by Jacob Pasternack *
State and federal governments have increasingly sought to regulate social media companies’ algorithm design to address concerns about content amplification and potential social media addiction. However, such efforts face significant First Amendment obstacles. This Contribution argues that current and proposed algorithmic laws are constitutionally dubious for four reasons: (1) Moody v. NetChoice’s protection of content-curation likely extends to companies’ decisions about all algorithmic recommendation systems, which involve human editorial judgments in training data selection and algorithm design; (2) algorithmic code possesses expressive purpose and constitutes protected speech; (3) despite platforms’ commercial objectives, their decisions about their algorithms do not qualify as commercial speech warranting lesser judicial scrutiny; and (4) “addictive” features remain presumptively protected speech.
The federal government and many state governments have either passed or are actively debating regulatory measures to minimize the perceived ills and excesses of social media algorithms,1 such as California’s “Protecting Our Kids From Social Media Act”2 and New York’s “Stop Addictive Feeds Exploitation for Kids Act.”3 However, as the Supreme Court warned in Moody v. NetChoice, LLC,4 these laws likely violate the First Amendment. This Contribution will seek to explain why, although Moody did not explicitly extend its reach to all social media algorithms, social media companies should receive First Amendment protection for all of their choices involving their algorithms.
First, Moody does not provide a principled reason for its carve-out of “automatic” and “personalized” algorithms, and the rest of the majority’s logic, alongside the mechanics of how algorithms work, indicates that social media companies should receive First Amendment protection for all of their editorial decisions regarding their algorithms. Second, all circuits to rule on the issue have held that “expressive” code is protected by the First Amendment,5 and if all algorithms are built on code that is expressive, as this Contribution argues, algorithms cannot be merely functional. Third, the choices social media companies make about their algorithms are not merely forms of commercial speech entitled to lesser protection. Last, “addictive” speech is not a category of expression that is exempt from First Amendment protections.
First, the First Amendment protections that Moody gave to social media companies’ editorial decisions regarding some kinds of algorithms reasonably extend to all algorithms, even those that are automatic and personalized. The First Amendment is the most well-known of the constitutional amendments: “Congress shall make no law . . . abridging the freedom of speech.”6 The First Amendment only protects expressive activity; thus, if something is not expressive, it is not protected.7 Accordingly, the key issue when applying the First Amendment to social media companies’ decisions regarding their algorithms is whether social media companies’ choices involving their algorithms involve constitutionally protected expressive activity.
Thus far, governments have employed two means of regulating algorithms: (1) regulating the kind of information, or content type, that an algorithm recommends and (2) regulating the method by which information is recommended.8 The former likely draws unconstitutional distinctions between certain messages, whether aimed at misinformation, deepfakes, or any other protected9 content that the government perceives as harmful.10 The latter—the focus of this Contribution—as partially established in the current landmark case on algorithms-as-speech, Moody v. NetChoice, LLC, unconstitutionally impacts the editorial decision-making of social media platforms.
Moody dealt with a Florida law that attempted to prevent social media companies from designing their algorithms in such a way that would cause them to “deprioritiz[e] posts by or about political candidates,” as well as a Texas law which imposed similar restrictions to prevent the “deboosting” or denial of “equal access or visibility to” content posted on social media platforms.11 Both of these measures were taken to counter what Florida and Texas perceived as the “skew[ing of social media feeds] against politically conservative voices.”12 While the main thrust of its decision was procedural, the Moody Court expressed skepticism of such laws because “the editorial judgments influencing the content of [social media] feeds are . . . protected expressive activity,” and states “may not interfere with [editorial] judgments simply because [they] would prefer a different mix of messages.”13 Expressive activity “includes presenting a curated compilation of speech originally created by others.”14 Thus, a platform directing its algorithm to promote, or demote, certain types of content is an “editorial choice” that is “expressive activity of its own.”15 The government “may not impose” upon how platforms “are selecting and moderating content,” as when platforms “decide which third-party content [social media] feeds will display, or how the display will be ordered and organized, they are making expressive choices.”16 Accordingly, laws impacting how an algorithm is designed burden the freedom of speech, implicating the First Amendment.
In a footnote, the Moody majority stated that its decision did not extend to “feeds whose algorithms respond solely to how users act online—giving them content they appear to want.”17 While the Court withheld its judgment on those kinds of algorithms, the logic of Moody reasonably extends to cover social media companies’ editorial decisions regarding all algorithms, and there is little reason to suspect that the excised algorithms should warrant any lesser First Amendment protection than those directly addressed in Moody.
The decision to withhold judgment seems to stem in part, if not in whole, from concerns raised by Justice Barrett in her concurrence. Justice Barrett posited that “fully” automated algorithms “may attenuate the connection between content-moderation actions and human beings’ constitutionally protected right[s].”18 It is unclear, however, why that should be the case. Barrett’s hypothetical simply is not how algorithms work: platform owners cannot just “hand the reins to an AI tool and ask it simply to remove ‘hateful’ content” divorced from human editorial judgment because algorithms do not pop into existence ex nihilo.19 Today, most algorithms are trained on datasets that allow them to learn “what content users will find compelling and surface it for them.”20 Thus, in Barrett’s example, any algorithm that is set loose on “hateful content” must have been trained on a dataset selected by the algorithm’s developers that inform it about what kind of content the developers consider “hateful.”
The same is true for algorithms that automatically present “to each user whatever the algorithm thinks the user will like.”21 Datasets establish the starting instructions for how the algorithm should function when analyzing any particular user’s data. No algorithm is independently “choosing [its] own goals and what substantive communications will achieve those goals.”22 Human developers will choose the initial training datasets that most accord with the type of content they want to promote.
While it is true that many algorithms incorporate users’ “data and then suggest content that the algorithms anticipate will be engaging[] or ‘interesting’ to users,” it does not follow that “these number-crunching capabilities” are not “expressive judgment[s].”23 Algorithms are not “passive receptacles of third-party speech . . . that merely emit what they are fed,” because if this were true, social media algorithms would predictably converge on showing the same type of content to the same person even accounting for the differences in the uses of each platform, but different platforms come to different conclusions: Facebook simply provides a user with content different from what X provides to that same user because each platform “is making all sorts of [different] judgments in determining what its customers want.”24 Algorithms may be “black boxes” where “even the programmers who built them do not know how they get from input to output,”25 but so “long as humans are making substantive editorial decisions,” which they do through setting “rules [that are] embedded in the program,” “inserting computers into the process does not eliminate the communication” that was intended at the outset.26
Even if Justice Barrett’s concerns about “fully” automated algorithms were accepted, this creates major line-drawing issues: at what point does an algorithm become fully automated? If any human involvement means that the algorithm is not fully automated, then no algorithm is fully automated, and Barrett’s concerns are misplaced. But if Justice Barrett means a situation where there is no human involvement after a certain point in an algorithm’s lifespan—say, at the time of deployment—why should that time, or any other, be the dividing line? And what happens if an algorithm is running automatically for many years, but then the developers decide they wish to tweak it? Is it no longer automatic? If it is not, can it become automatic again, and if so, when, and why? Ultimately, coding an algorithm to prioritize posts based on what is “most popular” is no different from finding “a trained monkey to perform this task” because “the same thing is communicated when the process is automated.”27 Dividing protected and unprotected algorithms based on their autonomy is riddled with such questions, and, as a policy matter, any chosen threshold could likely easily be circumvented by algorithm developers. Ultimately, it is “difficult to see how anything of constitutional significance could turn on this distinction.”28
Furthermore, the creators of an algorithm intend to “convey a particularized message” that has a great likelihood of being “understood by those who view[] it.”29 Social media companies intend to communicate that “we think whatever [we are showing you] is interesting.”30 While the algorithm might be the entity presenting content to a user, the type of content that the algorithm presents is predicated on editorial choices made by developers. Social media companies do not “merely rely[] on predictive modeling” to come to these decisions: they deliberately build or choose models that are best optimized toward what they think will present the most relevant content.31 This is an editorial decision. Furthermore, at least one circuit has recognized that the expressive nature of an algorithm does not change even if the algorithm is automated.32 And, even if one cannot divine a “clear expression” or “the values and viewpoints” of a platform from its personalized feeds for each of its millions or billions of users, “in the aggregate they give the feed a particular expressive quality.”33
Thus, Moody plausibly extends to automatic, personalized algorithms, as the development of these algorithms involves many of the same editorial decisions that the laws at issue in Moody encroached upon. And, if Moody extends to personalized or automatic algorithms, then many pending state laws may be doomed from the start, as they often seek to regulate decisions that companies make regarding their algorithms in unconstitutional ways.34
Second, algorithms are as expressive as they are functional, and their creation and development implicate editorial judgments that the First Amendment protects. In light of Moody, some have suggested that regulating social media algorithms is akin to regulating a product, rather than regulating speech—the “functional” argument.35
However, regulating any feature of an algorithm is regulating speech; it is not merely regulating “computer code that executes a function.”36 This is because the function of the computer code underlying algorithms “involves the expression or communication of ideas in a way that implicates the First Amendment.”37 Whether, and to what extent, code is speech is currently unsettled, but most courts that have faced this issue have decided that only “purely functional code with no expressive purpose, use, or intent” is “not covered by the First Amendment.”38 Because the code underlying algorithms shapes the type of content that will be prioritized and how it will be prioritized, it has an expressive purpose, use, or intent and comprises “editorial function[s]” that are themselves “an aspect of speech.”39 It is likely true that many algorithms are designed to “keep users ‘engaged’ based on the data they are constantly analyzing,” but this does not make such algorithms strictly “functional” because the deliberate “manipulation of listeners [or viewers] aimed at keeping them engaged and likely to pay the speaker[] is likely as old as the first songs or epic poems.”40
Third, and for similar reasons, social media companies’ choices about their algorithms are not a form of commercial speech. Commercial speech is expression that “does no more than propose a commercial transaction,” and laws burdening it are subject to a lower level of judicial scrutiny.41 Even if “the underlying motive” of social media companies in optimizing their algorithms to “spark[] more meaningful engagement is to . . . generat[e] more money for the company,” is this not true of many forms of communication?42 An author may write a book to “create public discourse and even political debate,” but if the author knew that his book would not be profitable relative to the cost of publication, he might not write it.43 And what of social media influencers, whose entire online persona, filtered through one’s “personal brand,” achieves both expressive and commercial aims? Would we dare say that, because these people have motivations that “mix[]” economic and communicative goals, the government can regulate their speech more easily?44 We would not, and the same reasoning should apply to social media algorithms. It “has not mattered in the past whether [a speaker] was merely responding to a market opportunity or was expressing [his] own subjective preferences,” and it should not matter now.45
Last, “addictive” speech is not a category of speech exempted from the First Amendment. Governments have also passed or are currently contemplating laws to limit the “addictive” features of social media algorithms. Or, to put it simply, governments seek to regulate certain elements of social media algorithms because “personalized feeds work too well.”46 This too is regulating speech, meaning that such laws must face some kind of First Amendment scrutiny.
Imagine that, instead of TikTok inundating your feed with content blessed by the algorithm to keep you “addicted,” you woke up one morning to your TikTok feed being comprised entirely of photos and videos of paint drying. TikTok retains all of its allegedly addictive features: infinite scrolling, push notifications, like counts, video autoplay, and the like. Even still, once you realize that TikTok is only providing you with content of paint drying, you would likely stop using TikTok, and if you realized you would only receive such content on this platform going forward, you would likely abandon the platform entirely, addictive features be damned.47 Thus, the “addictive” features of a social media algorithm are only as addictive as the quality, relevance, and visual appeal of the content allow them to be.48 This means that, in attempting to regulate “addictive” features of social media algorithms, what these laws really target is the content recommended by these algorithms.
Professor Eugene Volokh argues that the addictive aspects of social media algorithms should be treated no differently under the First Amendment than religious speech, which itself has many addictive aspects.49 As just one of his many examples, if “a ‘like’ button on social media makes people ‘kind of addicted to the feedback,’ one might equally say that positive reinforcement that formerly lonely people get from a new religious community” would have the same effect.50 From here, he argues that because “psychological addiction can’t justify restrictions that interfere with [religious] behavior,” it also can’t “justify restrictions on behavior presumptively protected by the Free Speech” clause.51
Volokh concludes that the addictive aspects of social media algorithms “are presumptively protected by the First Amendment” for both the platforms and their users, who also have “interests as listeners in the material produced using the design features.”52 Ultimately, what even counts as “addictive” is difficult to determine, as what one person might view as “addictive,” Volokh argues, another might view as “features [that] make [a platform] engaging and effective.”53
Some have suggested that narrower, piecemeal laws that target one addictive aspect at a time might survive First Amendment scrutiny based on the aforementioned functional argument.54 While Volokh does not close the door on all forms of regulation targeting algorithms, he contends that, even if such restrictions might not unconstitutionally discriminate between messages, they would only be permissible if a “significant amount of ‘addiction’ is indeed prevented by such restrictions.”55 This, clearly, requires significant factual evidence that simply has yet to be produced,56 and would also have to overcome the aforementioned difficulties with the functional argument. The recent settlement between Meta and multiple states admits of these challenges, as it requires that Meta implement many changes to Facebook’s and Instagram’s “addictive” features and algorithms while barring Meta from challenging “that any term [in the settlement] is unconstitutional or is preempted by, or in conflict with, any current or future law.”57 While this may be standard settlement language, it may also be a tacit acknowledgment of the dubious legality of such measures if they were implemented in legislation.
Ultimately, governments cannot and should not get involved in regulating social media algorithms. While it may seem beneficial to constrain how social media companies design their algorithms and present content, allowing the state this level of control further centralizes the power to decide how certain types of content are presented, which incidentally, if not invariably, impacts what type of content can be presented. Such laws would also further entrench the current market players in their cultural and economic positions by increasing the regulatory burden for competitors to enter the market, thus strengthening the very same “Big Tech” firms legislators claim they wish to weaken.58 In the end, this all serves to grant the state, both independently and through near-oligopolist actors, power over the “fundamental codes of a culture . . . [and] establish for every man . . . the empirical orders with which he will be dealing and within which he will be at home.”59
* Jacob Pasternack is a J.D. Candidate (2027) at New York University School of Law.
1. An “algorithm” will refer to a “technical means of sorting posts based on relevancy instead of publish time, in order to prioritize which content a user sees first according to the likelihood that they will actually engage with such content.” Maria Alessandra Golino, Algorithms in Social Media Platforms, Inst. for Internet & the Just Soc’y (Apr. 24, 2021), https://www.internetjustsociety.org/algorithms-in-social-media-platforms. This necessarily excludes sorting posts based on chronology or any content-agnostic grounds, as those kinds of feed orderings are not expressive and thus fall outside the scope of the First Amendment.
2. S.B. 976, 2023-2024 Reg. Sess., ch. 988 (Cal. 2024).
3. N.Y. Gen. Bus. Law §§ 1500-1506 (McKinney Supp. 2025).
4. 603 U.S. 707 (2024).
5. See infra note 38.
6. U.S. Const. amend. I.
7. Spence v. State of Wash., 418 U.S. 405, 409 (1974) (noting that conduct must be “sufficiently imbued with elements of communication” to “fall within the scope of the First . . . Amendment[].”).
8. Press Release, New Civil Liberties Alliance, NCLA Reaches Historic Settlement, Strikes Major Blow Against Government’s Social Media Censorship (Mar. 24, 2026), https://nclalegal.org/press_release/ncla-reaches-historic-settlement-strikes-major-blow-against-governments-social-media-censorship/ (describing a settlement prohibiting federal health agencies from “threatening social media companies into removing or suppressing constitutionally protected speech” and barring “these government authorities from directing or vetoing the companies’ social media content moderation choices”).
9. Some types of expressive activity, including but not limited to inciting others to commit imminent lawless action, making true threats, or activity deemed obscene, is not protected. See, e.g., Brandenburg v. Ohio, 395 U.S. 444 (1969) (incitement); Watts v. U.S., 394 U.S. 705 (1969) (true threats); Miller v. California, 413 U.S. 15 (1973) (obscenity); see also Free Speech Coalition, Inc. v. Paxton, 606 U.S. 461, 496–97 (2025) (holding that to protect children from seeing content obscene for children, adults’ access to legal First Amendment content can be burdened).
10. See Ira Rubinstein & Tomer Kenneth, Taming Online Public Health Misinformation, 60 Harv. J. on Legis. 220, 251–54 (2023) (noting that “laws restricting ordinary, private citizens from producing, receiving, or sharing” misinformation “would likely not survive First Amendment review” and focusing on methods of soft regulation as possible solutions); Moody, 603 U.S. at 738; see also Stuart M. Benjamin, Algorithms and Speech, 161 U. Pa. L. Rev. 1445, 1464 (2013).
11. Moody, 603 U.S. at 721.
12. Id. at 719.
13. Id. at 744.
14. Id.
15. Id. at 731–32.
16. Id. at 740, 743.
17. Moody, 603 U.S. at 736 n.5.
18. Id. at 746 (Barrett, J. concurring) (emphasis in original).
19. Id.; see also Benjamin, supra note 10, at 1479 (recognizing that “programs do not develop the models and analyses on their own”).
20. Chris Meserole, How Do Recommender Systems Work on Digital Platforms?, Brookings (Sept. 21, 2022), https://www.brookings.edu/articles/how-do-recommender-systems-work-on-digital-platforms-social-media-recommendation-algorithms/.
21. Moody, 603 U.S. at 746 (Barrett, J. concurring).
22. Benjamin, supra note 10, at 1481.
23. Order Denying Mots. for Prelim. Inj. at 8, Meta Platforms, Inc. v. Bonta, Nos. 5:25-cv-09792-EJD, 5:25-cv-09789-EJD, 5:25-cv-09795-EJD (N.D. Cal. Aug. 5, 2026), ECF No. 126, https://epic.org/wp-content/uploads/2026/08/0106.-08-05-2026-ORDER-DENYING-MOTIONS-FOR-PRELIMINARY-INJUNCTION.-Signed-by-Judge-Edward-J.-Davila-on-852026.-Rela.pdf; see also Bernstein v. U.S. Dept. of J., 176 F.3d 1132, 1142 (9th Cir. 1999) (finding that encryption code, which utilizes complex and lengthy numerical calculations (“number-crunching”) to scramble data, can be expressive and thus be protected by the First Amendment).
24. Order Denying Mots. for Prelim. Inj., supra note 23, at 8, 10; Benjamin, supra note 10, at 1470. See also Michelle Martin, Social Media Algorithms in 2026: How They Rank Content, Hootsuite (May 7, 2026), https://blog.hootsuite.com/social-media-algorithm/ (explaining how social media algorithms prioritize different things); Katie Fleeman, Different Platforms, Different Audiences, Knight Sci. Journalism @ MIT, https://ksjhandbook.org/social-media-reader-engagement/different-platforms-different-audiences/ (last visited Aug. 19, 2026) (exploring the differing most common uses of major social media platforms).
25. Marriah Binek, TikTok, the First Amendment, and Regulating Speech on the Internet, 31 Roger Williams U. L. Rev. 113, 119 (2026).
26. Benjamin, supra note 10, at 1491
27. Benjamin, supra note 10, at 1465.
28. Id. at 1469.
29. Spence, 418 U.S. at 411.
30. Order Denying Mots. for Prelim. Inj., supra note 23, at 9.
31. Id. at 8 (emphasis added).
32. See Anderson v. TikTok, Inc., 116 F.4th 180, 184 (3d Cir. 2024) (holding that recommendations from TikTok’s automated algorithm are “TikTok’s own expressive activity”).
33. Rubinstein & Kenneth, supra note 10, at 278; Moody, 603 U.S. at 738; see also Benjamin, supra note 10, at 1464 (noting that “many communications that the Supreme Court treats as speech do not express a clear viewpoint”).
34. See, e.g., N.Y. Gen. Bus. Law §§ 1500–1508 (2026); Cal. Health & Safety Code §§ 27000–27007 (2026).
35. See, e.g., Brett Frischmann & Peter Ormerod, Regulating Manipulative Design Is Not Preempted by CDA 230 or the First Amendment, 75 Emory L.J. 1101, 1140 (2026).
36. Niamh Kathleen Keane, Addictive by Design: Can States Regulate Social Media Algorithms in the Face of Free Speech Challenges?, 34 U. MIA Bus. L. Rev. 419, 438 (2026).
37. Def. Distributed v. Atty. Gen. of New Jersey, 167 F.4th 65, 82 (3d Cir. 2026).
38. Id.; see, e.g., Bernstein, 176 F.3d at 1142 (similar); cf. Junger v. Daley, 209 F.3d 481, 485 (6th Cir. 2000) (holding that source code is always protected by the First Amendment).
39. Moody, 603 U.S. at 731; see also Arvind Narayanan, Understanding Social Media Recommendation Algorithms, 23-01 Knight First Amend. Inst. (Mar. 9, 2023), https://knightcolumbia.org/content/understanding-social-media-recommendation-algorithms (finding that any shift in an algorithm “has major impacts . . . on the type of content that’s amplified”).
40. Eugene Volokh, Addiction to Constitutionally Protected Activity: Speech, Press, and Religion, 75 Emory L. J. 1175, 1193 (2026).
41. Virginia State Bd. of Pharm. v. Virginia Citizens Consumer Council, Inc., 425 U.S. 748, 776 (1976) (Stewart, J. concurring); see also C. Hudson Gas & Elec. Corp. v. Pub. Serv. Comm’n of New York, 447 U.S. 557, 573 (1980) (Blackmun, J. concurring) (agreeing that commercial speech should be examined with intermediate scrutiny).
42. Haley Bernstein, Reconciling Section 230 and the First Amendment: Should Social Media Companies be Held Liable for the Consequences of Their Recommendation Algorithms?, 19 J. Bus. & Tech. L. 373, 393 (2023); see also Benjamin Laufer & Helen Nissenbaum, Algorithmic Displacement of Social Trust, Knight First Amend. Inst., at 42 (Nov. 29, 2023), https://knightcolumbia.org/content/algorithmic-displacement-of-social-trust.
43. Bernstein, supra note 42, at 393.
44. See Ellen S. Yost, Social Support for Terrorists: Facebook’s “Friend Suggestion” Algorithm, Section 230 Immunity, Material Support for Terrorists, and the First Amendment, 37 Santa Clara High Tech. L.J. 301, 329–30 (2021) (examining whether social media algorithms are speech).
45. Benjamin, supra note 10, at 1475.
46. NetChoice, LLC v. Bonta, 152 F.4th 1002, 1009 (9th Cir. 2025).
47. Cf., e.g., Thomas Reimer, Environmental Factors to Maximize Social Media Engagement: A Comprehensive Framework, 75 J. Retailing & Consumer Servs., 103458 (2023), https://doi.org/10.1016/j.jretconser.2023.103458 (showing how content types impact social media engagement).
48. See Mufadhol Mufadhol et al., The Influence of Posting Frequency, Content Quality, and Interaction with Customers on Social Media on Customer Loyalty in a Start-up Business, 2(2) West Sci. Bus. & Mgmt. 582, 582–94 (2024), https://doi.org/10.58812/wsbm.v2i02.966 (discussing how the quality of content impacts user retention on social media platforms).
49. See generally Volokh, supra note 40.
50. Id. at 1184.
51. Id. at 1196.
52. Id. at 1201–02. The First Amendment protects both freedom of speech as well as the “right to receive information and ideas, regardless of their social worth.” Stanley v. Georgia, 394 U.S. 557, 564 (1969); see Pac. Gas and Elec. Co. v. Pub. Utilities Commn. of California, 475 U.S. 1, 8 (1986) (reaffirming that “the First Amendment protects the public’s interest in receiving information”). At least one court has contended that completely banning personalized recommendation algorithms would burden users’ First Amendment rights “by making it significantly more difficult to access speech a user wishes to receive, so a state probably could not constitutionally ban algorithms for the organization of speech (on social media or elsewhere) altogether.” NetChoice v. Griffin, 812 F. Supp. 3d 905, 925 (W.D. Ark. 2025); cf. NetChoice v. Murrill, 812 F. Supp. 3d 594, 643 (M.D. La. 2025) (finding an age-verification law “burdens prospective and current users’ ability to access and to engage in speech on covered platforms” (emphasis added)).
53. Volokh, supra note 40, at 1206.
54. See, e.g., Frischmann & Ormerod, supra note 35.
55. Volokh, supra note 40, at 1211.
56. Studies have consistently documented addictive-like behavioral patterns and linked them to anxiety, depression, and other negative outcomes, but the direction of the causal arrow has not yet been determined: longitudinal research has shown “addictive” social media use may worsen mental health, but preexisting poor mental health may drive “addictive” social media use. See, e.g., Christopher J. Ferguson, Do Social Media Experiments Prove a Link with Mental Health: A Methodological and Meta-Analytic Review, 14 Psych. Popular Media 201, 201–06 (2024), https://doi.org/10.1037/ppm0000541; Laura Lemahieu et al., The Effects of Social Media Abstinence on Affective Well-Being and Life Satisfaction: A Systematic Review and Meta-Analysis, 15 Sci. Rep. 7581 (2025), https://doi.org/10.1038/s41598-025-90984-3.
57. Proposed Consent Judgment and Settlement Agreement at 7, In re Social Media Adolescent Addiction/Personal Injury Prods. Liab. Litig., MDL No. 3047 (N.D. Cal. Aug. 26, 2026).
58. See, e.g., Mike Masnick, Meta’s Settlement Isn’t Tech’s “Big Tobacco” Moment, Techdirt (Oct. 6, 2026, 3:00 PM), https://www.techdirt.com/2026/10/06/metas-settlement-isnt-techs-big-tobacco-moment.
59. Michel Foucault, The Order of Things: An Archaeology of the Human Sciences xx (Vintage Books 1994) (1966).