The Class Action Megaphone

Empowering Class Members with an Empirical Voice

Alissa del Riego & Joseph Avery *


Class actions have been dominating headlines, both in mainstream outlets 1 See, e.g., Adam Liptak, Supreme Court Won’t Block Student Loan Class-Action Settlement, N.Y. Times (Apr. 13, 2023), and scholarly ones. 2 See, e.g., Nicholas Almendares, The Undemocratic Class Action, 100 Wash. U. L. Rev. 611 (2023); Peter N. Salib, Artificially Intelligent Class Actions, 100 Tex. L. Rev. 519 (2022); Xiyin Tang, The Class Action as Licensing and Reform Device, 122 Colum. L. Rev. 1627 (2022). Just one settlement that was recently reached on behalf of Facebook users impacts a diverse group of all U.S.-based Facebook users from 2007 to 2022. 3 Plaintiffs’ Notice of Motion and Motion for Final Settlement Approval, In re Facebook, Inc. Consumer Privacy User Profile Litig., 3:18-md-02843 (N.D. Cal. July 11, 2023), ECF No. 1145 at 3-4; see also Facebook Users in the United States of America, NapoleonCat., June 2023, On a surface level, lawsuits like this one appear to further one of the primary justifications for the class action mechanism: representation. Yet, if by representation we mean access to justice and furthering plaintiffs’ interests and aims, especially those pertaining to legislative or legal norm shaping, 4 Andrew Faisman, Note, The Goals of Class Actions, 121 Colum. L. Rev. 2157, 2175-77 (2021). then these lawsuits, in the end, might be self-defeating. After all, for those of us who were Facebook users, we may feel genuinely harmed that, without our permission, Facebook mined our data and made it accessible to third parties, including Cambridge Analytica. But are we satisfied with the pittance that we will receive through settlement? Are those pennies practically accepted on our behalf without our input access to justice?

There is a clear representational problem in class actions. Although designed to be a megaphone, amplifying the voices of isolated and disparate individuals who have suffered harms, the class action has come to stand as something of a sound deadener, silencing class members (while amplifying the voices of class counsel). Rather than creating efficiencies and increasing representation, class litigation often silences class members.

Consider the following hypothetical. 5 This hypothetical is loosely based on Boykin et al. v. Panera Bread Co., No. 1:18-cv-02461 (N.D. Ill.). Erica, a California resident, had her personal information compromised because of a restaurant chain’s negligence, and she had to pay a $500 ransom to secure the data. But the Chicago attorneys who, without Erica’s consent, filed suit on her behalf, had their suit dismissed because they could not allege an actual injury—a clearly unsatisfactory outcome. Of course, litigation and settlement agreements often produce unsatisfactory outcomes, but here the injured party did not even have the ability to consult with her attorneys, assist her representation, or agree to any compromise, as she was unaware of her own lawsuit. Even so, her attorneys did nothing unlawful, as they were prosecuting class claims, which do not, prior to class certification and settlement court approval, require consultation with class members. 6 See 3 Newberg and Rubenstein on Class Actions § 9:2 (6th ed. 2022) (“Rule 23 makes no explicit mention of communications with absent class members other than formal notice requirements”). Formal notice requirements, moreover, are scant, and the Supreme Court’s single decision on communication with absent class members (Gulf Oil Co. v. Bernard, 452 U.S. 89 (1981)) failed to create a firm directive for evaluating such communications, or lack thereof. Id. at § 9:3..

There are valid reasons for this seeming injustice: classes can contain hundreds, thousands, and often millions of class members, rendering meaningful consultation and traditional representation impossible. This has, in part, led to a reluctant acceptance of the validity of class actions, notwithstanding (or perhaps ignoring) class member disenfranchisement. 7 See David Rosenberg, Class Actions for Mass Tort: Doing Individual Justice By Collective Means, 62 Ind. L. J. 561, 567 (1987) (arguing that “bureaucratic justice implemented through class actions provides better opportunities for achieving individual justice than does the tort system’s private law, disaggregative processes”); cf Almendares, supra note 2, at 636-38, 659.

However, if this was true, it no longer is. The rise of artificial intelligence (“AI”), coupled with wide acceptance of and familiarity with empirical methods, has created an opportunity for a better version of class member representation. Specifically, in this Essay, we outline a two-step solution to the class action disenfranchisement problem: (1) the utilization of technology and empirical methods for class action attorneys to communicate effectively with class members; and (2) the introduction, via a slight modification of Rule 23 and the creation of “class member liaison counsel,” of a legal expectation and duty to communicate with class members. Implementing this solution would create a more transparent and accountable class action process, ensuring that class members have both a reasonable level of involvement and a voice in their own cases. It would forge a relationship between class counsel and class plaintiffs that more closely approximates the traditional, well-established, and almost sacrosanct attorney-client relationship.

I. Class Members: Disenfranchised “Clients”

Class members are sidelined at all points in the litigation process. Beginning with selection of counsel, unnamed class members do not actually “hire” their attorneys. 8 See 1 William B. Rubenstein, Newberg and Rubenstein on Class Actions §§ 3:72, 3:82 (6th ed. 2022). Rather, plaintiff class action attorneys essentially hire themselves or courts do if multiple attorneys seek to represent the same class. 9 Fed. R. Civ. P. 23(g)(2) (“If more than one adequate applicant seeks appointment, the court must appoint the applicant best able to represent the interests of the class.”). Unnamed class members typically are not even informed when litigation is initiated on their behalf. Attorneys simply file suit, and, when courts appoint counsel to represent the class, they are only required to find those attorneys “adequate.” 10 Fed. R. Civ. P. 23(g).

The unconsulted and unapproved assignment of class counsel would be less concerning if class counsel later communicated with class members. But they rarely do until there is a settlement or until class certification has been granted, typically occurring several years into the litigation, if ever. 11 See Fed. R. Civ. P. 23(c)(2)-(3), (e)(1). At those points, however, class members only have the opportunity to object to a settlement, which will likely be denied, or opt out of the class and incur the costs of pursuing their likely low-recovery claim independently, a financially burdensome and often unrealistic option. This lack of communication is particularly troubling because some injuries and remedies are less obvious than others, and class counsel, often a non-diverse group, may not prioritize the interests of all class members. 12 Alissa del Riego, Driving Diverse Representation of Diverse Classes, 56 U. Mich. J. L. Reform 67, 90-98 (2022). Furthermore, unless the action solely seeks injunctive relief, class counsel are incentivized to pursue economically quantifiable remedies upon which to base their compensation, potentially neglecting other remedies class members might favor, such as an admission of wrongdoing or a lower settlement that would increase in value if the offense were repeated. 13 John C. Coffee, Jr., Understanding the Plaintiff’s Attorney: The Implications of Economic Theory for Private Enforcement of Law Through Class and Derivative Actions, 86 Colum. L. Rev. 669, 712-20 (1986) [hereinafter Coffee, Understanding]; see John C. Coffee, Jr., Class Wars: The Dilemma of the Mass Tort Class Action, 95 Colum. L. Rev. 1343, 1365-84 (1995) [hereinafter Coffee, Class Wars]. This asymmetry has led some to argue that class counsel lack identifiable clients altogether. 14 See Coffee, Understanding, supra note 13 at 678.

Class counsel also have no duty to inform class members of rejected settlement offers. 15 As per Fed. R. Civ. P. 23(e)(1), only when a settlement proposal is considered does the court have grounds direct notice of the proposal if the parties show that the court will likely be able to approve the proposal under Rule 23(e)(2) and certify the class for purposes of judgment on the proposal. Only once class counsel agrees to accept a settlement are class members informed. At this point, objections to the settlement may be lodged. But, given the lack of communication throughout the litigation, class members lack the information needed to make reasoned, well-informed decisions and objections and thus lack the requisite knowledge to formulate arguments concerning whether an accepted settlement has been negotiated in good faith or is in their best interests, leaving courts, already inclined to accept class settlements, with little ammunition to deny class certification. 16 See Wal-Mart Stores, Inc. v. Visa U.S.A., Inc., 396 F.3d 96, 116 (2d Cir. 2005) (“We are mindful of the ‘strong judicial policy in favor of settlements, particularly in the class action context.’”); In re Advance Battery Techs., Inc. Sec. Litig., 298 F.R.D. 171, 174 (S.D.N.Y. 2014) (“The law favors settlement, particularly in class actions . . . where substantial resources can be conserved by avoiding the time, cost, and rigor of prolonged litigation.”).

II. Creating a Class Counsel-Class Member Relationship Via AI and Empirical Methods

While the realities of class actions demand that the class counsel-class member “relationship” be somewhat different than the traditional attorney-client relationship, these realities do not require that such a relationship be nonexistent. At the least, current technology enables a meaningful relationship with scant costs. In this Essay, we have focused on an issue essential to legal representation but largely swept aside in the class action context: attorneys’ obligations to communicate with clients. But what should such communication involve, and how can it be encouraged and enforced? In this Part, we answer these questions, starting with the practicable means of communication and moving towards a proposal for making it legally mandated.

A. A Means to Communicate Assisted By Technology, Empirical Methods, and AI

How can deliberation be incorporated into class actions, such that class members are afforded an informed voice, even if a limited one, throughout the litigation? Rule 23(g), which was enacted to ensure class members had adequate legal representation, was drafted two decades ago, when it was assumed that any communication would be limited and that class counsel would be paternalistic executors of dispersed and largely unidentifiable or unreachable class members. But this no longer has to be the case.

On the AI-end alone, scholars have shown that machine learning algorithms can be used to facilitate certification of otherwise uncertifiable classes. 17 See Salib, supra note 2, at 544-48 (showing how advanced machine learning algorithms can facilitate certification for innumerable currently uncertifiable classes by being trained to mimic the decisions of a jury in a particular case, expeditiously resolving individual questions without needing “mini-trials” and thus allowing common questions to predominate). If modern technology can contribute actual legal decisionmaking in class actions, then presumably classwide communication should also be feasible. In a different context, we know that artificial intelligence, such as OpenAI’s large language models, can be used to increase communication between legislators and their constituents, leading to increased trust between the parties. 18 Sarah Kreps & Maurice Jakesch, Can AI Communication Tools Increase Legislative Responsiveness and Trust In Democratic Institutions?, 40 Gov’t Info. Q. 101829, at 5-10 (2023), Relatedly, there are disciplines within business schools and departments within large companies devoted to identifying consumer preferences, with specialized experimental designs (“conjoint experiments”) existing for that purpose. 19 Sarah Kreps & Maurice Jakesch, Can AI Communication Tools Increase Legislative Responsiveness and Trust In Democratic Institutions?, 40 Gov’t Info. Q. 101829, at 5-10 (2023), Those same principles can be applied to determine the preferences of class members—and conjoint analyses are already being employed in class litigation to establish liability and quantify damages. 20 See, e.g., In re JUUL Labs, Inc. Mktg. Sales Pracs. & Prods. Liab. Litig., 609 F. Supp. 3d. 942, 974-75 (N.D. Cal. 2022) (finding conjoint analysis provided by class plaintiffs’ expert sufficient to survive class certification); In re FCA US LLC Monostable Elec. Gearshift Litig., 382 F. Supp. 3d 687, 697-99 (E.D. Mich. 2019) (denying exclusion of plaintiffs’ expert’s conjoint damages analysis). Finally, courts already have the power to query class members via questionnaires, 21 See Vinole v. Countrywide Home Loans, Inc., 571 F.3d 935, 947 (9th Cir. 2009) (holding that using tools “such as questionnaires” is “within the discretion of the district court”); see also In re Phenylpropanolamine (PPA) Prods. Liab. Litig., 227 F.R.D. 553, 621 (W.D. Wash. 2004) (using questionnaires); Millsap v. McDonnell Douglas Corp., No. 94-CV-633-H(M), 2003 WL 21277124, at *2 (N.D. Okla. May 28, 2003) (using questionnaires). so technology-assisted querying regarding core communication items, such as the claims pursued, damages suffered, relief sought, alternative acceptable relief, etc., we have identified should be expected.

Practically speaking, what might this look like? In a forthcoming paper, Professors del Riego and Avery actively survey mock class members, demonstrating how, at the beginning of a class action, class member preferences regarding attorney selection can be made known. 22 Alissa del Riego & Joseph Avery, Inadequate Adequacy? Empirical Studies on Class Members’ Preferences of Class Counsel, Utah. L. Rev. (forthcoming 2024). The authors speculate how a mechanism they term “representational notice” could be deployed to provide at least a subset of class members with notice of their claims and an opportunity to provide guided feedback. Class member responses, moreover, can be quickly and effortlessly analyzed and summarized by AI tools. 23 Peter Adebowale Olujimi & Abejide Ade-Ibijola, NLP Techniques for Automating Responses to Customer Queries: A Systematic Review, 3 Discover Artificial Intelligence 20 (2023) (“[Natural language processing (“NLP”)] usage and applications are…prompting widespread recognition and implementation in a variety of industries. […] Several studies have shown that NLP can be used to comprehend and interpret speech or text in natural language to accomplish the desired goals [of reaching previously unachievable levels of efficiency and quality].”).

While representational notice would yield class member input at the initial stages of litigation, it could not take into account new information uncovered during the discovery process, court rulings that change the scope of the litigation, settlement offers made by the defendant that offer some relief and eliminate the uncertainty of trial, and other data points that would, in a typical attorney-client relationship, precipitate communication with clients. When feasible and appropriate in such instances, class counsel could communicate with interested class members and deploy empirical tools to process feedback that keeps the attorneys abreast of class members’ preferences and interests.

In practice, here is what this would look like. Once class counsel has been appointed, they would employ reasonable efforts to expediently announce the existence of the litigation to class members, affording class members the ability to communicate with counsel prior to settlement or certification, when pleadings may no longer be able to be amended. 24 See cf. Order Denying Motions for Clarification, Reconsideration, and Vacatur, In re Takata Airbag Product Liability Litig., 1:15-md-02599 (S.D. Fla. May 6, 2021), ECF No. 4039 (denying plaintiffs opportunity to amend complaint in response to a dismissal order prior to class certification). This could include representational notice and/or targeted advertisements in both print, social, and electronic media and across airwaves. These communications should always include an email, physical address, and phone number class members can use to contact class counsel. Counsel must, of course, dedicate sufficient resources to responding to questions, inquiries, and requests for further communications sent by class members. With AI involved, however, these tasks become less onerous and more manageable. AI, of course, does not relieve counsel of their duties, but rather functions as a facilitator of such duties that they are ultimately responsible for supervising. 25 See Joseph Avery, Patricia Sánchez Abril & Alissa del Riego, ChatGPT, Esquire: Recasting Unauthorized Practice of Law in the Era of Generative AI, Yale J. L. & Tech. (forthcoming 2023). This problem is no different than that actively handled by elected officials, 26 See Kreps & Jakesch, supra note 18, at 2-3. and it echoes the role of triage nurses, who handle as much as they can before escalating certain pre-determined issues to attending physicians. 27 Lynn Sayre Visser, Anna Silva Monetjano, & Valerie G. A. Grossman, Fast Facts for the Triage Nurse : an Orientation and Care Guide in a Nutshell 1-34 (2015). In short, AI can be leveraged to respond to frequently asked questions and other anticipated inquiries, such as those requesting updates on the current status of the litigation. When it comes to issues that AI cannot handle or that counsel will want to handle personally, such as settlement objections and opt-outs, these inquiries from class members can be flagged by AI for counsel review. Counsel then can respond directly. Additionally, class members should be afforded the opportunity to subscribe to periodic litigation updates, which should be sent with reasonable frequency. Similarly, class members should be able to remotely attend periodic meetings with counsel.

It is important to note that the AI systems we propose do not include those systems that are beginning to offer legal advice. 28 Benjamin Weiser, Here’s What Happens When Your Lawyer Uses ChatGPT, N.Y. Times (May 27, 2023), Rather, the AI on which we are leaning are communication facilitation systems, such as those discussed by Professors Kreps and Jekesch. 29 See Kreps & Jakesch, supra note 18, at 1-2. On the survey end, these tools function like chatbots, soliciting answers to a designated range of questions (but, importantly in our case, not providing unscripted answers). They are cheap—if not free—as they can be developed using open-source packages. On the analysis end, these tools include natural language processing packages (also free and widely available), which can be used to parse and unpack class members’ communications. 30 See, e.g., William Leeson, Adam Resnick, Daniel Alexander, and John Rovers, Natural Language Processing (NLP) in Qualitative Public Health Research: A Proof of Concept Study, 18 Int’l J. Qualitative Methods 1 (2019). While there are valid fears surrounding AI use, including text-based AI’s propensity to hallucinate and break from reality, 31 See Weiser, supra note 28. most of those fears are not germane to the tools and use cases we are discussing.

But why would class counsel, which to date have enjoyed the luxury of managing litigation without the hurdle most attorneys face, want to expend financial and time-consuming resources to communicate with class members? The short answer is that they would not. Thus, in the succeeding subpart, we propose a modest change to Rule 23(g) that requires counsel to communicate with class members.

B. Creation of a Duty to Communicate Under Rule 23(g)(1)(A)(v) and Class Member Liaison Counsel

Outside of the class action environment, attorneys have a duty to communicate with their clients. 32 J. Nick Badgerow, Can We Talk? The Lawyer’s Ethical, Professional and Proper Duty to Communicate with Clients, Kan. J.L. & Pub. Pol’y, 1997-1998, at 105, 105. This includes a duty to honor and further their clients’ preferences. In the class action environment, we argue a modified duty to communicate should exist. This duty should require class counsel to make reasonable efforts to communicate with known class members regarding the existence of litigation, including claims, injuries, and damages pursued. More pointedly, this would include both a duty to inform class members of the status of their litigation and respond to class members’ inquiries, questions, and concerns. Below we show how such a duty can be enforced.

Rule 23(g)(1)(A) currently requires courts to evaluate attorneys’ adequacy for appointment as class counsel on the basis of their experience, knowledge of the law, time spent investigating potential claims, and the resources they are willing to commit to representing the class. To instantiate our proposed duty, Rule 23(g)(1)(A) should be amended to additionally require courts to consider the efforts and resources counsel has and is willing to commit to communicating with class members in assessing their adequacy to serve as counsel.

In practice, this would require counsel, at the beginning of litigation and prior to being appointed as interim counsel, to detail their past and intended communication efforts with class members. This would include the resources counsel would devote to communication efforts and their strategy for making those efforts. Later in the litigation, during class certification, counsel would have to specify the time, effort, and resources they spent communicating with class members in order for the court to determine whether they indeed adequately and fairly represented the interests of class members during the litigation. Like other adequacy and communication considerations in Rule 23, this duty would be flexible and reasonable (i.e., not absolute), taking into account financial feasibility, technical facilitations, the extent to which class members are identifiable, and other practicalities specific to each class action.

To ensure that such a duty is not diffused across counsel, courts should consider appointing counsel specifically charged with executing this duty. In larger class action lawsuits, notably multidistrict litigation suits involving class claims, 33 Contrary to popular notions, multidistrict litigation (“MDL”) and class actions are not mutually exclusive. Indeed, there is much overlap: a significant percentage of MDLs involve class claims subject to Rule 23 and its requirements. While it is increasingly common for larger class actions to be consolidated by the Judicial Panel for Multidistrict Litigation into MDLs, gaining MDL status does not remove such suits from Rule 23’s reach. That said, a subset of MDLs do not involve class claims, and this Essay is not directed at this subset of MDLs. for example, courts frequently appoint lead counsel, a lead plaintiffs’ committee, and liaison counsel. 34 See Dana J. Alvaré, Temp. Univ. Beasley Sch. L., Vying for the Lead in the “Boys’ Club” 2018 Update 3, 6 (2017) (discussing various tiers of appointments in class action multidistrict litigation suits); del Riego, supra note 12, at 84 (discussing the role of liaison counsel). General liaison counsel is responsible for administerial communications to other plaintiffs’ counsel and the court, and special liaison counsel are appointed to communicate with counsel in parallel state litigation or with the government. 35 del Riego, supra note 12, at 84. But courts have not appointed counsel to communicate with the largest and most important constituents in the litigation: class members. Thus, we propose the appointment of a new position—class member liaison counsel—that would be responsible for communication with class members.


Class actions strike an ironic pose: designed to provide justice for the unrepresented and built to increase judicial efficiencies, they appear to be struggling to fully fulfill these aims. While some scholars have argued that significant reform is necessary, 36 Richard Marcus, Revolution v. Evolution in Class Action Reform, 96 N.C. L. Rev. 903 (2018). we aver that nothing quite so disruptive is needed. Rather, by creating incentives for class counsel to communicate with class members and strategically leveraging the tools of the empirical sciences, counsel can adequately communicate with class members. This would lessen the representational dilemma, decrease class defections, and further efficiency aims.

This is not a panacea. Establishing communication channels between class members and class counsel will not solve all class counsel-class member agency problems. Nor will it guarantee that counsel will necessarily operate in class members’ best interests over the course of litigation. But it is a much-needed and eminently feasible route for forging class actions that better accord with traditional understandings of legal representation.

* Alissa del Riego is an Assistant Professor, University of Miami Herbert Business School. J.D., Harvard Law School; B.A., University of Miami. Joseph Avery is Assistant Professor, University of Miami Herbert Business School, University of Miami Department of Psychology. J.D., Columbia Law School; Ph.D., Princeton University; M.A., Princeton University; B.A., New York University.

“Megaphon” by floeschie is licensed under CC BY 2.0. To view a copy of this license, visit