incident reports and accompanying trip data reviewed as part of the preparation of both Safety Reports, especially given that many of these incidents are not related to sexual assault or misconduct. Instead, the Court will grant in part Plaintiffs' request for production of data and documents related to the (Redacted) With respect to incident tickets from 2018 through 2020, Uber shall produce the data and documents related to incidents occurring in the United States associated with any of the twenty- one categories of sexual assault and sexual misconduct from the Sexual Misconduct and Violence Taxonomy. For all incident reports that Uber produces, it shall also produce all underlying trip data (e.g., GPS information, trip date and time, etc.).
IT IS SO ORDERED.