Meta Defendants' Supplemental Responses and Objections to Plaintiffs' First Interrogatory

Raw text

Transcribed text as recorded, including headers and formatting fragments.

Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 1 of 14 AMENDED Exhibit 166 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT Case No.: 4:22-md-03047-YGR MDL No. 3047 In Re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation 1 2 3 4 Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 2 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) Ashley M. Simonsen (Bar No. 275203) COVINGTON & BURLING LLP 1999 Avenue of the Stars Los Angeles, California 90067-4643 Telephone: + 1 (424) 332-4800 Facsimile: + 1 (424) 332-4749 Email: asimonsen@cov.com 5 6 7 Attorneys for Defendants Meta Platforms, Inc. f/k/a Facebook, Inc.; Facebook Holdings, LLC; Facebook Operations, LLC; Facebook Payments, Inc.; 8 Facebook Technologies, LLC; Instagram, LLC; Siculus, Inc.; and Mark Elliot Zuckerberg Additional counsel listed on 9 signature pages 10 11 12 UNITED STATES DISTRICT COURT FOR THE NORTHERN DISTRICT OF CALIFORNIA OAKLAND DIVISION 13 14 15 16 17 18 IN RE: SOCIAL MEDIA ADOLESCENT ADDICTION/PERSONAL INJURY PRODUCTS LIABILITY LITIGATION THIS DOCUMENT RELATES TO: ALL ACTIONS MDL NO. 3047 Civil Case No. 4:22-md-03047-YGR: Honorable Yvonne Gonzalez Rogers META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY 19 20 21 22 23 24 25 26 Defendants Meta Platforms, Inc., Facebook Payments, Inc., Siculus, Inc., Facebook Operations, LLC, and Instagram, LLC, (collectively, “Defendants” or “Meta”), by and through their attorneys, Covington & Burling, LLP, submit their supplemental responses and objections to Plaintiffs' Interrogatory No. 1 (“Interrogatory") served on May 16, 2024. 27 28 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 3 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 enqueuing happens at around a_60-65% confidence level, with autodeletion when classifiers are at 2 3 II. approximately above 85-90%. Harm-Specific Classifiers 4 Classifiers run automatically and review content on Facebook and Instagram. Different classifiers 5 review and detect different types of policy-violating content, also known as “harm areas.” Those harm 6 7 8 areas include, but are not limited to, SSI, ED, bullying and harassment, graphic and violent content, adult nudity, and child safety. 9 A. SSI Classifiers 10 Meta first implemented a classifier to action SSI content at least as early as February 2019, and 11 12 13 Meta has continuously refined its SSI classifiers since then. The latest iteration of SSI classifiers were launched at different times in 2024.4 SSI classifiers are intended to detect suicide and self-injury content on Facebook and Instagram that may violate Meta's Policies.5 SSI classifiers consider the following 14 15 16 17 18 19 20 21 content policy-violating under Meta's Policies and subject to removal: “any content that encourages suicide, self-injury or eating disorders, including fictional content such as memes or illustrations, and any self-injury content which is graphic, regardless of context. as well as real time depictions of suicide or self-injury." However, not all policy-violating content is subject to removal, such as “[c]ontent about recovery from suicide, self-injury or eating disorders that is allowed, but may contain imagery that could be upsetting (such as a healed scar)."7 Meta removes content that encourages suicide, self-injury or eating disorders, including fictional 22 content such as memes or illustrations and any such content which is graphic, regardless of context. Meta 23 24 25 4 See infra Table 1. 26 5 Suicide & Self Injury, Meta: Transparency Center, https://transparency.meta.com/policies/community-standards/suicide- self-injury/(last visited Feb. 6, 2025). 27 6 Id. 7 Id. 28 11 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 4 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 also removes content that mocks victims or survivors of suicide, self-injury or eating disorders, as well as 2 real-time depictions of suicide or self-injury. Content about recovery from suicide, self-injury or eating 3 4 disorders that is allowed, but may contain imagery that could be upsetting, such as a healed scar, is placed behind a sensitivity screen.”8 5 Meta uses its AI technology, machine learning, and image-based technology (including pattern- 6 7 8 9 10 recognition signals, such as phrases and comments of concern) to proactively identify and take action on clearly violating SSI content. Meta refines its classifiers by providing both positive examples (actual SSI content) and contrasting negative examples (not actual SSI content—e.g., “I have so much homework, I want to kill myself”—so it can learn to distinguish the two.9 11 12 13 14 SSI Classifiers do not just review the content itself but instead may review other information— e.g., comments to a post where a user is seriously at risk, the classifier may detect and consider comments from human commenters such as “Tell me where you are” or “has anyone heard from [user]?”10 15 16 Training automated technology to identify and distinguish between SSI content that is violative or non-violative is incredibly complex. This is particularly the case on Instagram, where much of the content 17 is image-based and therefore the intended meaning substantially relies on subtle context. For instance, a 18 picture of train tracks may have to do with travel (and therefore be innocuous content), or it may relate to 19 20 SSI. Humans can quickly tell the difference between these images based on the context, but training AI to do so is challenging, particularly at scale. 21 The current threshold values for autodeletion and various soft actions on SSI content are listed 22 23 below.11 Content containing SSI is prioritized during human review because Meta has a vested interest 24 8 Id. 25 26 9 How Facebook AI Helps Suicide Prevention, Meta, https://about.fb.com/news/2018/09/inside-feed-suicide-prevention-and- ai/ (last visited Feb. 6, 2025). 10 Id. 27 11 See infra Table 1. 28 12 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 HIGHLY CONFIDENTIAL (COMPETITOR) Page 5 of 14 1 in preventing as much human harm as possible. Meta uses artificial intelligence to prioritize the order in 2 which potentially violating SSI content is reviewed to ensure efficient enforcement of policies and to get 3 resources to users in need quickly. 4 Table 1. SSI Classifiers and SSI Borderline Classifiers. 5 Plain- 6 Language Alpha- numeric Launch date of Threshold 7 Classifier code current value for autodeletion Threshold value(s) for soft action(s) Name iteration 8 SSI f590828509 08/12/2024 0.94 9 Classifier >P50 • 10 Ο Ο 11 • 12 13 Ο • Facebook: various enforcements EU Demotion on Newsfeed Suicide: 0.8 Self-injury: 0.8 EU Deboost on Newsfeed Ο Suicide: 0.3 Self-injury: 0.4 Non-EU Demotion on Newsfeed 14 Ο Suicide: 0.5 15 Ο Self-injury: 0.5 • Non-EU Deboost on 16 Newsfeed Ο Suicide: 0.3 17 Ο • 18 19 Ο 20 • 21 22 • Self-injury: 0.4 Age-based Demotion on Newsfeed Suicide: 0.3 Self-injury: 0.3 Age-based Deboost on Newsfeed Ο Suicide: 0.05 Ο Self-injury: 0.05 Age gating on Newsfeed 23 Ο Ο 24 25 26 27 Suicide: 0.1 Self-injury: 0.1 • Non-rec filtering Ο Suicide: 0.105 Ο Self-injury: 0.25 • Filtering demotion on Stories Ο Self-injury: 0.25 28 13 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 HIGHLY CONFIDENTIAL (COMPETITOR) Page 6 of 14 1 Plain- Alpha- Language numeric Launch date of 2 Classifier code current Threshold value for autodeletion Threshold value(s) for soft action(s) 3 Name iteration 4 • 5 • 6 Ο 7 8 • 9 Ο Ο 10 • Age gating on Stories Self-injury: 0.25 Age gating on Stories (reduce more) Self-injury: 0.1 Instagram: various enforcements > P40 Age gating Suicide: 0.89 Self-injury: 0.71 Age gating High COE 11 Ο Ο Suicide: 0.6 Self-injury: 0.24 12 • Non-rec filtering Ο Suicide: 0.6 13 Ο Self-injury: 0.24 • Hashtag filtering 14 Ο Suicide: 0.80 15 Ο Self-injury: 0.80 • Media SERP 16 17 Ο Suicide: 0.85 Ο Self-injury: 0.85 • Age-filtering Ο Suicide: 0.6 Self-injury: 0.24 18 19 20 21 22 23 24 25 SSI f551409441 04/15/2024 N/A Borderline 26 Classifier 27 28 Demotion Feed teen non- Ο • EU • Ο Ο Suicide: 0.7 Self-injury: 0.7 Demotion Feed teen EU Suicide: 0.85 Ο Self-injury: 0.85 • Demotion Stories teen Ο Suicide: 0.7 Ο Self-injury: 0.7 Facebook: various enforcements above >P10 • Age gating on Newsfeed: 0.1 • P-non MSI: 0.3 14 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY MD-03047-YGR CASE NO. 4:22- Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 HIGHLY CONFIDENTIAL (COMPETITOR) Page 7 of 14 1 Plain- 2 Language Classifier Alpha- numeric Launch date of code 3 Name current iteration Threshold value for autodeletion Threshold value(s) for soft action(s) 4 Non-rec filtering: 0.3 5 • • Age gating on Stories: 0.75 • Search demotion: 0.5 6 Instagram: various enforcements > P40 7 8 9 Hashtag filtering: 0.80 10 • Age gating: 0.91 • Age gating High CoE: 0.54 • Non-rec filtering: 0.34 • Media SERP: 0.85 • • Age filtering: 0.25 11 12 B. ED Classifiers 13 Meta launched a classifier to action ED content specifically at least as early as March 2021. Prior 14 15 to the launch of an ED-specific classifier, ED content fell under the purview of SSI policies and enforcement technology. The latest iteration of ED classifiers were launched at various times in 16 17 18 2024.12 ED classifiers are intended to detect ED content on Facebook and Instagram that may violate Meta's Policies.13 ED classifiers consider the following content as policy-violating and subject to removal: "[c]ontent that focuses on depiction of ribs, collar bones, thigh gaps, hips, concave stomach, or 19 20 protruding spine or scapula when shared together with terms associated with eating disorders.”14 Not all 21 policy violating ED content is subject to removal, such as content that “depicts ribs, collar bones, thigh 22 gaps, hips, concave stomach, or protruding spine or scapula in a recovery context."15 Such content is 23 24 25 12 See infra Table 2. 26 13 Suicide & Self Injury, Meta: Transparency Center, https://transparency.meta.com/policies/community-standards/suicide- self-injury/(last visited Feb. 6, 2025). 27 14 Id. 15 Id. 28 15 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 8 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 placed behind a sensitivity screen. Generally, content promoting or encouraging EDs are removed, but 2 3 users are allowed to post content that touches on their own experiences around self-image and body acceptance. 16 4 Because Meta's platforms see less ED content than SSI content, there are fewer datapoints for the 5 6 Al system to learn from in the ED sphere than for SSI. To combat this issue, Meta generates synthetic training data the classifiers can learn from. 7 8 ED Classifiers do not currently perform automatic deletions and therefore there is no threshold 9 value for autodeletion. All content with a value above P70 is demoted and enqueued for human review 10 and potential manual deletion. Any content at or below a value of P70 but higher than the respective 11 minimum values set forth below in Table 2 is demoted. By extension, Meta periodically re-assesses the 12 threshold value at which content is enqueued for human review, and the threshold value is subject to 13 change at any given time. 14 15 16 Plain- 17 language classifier Alpha- numeric Launch date of Table 2. ED Classifiers & ED Borderline Classifiers. Threshold value(s) for soft action(s) Threshold value for code current autodeletion 18 name iteration 19 ED Classifier f638461582 09/01/2024 N/A Facebook • EU Demotion on Newsfeed: 20 0.8 • EU Deboost on Newsfeed: 21 0.15 • Non-EU Demotion on 22 Newsfeed: 0.5 23 • Non-EU Deboost on Newsfeed: 0.15 24 • Age based Demotion on Newsfeed: 0.3 25 • Age based Deboost on Newsfeed: 0.05 26 27 16 Id. 28 16 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 HIGHLY CONFIDENTIAL (COMPETITOR) Page 9 of 14 1 Plain- 2 language classifier Alpha- numeric code Launch date of current Threshold value for autodeletion Threshold value(s) for soft action(s) 3 name iteration 4 • 5 6 7 8 Age gating on Newsfeed: 0.1 • Non-rec filtering: 0.061 Instagram: various enforcements >P40 Non-rec filtering: 0.81 Media SERP: 0.85 Demotion Feed teen non- 9 10 • • • Hashtag filtering: 0.80 • EU: 0.7 • Demotion Feed teen EU: 0.85 • Demotion Stories teen: 0.7 11 12 ED f551542423 04/15/2024 N/A Facebook 13 Borderline • Age gating on Newsfeed: 0.1 Classifier • P-non MSI: 0.3 14 • Non-rec filtering: 0.33 15 16 17 18 19 • Age gating: 0.907 • • Media SERP: 0.85 • • Age filtering: 0.26 (P40) • Search demotion: 0.5 Instagram: various enforcements >P40 Non-rec filtering: 0.7 (P60) Hashtag filtering: 0.80 20 21 C. Bullying and Harassment Classifiers 22 Meta first launched a classifier that actioned “bullying and harassment” (“B&H”) content at least 23 as early as May 2018. The latest iteration of B&H classifiers were launched at various points in 2022, 24 2023, and 2024.17 B&Hclassifiers are intended to detect bullying and harassmentcontentthat may violate 25 26 27 17 See infra Table 3. 28 17 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 10 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 Plain-language classifier name Alpha- numeric Launch date 2 code of current iteration Threshold value for autodeletion Threshold value(s) for soft action(s) 3 • 4 • 5 Youth filter on IFR: 0.1 Youth filter on Story: 0.5 6 7 IG Non-rec ANSA Photo f639524943 09/04/2024 N/A • Age gating: 0.623 • Age filtering: 0.159 Classifier on IG • 8 Age demotion: 0.104 9 IG Non-rec f641128648 09/10/2024 N/A Age gating: 0.623 10 ANSA video • classifier on IG Age filtering: 0.237 Age demotion: 0.158 11 12 13 F. Child Safety Classifiers 14 Meta first launched a classifier to action “child safety” (“CS”) content at least as early as 2018. 15 16 The current iterations of CS classifiers were launched at various times in 2020, 2023, and 2024.30 CS classifiers are intended to proactively identify and action content and behaviors that may violate Meta's 17 policies on Child Sexual Exploitation, Abuse, and Nudity. 31 Child Safety classifiers consider the 18 following content, inter alia, as violating: content, activity, or interactions that threaten, depict, praise, 19 20 support, provide instructions for, make statements of intent, admit participation in, or share links of the 21 sexual exploitation of children; content that solicits sexual content or activity depicting or involving 22 children, nude imagery of real or non-real children, and/or sexualized imagery of real or non-real children; 23 content that solicits sexual encounters with children; content that constitutes or facilitates inappropriate 24 25 26 30 See infra Table. 6. 27 31 Child Sexual Exploitation, Abuse, and Nudity, Meta: Transparency Center, https://transparency.meta.com/policies/community-standards/child-sexual-exploitation-abuse-nudity/ (last visited Feb. 7, 2025). 28 27 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 11 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 interactions with children; content that attempts to exploit real children by coercing money, favors or 2 intimate imagery with threats to expose intimate imagery or information or by sharing, threatening, or 3 stating an intent to share private sexual conversations or intimate imagery; content that sexualizes real or 4 non-real children; Groups, Pages, and profiles dedicated to sexualizing real or non-real children; content 5 that depicts real or non-real child nudity; videos or photos that depict real or non-real non-sexual child 6 7 8 abuse regardless of sharing intent; or content that praises, supports, promotes, advocates for, provides instructions for or encourages participation in non-sexual child abuse. 32 9 10 11 12 The names of the below classifiers generally align with their intended purpose. For example, the CSAM classifier is intended to identify CSAM, while the CSAM solicitation classifier is intended to identify the solicitation of CSAM. The Google Content Safety API classifier is intended to detect CSAM and is supported by Google's Application Programming Interface. The Child Sexualization (“CSx”) 13 14 classifier is intended to identify child sexualization content. The MetaGen for Child Safety classifier is intended to identify child safety violating content, generally. All of the classifiers identified in Table 6 15 16 are run simultaneously, with the exception of the Google Content Safety API classifier which is used to 17 prioritize human review queues. There are no borderline content detection classifiers in this harm area. 18 The threshold values for enqueueing CS content for human review and manual removal change frequently 19 to ensure we can enqueue as much potentially violating CS content for human review as possible. The 20 21 pIV ranking system works to ensure that potential CS content most likely to violate policy moves to the top of the review queue. 33 22 23 24 25 26 27 32 Id. 33 See discussion supra Section I.B. 28 28 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 Page 12 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) 1 Table 6. Child Safety Classifiers. 2 Plain-language 3 classifier name Alpha-numeric code Launch date of current iteration Threshold value for autodeletion 4 CSAM classifier FB: f588692710; 08/29/2024 N/A 5 IG: f565741344 • 6 7 8 9 10 CSAM FB: f636246456; 09/16/2024 N/A 11 Solicitation IG: f566561739 • classifier 12 • 13 14 15 16 17 CSAM-S f636721225 08/29/2024 N/A comment • 18 classifier 19 20 21 Child f590979030 08/28/2024 N/A 22 Sexualization (CSx) classifier 23 24 25 26 27 28 Threshold value(s) for soft action(s) Facebook Age filtering: 0.875 (P20) • Filtering for GenPop: 0.891 (P25) Instagram • Filtering for Teens: 0.943 (P25) Facebook Age filtering: 0.982 (P20) Filtering for GenPop: 0.984 (P25) Instagram • Filtering for Teens: 0.965 (P25) Facebook Filtering for GenPop: 0.9975 (P25) Instagram • IG: no soft actions Facebook • Age filtering: 0.942 (P20) • Filtering for GenPop: 0.945 (P25) Instagram • Filtering for Teens: 0.979 (P25) 29 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR Document 2885-65 Filed 03/26/26 HIGHLY CONFIDENTIAL (COMPETITOR) Page 13 of 14 1 Plain-language Alpha-numeric classifier name code 2 3 CSE-I Classifier f573679609 of current iteration 06/18/2024 N/A Launch date Threshold value for Threshold value(s) for soft action(s) autodeletion Facebook 4 • Age filtering: 5 • 6 7 8 9 Child subject to f589982602 objectionable 08/20/2024 N/A • 10 content classifier 0.83 (P20) Filtering for GenPop: 0.847 (P25) Instagram • IG: no soft actions Facebook Age filtering: 0.949 (P20) • Filtering for 11 12 GenPop: 0.951 (P30) Instagram 13 • 14 Filtering for Teens: 0.957 (P50) 15 16 Google Content N/A Safety API classifier 06/02/20 N/A No soft actions 17 18 III. Borderline Classifiers 19 20 21 In addition to its policies regarding clearly violating content, Meta also has had borderline content policies since 2020. A “borderline classifier” is an artificial intelligence tool that reviews and detects 22 content on Facebook and Instagram that is not strictly policy-implicating, but nonetheless considered 23 potentially problematic or low quality. For example, a borderline graphic and violent content classifier 24 would consider content of puss-filled or oozing wounds, people undergoing surgical operations, fight 25 videos, animal abuse, and fictional violence, not strictly policy-violating, but as “borderline content.” 26 27 28 30 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR Case 4:22-md-03047-YGR 1 Dated: February 21, 2025 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Document 2885-65 Filed 03/26/26 Page 14 of 14 HIGHLY CONFIDENTIAL (COMPETITOR) Respectfully submitted, COVINGTON & BURLING LLP /s/Ashley M. Simonsen Ashley M. Simonsen (State Bar No. 275203) asimonsen@cov.com COVINGTON & BURLING LLP 1999 Avenue of the Stars Los Angeles, CA 90067 Telephone: (424) 332-4800 Facsimile: + 1 (424) 332-4749 Emily Johnson Henn (State Bar No. 269482) ehenn@cov.com COVINGTON & BURLING LLP 3000 El Camino Real 5 Palo Alto Square, 10th Floor Palo Alto, CA 94306 Telephone: + 1 (650) 632-4700 Facsimile: +1 (650) 632-4800 Phyllis A. Jones, pro hac vice pajones@cov.com Paul W. Schmidt, pro hac vice pschmidt@cov.com Michael X. Imbroscio, pro hac vice mimbroscio@cov.com COVINGTON & BURLING LLP One CityCenter 850 Tenth Street, NW Washington, DC 20001-4956 Telephone: + 1 (202) 662-6000 Facsimile: + 1 (202) 662-6291 Attorneys for Defendants Meta Platforms, Inc. f/k/a Facebook, Inc.; Facebook Holdings, LLC; Facebook Operations, LLC; Facebook Payments, Inc.; Facebook Technologies, LLC; Instagram, LLC; Siculus, Inc.; and Mark Elliot Zuckerberg 33 META DEFENDANTS' SUPPLEMENTAL RESPONSES AND OBJECTIONS TO PLAINTIFFS' FIRST INTERROGATORY CASE NO. 4:22- MD-03047-YGR