Vol. 98, No. 5 Archives – Page 2 of 2 – Southern California Law Review

Pluralism and Listeners’ Choices Online

September 2025September 2025Alan K. Chen*

“The plain, if at times disquieting, truth is that in our pluralistic society, constantly proliferating new and ingenious forms of expression, ‘we are inescapably captive audiences for many purposes.’ ”¹

The speech and technology world has changed dramatically, even unimaginably, since Justice Powell penned these words about drive-in movie theaters. In attempting to grapple with this quandary in the contemporary era, James Grimmelmann offers us the provocative and original paper, Listeners’ Choices Online.² His contribution to this Symposium builds on earlier work in which he argues for a theoretical approach to free speech that makes listeners’ interests the central focus of First Amendment doctrine.³ As he argues in the earlier paper, freedom of expression involves what he calls a “matching problem”—ideally lining up speakers with listeners who want to hear their expression, but not with listeners who do not.⁴

The current paper is simultaneously both too complex and too nuanced to adequately summarize in this brief Comment, but here are a few of his main points as I interpret them, and that my comments will address.

Facilitating matching between willing speakers and willing listeners is the goal of a system of free speech. In that regard, “listeners’ choices matter more than speakers’.. . . A consistent

commitment to protecting these willing speaker-listener pairs results in a system of First Amendment law that regularly defers to listeners’ choices.”⁵

Applying that model resolves some of the important First Amendment questions arising from the regulation of contemporary electronic speech media.
It is useful to disaggregate communication media into four types, each of which presents distinct matching challenges: (1) Broadcast (television, radio, cable); (2) Delivery (telephone, email, messaging); (3) Hosting (providers of space for speech, but not engaged in speech themselves); and (4) Selection (directing listeners to specific content via algorithms based on the perception of listener preferences).⁶ Currently, hosting and selection functions are frequently combined, though that does not have to be so.⁷
Selection intermediaries play a key role in determining what listeners hear or see. This is an essential function because the sheer volume of speech available on the Internet creates otherwise insurmountable attention scarcity problems for listeners.⁸
This listeners’ choice model allows for limited regulatory interventions on the media’s selection functions that would not violate the First Amendment.
It would violate the First Amendment for regulators to prohibit intermediaries from offering listeners the ability to choose what speakers to listen to because that interferes with listeners’ right to listen.⁹
However, the government may permissibly intervene when a search engine (or, presumably, other selection intermediary) is dishonest or disloyal to its users, “when it shows them results that (objectively) differ from the engine’s own (subjective) judgment about what the users are likely to find relevant,”¹⁰ because that also interferes with listeners’ interests.
It would also be permissible to have a rule requiring pure selection intermediaries to treat first-party content evenhandedly with content posted by third parties.¹¹
“Seeing the Internet from listeners’ perspective is a radical leap. It requires making claims about the nature of speech and about where power lies online that seem counterintuitive if you are coming from the standard speaker-oriented First Amendment tradition. But once you have made that leap, and everything has snapped into focus again, it is impossible to unsee.”¹²

There is much to admire in Professor Grimmelmann’s paper. It makes a number of important and original contributions to thinking about the regulation of social media and is in many parts completely persuasive. First, consistent with the objective of this Symposium, it highlights listeners’ interests as a basis to evaluate the American system of freedom of expression. It is indisputable that the Supreme Court and legal scholars have underappreciated the role of listeners’ interests in articulating First Amendment doctrine.¹³ The primary context in which the Supreme Court expressly considers listener interests involves unwilling listeners as captive audiences, but those are the only cases that place listeners’ interests at center stage.¹⁴ The Court has upheld legal rules that bar speakers from imposing speech on unwilling listeners when the listeners’ “substantial privacy interests are being invaded in an essentially intolerable manner.”¹⁵ Even in captive audience situations, as Grimmelmann points out, under current doctrine the interests of willing listeners will sometimes outweigh the rights of unwilling listeners, particularly if it is easy for the latter to avoid the speech.¹⁶

Listeners’ Choices Online also offers us a way out of the ongoing effort to find the appropriate perspective through which to evaluate how First Amendment doctrine should apply to the contemporary media environment. Much recent scholarship has struggled with this question, with legal scholars sometimes seeking to find appropriate analogies from regulation of past communication technologies to justify a legal framework for thinking about the regulation of social media platforms.¹⁷ Is cable television like traditional television and radio broadcast media? Does regulation of telephone services offer any insight into how we ought to regulate digital communications? Is Facebook more like a parade or a shopping mall? Can social media companies be treated like common carriers, subjecting them to greater regulatory constraints than would otherwise be permissible to impose on private companies engaged in speech?¹⁸

None of the analogies work perfectly, however, because each different electronic speech medium bears some distinguishing features that complicate the analysis.¹⁹ Some, as the article points out, are mere vessels for delivery of content, while others engage in important speech-impacting selection decisions that help listeners sort through the onslaught of online content, but, in doing so, may affect listeners’ interests by providing them content they do not want to hear or directing them away from content they would welcome.²⁰

The Supreme Court has only just dipped its toes in the water, in its dicta in last term’s Moody v. NetChoice, LLC, with the majority opinion stating unequivocally that “[l]ike the editors, cable operators, and parade organizers this Court has previously considered, the major social-media platforms are in the business, when curating their feeds, of combining ‘multifarious voices’ to create a distinctive expressive offering.”²¹ But as Grimmelmann points out, that is looking at the challenged state laws exclusively from the platforms’ perspective, and not the listeners’.²²

Rather than attempting to argue purely by analogy with past regulations of earlier media technologies, Grimmelmann’s paper elegantly uses listeners’ interests and choices as an organizing principle that cuts across these different media to create a coherent First Amendment model for evaluating media regulations. He suggests that focusing on these interests allows us to see more clearly the competing speech interests involved in ways that the purely analogical approach simply cannot. His listeners’ choice theory emphasizes matching speakers to willing listeners, which can be accomplished by structural designs, by some content neutral government regulation, and, in part, by requiring the separation of hosting and selection functions in ways that maximize these speaker-listener connections.²³

While Professor Grimmelmann’s model is intriguing and helps us think about media regulation in useful ways, I offer three modest thoughts, two focused on whether, in some circumstances, prioritizing listeners’ rights may come at the expense of other important First Amendment values, and one questioning whether there is a need for further promoting listeners’ choices on social media given the increasing market for niche social media sites.²⁴

Prioritizing Listeners’ Choices May Diminish Public Discourse

First, permitting limited regulation of selection intermediaries to protect listeners’ interests could, in some cases, have deleterious effects on public discourse. Even the modest regulatory interventions that Grimmelmann suggests would be permissible to advance listeners’ interests could be leveraged to challenge selection intermediaries’ decisions to offer a more balanced, fact-checked feed to their subscribers. Or, even if those effects do not come to fruition, the very existence of regulatory interventions might deter selection intermediaries from experimenting with innovations to promote delivery of a greater diversity of content that does not cater purely to listeners’ interests.

Consider a hypothetical new platform calling itself Balanced Social Media (“BSM”). Following Grimmelmann’s model, let us assume that a different company is the host for BSM, which exclusively serves a selection function. BSM designs an algorithm that, for the most part, favors listeners’ choices of content, but adds three specific features that veer from the default rule. First, it builds in its own fact-checking mechanism that flags content posted by third-party users that may be objectively false or come from sources that have proven unreliable or inaccurate in the past. The BSM algorithm will still direct the user to that content, but the content will be marked with a red flag that warns the user that the factual foundation of the material may not be valid, and provides a link to a source that disputes the factual validity of the original post.

Second, the algorithm is designed to monitor users’ feeds to determine if they are seeking content that is unilaterally biased toward one particular ideology, for example, if a user reads only content posted by Fox News or MSNBC. If the algorithm identifies users who seek ideologically unbalanced content, it will occasionally feed such users some third-party content that comes from a dissimilar political perspective. This counter-ideological feed could come randomly or perhaps after the user has viewed ten consecutive stories from sources with their preferred ideological perspective.

Alternatively, BSM could instead offer a slightly less intrusive option under which, rather than posting counter-ideological content, BSM could give the user a warning or notice to the effect that the user has been reading content that is exclusively coming from sources with a specific political orientation and asking if the user would like to see something from a different perspective. This might operate in a manner like TikTok’s option for its users to set a daily screen time limit and be notified when they have reached that limit.²⁵ Under my hypothetical, however, users would not be able to turn off this setting.

Third, BSM occasionally posts its own independent content on the platform that discusses issues regarding the responsible use of social media and the importance of ensuring that information is factually accurate before posting it. As with the counter-ideological posts, it will feed periodically into all users’ feeds. BSM users cannot opt out of any of these functions; though, of course, they may decide they do not want to use BSM. When users sign up to use BSM, they are fully informed about the algorithm’s functions, which they agree to as part of the Terms of Service (“TOS”). The TOS even says, “BSM offers a new vision of social media, one that will deliver content that you did not ask for, or even that you do not want to see (of course, we cannot make you read it, that is up to you!). The goal of our model is to expose all people to a range of ideologically diverse content.”

Grimmelmann’s model seems to suggest that lawmakers might be able to forbid BSM to adopt these innovative features because they do not fully promote listeners’ choices. The fact-checking flags and counter-ideological feeds are content that many users may not wish to see; indeed, they may be viscerally repelled by these posts, particularly if this interferes with their ability to experience the emotional resonance associated with speech that highlights their own world views.²⁶ He suggests that regulators may be able to restrict selection intermediaries’ use of such algorithms to the extent that “it shows [users] results that (objectively) differ from the engine’s own (subjective) judgment about what the users are likely to find relevant.”²⁷ In fact, BSM’s model is designed to show user content they do not want to see. In Grimmelmann’s terms, the intermediary is being disloyal to its users (although because the algorithm’s functions are fully disclosed in the TOS, they can argue they are not being dishonest).²⁸

Moreover, the BSM-produced content (and maybe even the fact-checking posts) can be viewed as first-party content.²⁹ BSM is in some sense trying to compete in the social media market by offering a new way of delivering content. Would a pure listener-based approach result in such experiments being shut down by regulators because they are occasionally giving their first-party content priority over content posted by third parties?³⁰ That is, by feeding users first-party content in the form of sermons on the importance of truth in the responsible use of social media, has BSM interfered with listener choice? Because Moody holds that social media platforms are speakers when they make decisions about content moderation,³¹ they are unquestionably speakers if they are producing their own content. How would Grimmelmann’s model address the tension between a regulation prohibiting BSM from prioritizing first-party content to protect listeners’ choice and the platform’s First Amendment speech rights?

To the disloyalty argument, Grimmelmann might respond that because BSM is transparent about its algorithm, it is not actually being disloyal or dishonest to its users.³² Listeners who do not want this type of balanced approach can simply choose a different platform that better suits their listening tastes. However, while BSM is certainly giving listeners choice at the first level (platform selection), its model will inevitably result in some BSM users receiving speech at the second level (content selection) that they subjectively do not want to hear.

Elevating Listeners’ Choices Could Encourage Information Silos

A closely related concern with a system of electronic media regulation focusing primarily on promoting listeners’ interests is whether such an emphasis could have the broader systemic effect of exacerbating ideological information silos even more than under the current system.³³ An important function of a system of free expression is, of course, promoting robust public discourse. Public discourse is inherently oppositional—speakers of different viewpoints must be able to engage each other for it to meaningfully occur.

In many cases, speakers desire to reach listeners whom they believe will be persuaded by their messages if those listeners only had an opportunity to hear them. Anti-abortion advocates may sincerely believe that if women considering abortions only had more information, they would make different choices. Protesters concerned about the humanitarian crisis associated with Israel’s military actions in Gaza would like to reach those who are unconditionally sympathetic to Israel’s right to defend itself because they think, with additional information, these listeners may modify their positions. On social media as well, speakers try to convince unwilling listeners of the virtues of their political positions. Preaching only to the converted does not facilitate healthy discourse.

Outside of the captive audience context, which is almost exclusively applied to unwanted speech in one’s home,³⁴ a key function of the First Amendment is served by advancing the interests of speakers to influence those who are not inclined to agree with them.³⁵ As the Supreme Court has recognized:

[Speech] may indeed best serve its high purpose when it induces a condition of unrest, creates dissatisfaction with conditions as they are, or even stirs people to anger. Speech is often provocative and challenging. It may strike at prejudices and preconceptions and have profound unsettling effects as it presses for acceptance of an idea.³⁶

These express values are in direct tension with a purely listener-based approach. This may be particularly true of speech on social media, which the Court has argued is one of the “most important places . . . for the exchange of views.”³⁷

Thus, a second concern I have with a model prioritizing listeners’ choices over speakers’ is that its application, in many contexts, may impede what we might describe as lawful, but uncomfortable, speech that is intended to persuade.³⁸ If listeners can confine themselves only to speech they want to hear, even in the social media context, then prioritizing that interest can operate as a kind of quiet heckler’s veto. In a social media environment in which listeners’ choice prevails, it is hard to imagine how persuasion might work, either individually or collectively. Are there any situations involving such speech through media in which the default position is not valuing the listener over the speaker, and if so, how could that decision be implemented?

Perhaps our society is headed in this direction already given that, as Grimmelmann observes, even in the absence of regulation of selection intermediaries, listeners might deploy a combination of pure hosting platforms with middleware, a third-party software that allows them to customize their feeds at an even greater level of detail.³⁹ While this, too, would benefit listeners’ choices, it would move us in the direction of a more atomized speech universe—which is not necessarily a good thing, but at least it would not be the product of government intervention.

Market Responses Are Already Enhancing Listeners’ Choices

Finally, one could argue that market forces are already moving toward a listener-centric model with the proliferation of niche social media platforms, even in the absence of regulatory interventions.⁴⁰ There is some evidence that users are beginning to migrate from more general social media sites such as X (formerly known as Twitter), to special interest platforms where they can avoid the cacophony of hostile rhetoric in favor of sites where they can engage with a smaller cohort of people who share common interests. That development certainly enhances listener choice without risking the possible unintended consequences of regulations designed to promote listeners’ choice.

Especially during the 2024 election season, there seemed to be growing dissatisfaction with general social media sites because of the unavoidability of sometimes harsh political discourse. It was not uncommon to hear calls for platforms dedicated to only discussion of books, movies, music, gaming, and other mostly nonpolitical (or, at least, not primarily political) topics that listeners sought out to find some respite. Sports lovers initially were the exception to this rule, although even those users have now started fleeing X.⁴¹

Available statistics suggest that the market has responded to this interest and is already enhancing listener choice by serving its own matching function. About 115,000 users deactivated their X accounts on the day after the November 2024 Presidential Election.⁴² No matter how the total user base is measured, that is a very small percentage, which is unsurprising because network effects deter people from leaving even platforms with which they are dissatisfied. Of course, people can maintain active X accounts while still seeking out other outlets for speech. In comparison, niche social media platforms are still quite small. One of the largest, Goodreads, a platform to share book recommendations, had about 150 million users as of 2023.⁴³ Houzz, a home design social media platform, reportedly has about 70 million users.⁴⁴ A platform for movie lovers (especially indie) called Letterboxd now has about 17 million users.⁴⁵ Reddit, while open to a wide range of users, is well known for facilitating smaller communities to generate discussion of interest, and now has about 91 million daily active users.⁴⁶ Finally, Substack, a platform for distributing individualized newsletters to both paid and unpaid subscribers, now has approximately 50 million subscribers.⁴⁷

It may seem somewhat contradictory to fret about information silos while simultaneously celebrating the expansion of niche social media sites. To address this briefly, I would argue that the siloing problem is much more problematic on the larger, omnibus social media platforms than with niche social media platforms. Political discourse is one of the main features of the larger platforms, so cutting off speech that is ideologically diverse is truly undermining the opportunities for persuasion. In contrast, the niche social media sites are mostly excluding posts about other topics not because of any ideological commitments, but rather to help filter out what they regard as irrelevant information. That is not to say that political discourse cannot arise in the context of these niche sites,⁴⁸ but it is at least less likely to do so. And, of course, these users may be walling themselves off from any political speech, which could be problematic for public discourse in the long run. But there is nothing to suggest that these users might not still engage in political discourse on other platforms or in other contexts of communication in their non-online lives.

* * *

Notwithstanding my limited reservations and questions, I wholeheartedly welcome Professor Grimmelmann’s important and valuable contribution to thinking about the complex constitutional and social issues associated with regulation of electronic media in the current climate. Continued efforts to meaningfully apply standard First Amendment doctrine to new media allow us all to think critically about the best way forward.

98 S. Cal. L. Rev. 1387

Download

Alan K. Chen*

* Thompson G. Marsh Law Alumni Professor, University of Denver Sturm College of Law. Thank you to Erin Miller and to the editors and staff of the Southern California Law Review, and especially Simone Chu, for their efforts in organizing this fantastic Symposium. Thanks also to Nina Christensen and Charlotte Rhoad for their helpful research.

Unchosen Listening

September 2025September 2025Erin Miller*

INTRODUCTION

A century of developments in communications technology has done wonders for listeners.¹ In the not-so-distant past, the listener had few choices regarding the speech they heard. An urban listener might find a range of choice in libraries or bookstores, or among private associations’ meetings or periodicals. But mostly listeners heard what others happened to say, on the job, at church, in the neighborhood, and on the street-corner. Today, listeners have more choices than ever, and more ability to choose which speech to hear and which not to hear.

First Amendment doctrine, for its part, has occasionally shielded these listeners’ choices about whom to listen to, even over speakers’ choices about whom to speak to.² I largely agree with the authors in this symposium who argue that it should do so more often.³ The appeal is evident. Communication goes most smoothly with a willing speaker and a willing listener. And with listeners now awash in speech options, someone must choose among them. The listener seems obviously superior, as chooser, to many looming alternatives: the stalker, the fake-news purveyor, the oligopolistic media platform, or—most antithetical to the First Amendment—the government.

Yet this Commentary urges caution in so fully embracing listener’s choice that it becomes a paradigm of First Amendment jurisprudence. One can easily move from recognizing the advantages of willing listening (and speaking) to identifying the “core” of constitutionally protected speech as not just a “joint activity”—to use Ashutosh Bhagwat’s illuminating term⁴—but mutually consensual, affirmatively chosen by both speaker and listener.⁵ One might even begin seeing something constitutionally suspect about speech that lacks a consenting listener. On this interpretation, the fact that the speaker on his soapbox in Washington Square Park annoys some passersby is, while inevitable, still unfortunate—a sacrifice made for the sake of the other, more willing passersby.⁶

Yet in a not-too-distant future, this fact might be “correctible” by technology, assisted by artificial intelligence and mass data collection. Every webpage you visit might seamlessly filter out any content you have elected not to see; or software might elect for you, based on your eyes’ patterns of lingering. When you step onto the street, you might wear digital glasses to blur out all written speech that you have deemed unwelcome, and digital headphones to cancel all noise except oral speech you have deemed welcome.⁷

While this seems a paradise for listeners’ choice, it would not be one for listeners’ interests, another traditional First Amendment concern.⁸ The speech that we choose to hear does not always make us better off.⁹ Certainly, a listener’s choice is one aspect of their interests. And speech they vehemently reject—such as harassing speech—will rarely be in their interest. But a gray zone exists between chosen listening and coerced listening, which I will call unchosen listening. Some unchosen listening, I will argue, is a desirable part of being an autonomous person and citizen.

I will also argue that it is desirable for the broader society of which listeners are a part—including its collective knowledge, culture, and, especially, democratic institutions. These societal interests underlie the First Amendment, too, as Alan Chen also discusses in this symposium.¹⁰ Perhaps for these reasons, unchosen listening has been revered in First Amendment doctrine. As the Supreme Court declared in Terminiello v. Chicago, speech “may indeed best serve its high purpose when it induces a condition of unrest, creates dissatisfaction with conditions as they are, or even stirs people to anger. Speech is often provocative and challenging. It may strike at prejudices and preconceptions and have profound unsettling effects as it presses for acceptance of an idea.”¹¹

Indeed, the most fiercely protected First Amendment speech—perhaps its real “core” case—involves non-consensual listeners: speech in the public forum.¹² This means citizens trying to move or persuade fellow citizens, and especially those none too thrilled to listen. It is—for those familiar with the First Amendment landmarks—Abrams’s communist pamphlets thrown from the roof, Kovacs’s pro-labor commentary amplified from his truck, Mosley’s sign decrying race discrimination, Cohen’s jacket protesting the draft.¹³ And it is arguably the only speech government has a (currently recognized) positive obligation to protect—by preserving the openness of traditional public forums—under the First Amendment.¹⁴

While the Court has seldom expounded on the theme of unchosen listening, its silence invites less doctrinally bound theorizing. This Commentary thus offers a brief normative case for resisting a listener’s-choice paradigm (and maintaining more of a public forum paradigm).¹⁵

I. The Case for unchosen listening

What speech will a listener hear, if they mostly choose that speech themselves or have it selected for them based on algorithmic predictions of their preferences?¹⁶ Chances are, it will be fairly insular and homogeneous. That is, it will be familiar, comfortable, and unchallenging; and it will come from like-minded speakers and a small number of sources.¹⁷ Its insularity and homogeneity will be reinforced to the extent that the listener chooses with greater granularity—that is, chooses not just among sources or topics but among specific speech from each source or specific views on each topic.¹⁸ The result might resemble “The Daily Me” envisioned by Cass Sunstein.¹⁹ While not all listeners will choose like this, and few will do so all the time, enough listeners will likely do so in enough cases to affect the overall patterns of communication in society.

Assuming this is true, the following discussion identifies five significant interests that can be undermined when listening becomes too tailored to a listener’s choices: preference optimization, error avoidance, innovation, self-evolution, and tolerance.²⁰ Each interest has an individual dimension: for listeners themselves. Each also has a collective dimension: for a society in which the total pool of speech consumed is determined largely by listeners’ choices.

A. Optimizing Preferences

Even listener-chosen speech does not always satisfy listener preferences.²¹ Speech often defies expectations. Sometimes we have a “meh” reaction to a movie by our favorite director, or are spellbound by a movie we expected to bore. Our expectations are sometimes even insufficient for choice. I may have almost no sense of whether I would like a movie from an unfamiliar genre, because I have not experienced anything like it before. After all, the indicators of content that are available prior to hearing speech—for example, its speaker, venue, general subject, or reviews—are necessarily limited.

With ever more personalized listening technology, our mistaken choices—or mistaken predictions of our choices by algorithms—can lock us into inferior patterns of listening. Mistaken choices are easily corrected when they produce dissatisfaction, but less so when they produce merely suboptimal satisfaction. Say that I have watched and liked one Fellini film and one Spielberg film. Perhaps I am directed, by choice or algorithm, toward more Spielberg films rather than Fellini films. I continue to like Spielberg films and hence grow my love of Spielberg. I may never know the counterfactual––whether I would have preferred the Fellini films. The best chance of exiting a suboptimal equilibrium is exposure to unchosen speech, or at least less-precisely-chosen speech. Optimizing, therefore, involves reflective equilibrium between chosen and unchosen exposure.

A collective cultural danger lurks here, too, at least insofar as listeners rely on algorithmic services to make these finer speech selections for them.²² Because these services operate within a market economy, their corporate makers may face incentives to nudge listeners toward more popular, hence more easily satisfiable, preferences. Forced to rely on the same services to find listeners, speakers—including musicians, writers, filmmakers, journalists, and others—may also feel pressure to create content to win the algorithm’s game, as many have already.²³ The result may be a blander culture and public sphere, with fewer options over which to optimize our preferences.

B. Avoiding Error

Listening primarily to chosen speech may proliferate error. The first problem is that it may skew our available evidence. If, as assumed, we choose to listen to speakers who share our beliefs, including false ones, then they will be unlikely to offer evidence against those beliefs.²⁴ They may even pass on new false beliefs based on old ones. These harms are well-documented within “epistemic bubbles,” that is, discussion spaces in which certain views are systematically excluded.²⁵

A second problem is that too much chosen listening may objectionably narrow the range of subjects on which we possess evidence. People may choose to listen to little or no speech concerning decisions they will or must take, including voting. For instance, research suggests that the most promising ways to inform voters may involve incorporating speech that they do not choose alongside speech they do—such as entertainment.²⁶

Yet a third problem is that a steady diet of chosen speech may dull our capacity to think itself. As John Stuart Mill observed 166 years ago, reasoning is better stimulated and optimized by speech that is uncomfortable and challenging.²⁷ But chosen speech is, by assumption, typically the opposite. Chosen speech can exacerbate self-favoring and in-group-favoring cognitive biases.²⁸ Chosen speech may not even offer reasons, because it does not seek to persuade. Chosen speech can, by repeating or just assuming our beliefs, reinforce and bury them so deeply that we have trouble getting the “mental distance” necessary to scrutinize them—to question them and their justification.

C. Innovation

Unchosen listening can stimulate not just rational thought, but more creative and generative forms of thought that do not merely analyze evidence but propose new ideas or solutions.

Creativity is by its nature unexpected. It often involves making connections among facts, ideas, disciplines, cultures, and so forth that others—including oneself—had never thought to draw. We are thus sometimes likelier to arrive at an innovative solution not by continuing to think directly about the problem—constrained by our previous expectations—but by allowing our mind to wander and randomly associate.²⁹ The revolutionary inventions, discoveries, insights, and movements that have been prompted by contact with unusual contexts, other disciplines, or other cultures are countless. Consider an eclectic sampling. Clinical psychology has been seriously influenced by Buddhist religious practices. Impressionism in painting was largely inspired by Japanese ukiyo-e prints³⁰; and the painting of Jackson Pollock, darling of the Supreme Court’s own First Amendment cases, was likely inspired by Native American sandpainting.³¹ The QR code was suggested by the Go board,³² while the foldable shape of heart stents was suggested by Japanese origami.³³ Numerous engineering innovations, from aeronautics to robotics, have come from observing animals.³⁴ The latest monumental computer science invention, large language models, grew from insights in both linguistics and cognitive science. And this symposium’s own Aziz Huq uses observations in sociology for thinking about digital speech.³⁵

Creativity thus seems likeliest to be prompted by exposure to speech that is unexpected, or at least unguided by our own choices.³⁶ Unchosen speech can offer new concepts and frameworks of thinking that supplement or integrate with existing ones. Or it can simply interrupt habitual frameworks—generating another (creative) form of mental distance—and thus open the mind to the potential relevance of unexpected inputs. By contrast, chosen speech can lock in certain expectations of relevance.

One might expect collective, rather than individual, innovation to be fueled through highly personalized listening. It might segregate most listeners into speech subcultures, within which they could refine their own insular ideas—like members of a species separated among islands and continuing to evolve. But then a few innovators sampling from those radically diverse subcultures might produce innovations of even greater enormity. Yes, it seems just as plausible that, so long as enough listening is chosen, occasional sampling by a greater number and diversity of listeners would provide offsetting benefits of more sophisticated—if slightly less radically distinct—subcultures and more frequent innovation.³⁷

D. Self-Evolution

Persons arguably have a fundamental autonomy interest in being able not just to make up their minds, but to change them, too. This ability to change applies to their beliefs, as described above, but also their ends—their important goals, values, commitments, and so forth—and their tastes.³⁸ I do not mean the mere freedom to zig when one previously zagged. Certain factors beyond our control inevitably change over time: our external circumstances vary, and our set of experiences and beliefs expands. In response, it is often appropriate or even necessary to change ourselves in order to cohere, adapt, and grow. Otherwise, we become frozen into a course determined not by us, but by the “dead hand” of our past self.³⁹ Indeed, one might say that autonomous self-development just is the continual evolution of our self in response to new, external inputs.

It is almost always possible to change our ends, at least insofar as nothing physically prevents us. But pursuing our ends is often habitual, or even inertial. For us to have a meaningful opportunity to change,⁴⁰ we must occasionally gain mental distance—of yet another sort—sufficient to reflect directly on our existing ends and their continued desirability, with adequate understanding of alternatives.⁴¹

Listening only to chosen speech impedes this mental distance. Because we as listeners choose speech based on who we are, that speech is unlikely either to prompt critical self-reflection or to present an adequate range of alternatives for whom we could become. It may even passively reenforce our present ends and hinder our evolution.⁴² By contrast, consuming speech that is unchosen and hence more unfamiliar or even genuinely surprising is likelier to enable self-transformation.

A society that lacks such dynamism among its members, too, seems destined to remain trapped in outdated ways of life. As its membership grows and shifts, and the world changes, it may not evolve—culturally, politically, or legally—to meet new needs. As a staunch advocate of “[a] constitutional regime fearful of political entrenchment and dedicated to continual adaptation,”⁴³ Justice Oliver Wendell Holmes saw protection of specifically “the expression of opinions that we loathe” as indispensable.⁴⁴

E. Tolerance and Persuadability

In a pluralistic society, we must cooperate across irresolvable differences and disagreements. We must work with and take into account the interests of those with values, beliefs, and tastes significantly different from our own (let us call them our counterparts). Sometimes we must reach compromises with our counterparts, even at partial expense of our own ends. This demands tolerance toward our counterparts, that is, engaging and respecting them as rational agents and persons as we engage with them. We must, so far as possible, attempt to understand and make sense of their claims, identify issues of agreement, and persuade them on issues of disagreement.

Highly personalized listening can be expected to decrease tolerance. Humans have a well-documented proclivity to exhibit negative personal attitudes toward those who hold views—especially moral and political views—with which they strongly disagree.⁴⁵ We may also interpret the speech of our counterparts in uncharitable ways, thereby missing any merit in it.⁴⁶ But these effects can be countered, the more often we hear counterparts speak.⁴⁷ The more we so listen, the likelier we become to succeed in the social cooperative tasks mentioned above, and to empathize with our counterparts.

Tolerance is useful within any social group, but its usefulness is heightened within a democracy. For in a democracy, we exercise coercive power mutually over each other. As Jonathan Gingerich has argued, all participants in a democratic system, and especially potential losers, must have a genuine opportunity to persuade their fellow citizens on political issues.⁴⁸ This logically entails that citizens must open themselves, even ever so slightly, to the possibility of being persuaded⁴⁹, in light of their own fallibility and the basic rationality of other citizens—at least enough to actively listen.⁵⁰ For any opportunity for a speaker to persuade, however slight, dies if all listeners tune out.⁵¹

One might go further. Elected officials within a democracy owe a duty to their constituents to hear out their political concerns, no matter how much the officials would prefer not to. One might by extension argue that voters, as joint rulers over one another, have an obligation to hear out the political concerns of at least some of their counterparts.⁵² As Justice Louis Brandeis once declared, “public discussion is a political duty.”⁵³

II. Limitations and Implications

Having made the general case for unchosen listening, I offer a few brief remarks to situate it within contemporary free speech doctrine.

Hopefully it is clear that I do not mean to reject listener choice as a critical First Amendment concept. Listener choice is rightly viewed by courts as definitive in certain limited domains, such as the home.⁵⁴ Arguably courts should, even most of the time, defer to the listener’s choice, as the least objectionable, content-neutral proxy for their interests. After all, listeners’ choices usually converge with their personal interests. Respecting listeners’ choices also tends, through market mechanisms, to generate decently diverse media options, which are good for their democratic interests.⁵⁵

As mentioned at the outset, I am also inclined to believe that the First Amendment does not protect a right to coerce listeners.⁵⁶ The final scene of A Clockwork Orange is not a good look for the Constitution. Some speech, such as hundreds of creepy Facebook messages,⁵⁷ may be constructively coercive, because of its repetition, intimidation, or substantial disruption of normal activities. Other speech may be coercive because of time-constrained psychological vulnerabilities of a listener.⁵⁸ But listening is not coerced simply because unrequested, unaccepted, or even undesired. I am inclined to believe that even speech that ignores a listener’s affirmative assertion of unwillingness is not coerced, absent such exacerbating factors.⁵⁹ Speech is not sex. Nor is listening coerced if the asserter had some minimal exposure to the speech on a prior occasion—the glimpse before we “avert our eyes” or click to delete.⁶⁰

One of the main venues for speech of this sort is the public forum. Of course, speech in these spaces works better—to spark correction, creativity, transformation, or tolerance—when the listener is game. Indeed, the more actively a listener resists, the less likely speech will achieve any such benefits. But even if most listeners avert their eyes against unwanted speech in public, they at least gain minimal exposure to the message and the fact that someone wants to speak it. Even that regular exposure can heighten tolerance and reduce resistance toward other, perhaps more articulate, unchosen speech.

Virtual forums can prevent even such minimal exposure to speech. With digital listener’s-choice technology, you need not avert your eyes whenever you see an unwelcome message, but can often effectively swipe right to pre-screen all similar messages. Indeed, some messages can be silenced not because you affirmatively resisted them, but because messages you liked better were prioritized over them. Yet media platforms may increasingly turn to listener’s choice as a means of filtering principle that both satisfies users and evades responsibility.

Aggressive use of listener-chosen filtering may even portend the demise of the traditional public forum. The area of publicly owned space available for talking appears to have been shrinking for decades, as the Court has recognized.⁶¹ Listeners, willing and unwilling, have fled those spaces even faster since the digital era. Research shows that pedestrians now walk faster, and interact less often, in public spaces.⁶² Many factors seem to exacerbate this trend: ever more densely developed urban areas,⁶³ constrained public budgets,⁶⁴ perceived rises in crime, architectural features designed to make spaces less welcoming (likely targeted at the homeless),⁶⁵ and increasing stretches of everyday life spent online. When we talk to strangers, it is thus mostly in private, digitally moderated spaces, to the willing. Even when we do find ourselves conversing in physical spaces, severe geographical political segregation means that our interlocutors will often still be likeminded.⁶⁶ All of this stymies the serious engagement across difference that is critical to the democratic process.

Admittedly, the public forums of old were never ideal sites for reasoned discourse. Soapbox speakers are typically better at stirring passions than parsing arguments. Yet the solution should not be doctrinal abandonment of the public forum, and the commitment to unchosen listening that it represents.⁶⁷ Rather, courts should stand ready to approve the expansion of spaces in which we encounter at least some unbidden speech—if necessary, privately owned ones⁶⁸—and encourage reason-giving within them.⁶⁹ That would be the First Amendment paradise.

98 S. Cal. L. Rev. 1399

Download

Erin Miller*

* Associate Professor of Law & Philosophy, University of Southern California Gould School of Law. I am grateful to the participants in this symposium for inspiring this comment, to the student editors of the Southern California Law Review for their diligence at every step of the production process, and to Ashutosh Bhagwat, Rebecca Brown, James Grimmelmann, and Felipe Jiménez for detailed feedback.

The First Amendment of Fear

September 2025September 2025Nelson Tebbe*

Introduction

Fear can be a powerful silencer. Speakers may be thwarted not only by direct force but also when they check themselves because they anticipate adverse consequences. Some assessment of costs and benefits is involved whenever anyone decides to communicate, of course. That is normal and actually valuable. Yet acute anxiety, caused by the realistic prospect of violence or other grave harm, differs from ordinary consequential reasoning, even if both result in silence.

Today, speakers seem to be hesitating with concerning frequency. Their reticence is understandable, because disagreement and its consequences have become severe in certain settings. Acute fear of speaking has affected those on the right and on the left, though not perhaps in the same way or to the same degree. Consider an example at Columbia University. Reportedly, a truck with a billboard bearing the words “Columbia’s Leading Antisemites,” alongside the names and faces of students and faculty, appeared in Morningside Heights and drove slowly around campus.¹ The truck targeted Columbia affiliates who allegedly had signed a statement of solidarity with Palestinians in Gaza. The stunt was reported to be a project of a conservative media group, which also published a website that listed students and faculty members and was updated regularly.² People named on the website were said to be members of various campus groups that were engaged in protests against Israel’s military actions. The website asked readers to send messages to Columbia’s board of trustees urging them to “take a stand” against “these hateful individuals.”³ The group also purchased domain names that corresponded to the actual names of several students and faculty on the list, and it planned to send the truck to the targets’ homes. Two Columbia law students who were named had job offers withdrawn by law firms, according to news outlets covering the story.⁴

Although debates over Israel’s military actions in Gaza are particularly fierce, they are not unique. Fear of speaking seems to have intensified as political conflict has escalated in the United States and elsewhere.⁵ As differences have deepened and political identities have tribalized, speakers noticeably have withdrawn, nervous about the possible results. Some antagonism is inherent in healthy democratic discourse, and it is not cause for regret. Criticism is a feature of deliberation, and it is valued by First Amendment traditions. But anticipation of systematic violence is something of a different order, at least arguably. Intimidation like that can degrade democratic discourse and political cooperation—or that at least is the worry with respect to freedom of expression.

Technological changes have contributed to the climate of anxiety, of course. Although some of the activity at Columbia was analog—it took the form of a truck circling campus—other aspects have leveraged the efficiency of digital media. Today, any utterance can be preserved and disseminated, instantly and cheaply. Anonymity reduces accountability for the intimidation; though anonymity also can serve freedom of expression,⁶ its possible piercing can disincentivize debate. A feeling of surveillance can result—the sense that something you say can provoke reprisal that is utterly devastating.

Although this is hard to document, the university classroom itself shows signs of being impoverished by the effects of systemic fear. On questions of political controversy, students appear reluctant to volunteer views that even conceivably could expose them to retribution or stigmatization. Faced with a choice between the exploration of ideas that entails the risk of retribution and the safety of silence, many students opt for the latter, it seems. And that is true of at least some students on the right and left.

Professor Helen Norton’s insightful essay for this symposium explores a related dynamic surrounding the law of stalking.⁷ For purposes of this Essay, what her sophisticated analysis illustrates is a more general phenomenon, namely that speech generating fear can itself have a silencing effect. Although the constitutional debate surrounding stalking manifestly concerns the expressive rights of the stalker, its latent lesson is that there are speech interests on both sides—that the victims of harassment and intimidation themselves can become muzzled.⁸ Understanding that dynamic complicates any consideration of the First Amendment of fear.

This short Essay seeks to make modest progress on understanding and approaching the relationship between extreme fear and freedom of speech. Part I draws inspiration from Judith Shklar, who famously built a liberal political theory designed to shield citizens from fear.⁹ Though her theory was concerned solely with government oppression, and though she built up only a minimal kind of political morality, it can be complicated and complemented to include private sources of intimidation and, relatedly, to comprehend a positive government obligation to ensure the basic social and material conditions for a healthy speech environment. Part II then applies this political conception to the problem of doxing. State statutes regulating doxing already exist, though they mostly have not yet been tested for adherence to the First Amendment.¹⁰ Whether and how those laws comport with the right to freedom of expression is a complicated issue, one that must include consideration of the expressive interests of the targets or victims of doxing, as well as of the perpetrators, and it must involve the social and economic power relationships that constitute and distort the expressive environment.

I. The Politics of Fear

A place to start is with perhaps the most prominent political theorist of fear, Judith Shklar. Reading her later work today is bracing—it elicits a jolt of recognition. At the most basic level, Shklar seeks to organize a conception of liberalism around a summum malum, namely “cruelty and the fear it inspires, and the very fear of fear itself.”¹¹ Cruelty of the sort that concerns her is systematic; it is not haphazard but instead it entails the “deliberate infliction of physical, and secondarily emotional, pain” as an exercise of power by those in positions of strength against those in positions of weakness.¹²

Terror and acute anxiety are bad, supremely bad, partly because they interfere with freedom, understood as the ability to direct one’s own life. Here is a key passage: “Every adult should be able to make as many effective decisions without fear or favor about as many aspects of her or his life as is compatible with the like freedom of every other adult. That belief is the original and only defensible meaning of liberalism.”¹³ She explains that liberalism aims to lift the burden of political anxiety from the shoulders of adults, who then can order their lives according to their own beliefs, wants, and needs.¹⁴

Shklar’s chain of reasoning—from basic security, to the fear of its violation, to the fear of fear itself, and then to the connection of security to the exercise of basic freedom—resonates. She connects power differentials and their abuse to human emotion, and she connects psychological security to the ability of individuals to function as citizens in a political community.¹⁵ A commitment to personal security is political insofar as it concerns a necessary condition of the community—of the project of democratic cooperation.¹⁶

For theorists of free speech, it is a short step from Shklar’s liberalism to the realization that speakers cannot be free if they are fearful of physical violence or of power exercised against them in ways that threaten their safety. Is this silencing systematic? Shklar is not simply concerned with insecurity, again, but with insecurity that issues from power differentials and is patterned. Silencing that results from fear today could possibly be considered systematic, in a sense. Or you could say that some silencing is systematic in this way. In a radically polarized political climate, the content and viewpoint that risks retribution is foreseeable—and it is precisely this predictability that creates the conditions for censorship. If speakers did not know what positions or politics would endanger themselves or their families, they would not be able to avoid them. But because such viewpoints are foreseeable, and to the extent they are, they can be silenced.

Shklar is focused on public power, which she identifies with government action.¹⁷ And her sharp distinction between public and private power is recognizably liberal.¹⁸ To some degree, Shklar’s work nevertheless remains relevant here, even strictly construed, because it applies to certain violations of freedom of expression that are instigated by government actors. In this pattern, a public official identifies a political enemy with the knowledge and expectation that followers will harass and intimidate the targeted person, terrifying them into submission and silence. So, although the proximate harm is caused by private persons, the coordination and instigation come from politicians.

Yet taking a broader view, Shklar’s focus on state action neglects instances in which private actors threaten potential speakers without any apparent or actual coordination by government figures. Because the effect is often the same, and because the topics involved may well be matters of public concern, her neglect of nonpublic exercises of power to systematically silence people limits the usefulness of her insights in today’s speech environment.

It is true that Shklar acknowledges the relevance of some social and institutional conditions for the exercise of individual freedom. For example, she emphasizes some differences between her theory and Isaiah Berlin’s negative liberty, which is otherwise quite similar. She sees the importance of protecting not just negative liberty as such, but also the conditions that make its exercise possible, and she realizes that in this way negative liberty is a necessary but not sufficient condition for personal freedom.¹⁹ “No door is open in a political order in which public and private intimidation prevail,” she says, and therefore it is important to identify the “institutional characteristics of a relatively free regime,” including mechanisms for the dispersal of power—social as well as strictly political power.²⁰

In a prominent critique, Sam Moyn portrays the liberalism of fear as an instance of what he calls Cold War liberalism. Chastened and traumatized by the wars and totalitarianisms of the twentieth century, Shklar pessimistically “dropped any radical expectations of improvement” and retreated to defending minimal pluralism against the persistent threat of violence.²¹ Abandoning the aspirations of her earlier work, Shklar adopted a “ ‘survivalist’ approach to political theory,” one resigned to hope only for “damage control.”²² This form of liberalism offered few resources to resist the rise of libertarianism and neoliberalism, though it was distinct from both.²³

We need not accept the limits of the liberalism of fear. To the degree that Shklar herself is focused solely on precarity caused by government policing, we can expand her insights and apply them to nongovernmental sources of insecurity. Political and constitutional theory can assimilate the insight that freedom of speech, like other basic liberties, cannot be merely formal but must be real for the actual human beings living in historically specific social situations.²⁴ It is essential for the meaningful exercise of freedom of speech for certain essential social and economic conditions to obtain. Some of these conditions are egalitarian, and some are sufficientarian, as argued in other work.²⁵ Here, the specific point is that speakers do not have a meaningful ability to express themselves freely if they are subject to fundamental physical and psychological insecurity. This is not just the healthy fear of avoidable pain, as Shklar emphasizes, but the systematic circumstance of political polarization and power exertion that predictably suppresses particular viewpoints.

The First Amendment of fear qualifies as a political conception because it attends to social and economic power and locates solutions, ultimately if not exclusively, in the state. Although the closest causes of anxiety today often are other citizens, leveraging digital media and other technologies of terror, they are unlikely to be stopped solely by private means. Government has an obligation to ensure people’s security, and their sense of security, as a condition of meaningful political participation as cogovernors in a democracy. Whether that obligation itself has constitutional force is an interesting but different question, and regardless government ought to be constitutionally permitted to pursue the structural conditions for real

people’s meaningful exercise of the fundamental right to freedom of expression.

Is the political commitment to freedom from fear possible to implement in law, given existing First Amendment doctrine? There is some doubt. Part II explores that question in the context of state statutes criminalizing or otherwise regulating doxing.

II. Doxing and Silencing

In the little space that is left in this Essay, let’s consider attempts by law to guard against a particular cause of fear, namely doxing. These attempts face serious challenges under the First Amendment, since doxing typically constitutes speech that does not necessarily or obviously fall within an existing category of unprotected expression, and since it is regulated based on its content.²⁶ From the perspective of the First Amendment of fear, this legal circumstance could be seen as a matter of regret, insofar as doxing itself can have a powerful silencing effect on those it targets.²⁷ There may be no way to assimilate that insight into the existing structure of free speech doctrine. Yet this Part cautiously explores one possible pathway.

Doxing can be understood in several ways. The Oxford English Dictionary defines it as “[t]he action or process of searching for and publishing private or identifying information about a particular individual on the internet, typically with malicious intent.”²⁸ At root, and colloquially, the practice involves publishing identifying information about someone in order to facilitate harm of that person by third parties. Yet not every element of this understanding is essential; arguably, the trucks that circled campuses displaying the names and images of students constituted doxing even though they did not involve the internet.²⁹ And additional elements may be necessary, such as a particular level of mens rea, or a requirement that the information not already be publicly available, or that the target is not a public official.

California has an influential criminal statute that includes a specification of what constitutes doxing. It prohibits “electronically distribut[ing]” through various means “personal identifying information” of another person “with intent to place another person in reasonable fear for his or her safety, or the safety of the other person’s immediate family” and “for the purpose of imminently causing that other person unwanted physical contact, injury, or harassment . . . by a third party.”³⁰ Among the exceptions is the distribution of information with the target’s consent.³¹

Though the California criminal statute does not use the term doxing, it has been understood to regulate that activity. A newer state law provides a civil cause of action for “doxing,” which it defines using much the same language as the criminal provision.³² Legislative history shows that state lawmakers intended to provide a civil cause of action for doxing, referencing the criminal statute.³³ Also, Stanford University’s “Anti-Doxxing Policy” appears to be modeled on the California statutes and uses substantially the same definition of the prohibited activity.³⁴

Notable here is California’s use of the term “fear” to indicate the harm that it seeks to protect against. Apparently, the state believes that disabling anxiety on the part of victims is serious enough to warrant a criminal prohibition. And the statute recognizes that the electronic distribution of personal information has the power to generate a specific kind of harm, and to an extraordinary degree. Yet the statute also limits itself to fear of “physical contact, injury, or harassment,” not just anticipation of political criticism or even social ostracism, without more. Though there is considerable variation among state doxing statutes on this and other questions, the California approach is among the most straightforward.³⁵ Also notable is that the California statute does not exempt news reporting, perhaps because lawmakers reasoned that the intent requirement would not be satisfied and therefore an explicit exemption was unnecessary.

Could the California statute withstand a First Amendment challenge?³⁶ The difficulty, of course, is that the distribution of personal identifying information could itself be seen to be expression, or expressive conduct. And if the Speech Clause is implicated, then plausibly it requires strict scrutiny of the California statute, which regulates on the basis of content.³⁷ After all, the statute only prohibits a specific kind of speech, namely the distribution of certain identifying information using a particular medium. And whether the statute is narrowly tailored to a compelling government interest is unclear.³⁸

Under one theory, the California statute might be constitutional because it regulates a type of threat. On this approach, the regulated content would fall within a traditional category of unprotected speech. Compare California’s threat statute. It criminalizes threatening another person with “death or great bodily injury” with the specific intent that the statement be taken as a threat and under circumstances that convey a specific and immediate danger so that the target “reasonably . . . [is] in sustained fear for his or her own safety or for his or her immediate family’s safety.”³⁹ Threat statutes do typically guard against a particular kind of “fear.”⁴⁰ Conceivably, California’s doxing statute regulates a subset of threat—a specific kind of statement, made through a particular medium, that causes the target to feel insecure, particularly with regard to safety. Both statutes have an intent requirement and they both apply only to reasonable fears and imminent dangers. Neither requires the violence to eventuate because both recognize that the fear itself is harmful.

To be sure, there are differences that may be important. Crucially, the doxing statute shields against statements that risk harm not by the speaker, but by a third party. For this reason, it could be reasonably argued that incitement is the category of unprotected speech that is more closely analogous to doxing.⁴¹ But incitement does not centrally involve fear, which seems essential to the harm of doxing.

If it is correct that doxing can count as a type of threat, then its prohibition could be seen as having a kind of derivative constitutionality. Threats constitute a category of unprotected speech, under established Supreme Court doctrine.⁴² If California has criminalized a species of threat, then its doxing statute could survive under that First Amendment doctrine. Importantly, the state would have to require at least a recklessness level of mens rea to avoid chilling protected activity.⁴³ But because the doxing statute requires an “intent” and a “purpose,” it could be construed to clear that bar.⁴⁴

A possible doctrinal objection is that the Supreme Court protected an early form of doxing in NAACP v. Claiborne Hardware.⁴⁵ That decision is mainly known for its holding that a civil rights boycott of white-owned businesses was constitutionally protected.⁴⁶ Less well known is the Court’s holding that recording the names of customers who violated the boycott, reading those names at meetings, and publishing the names in a newspaper could not be punished.⁴⁷ If boycott enforcers from the NAACP intended to threaten these customers by causing them to fear that they would be identified to third parties who would harm them, then their activity of recording names looks like doxing. And the Court held that the organizers “admittedly sought to persuade others to join the boycott through social pressure and the ‘threat’ of social ostracism” but that “[s]peech does not lose its protected character . . . simply because it may embarrass others or coerce them into action.”⁴⁸ Violence did occur in the course of the boycott, but still, the Court held that publishing names of boycott violators could not be punished. This could be understood as precedent for protecting doxing.

Yet the Court also reaffirmed that both violence and the threat of violence are unprotected and could be punished.⁴⁹ What the First Amendment requires is precision, not protection of threatening speech.

When [violence or a threat] occurs in the context of constitutionally protected activity . . . ‘precision of regulation’ is demanded. Specifically, the presence of activity protected by the First Amendment imposes restraints on the grounds that may give rise to damages liability and on the persons who may be held accountable for those damages.⁵⁰

In the context of doxing, this may provide good reason to require an elevated level of mens rea.⁵¹ The lesson may be that while speech that identifies perceived wrongdoers cannot be punished, particularly when it addresses matters of public concern, speech that identifies individuals for the specific purpose of eliciting violence can be prohibited because criminalization with that elevated level of intent constitutes “precision of regulation.” Whether the “doxing truck” at Columbia could be criminalized under this approach would depend, in part, on whether it was operated with the requisite level of intent.

Another possible objection is that punishing only doxing that is specifically intended to elicit fear of violence, and that qualifies as a type of threat, does not match people’s common understanding of doxing. On this objection, it is undesirable for there to be a mismatch between the social meaning of a term like doxing and the legal prohibition that seeks to address the harm. That is reasonable. If the truck at Columbia were not motivated by the requisite intent, such that it did not constitute doxing as a legal matter, some would view that as a fault of the statute. One response is that not protecting against doxing at all, or doing so only through existing statutes designed for other purposes, also fails to match people’s reasonable expectation that the law should address serious harms. Another response is that many people seemingly do think that doxing involves “malicious intent” in the words of the Oxford definition, if not necessarily intent relative to violence.⁵² So maybe the mismatch is not so egregious.

Another mismatch is that the proposal here only prohibits doxing that creates a fear of violence, not also harassment.⁵³ That is because it is unclear whether a statute that guarded against fear of harassment, without more, would fall within the category of unprotected speech for true threats.⁵⁴ So here too, there is a potential mismatch between the proposal and colloquial understandings of doxing—an awkwardness that may simply be a cost of fitting this particular protection against disabling fear into the existing constitutional doctrine.

This proposal would bring the regulation of doxing within the unprotected category of threatening speech. Still, it is unfortunate that First Amendment law is being understood to require strict scrutiny of all speech regulations that fall outside a recognized category of unprotected speech, such as threats. As Genevieve Lakier and Evelyne Douek have argued, the Supreme Court has moved away from its traditional practice of evaluating speech regulations with respect to First Amendment values and competing considerations.⁵⁵ In this context, as noted, doxing regulation could serve important free speech values, especially by protecting victims against the sort of disabling hesitation that effectively silences them. So it could be said that in at least some cases there are expressive interests on both sides of a statute like California’s. A full consideration of values would take that symmetry into account.

Yet, for now, unless and until there is a meaningful change in the ideology or composition of the Roberts Court, the binary approach to speech doctrine must be taken as a fixed feature of constitutional law. And under that approach, a doxing statute like this one can best survive if it is understood to regulate a subset of threatening speech.

Conclusion

From the perspective of the First Amendment of fear, government ought to be at least permitted, if not required, to safeguard the conditions under which people have a meaningful and not just formal ability to participate in democratic discourse and otherwise express themselves freely. One obstacle to that freedom is the systematic apprehension of speaking on certain topics. Unfreedom of this kind is worth protecting against. Government has the ability and the responsibility to ensure expressive security—not freedom from fear of criticism or rebuke, but freedom from systematic fear of violence, at the very least. Arguably, this kind of safety is essential to expression. Doxing legislation may be one example of government protection that, despite facing constitutional hurdles, can be crafted so that it does more to promote than to frustrate First Amendment imperatives.

98 S. Cal. L. Rev. 1413

Download

Nelson Tebbe*

* Jane M.G. Foster Professor of Law, Cornell Law School. Thanks to Evelyn Douek, Eugene Volokh, participants in the Southern California Law Review symposium, and the staff of the law review for comments and suggestions. The author served as an observer for a committee of the Uniform Law Commission on doxing, but the arguments here do not reflect the views of the committee.

Protecting Listeners From Unwanted One-to-One Speech

September 2025September 2025Eugene Volokh*

I. The Value of the One-to-One vs. One-to-Many Line

“[N]o one has a right to press even ‘good’ ideas on an unwilling recipient,” the Supreme Court held in Rowan v. United States Post Office Department.¹ At the same time, “[t]he fact that society may find speech offensive is not a sufficient reason for suppressing it. Indeed, if it is the speaker’s opinion that gives offense, that consequence is a reason for according it constitutional protection.”² That is generally true even if the speaker’s opinion gives offense not just to “society” but to many of the speaker’s listeners.³

The best way to reconcile these principles, it seems to me, is to distinguish (1) one-to-one speech said to an unwilling listener from (2) one-to-many speech that reaches both potentially willing and unwilling listeners.⁴ Ashutosh Bhagwat well explains both the precedents and the policy arguments supporting the distinction. Most speech should generally be protected because it may persuade or inform some potentially willing listeners even if others are upset.⁵ But speech said solely to an unwilling listener, where it’s clear the listener is unwilling, is likely only to offend. The government can in many situations help protect listeners against such one-to-one speech, because that promotes the unwilling listener’s autonomy without interfering with communication to potentially willing listeners.⁶

And this helps explain the constitutionality of many common speech restrictions, including:

telephone harassment laws,⁷
do-not-call registries,⁸
harassment restraining orders that forbid speech to the protected person,⁹
application of university “hostile environment harassment” policies to people “following students around and yelling slurs or otherwise directing hostile speech at individual students who have demanded to be left alone,”¹⁰
application of workplace harassment law to one-to-one insults, or one-to-one repeated unwanted romantic advances,¹¹
residential picketing laws,¹² and more.

II. Must Restrictions on Unwanted One-to-One Speech Be Content-Neutral?

This general conclusion, however, raises subsidiary questions. A particularly important one is whether restrictions on one-to-one speech must be content-neutral.

There is precedent suggesting this, as well as broader First Amendment principles supporting such a view. Frisby v. Schultz upheld a content-neutral residential picketing ban on the grounds that such picketing is essentially speech targeted to the unwilling listener in the home.¹³ But Carey v. Brown had earlier struck down a residential picketing ban that excluded labor picketing because that exclusion made the law content-based.¹⁴ It was the content neutrality of the ban in Frisby that saved it.¹⁵

We see something similar in Rowan v. United States Post Office Department.¹⁶ Rowan upheld a statute that barred senders from sending material to householders, once the householder informed the post office that he “in his sole discretion believes [the mailings] to be erotically arousing or sexually provocative.”¹⁷ The statute was thus content-based on its face, but the Court stressed it was essentially content-neutral as enforced:

Both the absoluteness of the citizen’s right under [the statute] and its finality are essential; what may not be provocative to one person may well be to another. In operative effect the power of the householder under the statute is unlimited; he may prohibit the mailing of a dry goods catalog because he objects to the contents—or indeed the text of the language touting the merchandise. Congress provided this sweeping power not only to protect privacy but to avoid possible constitutional questions that might arise from vesting the power to make any discretionary evaluation of the material in a governmental official.¹⁸

Yet if content neutrality is indeed required in such situations, then many restrictions on one-to-one speech would be hard to defend. Telephone harassment laws, for instance, often specially target lewd or indecent harassing calls.¹⁹ Workplace harassment law ends up specially targeting one-to-one speech that is personally insulting.

Likewise, when various laws target one-to-one speech intended to “harass” or “abuse,” they must be treated as content-based. As the Court held in Reed v. Town of Gilbert, “[s]ome facial distinctions based on a message are obvious, defining regulated speech by particular subject matter, and others are more subtle, defining regulated speech by its function or purpose. Both are distinctions drawn based on the message a speaker conveys, and, therefore, are subject to strict scrutiny.”²⁰ When the “regulated speech” is defined by a purpose to harass or abuse, that definition generally targets speech that has a harassing or abusive “message.” The definition is therefore content-based.

More broadly, when even a “generally applicable law” is “directed at [a speaker] because of what his speech communicated”—when the speaker violates the law “because of the offensive content of his particular message”—that too is treated as a “content-based regulation of speech.”²¹ This would cover most harassment laws, at least when speech is found to be harassing because of its offensiveness rather than because it’s too loud or ties up telephone lines.

Indeed, relatively few of these laws actually set up Rowan-like rules that (1) require the listener to first tell a speaker, “stop speaking to me,” but then (2) make that order binding regardless of what the speaker wants to say. The laws are indeed aimed at “address[ing] the ‘first blow’ of curse words spoken only once.”²² At the same time, they aim to avoid giving someone an absolute veto on future communications: consider, for instance, workplace harassment, where the law can’t let employees categorically forbid any future communications (including on legitimate work-related topics) by coworkers.

Now perhaps that’s the wrong approach—perhaps the law should indeed insist on content neutrality even as to restrictions on unwanted one-to-one speech. Or perhaps content-based restrictions should indeed be subjected to strict scrutiny but might in some situations be upheld.

But I think it might be better to recognize that at least some such content-based restrictions are permissible when it comes to one-to-one speech, even if they wouldn’t be permissible as to one-to-many speech. The Court has acknowledged that content-based restrictions may be constitutional when “substantial privacy interests are being invaded in an essentially intolerable manner.”²³ Perhaps the “privacy interests” here should be read as not just focusing on privacy in the home, or true captivity of a sort where it is “impractical for the unwilling viewer or auditor to avoid exposure.”²⁴ Rather, perhaps they should also be seen as including intrusions on the listener’s autonomy rights that Professor Bhagwat rightly identifies: the targeting of a particular likely unwilling listener for one-to-one speech may be what is “essentially intolerable.”

R.A.V. v. City of St. Paul²⁵ may provide a helpful framework for dealing with this. The Court in R.A.V. held that content-based restrictions must generally be subject to strict scrutiny even when they are limited to subsets of unprotected categories of speech. For instance, a ban on racist fighting words would be presumptively unconstitutional even though a ban on all fighting words would be valid.²⁶ But the Court also held that this principle has certain exceptions, again where the content discrimination is entirely within an unprotected category; the relevant exceptions are:

“[w]hen the basis for the content discrimination consists entirely of the very reason the entire class of speech at issue is proscribable,”²⁷ for instance when the law restricts “only that obscenity which is the most patently offensive in its prurience,” or “only those threats” that are especially disruptive;²⁸
when “a particular content-based subcategory of a proscribable class of speech” is “swept up incidentally within the reach of a statute directed at conduct rather than speech”;²⁹ and
when “the nature of the content discrimination is such that there is no realistic possibility that official suppression of ideas is afoot.”³⁰

The same might apply with regard to subcategories of likely unwanted one-to-one speech, if Professor Bhagwat and I are right that such speech is essentially constitutionally unprotected. Indecent harassing phone calls, for instance, may well be especially likely to be unwanted, and a restriction on such calls may indeed be unlikely to involve “official suppression of ideas.”

Likewise, a prohibition of one-to-one speech intended to abuse or harass might be justified on the same theory, and might also be “swept up incidentally within the reach of a statute directed at conduct rather than speech,” given that such harassment laws often do target nonspeech conduct (such as physical stalking) as well as speech. R.A.V. itself gave hostile environment harassment law as an example of a law that may “incidentally” “swe[ep] up” “sexually derogatory ‘fighting words,’ among other words,” because it bans a wide range of conduct as well as speech.³¹ Likewise, the law may incidentally sweep up derogatory unwanted one-to-one speech more broadly. (For reasons I explain elsewhere, this rationale does not extend to offensive one-to-many ideological expression, even when it’s viewed as sexist, racist, and the like.³²)

III. When Must the Government Tolerate One-to-One Speech to Government Officials?

Though one-to-one speech to unwilling listeners may generally be forbidden, the analysis must be different when the speech is addressed to government employees on the job, especially public-facing employees. I agree with Professor Bhagwat that listeners generally have considerable autonomy interests in not hearing unwanted speech—interests that the government may protect. But when one works for the public,³³ one must accept the risk of disapproving speech from the public:

[R]eceiving mail from disgruntled constituents is usual for a politician. A person “who decides to seek governmental office must accept certain necessary consequences of that involvement in public affairs . . . [and] runs the risk of closer public scrutiny than might otherwise be the case.” Here, given Michael’s status as a selectman and the content of the letters, it cannot be said that Michael’s “substantial privacy interests [were] invaded in an essentially intolerable manner.”³⁴

This is particularly clear for elected officials,³⁵ candidates for office,³⁶ or high-level political appointees.³⁷ But it may be true for lower-level public-facing employees as well, such as police officers³⁸ or others.³⁹ Some cases do allow punishing offensive speech to such employees,⁴⁰ but I think they’re mistaken.⁴¹

IV. The Borders of “One-to-One”

Finally, “one-to-one” and “one-to-many,” like many such useful general phrases, may not fully capture the legal principles that courts should and do apply. To give one example, say someone is speaking simultaneously to three listeners, all of whom have asked the speaker to stop bothering them. That’s technically one-to-three speech, not one-to-one speech. But it should be restrictable as tantamount to one-to-one speech, precisely because it is addressed solely at unwilling listeners.

Likewise, say Wendy Smith’s ex-husband Harry Smith posts a Facebook message on his own page saying, “My ex @WendySmith is a slimy trollop.” (This @ syntax is specifically designed to notify the Facebook user WendySmith about the post; Twitter and Instagram have the same feature.) It is thus more or less like an e-mail to Wendy (one-to-one speech), coupled with a post about her to the author’s friends (one-to-many speech). If Wendy gets a harassment restraining order barring further correspondence from Harry, it would be constitutionally permissible for that order to be interpreted as banning such mentions; Harry would still be able to communicate with his friends by posting the same item without the @ (“My ex Wendy Smith is a slimy trollop”).⁴²

The hardest question arises when speech appears to be largely aimed at a particular unwilling listener but also reaches some other listeners. This is what the Court faced in Frisby v. Schultz, where it reasoned that residential “picketing is narrowly directed at the household, not the public”:

The type of picketers banned by the Brookfield ordinance generally do not seek to disseminate a message to the general public, but to intrude upon the targeted resident, and to do so in an especially offensive way. Moreover, even if some such picketers have a broader communicative purpose, their activity nonetheless inherently and offensively intrudes on residential privacy. . . .

Because the picketing prohibited by the Brookfield ordinance is speech directed primarily at those who are presumptively unwilling to receive it, the State has a substantial and justifiable interest in banning it.⁴³

Here the speech wasn’t “foisted (exclusively) upon unwilling listeners”⁴⁴—presumably at least some residential picketers also want to reach the resident’s neighbors.⁴⁵ Rather, the Court says the speech was targeted “primarily” at the resident and acknowledges that it might have also had “a broader communicative purpose.”

Distinguishing the “primary” audience from the “secondary” is of course subjective, plus it’s not clear why even secondary audiences should be ignored. For instance, if animal rights protesters are picketing outside a fur store, is their speech “directed primarily” at buyers, who are likely “unwilling to receive” the message (especially if the message is framed harshly)? After all, fur buyers presumably know well where the fur comes from—and like it. Or is the speech directed at least equally to neighbors and passersby, or to the likely relatively rare ambivalent customer?

Likewise, most people who go to abortion clinics are likely unwilling to hear from anti-abortion protesters and counselors, but some might be open to their arguments.⁴⁶ Most people who go to churches, synagogues, or mosques that are being picketed are unwilling to hear from protesters,⁴⁷ but again some might be persuadable.

I’m not sure how this line is to be properly drawn. Perhaps courts should view Frisby as limited to “residential privacy,” given its reliance on the precedents saying that, “[a]lthough in many locations, we expect individuals simply to avoid speech they do not want to hear, the home is different.”⁴⁸ On the other hand, there will always be arguments for extending this sort of extra protection beyond the home to medical facilities,⁴⁹ funeral homes,⁵⁰ high schools,⁵¹ places of worship,⁵² and more. Here, I just want to acknowledge the difficulty that this issue raises.

Conclusion

The one-to-one/one-to-many distinction is critical to understanding how and when unwilling listeners may be protected. I hope this short article has helpfully elaborated on a few questions the distinction raises.

98 S. Cal. L. Rev. 1427

Download

Eugene Volokh*

* Thomas M. Siebel Senior Fellow, Hoover Institution (Stanford); Gary T. Schwartz Distinguished Professor of Law Emeritus, UCLA.

Vol. 98 Editor’s Note

September 2025September 2025Southern California Law Review Admin

EDITOR’S NOTE

The articles in Volume 98, Number 5 of the Southern California Law Review were delivered as papers at a symposium on “The First Amendment and Listener Interests” in November 2024. The symposium was organized in coordination with Professors Erin Miller, Rebecca Brown, and Abby Wood, and with support from USC Gould School of Law.

Free speech rights affect the interests not just of speakers but also of listeners of speech. Listeners have interests as autonomous agents, thinkers, and voters, all of which are served by the information, ideas, reasons, and stimulus to thinking that listening to speech can provide. For this reason, many prominent free speech theories—and especially democratic theories—foreground listener interests.

Yet federal courts in constitutional free speech cases tend to focus on and privilege speakers, those who most clearly hold and exercise free speech rights, over listeners. Indeed, recent courts have increasingly turned away from earlier cases—involving, for instance, campaign finance, media regulation, and corporate and commercial speech—that explicitly recognized listener interests. Now, some First Amendment scholars are beginning to push back and suggest that this neglect of listener interests disserves both the doctrine and free speech theory.

Listener interests are at the heart of many high-profile legal questions, including ones involving commercial speech and telecommunications regulation. As the Supreme Court expands the number of entities that count as speakers, they may well set aside negative impacts on listeners. In Moody v. Netchoice, LLC, for example, the Court recently determined that social media platforms have free speech rights to “editorial discretion.” Such rights could, at least in theory, come into conflict with listeners’ access to information and discretion over what they consume.

Volume 98 of the Southern California Law Review is pleased to present this special symposium edition. This issue features a group of the nation’s leading scholars in First Amendment law providing various perspectives on contemporary issues in free speech rights from the perspective of listeners.

Editor-in-Chief & Managing Editor, Volume 98

Volume and Issue: Vol. 98, No. 5

Pluralism and Listeners’ Choices Online

98 S. Cal. L. Rev. 1387

Unchosen Listening

INTRODUCTION

I. The Case for unchosen listening

A. Optimizing Preferences

B. Avoiding Error

C. Innovation

D. Self-Evolution

E. Tolerance and Persuadability

II. Limitations and Implications

98 S. Cal. L. Rev. 1399

The First Amendment of Fear

Introduction

I. The Politics of Fear

II. Doxing and Silencing

Conclusion

98 S. Cal. L. Rev. 1413

Protecting Listeners From Unwanted One-to-One Speech

I. The Value of the One-to-One vs. One-to-Many Line

II. Must Restrictions on Unwanted One-to-One Speech Be Content-Neutral?

III. When Must the Government Tolerate One-to-One Speech to Government Officials?

IV. The Borders of “One-to-One”

Conclusion

98 S. Cal. L. Rev. 1427

Vol. 98 Editor’s Note

EDITOR’S NOTE

98 S. Cal. L. Rev. 761

Table of Contents