Digital Domination: Social Media and Contestatory Democracy

Abstract

This article argues that social media companies’ power to regulate communication in the public sphere illustrates a novel type of domination. The idea is that, since social media companies can partially dictate the terms of citizens’ political participation in the public sphere, they can arbitrarily interfere with the choices individuals make qua citizens. I contend that social media companies dominate citizens in two different ways. First, I focus on the cases in which social media companies exercise direct control over political speech. They exercise quasi-public power over citizens because their regulation of speech on social media platforms implies the capacity to arbitrarily interfere with citizens’ democratic contestation in the political system. Second, companies’ algorithmic governance entails the capacity to interfere with citizens’ choices about what mode of discursive engagement they endorse in their relationships with fellow citizens. By raising the cost of deliberative engagement, companies narrow citizens’ choice menu.

Introduction

On 27 April 2021, the US Senate Subcommittee on Privacy, Technology, and the Law held a public hearing on “How Social Media Platforms’ Design Choices Shape Our Discourse and Our Minds.” Lauren Culbertson (2021: 1), the head of US Public Policy at Twitter, asserted that they “remain committed to giving people the power to create and share ideas and information instantly with the world” and listed a number of proposed mechanisms by which they can grant their users expanded voice and greater individual choice on their platform. Social media companies’ emphasis on user voice and individual choice seems to be a response to an underlying public discontent about corporate power over the digital public sphere. The emerging socio-political problems in this field put greater public pressure on Big Tech regarding their extensive powers over users and other stakeholders.

I argue that social media companies’ power to regulate communication in the public sphere illustrates a novel type of domination that is particularly problematic from a democratic theory perspective. The idea is that, since social media companies can partially dictate the terms of citizens’ political participation in the public sphere, they can arbitrarily interfere with the choices individuals make qua citizens. Drawing on Philip Pettit’s neo-republican notions of domination and contestatory democracy, I demonstrate that digital domination in the public sphere is a threat to a crucial element in the political system. As such, even though it is private in origin, digital domination is not an instance of mere private domination. It is, rather, an instance of what I call quasi-public domination.

This article contains two theoretical innovations. First, it contributes to the attempts to analyze and assess the digital public sphere through the conceptual lens of domination by offering a novel argument (Bowman, 2021). Second, and perhaps more importantly, the article invites the reader to rethink the conventional categorization of forms of domination. As Pettit (2013: 3) famously puts it, public domination is often understood as the opposite of democratic political life and is perceived as a form of state power. On the other hand, private domination refers to the relations of power between private actors, which the state should address as a matter of social justice. I argue that this strict distinction is obsolete in our current circumstances, as some forms of private power have public functions, for example, regulating the limits and visibility of political speech. As a result, I propose to identify the type of domination by looking at its target domain rather than its agential origin, focusing on the type of choices an instance of domination characteristically jeopardizes. In this reading, the powers of social media companies amount to what I will call quasi-public domination. This is a form of domination that has a public target, although the power is exercised by private actors. By the target domain of a dominating relationship, I refer to what a type of domination characteristically does to our choice menu. It is a way of distinguishing one form of domination from another by looking at how each narrows individuals’ choice menu in relation to certain areas of activity, that is, potentially interfering with our social roles qua citizens, employees, or partners in an intimate relationship. Domination has a public target insofar as it implies the capacity to interfere with political choices individuals make qua citizens.

Why does it matter if social media companies’ powers amount to quasi-public domination instead of private domination? First, there is a straightforward answer to this question from within neo-republican theory. Public domination is a more urgent danger than private domination (Pettit, 2013: 24). This is because failure to address public domination multiplies the risk of private domination. In the absence of democratic control, the state’s protection of citizens from other citizens would be precarious. Hence, public domination means an increased risk of further domination. If social media companies’ powers amount to quasi-public domination, endangering the capacity of the political system to protect citizens against other types of domination, this would drastically change our perception regarding the urgency of the problem. Second, a more concrete payoff of conceptualizing quasi-public domination is the following: having a strict distinction between public and private domination creates an unduly state-centric bias in our imagination of what a plausible solution would look like (Watkins, 2015). In the neo-republican framework, private domination is often understood as a domain that needs to be regulated by the democratic state (Pettit, 2013). However, if the functioning of a political system is shaped or restrained by certain types of non-state domination, there is a good chance that this will have an impact on state capacity. Pettit’s procedural conception of a democratic state can be vulnerable to private actors’ power, as “the operation of procedures depends on the background organization of social actors and how they relate to the procedure” (Klein, 2021: 31). As a result, the state-centric outlook would overestimate the role of the state in addressing such power relations. Developing a conception of quasi-public domination allows us to categorize different types of domination without endorsing the state-centric bias. This is because the political system is understood not only through formal political bodies but also through their relationships with broader social dynamics such as the prospects of citizen contestation.

I contend that social media companies dominate citizens in two different ways. First, I focus on the cases in which social media companies exercise direct control over political speech. The vast array of content moderation practices conducted by social media companies is largely non-transparent and unaccountable (Flew and Wilding, 2021: 59). Although there is a degree of government regulation, and certain efforts to realize transparency, companies’ decisions about what speech is permissible are considerably discretionary (Flew et al., 2019: 43). This is a de facto exercise of uncontrolled public power because it arbitrarily regulates citizens’ participation in the practices of democratic contestation, which is a political right. Citizens’ ability to contest the existing policies and power relations is an essential element of a republican system of government. By virtue of their capacity to arbitrarily interfere with citizens’ role in the political system, social media companies dominate users qua citizens.

Second, in the absence of extensive democratic control, social media companies also have the power to dominate citizens and shape the public sphere in a more subtle way. Through their algorithmic decisions, companies can conduct automated regulation of speech that makes some discourses more visible than others (Crawford, 2016). Unlike the first mechanism, algorithmic control over visibility does not rely on censorship. Algorithmic decisions interfere with citizens’ choices about how they participate in the public sphere. Algorithmic intervention can reduce or increase the number of available perspectives a citizen can hear. For instance, algorithms that maximize user engagement often boost exposure to content from like-minded individuals, causing in-group ideological homogeneity and inter-group polarization (Himelboim et al., 2013; Hong and Kim, 2016; Langvardt, 2018). In a highly polarized environment, deliberative engagement with others becomes an increasingly unavailable option. As a result, managing the terms of visibility in the digital public sphere implies the capacity to interfere with citizens’ choices about what mode of communicative interaction, for example, more deliberative or antagonistic, they endorse in their relationships with fellow citizens.

The article proceeds as follows: in the first section, I provide an overview of certain basic notions from neo-republican democratic theory such as domination, democratic control, and contestatory democracy. This will lay the groundwork for the later discussion on social media domination and citizen contestation. In second section, I argue that social media companies have arbitrary power over citizens and this amounts to quasi-public domination. More specifically, I contend that it is not plausible to view social media companies as mere service providers who contract with their users on the basis of free consent. In the third section, I will review and reply to several objections.

Domination and Contestatory Democracy

Let me first explain why I choose to work from within Pettit’s neo-republican framework as opposed to other approaches to domination. The first advantage of Pettit’s framework is that it offers a sophisticated account of democratic society, highlighting how multiple political processes would interact with each other in a republican system of government. Second, the type of powers I focus on in this article neatly fit into Pettit’s (2002, 2013) agent-centric conception of domination. While radical republicans often emphasize diffused and structural domination, the powers of social media companies are rather centralized and under the intentional control of corporate bodies (Gourevitch, 2013; Rahman, 2017). Concerning the tasks I set for this article, the neo-republican framework works well, and, for the sake of parsimony, I do not wish to further complicate my basic model of political institutions.

In Pettit’s neo-republicanism, maximizing freedom as non-domination is the central goal. A dominates B to the extent that A has “the capacity to interfere on an arbitrary basis in certain choices” that B is able to make (Pettit, (2002: 52). Interference entails the restriction of one's choice menu by affecting one’s “use of the otherwise accessible objective or cognitive resources in virtue of which we say that you have the unvitiated capacity to do X or Y” (Pettit, 2013: 50). Furthermore, the notion of arbitrariness refers to one’s lack of control over the powers that impact one’s choice menu (Pettit, 2013: 58).

Pettit envisages a division of labor between social justice and political legitimacy, each addressing different forms of domination (Pettit, 2013: 24). Republican social justice focuses on citizens’ relationships with each other and aims to minimize relations of private domination, particularly observed in the economy, family, and workplace (Pettit, 2013: 114). In contrast, the neo-republican theory of democracy presents a conception of political legitimacy, highlighting the circumstances under which state power would not be dominating (Pettit, 2013: 3). Institutional features such as counter-majoritarianism, the rule of law, and dispersion of state power target different potential sources of public domination (Pettit, 2002: 172–177).

The elimination of public domination requires popular democratic control according to Pettit. Unless the citizenry determines the particular direction of public decision-making, there will always be plenty of latitude for political elites to exercise public authority at their own pleasure (Pettit, 2013: 240). Unlike mere influence, democratic control makes sure that the exercise of political power follows norms and interests that can plausibly be attributed to the citizenry. In the remaining part of this section, I will highlight how a contestatory conception of democracy is essential in realizing democratic control.

Contestatory democracy is the view that citizens should be able to effectively challenge government decisions (Pettit, 2000: 119). For an effective system of democratic control, “people must be on the watch for proposals or measures that are not suitably supported... and they must be ready to organize in opposition to such policies” (Pettit, 2002: 226). As political elites or tyrannical majorities can turn out to be unresponsive to other citizens’ interests and values, contestatory practices are conceived of as the cornerstone of ensuring non-domination. Citizens’ vigilance, civic engagement, and a culture of contestation compensate the shortcomings of mere electoral influence (Pettit, 1999: 174).

According to Pettit (2013: 260), practices of contestation and opposition in various sites of social and political life give rise to norms that each party is likely to endorse. To the extent that political power is dispersed and the citizenry is empowered, what Pettit calls the “acceptability game” prevails in the social world. The acceptability game requires actors to interact with each other on a more deliberative basis when they attempt to resolve their disputes, that is, offering fellow citizens reasons that they can be expected to find relevant (Pettit, 2013: 253). Insofar as citizens are repeatedly incentivized to play the acceptability game, “policy-making norms begin to emerge and crystallize at each site where contestation is brought or heard, answered or adjudicated” (Pettit, 2013: 261). Through the mechanisms of voice and contestation, “a social movement or other pressure creates a demand for state action; publicly supported legislation generates a statute embodying a new norm” (Pettit, 2013: 274). The emergence of such norms makes democratic control possible because policy proposals that violate the norms will be dropped off the agenda (Pettit, 2013: 265). As a result, contestatory democracy ensures that the people have democratic control over government, not because of their concerted intentional efforts but through the norms they create by engaging in decentralized and widespread practices of contestation at multiple sites.

Finally, I will highlight the importance of the public sphere in Pettit’s vision of contestatory democracy. Contestatory practices are fundamentally debate-based, “enabling citizens to raise the question” of whether decisions made by legislative, administrative, and judicial branches rely on proper considerations (Pettit, 2002: 188). Furthermore, mere representation is not sufficient for contestatory democracy. Citizens themselves should be in the position to protest and challenge the centers of power in the political system. Rights of association, petition, protest, and demonstration are all different mechanisms encouraging citizen vigilance against the danger of public domination. Social movements and citizen initiatives are two of the main vehicles of such contestatory practices (Pettit, 2002: 193). In order for these actors to effectively monitor and challenge public authorities, a contestatory democracy needs to guarantee a forum where contestations “can receive a proper hearing” (Pettit, 2002: 195). This vision of democracy requires a dynamic and resilient public sphere in which citizens consolidate their autonomous position and check the uses of public power. As Pettit (2003: 152) puts it, “there must be a dispensation of deliberation in place in the community as a whole.”

Defining the public sphere as “a realm of our social life in which something approaching public opinion can be formed,” contestatory democracy clearly necessitates the existence of autonomous and public forums of debate and opinion-formation (Habermas, 2006: 73). The venues of public debate and opinion-formation range from media institutions to local forums. As Pettit’s (2013: 275) model of democratic influence and control partly relies on citizen contestation and public debate, a free and egalitarian public sphere is required to enable marginalized voices to be heard. This is partly the source of Pettit’s concern about private media that is often in the service of sectional interests (Pettit, 2013: 234). For contestatory democracy to successfully eliminate public domination and ensure popular control over government, a healthy public sphere must be created and preserved.

Considering the central role of the public sphere in contestatory democracy, I will now explain how social media companies are a type of dominating agent by virtue of their quasi-public power over digital communication. Social media companies dominate individuals qua citizens because of their arbitrary power to interfere with citizens’ political participation in the public sphere.

Digital Domination Is Quasi-Public Domination

Can Private Actors Publicly Dominate?

Unlike Pettit, who draws a distinction between types of domination by looking at its agential origin, I propose to categorize domination by virtue of its target domain. The target domain of an instance of domination shifts the focus toward how domination relates to certain spheres of human activity. Our choice menus can be typically broken down into segments associated with our social roles in different domains, for example, our roles as citizens, economic actors, or family members. Rather than categorizing domination on the basis of agential origin, that is, whether power is exercised by the state or a private actor, my suggestion is to categorize it on the basis of how an instance of domination aims at choice menus associated with different social roles.

In this view, an instance of domination is quasi-public if it amounts to a private actor dominating others qua citizens. Consider a paradigmatic example of private domination. A husband dominates his wife by virtue of their patriarchal gender roles, making the latter unfree in her interactions with her husband structured by the family institution. The type of behavior regulated within the patriarchal household typically pertains to intimate relationships, women’s social visibility, and domestic labor that are all characteristic of the roles and actions within the family institution. Domination by virtue of family roles can be plausibly conceived of as private. This does not mean such domination is politically irrelevant. The republican public–private distinction is compatible with the claim that the personal is political. It is called private domination in the sense that it is relatively autonomous from power relations embedded in political institutions, that is, mechanisms of democratic influence and control.

However, the moment a husband has the capacity to effectively curb a wife’s political rights, the narrative turns into something different. ¹ This is because the roles and the relevant set of choices that are being narrowed partly changes. The individual is then dominated qua citizen in addition to being dominated by virtue of her roles within the family institution, for example, exploitation of her domestic labor. In other words, A is dominated qua X when the relevant set of choices that can be narrowed pertains to an area of activity in which A bears the role of X. When individuals are dominated qua citizens, the relevant set of choices that can be narrowed pertains to individuals’ role as citizens. If one has the capacity to arbitrarily interfere with others’ choices about political speech, democratic participation, or democratic influence, they are dominated as citizens.

Furthermore, quasi-public domination is not all about individual cases of narrowing choice menus. Their effects on the political system are a part of what makes these cases quasi-public. Quasi-public domination is more of a threat to democratic political life than a problem of social justice among private actors. Pettit’s category of social justice is meaningful to the extent that the political system is considerably isolated from private actors, whose relations have normative implications primarily for other social domains. However, in the cases of quasi-public domination, individuals’ domination qua citizens is at the very heart of the political system. When citizens’ choices about their political rights are potentially frustrated, this undermines the mechanisms of democratic influence and control Pettit outlines in his neo-republican model of government. ² A quasi-public dominator has the capacity to interfere with the ways citizen voice is transmitted to the political system because the former’s significant uncontrolled power decreases the latitude of citizens in fulfilling their political roles, for example, through protest, speech, or social mobilization. This implies a more direct control over citizens, that is, regulating the terms under which citizens can engage in contestatory practices.

Some radical republicans similarly discussed the political effects of concentrated private power (Rahman, 2017). However, the difference is that the paradigmatic examples of those discussions involve indirect exercise of power over citizens, for instance, influencing policy-making through corporate lobbying. In contrast, quasi-public domination amounts to the capacity to directly interfere with citizens’ political actions. That is why it partly resembles state domination. Quasi-public domination is more politically invasive. The difference between quasi-public domination and other democratically problematic private powers can be understood through this analogy: a private actor can narrow a slave’s choice menu by persuading or pressuring their master to do certain things to the slave. This resembles anti-democratic corporate power in politics in the abovementioned sense. In contrast, consider a second private actor who somehow obtained the capacity to torture or humiliate the slave directly, acting as a second master. This is the equivalent of quasi-public domination. While both cases are highly problematic from a republican perspective, quasi-public domination is ceteris paribus more undesirable. This is because it implies more independent entry points to domination, which makes it harder and more complicated to fight oppressive relations in the political system.

What are the practical implications of such a conceptual innovation? A strict distinction between private and public domination runs the risk of misleading our inquiry about how to fight against domination. As Pettit (2013: 3) puts it, private domination is primarily understood as a set of power relations against which the state “has to guard people.” I do not deny that a republican state has such a responsibility. However, this conceptual framework tends to overestimate what we can reasonably expect from a state as it assigns private actors to an inferior position of influence (Watkins, 2015). The problem is that the state is not omnipotent in relation to power relations between private actors. On the contrary, the state is both a structuring and structured institutional space, subject to the formative influence of private actors that transform, promote, or undermine political procedures (Klein, 2021). In other words, the power of certain private actors could be understood as an input determining how political procedures operate and translate into the exercise of state power, rather than a mere output of state regulation.

As a result, we need to think about political domination in non-state-centric terms. The concept of quasi-public domination enables one to theorize about the political system without granting the state a privileged status. I believe this makes it more difficult to endorse undue emphasis on state regulation, which I will further discuss in the final section, when we raise the question of how to practically address relations of domination. While state regulation is still an important tool of social change, this shift in attention is likely to create more space for alternative forms of political agency including social movements and other civil society actors. For instance, fighting the quasi-public domination of social media companies would involve several non-state strategies, for example, bottom-up direct action from digital platform “workers and users as a method of building our power and eroding the power of tech companies” (Muldoon, 2022: 169). Another possible form of political agency would be to focus on creating democratic and egalitarian alternatives to corporate digital platforms through crowdfunding and other types of social mobilization (Muldoon, 2022: 173).

The second, more theoretical, advantage of this conceptualization is that there is a better fit between our negative and positive evaluative terms. If the use of public domination serves to identify what a free political system should minimize, then it makes more sense to categorize relations of domination in terms of their impact on such an ideal of free state we are striving for. Quasi-public domination reduces the democratic quality of a political system, which also curbs the capacity of a republican state to address social injustices. Hence, unlike usual cases of private domination, such examples are not clearly differentiated from the basic functioning of a political system. This will have implications for democratic legitimacy. However, it does not sound conceptually coherent to suggest that democratic legitimacy is undermined by private domination, as the latter includes a wide variety of power relations that may not directly influence how the political system operates. By reconceptualizing the types of domination in the abovementioned way, one can make sure that democratic legitimacy and the social relations that undermine it, that is, varieties of public domination, are conceptual counterparts.

Social Media and Quasi-Public Domination

I will now demonstrate that social media companies’ power over the digital public sphere amounts to quasi-public domination. By social media companies, I mean corporate agents that govern social media platforms such as Facebook, Twitter, and YouTube. A social media platform is “a programmable digital architecture designed to organize interactions between users” (Van Dijck et al., 2018: 4). I believe there are two ways social media companies dominate individuals qua citizens. First, social media companies’ content moderation directly regulates users’ speech in the digital public sphere (Gorwa, 2019: 859). Community standards of social media platforms allegedly base their content moderation on identifying and removing content that involves hate speech, sexual violence, and nudity—which are all essentially contestable and politically controversial concepts. For instance, as one moves away from extreme examples, what counts as hate speech is difficult to decide on and subject to disputes. While some Black activists complain that they are being silenced on Facebook due to their anti-racist commentary, others highlight that the monitoring and evaluation procedures of the very same platform can get too permissive regarding misogynistic hate speech (Gillespie, 2018: 112; Guynn, 2019). Given that there is room for interpretative discretion and no democratic control over companies’ policies, content moderation of social media companies ultimately rests on their moderation teams’ arbitrary will. It is beside the point whether the moderation teams and/or corporate algorithms are making morally right decisions in these particular cases.

The possibility of arbitrary interference with online speech makes citizens’ political contestations dependent on the discretion of social media companies. A vivid illustration of this case occurred when the Twitter account of Lawyers for Civil Rights, a Boston-based nonprofit, was suspended after “posting support for the Black Lives Matter movement” (Dwyer, 2020). The companies have the capacity to keep citizens from effectively participating in the practices of contestatory democracy, for example, challenging a policy or a political view by discursive engagement and social mobilization. Such capacity to interfere might be less dangerous and tyrannical than government censorship backed by the threat of violence. After all, the strongest sanction that a social media company can impose is to suspend a user account. Nevertheless, given the increasing prevalence of social media in public political communication, exclusion from these platforms would deprive citizens of an significant resource that is needed for full-fledged participation in the public sphere.

Why do I hold that social media platforms are a significant element of the modern public sphere, crucial for citizens’ effective capacity for political participation and to contest the existing power relations? It is empirically well-established that participation in the digital public sphere is consequential for political actors and social movements (Breuer et al., 2015; Eltantawy and Wiest, 2011; Hong, 2013). Social movements amplify their voices and coordinate collective action by boosting their visibility on social media. Similarly, visibility on social media has significant consequences for politicians’ electoral success and even influences their performance in collecting donations (Hong, 2013; Shmargad and Sanchez, 2022). These platforms are also where public attention is extremely concentrated. For instance, it has been reported that social media platforms are the primary source of news for the majority of the US population (Martin, 2018).

Considering the high stakes of social media visibility, one needs to acknowledge that the costs of quitting social media are unacceptably high if one takes effective capacity for political participation seriously. Neo-republicans are not participatory democrats for sure. However, their vision of contestatory democracy requires a certain degree of civic engagement and citizen vigilance or at least significant opportunities to display such participatory activities, which is needed for popular democratic control (Pettit, 2002: 193). Hence, citizens’ effective capacity to participate in the public sphere is crucial even if this does not presuppose continuous and extensive citizen participation in democratic life. Exclusion from social media does not harm one’s formal political status, but can partially undermine one’s effective use of political rights. Furthermore, given the near-monopoly of the biggest social media companies over the central venues of socially networked communication, users do not have realistic exit opportunities (Flew and Wilding, 2021: 50). User consent does not seem to be normatively significant due to the high cost of exit and the lack of genuine alternatives, resulting from the quasi-monopoly of social media companies (Everett, 2018: 123). Citizens have to comply with or adapt to social media companies’ arbitrary interferences if they want to stay active in the digital public sphere. The possibility of interference applies to not all but to the most significant venues of social visibility where public attention is extremely concentrated. ³ Since the relationship between social media companies and citizens partially determines the latter’s effective use of some political rights, for example, the right to publicly contest powerful actors and engage in social mobilization, it implies domination of individuals qua citizens. This resembles dominating someone by the potential threat of depriving them of a significant resource: “If you depended on me for the resources needed to choose an option, then I would have a power of uncontrolled interference in the choice and would dominate you” (Pettit, 2013: 70). Citizens depend on social media companies’ goodwill for access to the effective venues of digital communication, which is a key political resource in contemporary societies. For them to effectively choose a practice of contestation, they should have access to social media platforms, which are monopolistically controlled by companies. By virtue of this vulnerability, companies’ regulatory actions imply arbitrary power to interfere with citizens’ political speech. ⁴

What about conventional media as an alternative? Can we say that the preexisting media outlets give citizens an exit opportunity which empowers them against social media companies? There are a number of difficulties with this idea. First, social media and conventional media are qualitatively different (Celikates, 2015). Most conventional media appoint citizens to a passive audience role. ⁵ Hence, it is not clear how the existence of such outlets would give citizens an alternative voice mechanism that is relevantly analogous to their participation on social media. Second, once social media platforms become the central venue of public debate and political communication, the importance of preexisting media outlets fades away or becomes dependent on social media (Meese and Hurcombe, 2021). Although switching to conventional media is an available option for citizens, such transition is not likely to foster the same public debate environment one is used to. Unless there is a concerted action resulting in the collective migration of public attention from social media back to conventional media, a single individual’s use of the latter is not likely to compensate for what one misses out by not being on the former.

There may still be some alternative venues of the public sphere, for example, local community forums, voluntary associations, and informal communication networks. However, these options might not be a perfect substitute for social media platforms that offer an equivalent opportunity set. Local and/or specialized public spheres like voluntary associations considerably depend on broader communicative structures such as conventional or social media to reach other parts of the democratic citizenry. Otherwise, the effectiveness of political speech might be curbed due to their regionally or functionally limited scope. This does not mean that it is impossible for citizens to achieve the same communicative outcomes in the absence of social media. However, it seems reasonable to hold that the cost of achieving these outcomes would be higher due to increased efforts, which is sufficient for my claim that citizens’ choice menu about political participation is being narrowed.

One might object to my argument by claiming that social media platforms have increased our latitude in many other ways, expanding our choice menu about effective citizen communication. Especially from the perspective of ordinary citizens who did not have substantial opportunities to voice their concerns via conventional media outlets, digitalization has revolutionized our effective capacity for communication. Following this, one might argue that social media companies do not dominate citizens because they have expanded citizens’ communicative choice menu much more than content moderation has narrowed them. However, this objection conflates how entrance into a social relation impacts our choice menu with how the social relation itself affects it later on. For instance, the debtor–creditor relationship clearly includes the potential for domination if appropriate measures are not taken, according to Pettit (2013: 114). Insofar as the creditor can effectively use her financial leverage to arbitrarily interfere with the debtor’s choices, this is domination. However, it is equally true that the debtor could have expanded her initial choice menu when she first entered into the relationship. This is simply because the payment she receives substantially improved her financial options. On the other hand, to the extent that the creditor gained a capacity to interfere with the debtor’s “expanded” choice menu, domination is still possible and likely. Following in the footsteps of this example, I contend that the positive impacts of social media technologies on citizens’ communicative capacity do not nullify our claims about social media companies’ quasi-public domination. Although the choice menu has been initially expanded in comparison with previous stages of the public sphere, the very same transformation has also given rise to a new social relation in which a small group of corporate actors have quasi-monopolistic control over the terms and conditions of citizens’ online speech and contestatory practices.

Let me note that social media companies have been making attempts to increase transparency and users’ voice in content moderation practices. For instance, the flagging system creates room for participation by allowing users to shift moderators’ attention to content they think violates the community standards (Crawford and Gillespie, 2016; Culbertson, 2021). Furthermore, Facebook has recently authorized an autonomous oversight board that can “uphold or reverse Facebook’s content decisions” (Oversight Board, 2020). However, these measures fall short of Pettit’s requirement that access to the mechanisms of contestation and influence are to be unconditional (Pettit, 2013). The digital public sphere is one indispensable mechanism in which citizens participate in public life and engage in contestatory practices. Instead of being unconditional and independent, the existence and functioning of these channels of influence still depend on the goodwill of corporate executive boards.

The second way social media companies dominate individuals qua citizens is the former’s indirect algorithmic control over citizens’ speech. Algorithms are “encoded procedures for solving a problem by transforming input data into a desired output” (Yeung, 2018: 506). Algorithms that are used on social media platforms determine the terms of visibility and interaction between users in a way that optimally fulfills the objectives of social media companies (Kuehn and Salter, 2020: 2592). Indirect algorithmic control over speech is different than content moderation in that it does not necessarily remove content or suspend user accounts. It rather arranges who can hear an individual user’s speech on what frequency. For instance, algorithmic control can determine one’s audience by algorithmically preselecting a group of users to whom content will be visible.

Why is such algorithmic control dominating? Corporate algorithmic control over online speech implies the capacity to interfere with how citizens participate in the public life of a political community. Say there are a number of available options citizens can choose when they decide on how they engage with fellow citizens and policy-makers: full deliberative mode, full antagonistic mode, and mixed mode (Forestal, 2021; Mouffe, 2005). In the deliberative mode, citizens make substantial, sincere attempts to deliberatively engage with others on the basis of an exchange of reasons. In the antagonistic mode, citizens prefer rhetorical discourses and appeal to their uncompromising normative values at the cost of cutting mutual deliberation short. In the mixed mode, citizens display a discursive attitude that is a balanced combination of deliberative and antagonistic modes. When citizens have access to the full scope of these options, they can reflect on which mode of discursive engagement is most reasonable in a particular context. For instance, a citizen can reasonably believe that a more uncompromising and conflict-driven attitude should be adopted against racism because we should not normalize racist positions as a minimally reasonable interlocutor in public discourse. In contrast, the very same citizen might think the public debate around the question of euthanasia deserves a more deliberative approach, as it is a normative puzzle with a large gray area. Depending on the mode of engagement citizens choose, the way they contest others’ views or the government’s decisions will change. Furthermore, depending on the style in which citizens talk with each other, public debates will be resolved differently. Hence, the mode of engagement is not only a matter of formality. Citizens’ choice of mode of engagement is integral to the exercise of their political participation rights. The choice of the mode of engagement will influence how and in what direction public opinion is formed, which also has significant implications for governmental policy-making (Hakhverdian, 2012).

Social media companies’ algorithmic control amounts to quasi-public domination because it arbitrarily interferes with citizens’ choices about modes of discursive engagement. One clear example is companies’ algorithmic preferences to maximize user engagement on social media platforms:

Outrage, catharsis, insecurity, and the need to belong to one’s tribe all seem to provide potent “triggers” and “rewards” for the habit-forming product. Recent studies have confirmed what is intuitively plain in social networks: appeals to group identification and negative emotion drive engagement more reliably than anything else. Algorithms optimized to drive engagement therefore tend to serve up cartloads of these reactive stories to users whose past engagement indicates they are predisposed to click on them (Langvardt, 2018: 361).

Social media companies are corporations that have an interest in maximizing user engagement for data extraction and other profit-maximizing objectives. Encouraging critical thinking and careful deliberation is not exactly the most lucrative strategy for that kind of business model. In contrast, they employ algorithms that reproduce users’ existing biases (Langvardt, 2018: 358). Algorithms that are optimal for social media companies’ business model highly personalize the content one is exposed to and boost the intensity of interaction among like-minded users, which induces greater addiction to social media platforms (Harris, 2021: 2; Kuehn and Salter, 2020: 2593). By utilizing human disposition to associate with group identities and to endorse bias-confirming narratives, algorithms systematically incentivize uncritical, one-sided, and reactive online behavior. This in turn creates a communicative environment with greater in-group ideological homogeneity and inter-group polarization (Himelboim et al., 2013; Hong and Kim, 2016).

Social media companies’ algorithmic control effectively interferes with citizens’ choices about modes of engagement in public debate. Citizens’ options are narrowed when, for instance, corporate algorithmic decisions make deliberative interactions with others increasingly inaccessible. The cost structure of one’s choice menu will change due to algorithmic governance. Deliberative engagement is likely to be more time-consuming and psychologically challenging when the institutional architecture systematically encourages hostile behavior. An important clarification is in order here. The normative problem I focus on is not about promoting anti-deliberation per se, but about narrowing citizens’ choice menu. For instance, if social media companies preferred an opposite kind of algorithm with which it is extremely difficult to interact with like-minded users, this would still count as domination. Perhaps the second algorithm could have positive qualities in terms of promoting desirable deliberative attitudes and civility in the public discourse. On the other hand, this algorithmic setting similarly interferes with citizens’ choices as it reduces their capacity to antagonistically interact with others and to form group identities with like-minded citizens on the basis of political contestations. Either way, social media companies narrow the choice menu of citizens in their decisions about how to contest others’ opinions or policies in the public sphere. Citizens do not have a sufficient level of democratic control over how these algorithms are designed or implemented, which I will further discuss in the next section (Langvardt, 2018). In the absence of democratic control over powers interfering with citizens’ choices, the arbitrariness condition holds.

One might disagree with this analysis by disputing its empirical presuppositions: the impact of social media on citizens’ mode of discursive engagement, for example, formation of echo chambers, has been overstated, according to some studies (Dubois and Blank, 2018). While an extensive analysis of the empirical literature is beyond the scope of this article, one should consider that other large-N studies that focus on tracking and experimental data show the significance of echo chambers and other polarizing tendencies (Allcott et al., 2020; Cinelli et al., 2021). Considering the mixed evidence on the subject, it would be prudent to moderate the claim about how effective social media companies’ algorithmic governance is. However, even with a more skeptical stance, it would be fair to say that the tendency to form echo chambers is substantial for a subset of social media users, as Dubois and Blank (2018) showed. In the scenario in which the empirical presuppositions are weakened, the dominating effects of algorithmic control would only apply to this subpopulation.

Another worry is that in-group homogeneity and inter-group polarization might be caused by the basic psychological tendencies of humans, and algorithms merely facilitate them. Some studies showed that “individuals’ choices played a stronger role in limiting exposure to cross-cutting content” (Bakshy et al., 2015: 1130). I believe this worry stems from a static understanding of human–algorithm interaction. In a more dynamic account, the distinct causal role of algorithms is easier to observe. Algorithms facilitate individual choice in that they rely on personalized input from users. However, algorithmic curation of online content turns user input at period t₁ into a choice menu-narrowing output at t₂. By overrepresenting cumulative past online behavior, it alters the cost structure of a user’s present choice menu, for example, raising the cost of reaching diverse information sources. Recent lab experiments confirm this insight by showing how algorithms can reinforce and polarize political opinions by iterating users’ past preferences (Cho et al., 2020). ⁶

If contestatory practices are an essential element of a democratic political system, and if contestation requires citizens’ latitude in deciding how they use their political rights, for example, political speech in a more deliberative (or antagonistic) mode, then social media companies dominate users qua citizens. Algorithmic control over speech considerably keeps citizens from fulfilling an essential role in the political system, namely free contestation, because it narrows the ways citizens can participate and contest power relations in the public sphere. Given that citizens’ choices about the mode of engagement have an impact on the direction of public-opinion formation, that is, more or less polarization, algorithmic power over speech disrupts citizens’ prospects of popular democratic control.

There is an important question I need to address at this point. Pettit’s conception of domination highlights that one’s choice menu is narrowed in the relevant sense if one’s use of “the otherwise accessible objective and cognitive resources” to do X is curbed (Pettit, 2013: 50). Then the question is whether citizens have the necessary objective and cognitive resources to reflect on and choose their mode of discursive engagement in the public sphere when algorithmic control is not around. In other words, do citizens normally think and make choices about their mode of discursive engagement? Afsahi (2021) convincingly argues that willingness to deliberate is often a function of one’s perceived self-interest. Actors strategically assess and respond to the relevant incentive structures when deciding to deliberatively engage with others. When one is incentivized to be perceived as a reasonable interlocutor by others, one is more likely to display deliberative attitudes. So we can say that there is a meaningful choice menu about modes of discursive engagement, in relation to which citizens often make decisions. Social media companies’ algorithmic control narrows the choice menu by unilaterally changing the relevant incentive structure and imposing their preferred configuration. For instance, when their algorithms polarize public opinion and make deliberative engagement increasingly inaccessible, one of the options in citizens’ choice menu becomes much more costly than the baseline level prior to the implementation of a particular algorithm.

As discussed in the previous section, in a contestatory conception of democracy, citizens’ individualized, unconditional, and efficacious influence on the political processes is a necessary condition for popular democratic control. Citizens’ speech and their decisions regarding different ways of participating in the public sphere are essential dimensions of contestatory practices. Social media companies’ powers undermine free citizen contestation in two ways: by regulating speech in a democratically unaccountable way and by interfering with citizens’ choices regarding their modes of discursive engagement. As contestatory citizenry is an informal institution that constitutes a building block of a democratic system, domination of citizens in their contestatory roles in the public sphere is a threat to democratic political life, rather than a matter of social justice. Social media companies’ powers are different from mere private domination as much as they are different from the traditional cases of public domination. The term quasi-public domination captures this peculiarity.

While this article presents a diagnostic argument identifying problematic power relations in the digital public sphere, it also directs our attention toward potential solutions, including a wide array of institutional interventions such as an increasing role for social movements, user control over algorithms, creating nonprofit alternatives to social media companies, and breaking up monopolistic companies (Kuehn and Salter, 2020; Langvardt, 2018). In the case of content moderation practices, some kind of collective democratic control seems plausible as these practices are often characterized by conflicts about what online community standards are desirable. Non-dominating content moderation practices would have to form a collective agreement without undermining contestation and pluralism. In the case of algorithmic control, one might opt for individualist solutions to minimize domination: offering a marketplace of different algorithms, and letting individual users choose what works best for them as advocated and partially implemented by some social media companies (Kastrenakes, 2021). It is important to note that the individual liberty of toggling between news feed algorithms is currently limited to a small number of options: ordering based on either chronology or popularity. Given that there are countless sorting possibilities based on different criteria, social media companies still enjoy significant control over what options are available to users. Full decentralization of algorithmic services by creating a genuine marketplace might shift the power away from social media companies. However, neo-republicans would still have reservations about this proposal. Further personalization of social media use through algorithmic individualization is likely to induce greater fragmentation in the public sphere. This might harm citizens’ mutual understanding and common awareness of their public affairs, which is a necessary condition for them to coordinate and engage in collective action to protect freedom as non-domination when necessary (Kurtulmus, 2022; Lovett and Pettit, 2019). Hence, collective but pluralist routes to democratic control remain a more promising direction. ⁷ As the article’s main argument is primarily diagnostic and conceptual rather than prescriptive, a more detailed account of how to address digital domination is a task for future research.

Some Further Objections

I will now discuss and reply to several objections. I have chosen to do this in a separate section as these objections can arguably apply to my theoretical framework as a whole, not just specific mechanisms of domination. The first objection might be the following: Citizens are not dominated by social media companies because democratic states extensively regulate digital platforms (Busch et al., 2018). States’ regulation of social media platforms is non-trivial, covering many areas including data protection, copyright, child pornography, public incitement to crime, and propaganda of unconstitutional organizations (Karaboga, 2018; Tworek and Leerssen, 2019). As democratic states are substantially controlled by citizens, the objection suggests that citizens are indirectly drawing the boundaries of social media companies’ powers. Whenever there is a sufficiently important issue on social media communication, democratic states step in and regulate social media companies’ conduct on digital platforms. Hence, social media companies’ powers are not uncontrolled or arbitrary in the relevant sense of the terms. Then one would contend that citizens are not dominated, at least in democratic states. Even when the state is non-democratic, extensive state regulation would imply that citizens’ loss of freedom is caused by state power, rather than social media companies, because the former is the ultimate determinant shaping social media communication.

The first problem with this objection is that it fails to acknowledge the role of discretion in the implementation of legal regulations. One well-known example is socialist critics’ argument about workplace domination, claiming that labor-empowering legal regulations still give managers extensive control of the interpretation and enforcement of rules (O’Shea, 2020: 558). As regulations are often vague and implemented by those who are at the top of the organizational hierarchy, socialist republicans hold that even extensive legal regulation is not sufficient to address managers’ arbitrary power in the workplace. Similarly, legal regulations regarding social media platforms leave a substantial degree of control in the hands of corporate actors as they are formulated in terms of broad and contestable notions such as hate speech or incitement to crime (Langvardt, 2018: 352). Furthermore, when the objectives of legal regulations are better specified, companies are incentivized to “engage in overkill if they are focused on penalties for underenforcement” (Langvardt, 2018: 352). Either way, insofar as the enforcement of legal regulations is managed by corporations, there is considerable room for discretion in which private interests prevail.

Furthermore, there is a systemic temporal gap between digital technologies and state regulation. While technologies develop “at an ever accelerating pace,” efforts at legal regulation are almost inherently slower due to the political dynamics of negotiation and compromise among different interest groups, with the possibility of a gridlock (Marchant, 2011: 19). Furthermore, social media and other information technology companies enjoy a significant degree of corporate agility that “far exceeds that of regulators seeking to constrain their operations” as a result of their dynamic organizational model informed by the fluid and constantly evolving nature of information technologies (Flew, 2019). Consequently, state regulation is likely to lag behind newly emerging or evolving aspects of social media communication.

The second objection could be that social media companies’ quasi-public domination is nothing new. One may hold that the increasing prevalence of corporate mass media enjoyed a similar dominating power throughout the twentieth century (Habermas, 1989). In this view, uncontrolled power of social media companies is just a remanifestation of conventional corporate media.

There are two reasons why social media companies’ powers are different than those of conventional media: one more practical reason, and one more conceptual. First, the inequalities of power caused by the corporate ownership structure of conventional media are basically a product of economic inequalities of wealth and income. As the sustainability of conventional media outlets requires a substantial income stream, these companies end up being unduly influenced by sectional interests due to their financial needs, business model, and ownership structure (Pettit, 2013: 234). However, the power asymmetry in this sector is not an inherent quality of the mass media technologies. Once public resources are provided for alternative media outlets, or non-mainstream political groups pool their resources, it is possible to create counter-hegemonic alternatives that can, at least partially, compete with corporate media. The problems with conventional mass media are largely a product of prior unequal distribution of resources.

In contrast, social media companies base their power on the network effects of their industries (Larson, 2020). The value of these platforms depends on how many people use them. As a result, the industry has an inbuilt bias toward monopolization. Unlike conventional media, the vast power asymmetries on social media platforms are not merely a product of preexisting economic inequalities. This makes dominant social media platforms an even more irreplaceable element of the public sphere and puts them into a greater position of leverage against their users or other stakeholders.

Second, even if the corporate and oligopolistic ownership structure of conventional media harms democracy in serious ways, it would be conceptually careless to say that traditional media outlets dominate individuals qua citizens. Undermining democratic equality and dominating individuals qua citizens are related but not identical. For sure, media bosses and editors can exercise arbitrary power over journalists due to their economic or organizational capacities. Furthermore, this can seriously harm democracy, as economic bias in the formation of public opinion hampers equal political influence of citizens (Pettit, 2013: 169). However, one can undermine the equal influence of a citizen without directly dominating that citizen. For instance, a wealthy individual who has more free time to engage in politics would constitute a challenge to equal influence. However, this does not mean a wealthy individual dominates less wealthy citizens by simply using his or her free time to engage in politics. Similarly, conventional corporate media undermines the equal influence of citizens, but it does not directly exercise power over what citizens can say or do. Conventional media’s power to interfere with choices primarily applies to journalists. The power of social media companies is much more intimate in that they directly regulate a large number of citizens’ behavior and speech.

The third objection is that my account of domination overstates the problem, as much of uncontrolled social media power is inconsequential in comparison with the most serious examples of domination. For instance, one may suggest that, when social media companies’ content moderation targets a citizen’s political speech, domination only causes minor harm to the citizen’s political status. After all, he or she can start another user account, create similar content, and go undetected for a substantial period of time. As content moderation teams often rely on other users’ reporting and complaints, previously banned citizens can avoid interference insofar as they fly under the radar (Crawford and Gillespie, 2016).

Even if we acknowledge that the dominating power of social media companies could be inconsequential at the individual level, this does not mean that their aggregate impact is unimportant. Contemporary republican theorists’ idea of micro-domination captures this insight (Lazar, 2021; O’Shea, 2018). For example, consider Facebook’s decision to take down users’ breastfeeding photos on the grounds of the prohibition of public nudity (Gillespie, 2018: 141). This could be a minor annoyance for a single individual. However, when multiplied by thousands of cases, such an intervention is likely to boost the preexisting patriarchal norms that control women’s bodies and their social visibility. Similarly, consider social media companies’ algorithmic power to determine what discourses will be visible to whom. I already discussed how this might influence individuals’ choice menu by changing the cost of different options, for example, deliberation. A single individual getting extremely anti-deliberative might not jeopardize the future of democracy. However, to the extent that social media companies’ algorithmic preferences create similar outcomes for large groups, this is likely to generate a serious impact on what kind of public discourse democratic societies face.

Conclusion

In this article, I have offered an account of domination on social media platforms. I first showed that citizens’ unconditional access to the public sphere is an essential element of neo-republican democratic theory. I then argued that certain types of domination exercised by private actors should be conceived of as quasi-public. In the next step, I showed that social media companies’ power to regulate speech amounts to quasi-public domination of citizens. Two main mechanisms of domination were identified: direct interference with speech, through content moderation practices, and indirect algorithmic control, narrowing citizens’ choice menu about modes of discursive engagement.