Digital Platforms and Epistemic Authority

Introduction

The Supreme Court has been confronted with several cases questioning social media platforms’ authority. In Murthy v. Missouri,^[1] two states and five social media users sued Executive Branch agencies and officials for allegedly pressuring social media platforms to restrict protected speech.^[2] This case involved social media platforms’ systems for restricting and labelling what many might call “health-related misinformation” during the COVID-19 pandemic.^[3] The plaintiffs argued that the social media platforms’ conduct injured them by censoring their speech and abridging their right to listen to third parties facing censorship.^[4] Their constitutional claim, which would not have been sustainable against private platforms, was brought against government agencies and officials for pressuring or encouraging platforms to suppress speech.^[5] Murthy v. Missouri raised questions of epistemic and political authority during a public health crisis and featured the major actors—the state, platforms, scientific experts, and publics—involved in the production and legitimation of authority. By epistemic authority, I refer to the capacity to shape or influence what others believe.^[6]

This Essay discusses online platforms’ power over the curation, presentation, and ordering of content and content creators, as well as their influence on the epistemic power of other institutions. I offer a relational framework for platforms’ power and corresponding authority, and suggest treating these as weak, proxy, or strong relative to other epistemically powerful entities and institutions like states, the press, scientific institutions, global civil society organizations, or the United Nations.^[7] Platforms’ relative power and authority might vary depending on the time and place in which they operate and make content decisions. For example, communicative platforms have weak or deferential epistemic power in China because political power shapes epistemic power and authority there.^[8] In contrast, Facebook had strong epistemic power in Myanmar, where users had access to few alternative sources of information and state actors had limited power.^[9] In the United States, communicative platforms have held borrowed or strong epistemic power because they are largely unregulated,^[10] with access to extensive First Amendment protection and immunity from intermediary liability through Section 230 of the Communications Decency Act.^[11] In Brazil, a new platform owner like Elon Musk overestimated X’s power and ignored the Brazilian supreme court’s orders to remove certain accounts from X, resulting in the platform being blocked nationwide.^[12]

Brian Leiter’s excellent work on platforms and epistemic authority examines the epistemic role of communicative platforms in detail, arguing that the internet has “undermined”—perhaps even “eradicated”—the credibility of epistemic authority.^[13] While this may be true of a particular time and jurisdiction, platforms’ authority and role can be configured differently elsewhere, or even under a different political regime in the United States. My more expansive, relational framing accounts for how power and authority fluctuate across states, places, times, and platforms. For example, a platform that holds great epistemic power in Myanmar may need to borrow epistemic power in the United States. Proxy authority from institutions carrying epistemic authority for health content might be seen as legitimate by groups who already trust Western medicine, and as propaganda by groups who do not.^[14]

Awareness of epistemic and political power, and of the relationships involved in constructing and recognizing authority, is a necessary step toward reinstating the epistemic authority of institutions. To do so, it may be necessary to address legitimate criticisms of institutionalized epistemic authority on the one hand and to limit the platforms’ exercise of their infrastructural power to manipulate users on the other.^[15] Merely requiring digital platforms to defer to outside institutions with trust deficits will not help users who treat these addictive platforms as knowledge infrastructure and whose distrust in institutions is exploited by extreme communities (who the platforms nudge them toward) that distrust epistemic authority on principle. Moreover, restricting platform manipulation limits how far platforms can dominate, erase, or shift institutionalized epistemic authority.

I. Communicative Platforms and Content Decisions

Communicative platforms play a curatorial role in determining what audiences see.^[16] This role is more visible on social media and search platforms, as it is possible to see what content is posted and how far it reaches. The curatorial role exists, albeit differently and discreetly, on generative AI platform-interfaces, which synthesize content into a single output rather than a ranked list.^[17] Private platforms make decisions about speech based on four sources of authority: their own, the state, the expert, and the public.^[18] In doing so, they can also validate or undermine these sources. I discuss social media platforms here because of the material accumulated about them, but similar negotiations of trust and authority take place over AI chatbots’ content.^[19]

This Part discusses these sources of political and epistemic authority over platform content, showing how they inform platform decision-making and where platforms draw authority from. For example, social media platforms went through phases: They initially claimed to exercise no authority over content, before claiming to exercise borrowed authority—from states and experts—over content.^[20] Digital communication platforms also invoke users’ rights to legitimize their exercise of curatorial authority, typically when their choices run counter to states and experts.^[21]

It is helpful to distinguish the role of platforms here. Social media and search platforms’ power and authority lie in the curation and ordering of content.^[22] Generative AI platforms also work with third-party data, but they process it differently and present it as their own synthesized output.^[23] Both are willing to discuss some external norms, such as removing non-consensual sexual media, that they will apply to content curation.^[24] However, they are rarely willing to discuss algorithmic features through which they hold users’ attention and target their vulnerabilities except when they are made to do so by courts or state regulators.^[25]

A. States and Platform Content

At the initial stages of the internet, web-based platforms seemed poised to claim they were outside the purview of state laws. This was the era of “cyberlibertarians,” and of John Perry Barlow declaring to states that they have “no sovereignty where we gather.”^[26] In this period, web-based platforms engaged in little to no content moderation, contesting states’ authority to require them to restrict content.^[27] Since the companies that owned these platforms were American, they were aided by the American state in this project of internet freedom.^[28] As Jack Goldsmith and Tim Wu predicted, this period of cyber liberty was short-lived, and states soon exerted power and authority over platform content.^[29]

However, as I have argued in prior work, while states can and do exercise authority over platform content, platforms choose whether or not and to what degree they accept states’ authority.^[30] In Thailand, Google might choose to follow the lèse-majesté law forbidding criticism of the king.^[31] In India, Facebook might choose to go beyond the law, proactively censoring content about Kashmiri freedom,^[32] speech protected by India’s constitution but stigmatized by the ruling party.^[33] In both cases, the platforms claim to act as proxies for the state, performing deference to the state’s authority when they characterize content as unlawful, in order to justify removing it. However, as the Indian example of overcompliance illustrates, this performative deference to the law belies the platform’s choice.^[34]

It is also evident from platforms’ engagement with states that platforms choose to comply with or contest the law depending on their perception of the state’s power, and the extent to which they see compliance as being in their interests. The platforms threaten to withdraw when Pakistan regulates them,^[35] but merely object when Australia does the same.^[36] In China, noncompliance is not an option.^[37] In Europe, which is composed of powerful states, American platforms push back through local litigation and lobbying, and seeking the intervention of the American government or other stakeholders when regulations threaten their business models and profits.^[38]

Whether a platform resists, complies or over-complies with states’ laws can also depend on the platform owner’s perception of his or her own power. Elon Musk, for example, tends to be aggressive in his willingness to challenge state power: He refused to comply with Brazilian law requiring X to block certain accounts accused of spreading misleading and harmful content, calling it censorship.^[39] In doing so, he claimed greater authority than the state over content decisions. However, it emerged that Musk was mistaken in his perception that he could overpower the Brazilian state, and he ultimately had to cede authority over content and pay the Brazilian fine.^[40] He is currently challenging the European Union (EU).^[41] Although he is unlikely to possess sufficient power to challenge the EU, the American government’s willingness to extend its power to him may weigh in his favor.^[42]

In sum, it appears that some states can successfully exercise political power over platforms. However, platforms’ economic and political power is growing. Authors like Rebecca MacKinnon and Paul Gowder have compared platforms to nation-states.^[43] I have shown in prior work that dominant platforms can successfully influence the framing and enforcement of law even in powerful states, since they have extensive monetary and political resources and can wait out unfavorable electoral cycles.^[44]

B. Experts and Platform Content

Platforms also seek out experts with epistemic power and authority to help them make or legitimize decisions about content. The widely known example is the Facebook Oversight Board.^[45] However, there are others, such as trusted flaggers and expert institutions like the World Health Organization during the COVID-19 crisis.^[46] Platforms often display their use of expert assessment in the context of controversial content or content likely to attract regulation. Unsurprisingly, this content includes threats to national security and child sexual abuse media,^[47] but extends to other content as controversies arise. Experts tend to form a part of platforms’ regulatory systems and have been included and featured in several platform self-regulatory strategies.^[48] The field of Trust and Safety, which develops, explains, and legitimizes platforms’ content decisions, is meant to professionalize social media content moderation.^[49]

Professor Jack Balkin explains that one might think of platform self-regulation in a triad in which platforms self-regulate when they fear state regulation.^[50] I have shown in prior work that although platforms are more fearful of powerful states and movements, cross-border solidarities between a group that might appear powerless (such as Myanmar’s civil society) and powerful actors (such as the United Nations and the New York Times) can push platforms to self-regulate.^[51] It appears that platforms might also self-regulate differently, allowing speech they previously restricted, in response to political power.

Social media platforms developed strategies leveraging the authority and legitimacy of experts, such as trusted-flagging, consultation, and oversight boards, to legitimize their self-regulatory models for user-content. They even hired experts full-time or commissioned reports from them. At their weakest, platforms publicly defer to experts to legitimize their content decisions, seeking epistemic authority by proxy. Two prominent ways in which platforms have done so are Twitter’s Trust and Safety Council and Facebook’s Oversight Board. Although these efforts were ultimately self-regulation, the public inclusion of experts was an effort to draw on the experts’ epistemic power and authority.

C. The Public and Platform Content

The public also has rights, at least in theory, over platform content.^[52] This authority presents itself in three significant ways. Two are visible and direct, with platforms performing deference to their users’ authority. This invocation of users’ authority implies that platforms recognize, again, at least in theory, that the public has some authority. The third indirect and less visible way is that politically empowered users and other affected parties exercise power indirectly over platform content. They can do so through other epistemically powerful institutions such as the state, expert bodies, the media, and civil-society institutions.

The first category of visible referencing of users’ (and future users’) democratic authority is by communicative platform owners who cite users’ rights to justify their company’s speech decisions.^[53] Since the public’s power and authority have to be mediated through institutions, it is difficult to measure. For example, when platforms reference their users’ rights or opinions, it is unclear whether this is actually connected with what the majority of users desire or need, or whether it is performative. It is a concern worth considering, given that platforms seek to legitimize their resistance to regulation and self-regulation by invoking the powerful claim of users’ rights to free expression. Although Elon Musk has recently been the most vocal example of this approach, Mark Zuckerberg and Jack Dorsey have also made public statements suggesting that their hands-off approach to content moderation is in the interests of their community, or that particular choices are meant to serve their communities.^[54] Arguably, platforms can attribute whatever goals they wish to users as users gradually lose access to institutions through which they can articulate and pursue their collective goals.

The second visible recognition of users’ authority is when major platforms involve users in content governance, offering them opportunities to voice their opinions and dissatisfaction with platforms’ decisions, arguably to vest their content decisions with sociological legitimacy.^[55] It is difficult to say to what extent these opinions impact policy but offering a pathway to express them and to contest platform decisions is inherently valuable.^[56] Most recently, the use of community notes has given users a voice, privileging their opinions over those of experts. Meta for example, has changed its practice of using external fact checkers to eliciting community notes.^[57] Engagement with users in this manner—placing the platform at the center and atomizing user engagement that would be more effective when exercised collectively—allows the platform to control the terms of engagement and who prevails.

The third way is when users and others access collective political power through institutions or movements, or institutions that can influence platforms.^[58] This is the only method that externalizes power from the platform, usually making users’ opinions and platforms’ responses transparent to the public. For example, when the press and advocates have publicly raised concerns about racism and sexism on platforms, they have sometimes triggered change.^[59] This form of influence is ideologically neutral. As they do with states, platform company-leaders choose which points of view are empowered to influence platforms’ content moderation.^[60] However, when dominant platforms gain sufficient political power—for example, through alignment with political leaders—they disregard movements, especially when those movements oppose the political leaders from whom platforms’ political power derives.

For publics to exercise relatively democratic, collective power, they need access to democratically oriented institutions that can mediate change. By this, I refer to institutions that are designed to hear, consolidate, and act on the publics’ voice. An example of why this is critical is that Meta, while choosing which publics’ voices influence it, initially ignored concerns raised by civil society about Facebook’s role in the Myanmar genocide.^[61] Once civil society mobilized the support of institutions like the international press and the United Nations, the company had to respond to the public’s epistemic authority on its impact on Myanmar.^[62]

II. Communicative Platforms and Epistemic Power

The different sources of authority—states, experts, and publics— platforms invoke to legitimize content decisions carry epistemic power. By epistemic power, I refer to the “privilege . . . given operative diagnoses of one’s standing with respect to knowledge possession, knowledge attribution, and/or knowledge production.”^[63] These actors’ influence fluctuates with their power relative to the platforms. Platforms comply, ignore, or challenge their authority depending on how they perceive each source of authority’s power. A relational framework accounts for this variation, treating platforms’ epistemic power not as a fixed property but as contingent on the support of more powerful actors, whose power and authority in turn fluctuate.

In his insightful essay, Professor Brian Leiter explained how the internet has discredited expertise and “eradicated” epistemic authority.^[64] I build on his work, shifting the frame to the platforms’ relationships to show how platforms’ power and claims to authority interact with and co-opt other sources of power and authority. This might include the legacy media, for example, which displaced other sources of epistemic power.^[65] Although platforms threaten expertise as Professor Leiter argues they do (especially for reasons I discuss in Part III), they also draw on it for proxy authority. However, they undermine this authority—even as they co-opt it—using their own infrastructural power. This undermining is enabled by the distrust institutions incur through exclusion and injustice.^[66]

A platform’s epistemic power may not be acquired through processes recognized and accepted by society, but it exists as a consequence of its infrastructural power. Epistemic power or status is always relative to someone else’s.^[67] In other words, it is relational.^[68] A relational account of epistemic power and authority permits acknowledgment that some groups have good reasons to distrust democratic institutions and institutional epistemic authority in the form of expertise. In my policy solutions, I offer a way for institutions to earn these groups’ trust by hearing them and accounting for their concerns and goals. In this Part, I discuss epistemic power, epistemic authority, algorithmic and infrastructural authority exercised by platforms, and how different actors’ power and authority interact with and co-opt each other.

A. Epistemic Power and Authority

The debates about the COVID-19 pandemic illustrate the significance of epistemic authority. Broadly, we refer to epistemic authority as acceptance of someone else’s validation of a factual assertion, using it as an analytical tool to resolve uncertainty.^[69] Someone who possesses epistemic authority can tell others what to believe and can expect to be believed because of their authority.^[70] For example, the world initially accepted the scientific classification of Pluto as a planet, and then its demotion to a dwarf planet, without requiring public investigation or consensus.^[71]

Professor Leiter treats epistemic authority as absolute, arguing that it helps “subjects believe what is true (or more likely to be true), and without that help, those subjects would be more likely to end up believing falsehoods or partial truths.”^[72] This is a necessary move given that his concern is about the internet discrediting institutional expertise.^[73] My goal is to more closely examine the process of discrediting expertise by identifying the external factors that enable it and distinguishing between the wholesale discrediting of all expertise and the arguably reasonable distrust of epistemic authority that arises within communities whose knowledge, experiences, and concerns are excluded from institutions. There is room between the extremes of treating institutionalized epistemic authority as irredeemably biased or as entirely objective and unsullied by ideology or politics.

Legal scholars have also taken a critical view of expertise, recognizing that it can be shaped by power, privilege, and performance.^[74] Most prominently, Professor Mark Tushnet has written that expertise has ideological and political content, and that consensus in science results from processes that involve the exercise of power.^[75] Professor Matt Lawrence has pointed out that industry can translate its economic and political power into epistemic power, “bending science.”^[76] None of this is surprising to scholars in any industry-dominated field in which funding and patronage encourage and spotlight some ideas while rendering others less visible.^[77] Kristie Dotson’s discussion of epistemic power permits a conversation about epistemic privilege and standing in the context of “knowledge possession, knowledge attribution, and/or knowledge production.”^[78]

Platforms can and do exercise epistemic power, as Professor Leiter worries that they might.^[79] This power is why people believe what chatbots or QAnon tell them.^[80] Epistemic power can exclude knowledge or order it in ways that are more political than scientific. In the context of platforms, it is also helpful to understand the forces that displace epistemic authority and the direction(s) in which they displace it.

B. Expertise and Trust

Entire bodies of work critique the way in which expertise is constructed and influenced.^[81] Institutional battles about discriminatory scientific theories are as old as institutions.^[82] When African-American communities distrust experts who tell them to get the COVID-19 vaccine, they distrust institutional expertise that has excluded them and experimented on them.^[83]

Socialization processes shape epistemic authority, and it is helpful to understand them.^[84] The boundaries of epistemic communities are “shaped in important part by social processes of exclusion and inclusion, and within the processes of inclusion, there are social processes creating hierarchy.”^[85] These processes are acted on by external forces such as political authority, whether institutional or social-movement driven.^[86] If a state does not allow people of a particular gender or race to go to university or study science, the epistemic community and field will form without their point of view. The same result might emerge from structural inequality excluding certain groups. Professor Charles has discussed how, in Virginia v. Black,^[87] Justice Clarence Thomas played a critically influential role, commanding epistemic deference as the only African-American Justice on the Supreme Court at the time.^[88]

It can also be helpful to think about the other kinds of authority that can substitute for, influence, and be mistaken for epistemic authority. Epistemic authority may be based on knowledge, but it can be at odds with and shaped by political authority, paternalistic authority, operative authority, charismatic authority, moral authority, and religious authority.^[89] This is the source of the debate between creationism and evolution: Epistemic authority conflicts with religious and sometimes political authority.^[90] Scientists have correlated resistance to vaccination, and presumably to speech favoring vaccination, with distrust of authority or distrust of the epistemic institutions producing and favoring vaccination. ^[91] They have also correlated this resistance with religion, free-market ideology, and pseudo-science and alternative epistemology, among other epistemic leanings.^[92]

This implies that although communicative platforms threaten epistemic authority, they can do so in part because of existing fissures in, and distrust of, institutionalized epistemic authorities. One answer is certainly to expect communicative platforms to defer to and amplify institutionalized epistemic authority. However, addressing the structural exclusion and elitism in expertise, which triggers the distrust, should also be a part of the solution.

C. Platforms’ Power and Authority

Platforms (with some exceptions) seek epistemic power in the interests of their core profit-making function: targeting advertisements to users.^[93] Their power and authority, being relational, is contingent on more powerful actors’ willingness to allow, recognize, and support their exercise of power.^[94] Platforms claim authority for their content curation based on three kinds of epistemic power: algorithmic, infrastructural, and human expert-driven. I show below how each of these is relational and fluctuates.

Algorithmic or data-driven authority derives from the platforms’ access to large datasets and the computational power to process the data, usually personalizing it for the user. ^[95] It is the platforms’ value proposition, upon which the market for targeted advertisements relies.^[96] The Supreme Court is verging on treating an algorithmically curated feed as an expressive product.^[97] Legal recognition of algorithmic authority can confer epistemic status, just as it has done for the press. States validate platforms’ algorithmic authority when they attempt to co-opt their algorithmic power, tacitly recognizing platforms’ authority to regulate content.^[98] By treating platforms as authoritative curators, law confers epistemic status on them, making it possible for platforms’ content curation to carry authority. But algorithmic authority is being called into question.^[99] Algorithmic outcomes and goals are being challenged in courts and investigated.^[100] If these challenges succeed and open algorithmic curation to question, platforms’ authority arising from algorithmic power may weaken as states stop recognizing and adapting to it.

Platforms also exercise authority derived from their infrastructural power and from their increasing intermediation of what content reaches society and in what form it reaches it.^[101] Aziz Huq has argued that platforms’ “unique ability to control and direct users’ attention,”^[102] or what Julie Cohen might call their infrastructural role,^[103] allows them to exercise a distinct form of epistemic power. This epistemic power extends beyond the platforms and affects knowledge institutions and industries like the legacy media that depend on online platforms to reach consumers.^[104] Natalie Helberger has argued that platforms acquire “systematic opinion power” and influence other actors, including the media.^[105] Because of their infrastructural role, the epistemic power of the companies that own major online platforms extends beyond the platforms themselves to the industry, influencing what the media reports and how politicians frame and shape public engagement.^[106]

Platforms’ epistemic power is growing because of their domination of communicative infrastructure, their gradual capture of culture, and their consequent capacity to render issues salient.^[107] The increasingly infrastructural role they play in communication further increases their capacity to render issues salient.^[108] Their epistemic power or status is always relative to someone else’s and is relational.^[109] It is clear, for example, that social platforms are accumulating power over the legacy-media industry.^[110] Although anti-trust efforts and structural proposals like the middleware idea might restrict platforms’ infrastructural power and influence, they have had a limited effect so far.^[111]

The third is human expertise, which platforms publicly reference as a proxy to justify the systems they use to curate content.^[112] Platforms rely on other institutions’ epistemic authority as proxies to validate or legitimize their epistemic authority over content decisions, through relationships with external experts.^[113] They seek credibility and legitimacy through processes and external experts whose institutional roles give them epistemic power.^[114] A prominent example is Facebook’s reliance on its expert Oversight Board to make politically fraught content moderation decisions and more generally to legitimize the platform’s content decision-making system.^[115] External experts, even those drawn into a company’s self-regulatory framework, have their own reputations and credibility to consider.^[116] They are not always compliant, as evident from the Oversight Board’s unexpected response to Facebook’s decision to deplatform President Trump and refer the matter to the Board.^[117] However, external experts can legitimize companies’ self-regulatory processes when the platforms come under state or public scrutiny.

Platforms claim authority over algorithmic decisions.^[118] They claim a sort of contingent or proxy authority for their content policies through their internal and external experts.^[119] While many of these experts possess epistemic authority because of their training or long-term affiliation with external institutions, the field of Trust and Safety is an internally constructed system aspiring to epistemic authority, which now exists within and outside communicative platforms.^[120]

Platforms’ systems for content decisions can depend on their relationship with political power. The Trump administration has recently questioned the authority of foreign fact-checkers to make decisions about American content and appears not to favor expert-driven content curation.^[121] In view of the political shift away from supporting expert-curation of platform content, major social media platforms are moving away from experts leveraging their epistemic authority to legitimize content decisions. Instead, they are encouraging outspoken users to weigh in on content decisions as a different way to legitimize them and to preserve the platforms’ epistemic power to curate content.^[122] Human expertise is also being challenged and replaced by algorithmic authority.^[123] As may be evident from the examples above, communicative platforms exert power and authority in relationships with others.

III. Platform Power and Epistemic Authority

Users may trust content on and from the platforms because they do not see themselves as having access to better sources. They might also distrust and resent the content if it runs counter to their intuitions or appears to take political positions that they do not agree with.^[124] This distrust might be validated by political leaders who are skeptical of either the platforms or the experts they cite. If the dissatisfied political leaders are in power, they might try to discipline platforms using the law.^[125]

This triggers false binaries in debates about whether platforms or states should exercise epistemic power. These binaries undermine the critical role of the institutionalized epistemic authority that Professor Leiter defends in his paper.^[126] The plaintiffs in Murthy v. Missouri subscribed to the binary of state-or-platform power, instead of allowing for epistemic authority.^[127] Although institutionalized epistemic authority is important, it often fosters mistrust for reasons I discussed in Part II.B. A relational account of platforms’ epistemic power and authority can enable policymakers to regulate platforms’ disruptive epistemic power more effectively. Regulators need to restrict the stronger, less visible sources of platforms’ epistemic power, namely their algorithmic and infrastructural power. This power, along with the contingent reliance on institutional expertise, needs to be disciplined into processes that serve democracy and which the platforms cannot bend entirely to their business model.

A. Platforms’ Epistemic Status

In some states, platforms are so powerful that they are difficult to regulate.^[128] In other states, they can be regulated for a time but have the resources to introduce friction into regulation through litigation, negotiation, lobbying, and pressure from other actors.^[129] There are also states in which they test and negotiate their power in relation to the state. ^[130] This is consistent with the idea that platforms’ power is relational and fluctuates. Since platforms carry more epistemic power than most individuals, a wealthy individual might purchase a platform to boost his own epistemic power.^[131] This is similar to, albeit at an arguably greater scale than, individuals purchasing newspapers for their epistemic power.

Platforms alter their processes of justifying their decision-making based on what powerful states value. This is clear from Meta’s shift from fact-checking to community notes after the 2024 presidential election.^[132] During the Biden Administration, Mark Zuckerberg was eager to defer to experts and display Meta’s proxy epistemic-authority in its management of pandemic-related content.^[133] This might have been directed at addressing concerns about social media platforms disrupting the influence of epistemic authority on the public during a public health crisis. However, more recently, Mark Zuckerberg has attributed his actions during the pandemic to pressure from the political leadership.^[134] If this is true, then the previous administration’s political power shaped Meta’s recognition and amplification of institutionalized epistemic authority. If it is false, then Mark Zuckerberg likely meant to distance Meta from the proxy epistemic-authority that the current political leadership does not recognize.

This is arguably a display of relative weakness. ^[135] However, it is also arguably an exercise of power, whereby platforms, through platforming and deplatforming viewpoints, stay aligned with political power while remaining indifferent to the source and goals of the political power the platform is complying with.^[136]

There is a reputational cost to platforms’ visible willingness to bend to power. However, this compliance is intended, presumably, to preserve the platforms’ algorithmic and infrastructural power, even if at the expense of legitimacy deriving from expertise. In other words, the platforms are flexible about their human or institution-driven content decisions but relatively resistant to regulation of their algorithmic and infrastructural power. This suggests that they are willing to assume weak status in the context of one kind of power in order to preserve and accumulate a different, more opaque form of power.

B. Democratizing Platforms’ Epistemic Power

When platforms rely on proxies to avoid regulation, they prefer to do so on their own terms.^[137] They incorporate third-party experts into their content decisions in ways that do not quite introduce accountability, but offer certification or an expert opinion, such as a health agency’s public statements, trusted flaggers, or an oversight board’s opinions.^[138] In other words, the third-party experts are intended to strengthen the platforms’ epistemic power by lending their decisions proxy epistemic authority from other institutions.

This relationship between platforms and third-party expertise is critical because platforms can also influence experts’ epistemic power and reach. When Elon Musk changed the Twitter algorithm, the changes accidentally spotlighted a menswear writer’s social media account, amplifying it to all Twitter users.^[139] As a result, Derek Guy has become a very influential commentator on men’s fashion such that even I, who know very little about how long women’s jackets ought to be, have absorbed his strictures for men’s jackets. Web-based platforms can influence the audiences of journalists, scientists, and their institutions and exert influence over the institutions whose authority supported their accumulation of epistemic power.

The platforms’ value proposition for advertisers is that they engage and persuade amenable users on advertisers’ behalf.^[140] If platforms are marketing machines, then this engagement and persuasion is a part of their design. It is a form of infrastructural power, not authoritative but persuasive and manipulative, that is leveraged by other actors, such as Russian propagandists and QAnon.^[141] Although platforms tend to be flexible about their content rules, they are rarely willing to negotiate over their infrastructural and algorithmic power and the authority they derive from both.

It is clear that despite platforms’ reliance on experts and their epistemic authority, the experts’ recommendations do not always make it to the level of algorithmic design.^[142] Platforms’ epistemic power through their algorithms is now being challenged through litigation in the United States, and the EU’s new laws requiring major platforms to conduct risk assessments.^[143] The EU has questioned TikTok’s harmful addictiveness and capacity to exploit users’ cognitive vulnerabilities.^[144] If states find a way to restrict platforms’ efforts to engage and manipulate their users, they will be addressing a significant dimension of platforms’ epistemic power.

Ensuring that platforms’ epistemic power is restricted such that it cannot be used in harmful ways by the platform, states, or third-party actors requires structural reform that prevents a few actors from exerting infrastructural control over the knowledge and conversations accessible to most of society. Platforms also need a democratic or legitimate process-revolution that stabilizes and democratizes how they wield their epistemic power. Proposals to treat them as fiduciaries are just the beginning.^[145] There need to be additional norms, including ones that require deference to institutions created for certain kinds of knowledge production.

Journalism legitimized itself by professionalizing and developing a norm of objectivity.^[146] However, this was clearly insufficient as the press has long had its epistemic power co-opted by authoritarian states and is now struggling to survive in most countries across the world. However, many models of professionalization and accountability were developed for the press—unions, development of reporting norms, press councils, editorial independence and more—that suggest that infrastructural epistemic power can be disciplined.

C. Legitimizing Expertise

For the reasons I discussed above, Professor Leiter is probably right when he writes that web-based platforms threaten institutionalized epistemic authority.^[147] Allowing platforms to accumulate strong epistemic power relative to other political and epistemic institutions undermines access to knowledge, information, and accountability. However, the project of restoring epistemic power to these institutions must include addressing distrust arising from legitimate concerns about their exclusionary history. A relational account of platforms’ epistemic power and authority allows us to acknowledge that some groups have good reasons to distrust institutional epistemic authority, in the form of expertise, and the democratic institutions that draw on it.^[148] These groups continue to feel alienated from the systems that create and validate institutional epistemic authority.^[149]

The longer-term response to these concerns would, of course, include addressing the problem of exclusion through bursaries, mentoring schemes, and institutional change through which epistemic institutions designed for a few groups might welcome and adapt to traditionally excluded groups.^[150] The goal would be to create a pathway to these institutions for excluded groups, and then ensure that they have the support necessary to flourish, join, and heterogenize previously exclusionary epistemic communities.

Scientific communities can mistake their political values for facts.^[151] It is possible to question the values and assumptions that they trigger by consulting the public about threats and risks that they are experiencing. This also helps with sociological legitimacy through procedural justice: Reorienting expert institutions and communities to treat the public and their subjects fairly, ethically, and respectfully might increase excluded communities’ trust in the institutions.^[152] What I propose is not a performative exercise. Expert institutions may be blinkered by their own composition. For example, if medical research is largely led and directed by men, they would need to listen to women to identify areas of research concerned largely with women’s bodies. Democratization builds trust in expertise and scientific institutions.^[153]

However, the most effective way for marginalized groups to reshape blinkered institutional consensuses is collective action. Communities have successfully persuaded epistemic experts to hear them and adapt to their concerns.^[154] Aziza Ahmed has written a detailed account of how feminists successfully reshaped the scientific and legal consensus about AIDS.^[155] Justice movements empower communities, emphasize leadership “from the bottom,” and ensure that experts account for communities’ experiences, priorities and frameworks.^[156]

Social media platforms can help experts understand communities’ concerns and help communities to mobilize for change. Experts, or others translating their findings, may be accustomed to relying on their epistemic authority to be believed. However, a more persuasive and inclusive approach will mitigate the trust deficit. Studies have found, for example, that working with community leaders can be critical to reaching underserved communities.^[157] Here, the community leaders’ relational authority serves as a proxy to translate epistemic authority to communities that might not trust the institutions that produce it. In the spirit of health-justice movements, they can communicate and advocate for the communities when the institutions fail to grasp their concerns. In other words, instead of permitting social media platforms to leverage distrust of institutions to claim epistemic power, the platforms can be leveraged to understand and address distrust, strengthening the institutions.

Conclusion

Whether a platform defers to the state, borrows legitimacy from experts, or asserts its own epistemic power depends on the political context in which it operates and the relative power of the actors surrounding it. This relational account matters because it reveals that platforms’ epistemic power is contingent and therefore susceptible to influence. The need to address this power is critical in the context of AI platforms, which convert epistemic power into authority by synthesizing knowledge into a single persuasive output, unlike social media platforms which offer alternatives.

Restoring epistemic authority to the institutions built to sustain it requires platform regulation as well as addressing the trust deficit that alienates people who ought to rely on institutional expertise. Regulators should target not just platforms’ algorithmic power and infrastructural power, in addition to their content rules, but also discipline how platforms exercise their power. This might include deference to democratically recognized institutions designed for certain kinds of knowledge production. However, efforts must also be made to enable the public—especially marginalized groups—to trust epistemic institutions. Without this trust, it will be relatively ineffective to require platforms to defer to these institutions.

1. . 144 S. Ct. 1972 (2024). ↑
2. . Id. at 1981. ↑
3. . Id. at 1981–82 (describing how social-media platforms applied their respective misinformation policies during the COVID-19 pandemic and the 2020 Presidential election season); see also World Health Organization, Infodemics and Misinformation Negatively Affect People’s Health Behaviours, New WHO Review Finds (Sep. 1, 2022), https://www.who.int/
   europe/news/item/01-09-2022-infodemics-and-misinformation-negatively-affect-people-s-health-behaviours–new-who-review-finds [https://perma.cc/8DQ3-68KY] (explaining that social media can accelerate spreading “health-related misinformation” due to overproduction of data). ↑
4. . Murthy, 144 S. Ct. at 1986. ↑
5. . Id. at 1984; see also Derek E. Bambauer, The Jawboning Cases End with a Bang Disguised by a Whimper, 2023–2024 Cato Sup. Ct. Rev. 157, 162 (2023) (“According to the plaintiffs, the Biden administration, writ large, had pressured these platforms to censor speech.”); Blake E. Reid, Jawbreaking and Counterboning, 58 Conn. L. Rev. (forthcoming May 2026) (manuscript at 13) (on file with author) (underscoring the difficulty of pleading a causal connection between jawboning and a private intermediary’s suppression of speech). ↑
6. . Brian Leiter, The Epistemology of the Internet and the Regulation of Speech in America, 20 Geo. J.L. & Pub. Pol’y 903, 906 (2022). ↑
7. . This relational framing is inspired by Jack Balkin’s writing. See generally Jack M. Balkin, Free Speech Is a Triangle, 118 Colum. L. Rev. 2011 (2018) (describing the twenty-first-century pluralistic model of speech regulation, and best practices for government regulation of modern digital infrastructure). ↑
8. . See Alex Chan, Guiding Public Opinion Through Social Agenda-Setting: China’s Media Policy Since the 1990s, 16 J. Contemp. China 547, 547–49 (2007) (explaining that in China’s media policy, “the guidance of public opinion” is a more subtle form of state-sponsored propaganda, whereby the party state enlists media outlets to shape what people think about); Wilfred Yang Wang & Ramon Lobato, Chinese Video Streaming Services in the Context of Global Platform Studies, 12 Chinese J. Commc’n 356, 361–62 (2019) (discussing how the Chinese government treats communication technologies as part of its political apparatus, and the media as a site for collective reform and change, rather than a private enterprise for free communication). ↑
9. . Victoire Rio, The Role of Social Media in Fomenting Violence: Myanmar 5 (2020), https://toda.org/assets/files/resources/policy-briefs/t-pb-78_victoire-rio_role-of-social-media-in-fomenting-violence-myanmar.pdf [https://perma.cc/XEK9-R8RY] (noting the ubiquity of Facebook in Myanmar and that the country’s parliament even used the platform to organize an election in 2018). See also Chinmayi Arun, Facebook’s Faces, 135 Harv. L. Rev. F. 236, 253 (2022) (discussing Facebook’s role in facilitating genocide in Myanmar). ↑
10. . See Cecilia Kang, Ryan Mac & Eli Tan, Meta and YouTube Found Negligent in Landmark Social Media Addiction Case, N.Y. Times (Mar. 25, 2026), https://www.nytimes.com/
    2026/03/25/technology/social-media-trial-verdict.html [https://perma.cc/36WQ-AD8W] (“[M]ost efforts to regulate social media in the United States have failed.”). ↑
11. . See Jack M. Balkin, The Future of Free Expression in a Digital Age, 36 Pepp. L. Rev. 427, 434 (2009) (explaining how Section 230 has created a vibrant culture of Internet freedom by giving intermediaries a privilege against certain lawsuits based on third-party content); Jack M. Balkin, Free Speech Versus the First Amendment, 70 UCLA L. Rev. 1206, 1226, 1236 (2023) (stating that First Amendment doctrine seems to protects digital companies’ content moderation, but not users’ speech affected by this content moderation); Danielle Citron & Benjamin Wittes, The Problem Isn’t Just Backpage: Revising Section 230 Immunity, 2 Geo. L. Tech. Rev. 453, 460 (2018) (arguing that courts have constructed Section 230’s immunity provision to provide “an immunity from liability far more sweeping than anything the law’s words, context, and history support”). ↑
12. . See Ana Ionova, Brazil Kept Tight Rein on Big Tech. Trump’s Tariffs Could Change That., N.Y. Times (Aug. 11, 2025), https://www.nytimes.com/2025/08/11/world/americas/brazil-big-tech-trump-tariffs.html [https://perma.cc/7UAY-F5N2] (“When Elon Musk ignored court requests . . . to remove accounts from X, Justice Moraes blocked the platform in Brazil.”); Ben Derico & Ione Wells, Brazil Lifts Ban on Elon Musk’s X after It Pays $5m Fine, BBC News (Oct. 8, 2024), https://www.bbc.com/news/articles/c5y06vzk3yjo [https://perma.cc/HVA4-H2GH] (reporting that X ultimately complied with Justice Moraes’s requests after the platform was blocked). ↑
13. . Leiter, supra note 6, at 935. ↑
14. . See Dawn L. Holford, Angelo Fasce, Thomas H. Costello & Stephan Lewandowsky, Psychological Profiles of Anti-Vaccination Argument Endorsement, 13 Sci. Reps., July 2023, at 2 (discussing the underlying psychological attributes contributing to vaccine hesitancy, including conspiratorial ideation); Angelo Fasce, Phillip Schmid, Dawn L. Holford, Luke Bates, Iryna Gurevych & Stephan Lewandowsky, A Taxonomy of Anti-Vaccination Arguments from a Systematic Literature Review and Text Modelling, 7 Nature Hum. Behav. 1462, 1464 (2023) (documenting vaccine skeptics’ mistrust in “various authorities and communicators about vaccines”); Amudalat Ajasa, Amanda Paule & Ian S. Brundige, For Some Black Americans, Vaccine Hesitancy Is Just One Part of a Legacy of Mistrust, Guardian (Aug. 20, 2021), https://www.theguardian.com/society/2021/aug/20/black-americans-covid-19-vaccine-hesitancy-mistrust [https://perma.cc/2NLY-KX5Y] (contextualizing vaccine hesitancy among Black Americans in the history of their mistreatment by medical institutions). ↑
15. . See Ryan Calo, Digital Market Manipulation, 82 Geo. Wash. L. Rev. 995, 1033 (2014) (explaining how firms exploit consumers’ data to target their vulnerabilities). See generally Neil M. Richards, Intellectual Privacy, 87 Texas L. Rev. 387 (2008) (arguing intellectual privacy is essential to free thought and expression). ↑
16. . Tarleton Gillespie, Platforms Are Not Intermediaries, 2 Geo. L. Tech. Rev. 198, 201–02 (2018) (framing content moderation as a service offered by platforms to users). ↑
17. . See Ted Selker & Yunzi Wu, Generative AI’s Aggregated Knowledge Versus Web-Based Curated Knowledge 2 (Oct. 15, 2024) (unpublished manuscript) (on file with author), https://arxiv.org/pdf/2410.12091 [https://perma.cc/ZSV5-TBVQ] (describing AI search engines as synthesizing perspectives as opposed to traditional search engines that found information). ↑
18. . See Brenda Dvoskin, The Illusion of Inclusion: The False Promise of the New Governance Project for Content Moderation, 93 Fordham L. Rev. 1315, 1318 (2024) (identifying the stakeholders who play a role in online content moderation). ↑
19. . See Steven Adler, I Led Product Safety at OpenAI. Don’t Trust Its Claims About ‘Erotica.’, N.Y. Times (Oct. 28, 2025), https://www.nytimes.com/2025/10/28/opinion/openai-chatgpt-safety.html [https://perma.cc/26AP-RLQ6] (discussing mental health risks associated with OpenAI); Claudia E. Haupt, Misplaced Trust in Artificial Professional Advice, B.U. J. Sci. & Tech. L. (forthcoming) (manuscript at 2) (on file with author), https://papers.ssrn.com/s
    ol3/papers.cfm?abstract_id=6328899 [https://perma.cc/M9D9-VZEV] (arguing that the displacement of professionals by AI undermines confidence in human professionals). ↑
20. . See Kate Klonick, The New Governors: The People, Rules, and Processes Governing Online Speech, 131 Harv. L. Rev. 1598, 1650 (2018) (“[P]latform architecture has also been informed by and subject to government interference”); see, e.g., Removal of COVID-19 Misinformation, Oversight Bd. (Apr. 20, 2023), https://www.oversightboard.com/decision/pao-sabu4p2s/ [https://perma.cc/3RP5-86L6] (advising Meta to continue removing COVID-19 misinformation after consulting individuals “representing fact-checking organizations, public health bodies and experts, misinformation researchers, digital literacy and communication experts and human rights advocates”). ↑
21. . See, e.g., Hafsa Naeem Baig, Musk Says, He Prefers “Free Speech” Over Censorship, News Int’l (Feb. 4, 2026), https://www.thenews.com.pk/latest/1391040-musk-says-he-prefers-free-speech-over-censorship [https://perma.cc/Q3NW-9R5A] (profiling Musk’s vocal support of free speech); Tabby Wilson, Musk Says Outcry over X’s Grok Service Is “Excuse for Censorship,” BBC (Jan. 10, 2026), https://www.bbc.com/news/articles/ce3kqzepp5zo [https://perma.cc/7TGM-JY7P] (reporting Musk’s pushback against UK regulatory authorities and critics of X, claiming they “just want to suppress free speech”); Joel Kaplan, More Speech and Fewer Mistakes, Meta Newsroom (Jan. 7, 2025), https://about.fb.com/news/2025/01/meta-more-speech-fewer-mistakes/ [https://perma.cc/C26S-28EL] (reporting that Facebook ended third-party fact checking over concerns it infringes on freedom of expression). ↑
22. . See Tarleton Gillespie, Custodians of the Internet 13 (2018) (noting that content moderation is “the commodity” that platforms offer). ↑
23. . See Selker & Wu, supra note 17 (describing AI search engines as synthesizing perspectives as opposed to traditional search engines that found information). ↑
24. . See Ignacio Cofone, Grok, Deepfakes, and the Collapse of the Content/Capability Distinction, Just Security (Feb. 9, 2026), https://www.justsecurity.org/130630/grok-deepfakes-content-capability/ [https://perma.cc/2GGZ-DQZU] (observing that the integration of generative AI into social media platforms challenges existing content moderation policies). ↑
25. . For examples of information unearthed by courts, see Peter Chapman & Matt Steinberg, Measuring Risk: What EU Risk Assessments and US Litigation Reveal About Meta and TikTok 28 (2026), https://kgi.georgetown.edu/wp-content/uploads/2026/02/
    Measuring-Risk_-What-EU-Risk-Assessments-and-US-Litigation-Reveal-About-Meta-and-TikTok-_KGI_2026.pdf [https://perma.cc/2TNG-VPJQ] (“Materials released in US litigation provide granular detail related to the design of Meta’s algorithmic recommender systems.”). ↑
26. . John Perry Barlow, Declaration of the Independence of Cyberspace, Elec. Frontier Found. (Feb. 8, 1996), https://www.eff.org/cyberspace-independence [https://perma.cc/FG8H-F4MM]; see also Andrew D. Murray, Nodes and Gravity in Virtual Space, 5 Legisprudence 195, 195 (2011) (discussing Barlow and other cyberlibertarians). ↑
27. . See, e.g., Marc H. Greenberg, A Return to Lilliput: The LICRA v. Yahoo! Case and the Regulation of Online Content in the World Market, 18 Berkeley Tech. L.J. 1191, 1194–95, 1207 (2003) (noting that Yahoo did not regulate terms of sales and disputed French jurisdiction to regulate such sales). ↑
28. . See Jack Goldsmith & Tim Wu, Who Controls the Internet?: Illusions of a Borderless World 23–25 (2006) (underscoring how the Internet’s founders, supported financially by the U.S. Defense Department, institutionalized their libertarian views of online communications through the Internet Engineering Task Force); Joseph Marks, Hillary Clinton: “Internet Freedom” Activist?, POLITICO (Aug. 10, 2015), https://www.politico.com/
    story/2015/08/hillary-clinton-2016-internet-freedom-121229 (detailing State Department’s efforts to promote internet freedom during Arab Spring). ↑
29. . Goldsmith & Wu, supra note 28, at vii–viii. ↑
30. . See Arun, supra note 9, at 247–50 (describing social media platforms’ options as including interpreting, following or contesting local laws). ↑
31. . See Klonick, supra note 20, at 1623 (discussing Google’s agreement to remove certain videos that violated Thai law). ↑
32. . See Max Fisher, Inside Facebook’s Secret Rulebook for Global Political Speech, N.Y. Times (Dec. 27, 2018), https://www.nytimes.com/2018/12/27/world/facebook-moderators.html [https://perma.cc/52TC-HHDD] (reporting that Facebook instructed moderators to censor posts advocating for Kashmiri freedom). ↑
33. . See Aijaz Hussain, Indian Authorities Release Kashmiri Journalist Fahad Shah After 21 Months in Prison, AP News (Nov. 24, 2023), https://apnews.com/article/kashmir-india-journalists-media-freedom-dc46f76be20d69f7d53d37abf2b3ea8b [https://perma.cc/2LRQ-CG7G] (noting that the Constitution of India guarantees free speech, and the Modi government has cracked down on free expression in Kashmir). ↑
34. . See Arun, supra note 9, at 247–50 (showing how platforms choose to ignore or follow local laws, and do so under their own interpretations). ↑
35. . Vindu Goel & Salman Masood, Facebook, Google and Twitter Rebel Against Pakistan’s Censorship Rules, N.Y. Times (Feb. 27, 2020), https://www.nytimes.com/2020/02/27/
    technology/pakistan-internet-censorship.html [https://perma.cc/A9JQ-KX6B]. ↑
36. . Paul Karp, Australia Passes Social Media Law Penalising Platforms for Violent Content, Guardian (Apr. 4, 2019), https://www.theguardian.com/media/2019/apr/04/australia-passes-social-media-law-penalising-platforms-for-violent-content?CMP=share_btn_url [https://perma.cc/E8QE-C2XK] (explaining that the platforms protested but ultimately complied with the broadly framed Australian law, which threatened the platforms’ employees with imprisonment if they failed to remove violent content). ↑
37. . See Claire Fu & Daisuke Wakabayashi, There Is No TikTok in China, but There Is Douyin. Here’s What It Is., N.Y. Times (Apr. 25, 2024), https://www.nytimes.com/2024/04/25
    /business/china-tiktok-douyin.html [https://perma.cc/RG9F-ZEE6] (“Like other social media services in China, Douyin follows the censorship rules of the Chinese Communist Party.”); see also Chan, supra note 8 (describing the Chinese model as one in which the government guides public opinion with the help of the media, including social media). ↑
38. . See Vivienne Walt, Inside the Big Tech Lobbying Machine Aiming to Halt Social Media Bans, N.Y. Times (Feb. 24, 2026), https://www.nytimes.com/2026/02/21/business/dealbook/big-tech-lobbying-social-media-bans.html [https://perma.cc/3RPD-4J8B] (detailing Silicon Valley’s lobbying efforts in the European Union); Cynthia Kroet, Big Tech Spending on Brussels Lobbying Hits Record High, Report Claims, Euronews (Oct. 29, 2025), https://www.euronews.com/next/
    2025/10/29/big-tech-spending-on-brussels-lobbying-hits-record-high-report-claims [https://perma
    .cc/6F37-5QDH] (reporting on increased spending by Meta, Microsoft, and Apple on lobbying in the European Union); Humeyra Pamuk, Exclusive: Rubio Orders US Diplomats to Launch Lobbying Blitz against Europe’s Tech Law, Reuters (Aug. 7, 2025), https://www.reuters.com/
    sustainability/society-equity/rubio-orders-us-diplomats-launch-lobbying-blitz-against-europes-tech-law-2025-08-07/ [https://perma.cc/WEA3-3CYX] (describing instructions given to American diplomats to further U.S. technology company interests in Europe); Siladitya Ray, JD Vance Knocks EU’s Regulation of US Tech Giants: ‘America Cannot Accept That,’ Forbes (Feb. 11, 2025), https://www.forbes.com/sites/siladityaray/2025/02/11/jd-vance-knocks-eus-regulation-of-us-tech-giants-america-cannot-accept-that/ [https://perma.cc/8BNF-XXRE] (noting Vice President J.D. Vance’s warning to European governments against interfering with U.S. technology company interests). ↑
39. . Derico & Wells, supra note 12 (“A self-declared ‘free-speech absolutist,’ the billionaire entrepreneur had described Justice Moraes’ move to ban several dozen accounts as an abuse of power and a violation of free speech.”). ↑
40. . See id. (“[T]he platform began to comply with the court’s orders in an apparent U-turn.”). ↑
41. . See Laura Cress, EU Investigates Elon Musk’s X over Grok AI Sexual Deepfakes, BBC (Jan. 26, 2026), https://www.bbc.com/news/articles/clye99wg0y8o [https://perma.cc/K8T7-RRQL] (describing Musk’s mocking of a European Union investigation into the creation of sexual deepfakes on X and his criticism of governments scrutinizing X’s image editing features). ↑
42. . See Ionova, supra note 12 (reporting that President Trump imposed tariffs against Brazil); Cristiano Lima-Strong, Trump Administration Locks Arms with Musk’s X Against EU Tech Enforcement, Tech Pol’y Press (Dec. 5, 2025), https://techpolicy.press/trump-administration
    -locks-arms-with-musks-x-against-eu-tech-enforcement [https://perma.cc/9EFW-BFTH] (describing Trump Administration officials’ denouncement of the EU’s decision to fine X). ↑
43. . See Paul Gowder, The Networked Leviathan: For Democratic Platforms 39–40 (2023) (analogizing city and state governments to digital platforms); see generally Rebecca MacKinnon, Consent of the Networked (2012) (arguing that private networks and platforms increasingly share in the governance functions traditionally carried out by nation-states). ↑
44. . See Chinmayi Arun, The Silicon Valley Effect, 61 Stan. J. Int’l L. 1, 75–79, 93–99 (2025) (positing that major tech companies are not subject to “the temporality of election cycles” and noting the large sums of money they spend lobbying the United States government). ↑
45. . See Kate Klonick, The Facebook Oversight Board: Creating an Independent Institution to Adjudicate Online Free Expression, 129 Yale L. J. 2418, 2426–27 (2020) (suggesting Facebook needed the Oversight Board to confer legitimacy on the content moderation policies of its platforms). ↑
46. . See Emma Bowman, Facebook Steps Up Efforts to Combat the Spread of Coronavirus Misinformation, NPR (Mar. 26, 2020), https://www.npr.org/sections/coronavirus-live-updates/2020/03/26/822245048/facebook-steps-up-efforts-to-combat-the-spread-of-coronavirus-misinformation [https://perma.cc/LWR9-UD3E] (quoting Facebook’s vice president of global affairs and communications with saying “it’s easier for the company to act under the ‘strict expertise and guidance’ from institutions like WHO and CDC”); see also Naomi Appelman & Paddy Leerssen, On “Trusted” Flaggers, 24 Yale J.L. & Tech. (Special Issue) 452, 453 (2022) (arguing that “trusted flaggers” are taking on an influential role in “platform governance”); Amre Metwally, The Governors’ Advisors: Experts and Expertise as Platform Governance, 24 Yale J.L. & Tech. (Special Issue) 510, 512–13 (2022) (listing proposals for and examples of platforms outsourcing content-moderation decisions to experts). ↑
47. . See, e.g., Dangerous Organizations and Individuals, Meta, https://transparency.meta.com/
    policies/community-standards/dangerous-individuals-organizations/ [https://perma.cc/6C5M-N39Q] (deferring to the United States government’s designation of terrorist organizations); Child Sexual Exploitation, Abuse, and Nudity, Meta, https://transparency.meta.com/policies/
    community-standards/child-sexual-exploitation-abuse-nudity/ [https://perma.cc/P7EQ-LHER] (explaining that Meta works with “external experts . . . to discuss and improve” its content-moderation policies with respect to child sexual exploitation). ↑
48. . See Appelman & Leerssen, supra note 46 (“[T]rusted flagging outsources part of the responsibility for content moderation from platforms to third parties.”). ↑
49. . Tom R. Tyler, Tracey L. Meares & Matt Katsaros, New Worlds Arise: Online Trust and Safety, 8 Ann. Rev. Criminology 171, 180 (2025) (pointing out that more than 100,000 people were employed in Trust and Safety). ↑
50. . See Balkin, supra note 7, at 2014–16 (visualizing the twenty-first century model of speech regulation as a triangle, where governments pressure digital platforms to regulate individual speakers’ speech). ↑
51. . See Arun, supra note 9, at 253 (“[O]nce the role of Facebook in enabling incitement to violence made it to The New York Times, and the United Nations Independent International Fact-Finding Mission on Myanmar publicly chastised the company for its role, the Burmese people had the company’s attention.”). ↑
52. . See Moody v. NetChoice, LLC, 144 S. Ct. 2383, 2407–09 (2024) (acknowledging that it is “a fundamental aim” of the First Amendment to have a well-functioning public sphere where citizens have access to information from many sources); see also James Grimmelmann, Listeners’ Choices, 90 U. Colo. L. Rev. 365, 378–379, 382 (2019) (arguing that “media companies of the Internet age” arrange speaker-listener pairings, and that interference with the “joint communicative project” between speaker and listener is inherently suspect); Jack M. Balkin, Digital Speech and Democratic Culture: A Theory of Freedom of Expression for the Information Society, 79 N.Y.U. L. Rev. 37, 41–46 (2004) (pointing out that digital technology enables mass cultural participation, which is critical to democracy and liberty); Erin L. Miller, Amplified Speech, 43 Cardozo L. Rev. 1, 10, 61 (2021) (arguing that editorial discretion rights exist in service of democratic discourse). ↑
53. . See, e.g., Baig, supra note 21 (quoting Musk as saying on behalf of X, “We will support free speech rather than agree to be censored”); Kaplan, supra note 21 (justifying Facebook’s changes to its content-moderation policies as necessary to comply with the platform’s commitment to free expression); see also Tyler et al., supra note 49, at 176–77 (contrasting Musk’s free-speech absolutism with his restriction of content threatening his personal safety to underscore how the values of those controlling platforms are directly reflected in how those platforms are governed). ↑
54. . See Justin Hendrix, Transcript: Mark Zuckerberg Announces Major Changes to Meta’s Content Moderation Policies and Operations, Tech Pol’y Press (Jan. 7, 2025), https://www.techpolicy.press/transcript-mark-zuckerberg-announces-major-changes-to-metas-content-moderation-policies-and-operations/ [https://perma.cc/EA47-3BKG] (“[W]e’re going to get back to our roots and focus on reducing mistakes, simplifying our policies, and restoring free expression on our platforms.”); Katherine Tangalakis-Lippert & Hannah Getahun, Jack Dorsey Apologizes for His Twitter-Moderation Choices (Dec. 13, 2022), https://www.businessinsider.com/
    jack-dorsey-twitter-wrong-thing-for-the-internet-and-society-2022-12 [https://perma.cc/65KW-RTVS] (reporting that Dorsey believes content moderation on Twitter disserved the Internet and society). ↑
55. . See Arun, supra note 9, at 243 (describing the appeals process for users who disagree with Facebook’s content-moderation decisions); Ben Bradford, Florian Grisel, Tracey L. Meares, Emily Owens, Baron L. Pineda, Jacob N. Shapiro, Tom R. Tyler & Danieli Evans Peterman, Just. Collaboratory, Yale L. Sch., Report of the Facebook Data Transparency Advisory Group 33–34 (2019), https://law.yale.edu/sites/default/files/area/
    center/justice/document/dtag_report_5.22.2019.pdf [https://perma.cc/H2GR-U6FX] (concluding that when users are treated with dignity and given an opportunity to share their voice, they are more likely to perceive authority and rules as legitimate); Tom R. Tyler, Procedural Justice, Legitimacy, and the Effective Rule of Law, 30 Crime & Just. 283, 284 (2003) (arguing that people’s subjective judgments about procedural fairness powerfully influences law-related behaviors). ↑
56. . See Tom R. Tyler, What Is Procedural Justice?: Criteria Used by Citizens to Assess the Fairness of Legal Procedures Paper, 22 L. & Soc’y Rev. 103, 117 (1988) (concluding from the results of a regression analysis that fairness judgments do influence satisfaction and evaluation). ↑
57. . Kate Conger, Meta Turns to Community Notes, Mirroring X, N.Y. Times (Jan. 7, 2025), https://www.nytimes.com/2025/01/07/business/meta-community-notes-x.html [https://perma.cc/73TM-FAN8]. ↑
58. . See Arun, supra note 9 (arguing that leveraging influential institutions like the New York Times and the United Nations helped bring Facebook’s role in the Rohingya genocide to light); Rachel Griffin, How Public Pressure Forced Facebook to Change Its Policies on Hate Speech, SciencesPo (July 9, 2020), https://www.sciencespo.fr/public/chaire-numerique/en/2020/07/09/
    how-public-pressure-forced-facebook-to-change-its-policies-on-hate-speech/ [https://perma.cc/59SN-8BME] (discussing the role the Black Lives Matter movement played in bringing attention to hate speech on Facebook and other social media). ↑
59. . See, e.g., Shannon Bond, Facebook Will Examine Whether It Treats Black Users Differently, NPR (Nov. 18, 2021), https://www.npr.org/2021/11/18/1056916140/facebook-to-study-black-users-experience [https://perma.cc/GL9W-M3TA] (noting that the Anti-Defamation League organized an advertiser boycott of Facebook over hate speech); Jon Keegan, Facebook Got Rid of Racial Ad Categories. Or Did It?, Markup (July 9, 2021), https://themarkup.org/citizen-browser/2021/07/09/facebook-got-rid-of-racial-ad-categories-or-did-it [https://perma.cc/ZC4R-FS4Z] (mentioning that after a 2016 ProPublica investigation, Facebook eliminated racial categories for online advertisements); Alaina Demopoulos, Free the Nipple: Facebook and Instagram Told to Overhaul Ban on Bare Breasts, Guardian (Jan. 18, 2023), https://www.theguardian.com/technology/2023/jan/17/free-the-nipple-meta-facebook-instagram [https://perma.cc/ZA96-WZV9] (connecting a 2023 decision of Meta’s oversight board calling for the overhaul of the company’s rules related to bare-chested images of women to the work of “lactivists” in the 2000s). ↑
60. . See Arun, supra note 9, at 251 (“Facebook’s staff, community, and people affected by it include influential groups and individuals as well those who are less able to make themselves heard and understood through the communicative process made available to them.”). ↑
61. . See Victoria Milko & Barbara Ortutay, ‘Kill More’: Facebook Fails to Detect Hate Against Rohingya, AP News (Mar. 21, 2022), https://apnews.com/article/technology-business-bangladesh-myanmar-united-nations-f7d89e38c54f7bae464762fa23bd96b2 [https://perma.cc/U32N-T9J6] (reporting that Facebook ignored warnings from local internet policy advocates that anti-Rohingya hate speech was spreading on the platform). ↑
62. . See Arun, supra note 9 (discussing how it took the flexing of the New York Times and the United Nations’ epistemic authority to get Meta’s attention). ↑
63. . Kristie Dotson, Accumulating Epistemic Power: A Problem with Epistemology, 46 Phil. Topics 129, 139 (2018). ↑
64. . See Leiter, supra note 6, at 905, 918–21, 935 (surveying recent events indicating the depreciation of expertise and epistemic authority in the United States). ↑
65. . See Yochai Benkler, Robert Faris & Hal Roberts, Network Propaganda: Manipulation, Disinformation, and Radicalization in American Politics 359 (2018) (“[P]rofessional mainstream media continue to play an enormous role for most Americans.”); see also Mark Tushnet, The Next Generation of Free Expression Scholarship: A Very Short Manifesto (in Memory of Fred Schauer), 33 Wm. & Mary Bill Rts. J. 999, 1018 (2024) (explaining how the legacy media’s pursuit of objectivity allowed them to claim epistemic authority). ↑
66. . See Chad Roat, Kashica J. Webber-Ritchey, Roxanne S. Spurlark & Young-Me Lee, Black Americans Receiving the COVID-19 Vaccine and Effective Strategies to Overcome Barriers: An Integrative Literature Review, 10 J. Racial & Ethnic Health Disparities 2577, 2583 (2022) (laying out the reasons for Black Americans’ mistrust of the medical establishment); Ajasa et al., supra note 14 (identifying historical memories of unethical medical experiences as a source of distrust among some Black Americans toward COVID vaccines). ↑
67. . See Dotson, supra note 63 (grounding epistemic authority in one’s “standing” and “status” with respect to knowledge possession, knowledge attribution, and/or knowledge production). ↑
68. . See Richard T. De George, The Function and Limits of Epistemic Authority, 8 S.J. Phil. 199, 199 (1970) (“Every type of authority is relational in nature in the sense that it follows the form: x is an authority for y over a particular field z and (if legitimate) in virtue of some justification.”). ↑
69. . See Guy-Uriel E. Charles, Colored Speech: Cross Burnings, Epistemics, and Triumph of the Crits?, 93 Geo. L.J. 575, 610 (2005) (explaining that epistemic authority is invoked when a factual assertion is accepted as true because someone with such authority says it is true). ↑
70. . Leiter, supra note 6 (describing the influence of epistemic authorities on what people think). Leiter’s analysis is based on the work of Joseph Raz. Id. at 906; see generally Joseph Raz, Authority, Law and Morality, 68 MONIST 295, 296 (1985). ↑
71. . Pluto Facts, NASA, https://science.nasa.gov/dwarf-planets/pluto/facts/ [https://perma.cc/
    US4B-WBZK]. It is worth noting that William Shatner is lobbying for a public reclassification, albeit playfully. Andrea Leinfelder, Star Trek’s William Shatner Wants to Use Elon Musk’s Political Power to Make Pluto a Planet Again, Hou. Chron. (May 16, 2025), https://www.houstonchronicle
    .com/news/houston-texas/space/article/musk-shatner-trump-pluto-planet-20330792.php [https://perma.cc/NA4U-T23H]. ↑
72. . Leiter, supra note 6, at 907. ↑
73. . See id. at 905 (warning that the Internet has upended the idea of expertise and ensured that hundreds of millions of people will have false beliefs “about the causal order of nature”). ↑
74. . See, e.g., Mark Tushnet, Epistemic Disagreement, Institutional Analysis, and the First Amendment Status of Lies, 4 J. Free Speech L. 651, 677–78 (2023) (observing that some groups have more power to determine epistemic authority than others). ↑
75. . See id. at 681 (“Consensus standards in science result from socialization processes that involve exercises of power.”). ↑
76. . Matthew B. Lawrence & Haomin Yan, Moral Panic or Public Health Crisis? Lessons from Drugs and Gambling for “Addictive” Design, 48 Seattle U.L. Rev. 623, 638 (2025). ↑
77. . See Lina M. Khan, Sources of Tech Platform Power, 2 Geo. L. Tech. Rev. 325, 328 (2018) (discussing the dangers associated with tech platforms having too much power over the distribution and content of news); K. Sabeel Rahman, Democracy Against Domination 45 (2017) (detailing how the financial industry influenced policy makers’ views “through campaign contributions and interest group politics”). ↑
78. . Dotson, supra note 63. ↑
79. . See Leiter, supra note 6, at 905 (discussing how the Internet enables the dissemination of false beliefs even where experts agree); see also David Nieborg, Thomas Poell, Robyn Caplan & José Van Dijck, Introduction to the Special Issue on Locating and Theorising Platform Power, 13 Internet Pol’y Rev. (Special Issue) 1, 6 (2024) (explaining how platforms exercise power by governing users and moderating content); Robyn Caplan & Danah Boyd, Isomorphism Through Algorithms: Institutional Dependencies in the Case of Facebook, Big Data & Soc’y, Jan.–June 2025, at 1, 1, 5 (noting that Facebook redefines the value or relevance of information and news media because of its outsized role in how people access content); Natali Helberger, The Political Power of Platforms: How Current Attempts to Regulate Misinformation Amplify Opinion Power, 8 Digit. Journalism 842, 843 (2020) (arguing social media platforms are political actors with political power). ↑
80. . See, e.g., Ryan Mancini, Rep. Marjorie Taylor Greene Says She No Longer Believes in QAnon Conspiracy Theory, Hill (Nov. 5, 2025), https://thehill.com/homenews/house/5591006-marjorie-taylor-greene-qanon-rejection/ [https://perma.cc/9KBN-VDFN] (noting that Representative Marjorie Taylor Greene fell victim to QAnon conspiracy theories on social media); Teddy Rosenbluth, Health Advice From A.I. Chatbots Is Frequently Wrong, Study Shows, N.Y. Times (Feb. 12, 2026), https://www.nytimes.com/2026/02/09/well/chatgpt-health-advice.html [https://perma.cc/VSK5-4C2N] (discussing how people rely on incomplete information from chatbots for health advice). ↑
81. . See, e.g., Sheila Jasanoff, Ordering Knowledge, Ordering Society, in States of Knowledge: The Co-Production of Science and Social Order 39–41 (Sheila Jasanoff ed., 2004) (discussing the role of social practices in constructing knowledge institutions and scientific discourse); Maria J. Azocar & Myra Marx Ferree, Engendering the Sociology of Expertise, 10 Socio. Compass 1079, 1079–80 (2016) (examining processes of gendering expertise). ↑
82. . These discriminatory theories are typically based on ideological empirical research or conclusions. See, e.g., Halford H. Fairchild, Scientific Racism: The Cloak of Objectivity, 47 J. Soc. Issues 101, 108–09 (1991) (“The omnipresence of racist ideology—popularly and in academe—provides a cultural impetus for the generation of racist conclusions in social science.”); Aziza Ahmed, Risk and Resistance: How Feminists Transformed the Law and Science of AIDS 34–60, 147 (2025) (showing that advocacy and testimonies shifted the scientific consensus to a more inclusive definition of AIDS; relating W.E.B. Du Bois’ challenges to racialized scientific expert belief that African Americans suffered from poor health because of their race instead of the socio-economic conditions that affected African Americans’ health). ↑
83. . See Ajasa et al., supra note 14 (recounting the historical mistreatment of Black Americans by the medical establishment, including the Tuskegee Study); Claire Laurier Decoteau & Paige L. Sweet, Vaccine Hesitancy and the Accumulation of Distrust, 71 Soc. Probs. 1231, 1231–33, 1243 (2024) (demonstrating that marginalized groups’ distrust of the medical establishment results from both epistemic harm and corporeal harm). ↑
84. . See Tushnet, supra note 65, at 1011 (observing that epistemic communities are shaped by social processes of exclusion and inclusion); see, e.g., Kimani Paul-Emile, The Regulation of Race in Science, 80 Geo. Wash. L. Rev. 1115, 1120 (2012) (arguing that race should be used as a variable in biomedical research because of its social, not biological, salience). ↑
85. . Tushnet, supra note 65, at 1011. ↑
86. . See id. (noting that social process are “titled with respect to politics . . . ordinarily, in favor of the status quo”); Dotson, supra note 63, at 145–6 (pointing out that structural considerations like components of the social, cultural and political environment shape “epistemological high ground”); Ahmed, supra note 82, at 34–35 (discussing how feminists shaped the science and law pertaining to AIDS). ↑
87. . 538 U.S. 343 (2003). ↑
88. . See Charles, supra note 69, at 610–11 (arguing that Justice Thomas’s background and life experiences commanded epistemic deference). ↑
89. . See George, supra note 68 (distinguishing epistemic authority from political authority, paternalistic authority, operative authority, charismatic authority, moral authority, and religious authority). ↑
90. . See Religious Groups’ Views on Evolution, Pew Rsch. Ctr. (Feb. 3, 2014), https://www.pewresearch.org/religion/2009/02/04/religious-groups-views-on-evolution/ [https://perma.cc/T43M-KLZM] (listing the views of various major religions on evolution). ↑
91. . See Holford et al., supra note 14 (enumerating epistemic leanings in the context of vaccine hesitancy, including distrust, unwarranted beliefs, worldview and politics, religious concerns, and epistemic relativism); Fasce et al., supra note 14, at 1464–65 (discussing the attitude roots of vaccine hesitancy). ↑
92. . See id. ↑
93. . See Julie E. Cohen, Infrastructuring the Digital Public Sphere, 25 Yale J.L. & Tech. (Special Issue) 1, 29 (2023) (explaining how platforms elicit information from advertisers to optimize ad targeting and ad content); Tarleton Gillespie, The Politics of ‘Platforms,’ 12 New Media & Soc’y 347, 353–54 (2010) (noting YouTube is funded almost entirely by advertising). It is also possible that some platform owners, such as Elon Musk, seek epistemic power for its own sake or because of its relationship with political power. See Killian Fox, Jeff Jarvis: ‘Elon Musk’s Investment in Twitter Seemed Insane, but It Gave Him This Power,’ Guardian (Nov. 23, 2024), https://www.theguardian.com/technology/2024/nov/23/jeff-jarvis-elon-musks-investment-in-twitter-seemed-insane-but-it-gave-him-this-power [https://perma.cc/U64Y-3C5N] (discussing how Musk’s investment in Twitter gave him power); Sheila Dang, Elon Musk Manages Free Speech Versus ‘Hellscape’ at Twitter, Reuters (Oct. 29, 2022), https://www.reuters.com/technology/elon-musk-takes-over-twitter-free-speech-limits-tested-2022-10-28/ [https://perma.cc/8V29-H4VX] (explaining Musk’s intention to restore free speech on Twitter); Barbara Ortutay, How Elon Musk Uses His X Social Media Platform to Amplify Right-Wing Views, AP News (Aug. 13, 2024), https://apnews.com/article/elon-musk-x-influence-uk-us-politics-48cc1870cbdb5df03829f5ab303b579d [https://perma.cc/8NUN-4H4V] (noting that Musk purchased Twitter because he did not believe it was living up to its potential as a free-speech platform); Nell Clark, Trump’s Social Media Site Hits the App Store a Year after He Was Banned from Twitter, NPR (Feb. 22, 2022), https://www.npr.org/2022/02/22/1082243094/trumps-social-media-app-launches-year-after-twitter-ban [https://perma.cc/8T67-3PHK] (explaining that President Trump’s Truth Social is supposed to promote “free speech” and avoid “political manipulation”). ↑
94. . See Lima-Strong, supra note 42 (reporting on the Trump Administration’s public support of X); Ray, supra note 38 (writing about Trump administration’s warning that it will not accept efforts by foreign governments to tighten regulations on American tech companies); Pamuk, supra note 38(describing the Trump Administration’s lobbying campaign against tech regulation). ↑
95. . See Rob Kitchin, Big Data, New Epistemologies and Paradigm Shifts, Big Data & Soc’y, Apr.–June 2014, at 1, 2 (“Big Data analytics enables an entirely new epistemological approach for making sense of the world; rather than testing a theory by analysing relevant data, new data analytics seek to gain insights ‘born from the data.’”). ↑
96. . Julie E. Cohen, Between Truth and Power 42–43 (2019) (explaining that platforms strive “to become and remain the indispensable point of intermediation for parties in [their] target markets” in order to extract monetary value from the accumulation of user data); see also Calo, supra note 15 (“[A] firm with the capacity and incentive to exploit a consumer could, for instance, monitor the number of decisions she makes on her phone and target the customer most intensely at the moment she is most depleted.”). ↑
97. . Moody v. NetChoice, LLC 144 S. Ct. 2383, 2398, 2402 (2024) (asserting that algorithmically curated feeds sometimes create expressive products, triggering First Amendment scrutiny). ↑
98. . See Hannah Bloch-Wehba, Automation in Moderation, 53 Corn. Int’l L.J. 41, 43 (2020) (arguing that laws holding platforms responsible for content moderation expand platform authority to police online content). ↑
99. . See Solon Barocas & Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671, 673–74 (arguing that algorithms can exhibit discriminatory tendencies even if they are not programmed to do so). ↑
100. . See Chapman & Steinberg, supra note 25, at 28 (pointing to materials released in U.S. litigation against Meta, which reveal that the company sought to optimize engagement at the risk of exacerbating users’ addiction problems). ↑
101. . See Cohen, supra note 93, at 8 (“Different choices about ownership and management of infrastructural resources shape the patterns of downstream relationships and uses in ways that produce large external effects.”); Julie E. Cohen, Law for the Platform Economy, 51 U.C. Davis L. Rev. 143–45 (discussing how platforms, by providing infrastructural access, set the termsof this access, thereby wielding political and technological authority through their infrastructural role). ↑
102. . Aziz Z. Huq, Military Democracy Comes to the Metaverse?, 72 Emory L.J. 1105, 1116 (2022). ↑
103. . See Cohen, supra note 96, at 39 (arguing that digital platforms’ media infrastructures have changed “the conditions of access and the need for access”). ↑
104. . See Chinmayi Arun, Manipulative Infrastructures: Platform Power and Freedom of Thought, 111 Corn. L.R. (forthcoming 2026) (manuscript at 6) (“[S]ocial media is eroding legacy media’s agenda-setting and threatening its business model.”); Caplan & Boyd, supra note 79, at 1 (explaining how news outlets have no choice but to alter their content with algorithms in mind); Tom Crowley, Facebook, Google to Fund Australian Journalism under New Tax Plan, ABC News (Dec. 12, 2024), https://www.abc.net.au/news/2024-12-12/government-forces-facebook-google-to-pay-for-australian-news/104717608 [https://perma.cc/79SJ-DQDN] (reporting that the Australian government will impose a new tax requiring tech platforms like Facebook, Google, and TikTok to create funding agreements with news organizations); Nadine Yousif, Meta’s News Ban in Canada Remains as Online News Act Goes into Effect, BBC (Dec. 19, 2023), https://www.bbc.com/news/
     world-us-canada-67755133 [https://perma.cc/Q62U-VWKG] (discussing Canada’s mandate that tech giants pay news outlets for their content). ↑
105. . Helberger, supra note 79, at 846. ↑
106. . See generally Caplan & Boyd, supra note 79, at 1 (stating that digital media has changed the landscape with respect to the news); David B. Nieborg & Thomas Poell, The Platformization of Cultural Production: Theorizing the Contingent Cultural Commodity, 20 New Media & Soc’y 4275, 4276 (asserting that platforms exercise power by moderating users and moderating content). ↑
107. . See Julie E. Cohen, Tailoring Election Regulation: The Platform Is the Frame, 4 Geo. L. Tech. Rev. 641, 647 (2020) (highlighting how some platforms’ market dominance has contributed to tribalism); Joan Donovan & Danah Boyd, Stop the Presses? Moving from Strategic Silence to Strategic Amplification in a Networked Media Ecosystem, 65 Am. Behav. Scientist 333, 343 (2021) (discussing how the journalistic tactic of strategic silence has been rendered obsolete by networked media’s amplification practices); Caplan & Boyd, supra note 79, at 5 (showing how Facebook changed the journalism industry by redefining the value and relevance of information and the economics of journalism). ↑
108. . Donovan & Boyd, supra note 107, at 335–36 (“The complexity of the networked communication system has led to the erosion of traditional editorial control; news media’s ability to actively balance public interest against public harm in publishing or broadcasting has come undone.”). ↑
109. . See supra note 67–68 and accompanying text. ↑
110. . See supra note 104 and accompanying text. ↑
111. . See generally Lina Khan, The Separation of Platforms and Commerce, 119 Colum. L. Rev. 973 (2019) (discussing the structural separation of dominant digital platforms); Francis Fukuyama, Barak Richman, Ashish Goel, Roberta R. Katz, A. Douglas Melamed & Marietje Schaake, Middleware for Dominant Digital Platforms 3 (2020) (introducing middleware as a solution to digital platforms’ control over communication and political discourse). ↑
112. . See Metwally, supra note 46, at 513–14 (using YouTube as an example of a digital platform which publicly announced that it developed content-moderation policies in consultation with third-party experts); Klonick, supra note 20, at 1619–21 (pointing to digital platforms that have used lawyers trained in American free speech and First Amendment law to develop content-moderation policies); Tyler et al., supra note 49, at 172 (detailing how platforms have recruited experienced police leaders, prosecutors, or lawyers to design online governance systems). ↑
113. . See Klonick, supra note 20, at 1655 (noting that platforms have worked with outside groups to discuss content-moderation policies); Metwally, supra note 46, at 515–16 (observing that outside experts, including third-party consultants, research institutes, and companies offering formal services, work for social media companies). ↑
114. . See Klonick, supra note 20, at 1656–58 (providing examples of how Facebook has encouraged engagement with civil society groups, government officials, and reporters to develop content-moderation policies). ↑
115. . See Arun, supra note 9, at 245 (arguing that the Facebook oversight board was meant to lend sociological legitimacy to Facebook’s content decisions); Klonick, supra note 45, at 2425–26, 2438 (pointing out that one view of the oversight board is that it is a public relations move to allay concerns about the company) See generally Paul Gowder, The Networked Leviathan (2023) (arguing oversight bodies like Meta’s Oversight Board can provide limited rule-of-law constraints on platform governance but cannot fully democratize platform decision-making). ↑
116. . See Arun, supra note 9 at 244 (“The Oversight Board is an expensive and time-consuming endeavor that affects the reputations of the experts on the Board itself as well as the highly regarded experts in its secretariat.”). ↑
117. . See id. at 256 (noting that while the Oversight Board upheld Facebook’s decision to restrict President Trump’s accounts, it chastised the company for failing to come up with appropriate rules, leading the company’s Vice President of Global Affairs and Communications to publicly voice his disappointment with the decision). ↑
118. . See About Recommendations on Facebook, Facebook, https://www.facebook.com/help/
     1257205004624246 [https://perma.cc/J5UQ-NPLT] (noting Facebook “sought input from 50 leading experts” to develop algorithmic guidelines); Our Approach to Facebook Feed Ranking, Meta (June 11, 2025), https://transparency.meta.com/features/ranking-and-content/ [https://perma.cc/L336-YTBQ] (describing how Meta personalizes each user’s feed based on user behavior); Michael Liedtke & Chris Megerian, TikTok’s Algorithm to Be Licensed to US Joint Venture Led by Oracle and Silver Lake, AP News (Sep. 22, 2025), https://apnews.com/
     article/trump-tiktok-china-d5d8a1d56b5185778536874d7fc1ee62 [https://perma.cc/W9PN-PLUX] (reporting that Oracle will lead U.S. oversight of TikTok’s algorithm); Liv McMahon, Instagram Courts TikTok Users with Algorithm Revamp, BBC (May 3, 2024), https://www.bbc.com/news/articles/cmm3yn4pr17o [https://perma.cc/CS92-P2YC] (including Instagram’s announcement that it would “correct” its previous approach to recommending content); Evelyn Douek, What Kind of Oversight Board Have You Given Us?, U. Chi. L. Rev. Online (2020) [https://perma.cc/ZZN9-3LU3] (observing that “the way Facebook’s algorithms rank content for display to users is conspicuously absent from the [bylaws] altogether”). ↑
119. . See Klonick, supra note 45, at 2459 (positing that the Charter of the Oversight Board reflects an adjudicatory rather than a representative model); Jonathan Shieber, Zuckerberg Details the Ways Facebook and Chan Zuckerberg Initiative Are Responding to COVID-19, TechCrunch (Mar. 3, 2020), https://techcrunch.com/2020/03/03/zuckerberg-details-the-ways-facebook-and-chan-zuckerberg-initiative-are-responding-to-covid-19/ [https://perma.cc/AP3L-3W9U] (describing Facebook’s consultation with global health experts during the COVID-19 pandemic). ↑
120. . See Metwally, supra note 46, at 519 (describing how former Trust & Safety team employees from digital platform companies have organized “their own initiatives, groups, and organizations” to professionalize and institutionalize Trust & Safety). ↑
121. . See Richard Luscombe, Trump Administration Moves to Deny Visas to Factcheckers and Content Moderators, Guardian (Dec. 5, 2025), https://www.theguardian.com/us-news/2025/dec/05/trump-administration-us-visa-crackdown [https://perma.cc/U7YY-2GDM] (reporting that American diplomats were instructed to reject visa applications from individuals employed as fact-checkers); Lima-Strong, supra note 42 (observing that the State Department is seeking to restrict immigration by those the Trump Administration believes have censored Americans). ↑
122. . See Conger, supra note 57 (“Social media companies are increasingly relying on fact-checks written by their users, allowing companies to step back from politically loaded decisions about what content to take down.”); Chris Vallance, Meta Is Ditching Fact Checks for X-Style Community Notes. Will They Work?, BBC (Jan. 25, 2025), https://www.bbc.com/
     news/articles/c4g93nvrdz7o [https://perma.cc/3AGG-DTP3] (describing Meta’s pivot from using independent third-party fact-checkers to community notes). ↑
123. . See Tyler et al., supra note 49, at 178 (noting that hate speech on Facebook is now flagged primarily by the platform’s algorithm, as opposed to user reports); see also Dave Willner & Samidh Chakrabarti, Using LLMs for Policy-Driven Content Classification, Tech Pol’y Press (Jan. 29, 2024), https://techpolicy.press/using-llms-for-policy-driven-content-classification [https://perma.cc/SW3H-49PA] (listing the advantages of LLMs over people in content labeling). ↑
124. . It is, however, worth noting that algorithmic curation allows platforms to tailor content to users, and therefore users’ dissatisfaction is of a different nature than whatever dissatisfaction they may have with the print or broadcast media. ↑
125. . See Daphne Keller, Texas, Florida, and the Magic Speech Sorting Hat in the NetChoice Cases, Lawfare (Feb. 21, 2024), https://lawfaremedia.org/article/texas-florida-and-the-magic-speech-sorting-hat-in-the-netchoice-cases [https://perma.cc/UR5Q-C5RL] (describing laws in Texas and Florida that sought to curb the power of social media platforms over content moderation). ↑
126. . See Leiter, supra note 6, at 907 (arguing that epistemic authority depends on institutions instilling reliable norms about whom to believe). ↑
127. . See Murthy v. Missouri, 144 S. Ct. 1972, 1984 (2024) (explaining plaintiffs’ claims, namely that the Government violated the First Amendment by pressuring platforms to censor their speech). ↑
128. . See Jenifer Whitten-Woodring, Mona S. Kleinberg, Ardeth Thawnghmung & Myat The Thitsar, Poison If You Don’t Know How to Use It: Facebook, Democracy, and Human Rights in Myanmar, 25 Int’l J. Press/Politics 407, 410–12 (2020) (describing Facebook’s extraordinary influence in Myanmar in the 2010s); Goel & Masood, supra note 35 (reporting that global internet companies leveraged their influence in Pakistan to force the government to retreat from imposing sweeping rules on internet censorship). ↑
129. . See Arun, supra note 44, at 90, 95–96 (describing strategies used by Silicon Valley companies to exert political influence); supra note 38 and accompanying text. ↑
130. . See, e.g., Laura Cress, X Could ‘Lose Right to Self Regulate’, Says Starmer, BBC (Jan. 12, 2026), https://www.bbc.com/news/articles/cq845glnvl1o [https://perma.cc/3KRZ-XP8M] (discussing the UK’s threat to revoke X’s right to self-regulate); Derico & Wells, supra note 12 (reporting that Brazil lifted its ban on X after the platform paid a five-million-dollar fine); Matthew Broersma, Ireland Investigates X Over Content Moderation, Silicon (Nov. 13, 2025), https://www.silicon.co.uk/e-regulation/legal/ireland-x-probe-627518 [https://perma.cc/2FWD-RDBU] (describing Ireland’s investigation into X over its content moderation mechanisms); Cress, supra note 41 (detailing the European Commission’s investigation into X over its AI tool, Grok). ↑
131. . See Kari Paul, Elon Musk Reportedly Forced Twitter Algorithm to Boost His Tweets after Super Bowl Flop, Guardian (Feb. 15, 2023), https://www.theguardian.com/technology/2023/
     feb/15/elon-musk-changes-twitter-algorithm-super-bowl-slump-report [https://perma.cc/KB7Z-MY98] (pointing to Musk’s efforts to reconfigure Twitter in order to elevate the visibility and reach of his tweets). ↑
132. . Conger, supra note 57 (mentioning Meta’s pivot toward community notes a few months after the 2024 presidential election). ↑
133. . Mark Zuckerberg Announces Facebook’s Plans to Help Get People Vaccinated Against COVID-19, Meta Newsroom (Mar. 15, 2021), https://about.fb.com/news/2021/03/mark-zuckerberg-announces-facebooks-plans-to-help-get-people-vaccinated-against-covid-19/ [https://perma.cc/G7XG-33B5] (discussing connecting people with “authoritative Covid-19 information” and working closely with “national and global health authorities”). See also Ryan Tracy, Facebook Bowed to White House Pressure, Removed Covid Posts, Wall Street Journal, (July 28, 2023), https://www.wsj.com/politics/policy/facebook-bowed-to-white-house-pressure-removed-covid-posts-2df436b7 [https://perma.cc/6H9N-BFAT] (discussing internal Meta emails about COVID-19 content removed as a result of pressure from the White House). ↑
134. . Mark Sweney, Mark Zukerberg Says White House ‘Pressured’ Facebook to Censor Covid-19 Content, Guardian (Aug. 27, 2024), https://www.theguardian.com/technology/article/
     2024/aug/27/mark-zuckerberg-says-white-house-pressured-facebook-to-censor-covid-19-content [https://perma.cc/PRC9-6MBC]. ↑
135. . By “weakness” I am not suggesting that platforms are unable to influence other actors, but rather that they feel required to comply, at least in the interim, with what the “more powerful” actor asks of them. ↑
136. . Compare Bowman, supra note 46 (detailing Facebook’s efforts to curb pandemic-related misinformation amidst criticism from lawmakers it was not doing enough), with Sweney, supra note 134 (reporting that Zuckerberg felt pressured to “censor” content related to COVID-19 by the government during the pandemic). ↑
137. . See Balkin, supra note 7, at 2019 (pointing out that platforms moderate their own content to avoid nations from imposing liability on them through law); Chinmayi Arun, The Facebook Oversight Board: An Experiment in Self-Regulation, Just Sec. (May 6, 2020), https://www.justsecurity.org/70021/the-facebook-oversight-board-an-experiment-in-self-regulation/ [https://perma.cc/B9AB-7BAH] (“Facebook may make its own rules, but . . . it is likely to be more willing to comply with them than it might be to abide by external standards.”). ↑
138. . See Dvoskin, supra note 18, at 1340–41 (discussing how Meta’s Oversight Board has relatively little power, and focuses on “concerns that the industry can easily absorb without putting business interests on the line”); Metwally, supra note 46, at 513, 515–16 (noting that social media companies “flaunt” their relationship with third-party experts to highlight their influence on policy and enforcement but are opaque about which expert actors and world views they listen to and which ones they ignore). ↑
139. . Alexa Corse, The ‘Menswear Guy’ May Be the Twitter Algorithm’s Favorite Person, Wall Street J. (Jan. 28, 2023), https://www.wsj.com/articles/elon-musk-twitter-feed-algorithm-menswear-guy-11674933231 [https://perma.cc/47N5-34PR]. ↑
140. . See Tarleton Gillespie, The Politics of ‘Platforms,’ 12 New Media & Society 347, 345–55 (2010) (platforms like YouTube convince advertisers to use them to reach consumers); Ryan Calo, Digital Market Manipulation, Geo. Wash. L. Rev. 995, 1016 (2013) (arguing that digital advertisement is about matching the right advertisement with the right user or consumer); Cohen, supra note 93, at 25 (stating that platforms’ value proposition for advertisers is infrastructural). ↑
141. . See Olivia Solon & Sabrina Siddiqui, Russia-Backed Facebook Posts ‘Reached 126m Americans’ During US Election, Guardian (Oct. 31, 2017), https://www.theguardian.com/
     technology/2017/oct/30/facebook-russia-fake-accounts-126-million [https://perma.cc/9YPU-G2CF] (detailing Russian interference in the 2016 election through digital platform manipulation); Calo, supra note 15, at 1032 (observing that tailored content provides third-parties with “powerful tools of persuasion”); Brett Frischmann & Peter Ormerod, Regulating Manipulative Design Is Not Preempted by CDA 230 or the First Amendment, 75 Emory L.J. (forthcoming 2026) (manuscript at 3) (on file with author), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5587430 [https://perma.cc/5NC6-PL5F] (explaining that social media can manipulate users by hijacking attention, undermining self-governance, and engineering addiction); Daniel Susser, Beate Roessler & Helen Nissenbaum, Online Manipulation: Hidden Influences in a Digital World, 4 Geo. L. Tech. Rev. 1, 9–10 (discussing allegations that Cambridge Analytica improperly interfered with the 2016 election). ↑
142. . See, e.g., Adler, supra note 19 (writing that despite receiving “troubling reports,” OpenAI continued to enable users’ downward mental health spirals); Helberger, supra note 79, at 850 (noting media law and policy will need to shift to diffusing control over opaque algorithms); Ruairí Harrison, Tackling Disinformation in Times of Crisis: The European Commission’s Response to the Covid-19 Infodemic and the Feasibility of a Consumer-Centric Solution, Utrecht L. Rev. Oct. 2021, at 1, 24–25 (comparing Facebook’s response to the European Commission’s calls for an active approach to COVID-19 with its treatment of other misinformation). It is worth noting, however, that social media platforms developed ways to remove all child sexual abuse media and non-consensual sexual media from their platforms (even if X appears to have undermined Twitter’s systems for this removal). ↑
143. . Chapman & Steinberg, supra note 25, at 1–4 (discussing the EU’s Digital Services Act and platform litigation in the U.S.). ↑
144. . See Press Release, Eur. Comm’n, Commission Preliminarily Finds TikTok’s Addictive Design in Breach of the Digital Services Act (Feb. 5, 2025), https://ec.europa.eu/commission/
     presscorner/detail/en/ip_26_312 [https://perma.cc/D8DU-QBMT] (“[T]he European Commission preliminarily found TikTok in breach of the Digital Services Act for its addictive design.”). ↑
145. . See generally Jack M. Balkin, Information Fiduciaries and the First Amendment, 49 U.C. Davis L. Rev. 1183 (2016) (setting forth the concept of an information fiduciary, those with special duties to act without harming the interests of people they collect information from). ↑
146. . See Mike Ananny, Networked Press Freedom 75 (2018) (“[P]ress autonomy emerged from separations and dependencies in the service of objectivity.”). ↑
147. . Leiter, supra note 6, at 905. ↑
148. . See supra note 66 and accompanying text. ↑
149. . See Brian A. O’Shea & Michiko Ueda, Who Is More Likely to Ignore Experts’ Advice Related to COVID-19?, 23 Preventive Med. Reps., 2021, at 1, 3 (2021) (pointing to structural reasons for some vaccine-hesitant individuals’ continued distrust, including discrimination). ↑
150. . See P. Wesley Schultz, Paul R. Hernandez, Anna Woodcock, Mica Estrada, Randie C. Chance, Maria Aguilar & Richard T. Serpe, Patching the Pipeline: Reducing Educational Disparities in the Sciences Through Minority Training Programs, 33 Educ. Evaluation Pol’y Analysis 95, 96 (2011) (discussing how to diversify scientific research communities to better reflect the U.S. population). But see Matthew A. Cannady, Eric Greenwald & Kimberly N. Harris, Problematizing the STEM Pipeline Metaphor: Is the STEM Pipeline Metaphor Serving Our Students and the STEM Workforce?, 98 Sci. Educ. 443, 454 (2014) (arguing that the pipeline metaphor undermines efforts to increase diversity in STEM fields); Angela P. Harris & Aysha Pamukcu, The Civil Rights of Health: A New Approach to Challenging Structural Inequality, 67 UCLA L. Rev. 758, 789–92 (2020) (arguing that the Supreme Court’s individualist view of discrimination has limited the scope and efficacy of legal remedies for discrimination). ↑
151. . See Michael S. Carolan, Science, Expertise, and the Democratization of the Decision-Making Process, 19 Soc’y & Nat. Res. 661, 662 (2006) (observing how facts and values get tangled together in the field of environmental science); Tushnet, supra note 74, at 681 (noting expertise ordinarily has some ideological and political content). ↑
152. . See Tyler, supra note 56, at 128–29 (describing sociological legitimacy as necessitating fairness and ethicality, including respectful treatment). ↑
153. . See Peter Weingart, Trust and Distrust of Scientific Experts and the Challenges of the Democratization of Science, in The Oxford Handbook of Expertise and Democratic Politics 29, 37 (Gil Eyal & Thomas Medvetz eds., 2023) (tracing the change from an elitist hierarchical model toward a participatory, egalitarian arrangement).  ↑
154. . See id. (showing the impact AIDS activists had on the protocols of clinical tests of AIDS drugs and the scientific consensus recognized by law). ↑
155. See generally Ahmed, supra note 82 (showing how feminist lawyers and activists changed the attitudes of the legal and public health professions to the AIDS crisis). ↑
156.  See Harris & Pamukcu, supra note 150, at 810–11 (arguing in favor of justice movements’ empowerment of communities, advocating for a bottom-up approach to leadership, and asserting that front-line communities can serve as a check on expert-driven policies). ↑
157.  See generally Phrashiah Githinji, Alexandra L. MacMillan Uribe, Jacob Szeszulski, Chad R. Rethorst, Vi Luong, Lucy Xin, Laura J. Rolke, Miquela G. Smith & Rebecca A. Seguin-Fowler, Public Health Communication During the COVID-19 Health Crisis: Sustainable Pathways to Improve Health Information Access and Reach Among Underserved Communities, 11 Human. & Soc. Scis. Commc’ns, 2024, at 1 (finding that community leaders helped temper distrust and served as intermediaries for health information). ↑;