A Libertarian Vision for Digital Expression

A generation ago, the American publishing environment—and, consequently, the expressive capacity of most Americans—looked very different. Before the widespread adoption of the internet, the press looked less like a set of universally accessible technologies and more like a privileged class. For most Americans, the opportunity to speak to a wide audience was limited. Media publication and distribution tools were still quite expensive or labor intensive. Listeners had limited means of searching for speakers they wished to hear. Sure, you could pump out mimeographed zines in your basement and pass them out on street corners (and many did), but technological limits placed practical restrictions on who could speak and how far their speech could travel. In turn, these natural limits created opportunities for state action that served to further fetter speech.

The advent of radio and television in the first half of the 20th century brought a wider array of speakers into American homes, but both new media forms required expensive infrastructure and the exclusive use of certain broadcast frequencies. This centralization empowered gatekeepers—from Walter Cronkite to your local newspaper editor—to decide which perspectives received wide circulation.

The internet changed all of that, throwing open the floodgates to a deluge of speech. The early internet contained few gatekeepers, although it required a fair amount of specialized knowledge to host and operate your own website. Forums, message boards, chat services, and eventually social media emerged to host a diverse cacophony of voices, allowing everyday Americans to communicate in one-​to-​one, one-​to-​many, and, for the first time, many-​to-​many formats, instantaneously and at a low price.

This internet speech ecosystem came with its own set of new gatekeepers. Social media platforms enlisted content moderators to keep their digital forums consumer friendly; speech middlemen deeper in the internet stack—such as Cloudflare, a content delivery network that sits between Web users and the servers of websites they visit—realized that they possessed incredible power over who could maintain a presence on the Web. ¹

With the doors (mostly) thrown open to speech, internet payment processing became a new locus of gatekeeping. Although it might have become harder to prevent given speakers from publishing their thoughts, if it could be made difficult for their fans and supporters to fund their expressive activities, the amount of time they could realistically devote to speaking would be limited. This factor has increasingly challenged speakers who toe the legal limits of free expression.

These new chokepoints—regardless of whether they are “architected” or “natural”—may be inclined to limit expression and can be influenced, or even captured, by illiberal advocacy organizations and foreign states. Some of these constraints are acceptable and unavoidable. Not every node in every network must host every sort of speech. However, understood as an ecosystem, an internet that is friendly to free expression ought to provide some means of allowing consenting speakers and listeners to interact with one another, ideally without great expense or specialized knowledge.

Predicting the future of speech is difficult. Unlike many other policy areas, freedom of expression vis-​à-​vis the state does not guarantee speakers an audience, or, more importantly, access to the communicative tools necessary to reach receptive listeners. Much of what we think of when we envisage a regime of expressive freedom relies on a tolerant, liberal culture willing to give dissenting voices a hearing. That being said, the technological innovations discussed in this chapter will help individuals to more easily express themselves and will limit the effective impact of both illiberal cultural turns and state exploitation of chokepoints in systems otherwise legally supportive of free speech.

The following sections examine how state censorship laundered through private platforms and the erosion of intermediary liability protections will spur the decentralization of internet forums. In turn, this development will recenter speech freedom disputes on the individual speaker while exacerbating the internet-​borne challenge to authority.

As a speech delivery mechanism, the internet is far more complex than its predecessors. Whenever you speak over the internet, whether publishing state secrets or sending a photo of your cat, you use the services of countless intermediaries that you share with billions of other speakers. From the physical device you use to tap out your message, and the wires or antennas through which your speech travels, to the multitude of programs and protocols that govern interactions between various layers of the internet, hundreds of hands help your speech reach its intended audience. Compared with the process of publishing a handbill, internet speech requires the participation of a wider array of intermediaries. Whereas these interreliant middlemen mostly serve to increase the range and velocity of speech, many have the capability to act as gatekeepers, refusing disfavored speakers the use of their services.

Not all intermediaries act as gatekeepers; generally, the more important the intermediary, the less gatekeeping it does. Physical infrastructure providers like AT&T or the providers of low stack functions like internet protocol address assignation rarely engage in gatekeeping. Furthermore, not all gatekeeping limits the public’s publishing capacity, and some gatekeeping functions, such as spam filtering, are universally appreciated and infrequently implicated in freedom-​of-​speech concerns.

Nevertheless, because internet speech depends on this battery of middlemen, gatekeeping authority can be used to suppress speech. Sometimes, this is done for purely private reasons. Forums dedicated to specific functions exclude off-​topic submissions. Advertising-​funded video hosting platforms like YouTube exclude content that concerns advertisers. From Dove soap to Hasbro, no one likes their adverts to run in front of ISIS videos. Community formation and maintenance will always require some gatekeeping. Private speakers can usually find alternatives to social media platforms that have rule sets they find too restrictive, although those substitutes may not provide their speech with as wide an audience as larger platforms.

State interference with this ecosystem of speech facilitators is particularly concerning. Because states may coerce intermediaries to refrain from carrying disfavored speech, their prohibitions are totalizing and offer few avenues for exit. A few private platforms may gravitate toward common rule sets or embrace cartel-​like cross-​platform speech compacts; however, without a means of preventing defection, the power to completely exclude evades these arrangements.

Government regulation of internet speech takes many forms. Formally, states may prohibit the publication of certain words or phrases, targeting either the speaker, intermediaries, or both. They may increase the cost of participation in this ecosystem by imposing costly regulations on intermediaries or by taxing social media users directly. ²

Increasingly, Western governments have preferred tinkering with intermediary liability—the extent to which platforms are held legally responsible for their users’ speech—to regulate speech indirectly. Germany’s NetzDG (Network Enforcement Act) and the United States’ FOSTA (Fight Online Sex Trafficking Act) render platforms liable for certain forms of user-​generated content, requiring private firms to make legal judgments on the fly. ³ Regulation can also tip the scales in favor of takedowns by shortening private content moderation windows or by mandating preemptive filtering rather than post hoc complaint and review procedures.

Less explicit regulation is accomplished through “jawboning,” when governments alter private behavior through threats of regulation or legal sanction. Jawboning efforts may exhort current gatekeepers to shift the focus of their efforts, like the European Commission’s hate speech Code of Conduct, a voluntary agreement presented as an alternative to regulation that privileges the removal of hate speech. The Code of Conduct not only requires platforms to prioritize the removal of hate speech over other goals, it also standardizes their definitions of “hate speech.” ⁴

Definitional uniformity prevents the speech ecosystem from finding even a seedy outlet for controversial or disfavored speakers. Given that most platforms are international, a definitional concession in one part of the world may affect the speech capacity of internet users worldwide. As platforms race to engage network effects by connecting ever-​greater numbers of people, they risk combining the particular prohibitions of many different states and societies into a regulatory whole that is even more restrictive than the sum of its parts. As the second half of the globe’s population comes online, the diversity of viewpoints and local taboos will increase, at least at first. In the long term, speech norms are likely to converge somewhat, though the maturation of translation software will exacerbate tensions in intercultural moderation. ⁵

Jawboning may also impel intermediaries that have thus far refrained from exercising gatekeeping authority to begin doing so. Financial services, particularly payment processing, are increasingly being dragooned into policing speech. Beyond their use by state censors, payment processing chokepoints also serve as beacons for cultural conflict. These chokepoints provide a means of effectively politicizing commercial relationships made legible and public by their presence on the Web. Digital payments give speakers access to a far wider set of potential supporters, but the relationships between specific speakers and the services upon which they rely are made explicit by payment plug-​ins, searchable fundraising platforms, and metadata.

Other commercial aspects of the speech ecosystem—from the provision of domain names to cloud storage—are similarly legible, though historically less politicized. And both state actors and private activists are likely to become more sophisticated in their selection of pressure points as time goes on. ⁶ Internet speech is, for the moment, reliant on this chokepoint-​ridden ecosystem. Current regulation, formal and informal, public and private, uses these chokepoints to limit the distribution of disfavored speech. As such, most near-​future speech conflicts will occur under a regime of confusingly muddled public and private governance. In the long run and in an increasingly global speech ecosystem, expansive expressive capacity will only be secured through decentralization.

On one front, speech-​enhancing technologies continue to advance within the platform paradigm, offering users novel ways to communicate with one another. VRChat offers new forms of gesture-​based speech, the benefits of which have already been seen. ⁷ However, as long as these applications are centrally distributed and administered via the Oculus or SteamVR stores, they can be bound by regulation and jawboning just as older platforms have been.

Free, encrypted, cross-​platform messaging apps allow groups of people to communicate privately, establishing their own standards for group admission and acceptable speech. This development allows individuals to exit public platforms, avoiding their vulnerability to government jawboning and market-​driven moderation demands in exchange for fewer opportunities for discovery. Within such closed chat networks, it is harder for speakers and listeners to find one another. As well, though these networks are encrypted to varying degrees, they still run on centralized infrastructure. They are operated by governors incapable of restraining themselves and therefore are ultimately incapable of resisting pressure to “do something” about disfavored speech.

Eschewing reliance on market forces or protective legal regimes, a structurally decentralized internet would resist state regulation and the ever-​shifting tides of public opinion by eliminating chokepoints that allow single actors to limit a speaker’s use of the system as a whole. From the physical antennas and receivers to the provision of microblogging services, an idealized decentralized internet would replace capture-​subject intermediaries with open and secure protocols that are accessible to everyone.

Rather than being the property of a single institution, protocols need not be owned by anyone and may be used by any two parties willing to adopt the same standards. In practice, nonprofits make most decisions concerning the management of widely used protocols. Existing protocols have largely resisted state jawboning because, unlike edge publishing platforms, they are viewed as common infrastructure.

Much of the modern internet is built on top of open-​source protocols; HTTP delivers our webpages, while POP3 and IMAP carry our email. But when we use the internet to interact with the rest of the world, when we buy, bank, and speak, we do so through layers of platforms. Even our email, carried by open protocols, is most often accessed via centralized webmail platforms. ⁸

Although the widespread adoption of decentralized protocols to manage social identity, payments, and publishing would dramatically enhance the expressive capacity of individuals, decentralized alternatives to platforms have always struggled to gain a broad base of users. Platforms trade control for convenience. The platform takes care of setup, security, and conflict resolution, often at the expense of free speech. Decentralized services require much more of users. They may be asked to shoulder some of the load of network management or, at the very least, to take responsibility for their own security, keeping track of public and private keys rather than relying on passwords stored on platforms.

Traditionally, decentralized internet proposals, and the open-​source software community more generally, have found it hard to effectively compensate those responsible for operating and improving free software. In an open ecosystem, it is difficult to stop free riders from using services without contributing to their upkeep.

Blockchain technology—a form of peer-​to-​peer network popularized by Bitcoin as a means of recording transactions and preventing the double-​spend problem—offers a framework for remunerating network facilitators without establishing an authority capable of controlling network traffic flows. A blockchain is “an open, distributed ledger that can record transactions between two parties efficiently and in a verifiable and permanent way.” ⁹

Within the Bitcoin network, miners lend their computers’ processing power to the network, solving incredibly difficult equations to verify the integrity of each new block of stored transactions. In turn, the network compensates them with newly created Bitcoins, distributed in a pseudorandom fashion to miners, based on the amount of work they do for the network. While the Bitcoin network prioritizes the stable creation of money, other blockchains may provide public ledgers that offer decentralized speech, social, and identity-​verification services.

The InterPlanetary File System (IPFS) and the Filecoin token offer a glimpse of how blockchain-​driven decentralization can radically expand and secure the individual’s publishing ability. IPFS allows a file to be stored, in whole or in part, across multiple locations. An Ethereum blockchain maintains a public ledger of the file’s locations. You can access any version of the file recorded on the ledger, and IPFS directs you to the nearest copy. In the event of a denial-​of-​service attack, or the seizure of a server holding one copy of the file, IPFS will simply direct a searcher to another copy of the file.

This system works well in principle, but it doesn’t create much incentive to offer storage space to the IPFS network. That’s where Filecoin comes in. Sitting on top of the IPFS network, Filecoin is a cryptocurrency used to purchase storage space on the IPFS network and remunerate users who allow IPFS files to be stored on their hard drives. ¹⁰

The location, or locations, of a file stored via IPFS are invisible to the end user. The identity of the host is likewise unknown, and the host himself is unaware of the contents of the file fragments he stores in the unused parts of his hard drive. As such, the application of any sort of censorial pressure, by either states or civil society groups, will be extremely difficult. Discovering the physical location from which a file is hosted presents a real challenge for state authorities. Even in the event of an initial takedown of one copy of a file, it can be immediately and seamlessly rehosted elsewhere without compromising its ability to be discovered via search.

The IPFS system may feel utopian, but it has already proved its worth as a means of censorship resistance. When Turkey blocked its citizens’ access to Wikipedia in 2017, a version of the site was hosted via IPFS. Turkey could identify and bar connections to Wikipedia’s servers, but it could not prevent Turks from accessing copies of the site indexed through the IPFS system. ¹¹ Although this example serves as a valuable proof of concept, the establishment of a functional ecosystem of decentralized speech protocols will face many challenges.

The timelines of both ratcheting censorial pressure and the development of user-​friendly, decentralized publishing tools will determine the makeup of the community of users that composes the decentralized Web. Early adopters—who will do the most to shape its architecture and norms—are unlikely to be ordinary in either outlook or technical competence. A singular mass exodus from the platform internet is improbable; instead, specific groups of users, often from preexisting communities, will find cause to look for the exits.

Adoption will likely be piecemeal, in response to specific content moderation controversies. This response may be driven by either state or private action. The passage of FOSTA—a law that renders platforms responsible for prostitution and sex trafficking—drove sex workers to establish their own instance of Mastodon, a decentralized Twitter clone. ¹² Fearing private bias in Twitter’s content moderation, right-​wing Twitter users have jumped ship for Gab, a more liberally governed microblogging platform. While Gab is not decentralized, its pillorying as a den of far-​right villainy illustrates how the first mover problem could hobble decentralized speech protocols.

If decentralized communication networks are initially viewed as subversive or dangerous, home to speech exiled from respectable platforms, their users may suffer reputational harm, slowing adoption. These concerns may also form the basis for regulation that could strangle decentralized innovations in the crib: think licensure of, or the imposition of liability upon, network participants. Once these services are adopted widely, such regulation would become politically infeasible and practically unenforceable. Regulating niche hobbyists and drug sellers off the internet, however, is unlikely to provoke much backlash. The cryptocurrency community has already done a great deal to normalize the use of blockchain technology, but beyond their contestation of state monopolies on the creation of money, most blockchain uses have not raised politically salient concerns.

To prevent censors from simply moving their efforts lower on the internet protocol stack or using physical chokepoints to cut off messages between decentralized network nodes, at least some of our physical internet infrastructure must be decentralized. Rather than connecting to other machines using fiber-​optic cables owned by a singular utility, or via wireless infrastructure managed by a few cell service providers, users of a truly decentralized internet will avoid these chokepoints by using large-​scale mesh networks. In the past, when internet connectivity was limited to desktop computers, the possibilities of mesh networks were limited. A few local routers could be connected to one another, but broad coverage was difficult to achieve.

Today, the ubiquity of internet-​connected handsets and the internet of things makes large-​scale mesh networks feasible. ¹³ In the future, you will receive internet access through your neighbor’s thermostat, or one of countless other internet-​connected devices. Should one device fail, or should your neighbors dislike the manifestos you publish via their thermostats and cut you off, your traffic can be routed elsewhere. This web of devices may, in turn, connect to either traditional cable networks or novel wireless networks made up of drones, balloons, gliders, and satellites.

While past satellite internet systems relied on larger ground-​based gateways and faced expensive launch constraints, the falling cost of orbital delivery services and the miniaturization of satellite receivers are opening the skies to communications infrastructure like never before. It is now possible to send a message to a satellite and, using Bitcoin, pay for it to be broadcast back to Earth. ¹⁴ Though these data are broadcast, and therefore can technically be received by anyone, encrypting them ensures that only recipients with the correct keys can decrypt the encoded messages. The protocols that receive, process, and retransmit this message are content agnostic. Such a system could also host Mastodon instances or other decentralized publishing applications.

Unfortunately, taking advantage of this cutting-​edge means of publication is technically demanding at present; even configuring a computer to receive this type of message is beyond the capabilities of the average internet user. That is to be expected; it will take time to make this suite of technologies user friendly. The internet was once wholly the domain of hobbyists. Demand for expressive freedom will drive adoption to some extent, but such a system of satellites, mesh networks, and decentralized software will also offer privacy benefits and expanded coverage, especially when compared with legacy infrastructure.

Although early adoption may be piecemeal, even partial adoption will greatly benefit the users who have cause to opt in to this kind of system. In time, as these technologies mature and their benefits are better appreciated, more users will join, in some cases without realizing they are doing so. When the next TikTok or Discord is natively decentralized, simultaneously used and hosted by hundreds of millions of people, the distributed anti-​fragility of these communication tools will render the individual’s expressive potential truly secure.

With interpersonal communication thus shielded against the vagaries of law and custom, state censors and media critics will find themselves forced to refocus on individual speakers rather than the communication systems that give them voice. Internet users—either liberated or estranged from the governance of platform content moderators—will establish new, individualized means of tracking reputation and filtering out unwanted content.

Secure, decentralized, publishing protocols will not obviate the demands of censors. Speech will still be seen to cause distinct harms; some speech and media, like libel and nonconsensual pornography, will rightly remain illegal. The presence of irrepressible communication tools will force censors to refocus their efforts on individual speakers, rather than attempting to suppress undesirable speech by press-​ganging intermediaries into playing censor.

This approach will spur greater focus on the tensions between individual anonymity and accountability. If instead of relying on platform governance to suppress disfavored speech, state authorities must identify and punish bad actors, software that provides for anonymous communication will face increased political scrutiny. Identity verification will gain newfound importance, as will mechanisms for tracking reputation. Even as the decentralized internet becomes more user friendly, these challenges will be, to some extent, baked into the medium. In rejecting centralized administration capable of succumbing to censorial state or social pressure, a greater onus will fall on users to ensure that they receive only the information they desire and that their sources are trustworthy.

Instead of relying on platforms to provide human or algorithmic moderation and curation, users will assemble their own sets of reputation trackers, block lists, and content curation algorithms. Many internet users approach algorithmic filtering with skepticism, largely because of the opaque nature of platform recommendation algorithms. In the hands of individuals, curation algorithms would lose their totalizing impact and could be adjusted over time to prevent either over- or undermoderation. In a decentralized environment, most users will adopt some combination of these tools to manage their social experience. Different tools will appeal to different moderation demands without affecting the internet experiences of those with different preferences.

Collaborative Twitter block lists offer one example of how these tools might function, but, freed from platform-​based API (application programming interface) restrictions, many individualized moderation plug-​ins will offer more fine-​tuned content discrimination options. ¹⁵ Imagine deciding to see only family members’ vacation photos while entirely avoiding their political rants or modifying your social feed with an algorithm that draws on past interactions to prescreen harassment without expunging the jovial use of foul language by a friend. This style of moderation will not satisfy proponents of militant democracy concerned with their neighbors’ information consumption habits, but it will effectively address the concerns of everyday internet users. For users who desire a highly curated experience, the decentralized internet might look a lot like current social media platforms.

Platform-​style governance disputes will not disappear entirely, but their stakes will be lowered by reduced exit costs. Not all nodes within a decentralized internet will choose to accept connections from all other nodes. The precise architecture of decentralized services will surely be a subject of continuing debate among users and stakeholders. In cases of intractable disagreement among stakeholders, the same digital protocol could be replicated under a new governance structure, or, in certain blockchain disputes, “forked” from the existing protocol with limited impact on the experience of users on either side of the conflict.

States will make use of decentralized publishing tools as well, often for malicious purposes. Just as some condemn encrypted messaging for its utility to criminal elements, some liberals may regret the loss of Mark Zuckerberg’s ability to limit the voice of Myanmar’s generals. The broader effects of widespread, censorship-​resistant publishing tools may place new stresses on even liberal political regimes.

Current social media platforms certainly appear to be a boon to dissidents across the political spectrum. However, liberals should take solace in the fact that these innovations will merely reveal existing, albeit suppressed, discontent. Novel publishing tools do not drive otherwise satisfied citizens to hoist the black flag and turn against their governments. Instead, they provide for the perfection of liberal democracy, allowing governors to truly engage with, and respond to, the full spectrum of public sentiment, unfiltered by gatekeepers and censors both public and private.