Democracy and the Epistemic Problems of Political Polarization

Abstract

Political polarization is one of the most discussed challenges facing contemporary democracies and is often associated with a broader epistemic crisis. While inspiring a large literature in political science, polarization’s epistemic problems also have significance for normative democratic theory, and this study develops a new approach aimed at understanding them. In contrast to prominent accounts from political psychology—group polarization theory and cultural cognition theory—which argue that polarization leads individuals to form unreliable political beliefs, this study focuses on system-level diversity. It argues that polarization’s epistemic harms are best located in its tendency to reduce the diversity of perspectives utilized in a democratic system and in how this weakens the system’s ability to identify and address problems of public concern. Understanding such harms is also argued to require a greater consideration of the political dynamics of polarization and issues of elite discourse, alongside political psychology.

INTRODUCTION

Political polarization has become one of the most discussed challenges facing contemporary democracies and has been associated with a range of ills, from political instability to democratic backsliding (Iyengar et al. 2019). It is also often linked to a broader “epistemic crisis” of democracy involving concerns for misinformation, a breakdown of public debate, and a loss of political problem-solving (Benkler, Faris, and Roberts 2018). While producing a broad literature within political science, polarization and its associated epistemic harms have significant implications for normative democratic theory. Democratic institutions are rightfully valued for their procedural fairness but must also be able to effectively identify and address pressing problems of public concern, a point recently reemphasized by several democratic theorists (Anderson 2006; Benson 2019b; Landemore 2013; Warren 2017). This study is interested in the epistemic problems political polarization may pose to democracy and develops a new account aimed at understanding them. Stated briefly, I argue that polarization has the tendency to reduce the diversity of perspectives utilized in a democratic system and that this weakens its capacity to identify and address problems of public concern.

By locating political polarization’s epistemic harms in a system-level reduction in diversity, my account differs from previous approaches. Two prominent candidates are group polarization theory, associated with Sunstein (2000; 2018) and Talisse (2019; 2021), and cultural cognition theory associated with Kahan (2012; 2016). ¹ These accounts focus on certain psychological mechanisms involved in polarization—particularly issue and affective polarization—and their tendency to lead individual citizens to form unreliable or false political beliefs. I argue that there are two problems with these current approaches. Firstly, these psychological mechanisms are not necessarily dysfunctional in how they influence individual beliefs; secondly, the individual harms they are said to create are often mitigated by other institutions in a democratic system. These accounts therefore struggle to explain the epistemic problems polarization creates for democracy.

Given these limitations, I argue that the epistemic problems of polarization are best located in a system-level reduction in diversity, rather than in the formation of individual false beliefs. The positive argument of the paper proceeds in two steps. In the first step, I develop an epistemic model of democracy, which places emphasis on a democratic system’s ability to utilize a diversity of perspectives. Without perspectival diversity, I argue that we cannot have confidence that such a system will effectively identify and address the full range of problems of public concern. In the second step, I show how polarization tends to reduce such diversity and how this requires giving greater attention to the political dynamics of polarization, alongside political psychology. While group polarization and cultural cognition struggle to explain reductions in perspectival diversity, I combine them with a discursive account of polarization, which emphasizes the role of elite framing in aligning political identities along an “us vs them” cleavage and encouraging the adoption of a limited range of perspectives.

My central claim, then, is that the epistemic problems of polarization are best located in a system-level reduction in diversity, rather than in individual false beliefs, and that appreciating this requires greater consideration of polarization’s political dynamics. Before continuing, I wish to clarify the scope of this study. Its primary interests are normative in aiming to identify the epistemic harms that political polarization creates for democracy. In doing so, however, it also aims to identify the mechanisms involved in polarization that produce these harms. My discussion of alternative approaches is therefore in part normative (can they identify a particular epistemic harm for democracy?) and in part explanatory (can they explain how polarization produces this harm?). Importantly, the explanatory component is limited to polarization’s epistemic problems, rather than a broader account of its origins, causes, or dynamics. I therefore focus on identifying and explaining the epistemic problems polarization presents to democracy, rather than defending a larger explanatory account of polarization itself.

THE PSYCHOLOGY OF POLARIZATION

Political polarization is a multifaceted phenomenon, but three often connected forms are particularly relevant to this study. The first, which I will refer to as sorting polarization, refers to the greater alignment of social divisions and identities (McCoy, Rahman, and Somer 2018). In the United States, for instance, ideological, religious, and ethnic divisions have increasingly aligned with the partisan divide (Mason 2015). The second is issue polarization, which refers to the distance between the policy preferences of opposing partisans, and the third is affective polarization, which refers to the gap between in-group and out-group sentiments (Iyengar et al. 2019; McCarty, Poole, and Rosenthal 2016). The epistemic problems that political polarization creates for democracy may then be associated more or less with any one of these different forms of polarization.

Two prominent candidates for understanding such problems are group polarization theory (GPT) and cultural cognition theory (CCT). Both accounts focus on certain psychological mechanisms involved in issue and affective polarization at the mass level and are argued to lead individuals to form more unreliable or false political beliefs. To the extent that these distorted beliefs adversely impact voting at the ballot box and deliberation in the public sphere, political polarization can be said to reduce the epistemic quality of democracy.

GPT argues that like-minded groups tend to adopt more extreme beliefs (Sunstein 2000). It suggests that enclave deliberation will lead to a convergence on more extreme versions of a group’s initial disposition. For example, groups of conservatives are found to become less supportive of same-sex unions after discussion, and liberals are more supportive of climate change treaties (Sunstein 2018). Extremity can refer to increases in confidence (e.g., more confident that taxes should be cut by 5%) and/or a change in the content of that belief (e.g., taxes should be cut by 10% rather than 5%). This effect is often linked to political polarization due to the social segregation of partisans. Sunstein (2018), for instance, argues that the Internet has led individuals into politically homogeneous bubbles, while Talisse (2019) argues that partisans have moved geographically into politically segregated communities. ² This then produces group polarization on a large scale where opposing partisans become and perceive each other to be “extreme versions of themselves,” increasing both issue and affective polarization (Talisse 2019, 159).

Group polarization is associated with epistemic problems due to the mechanisms that produce it. Firstly, like-minded groups tend to produce a biased argumentative pool giving group members greater reasons to adopt the dominant position. Secondly, individuals face social pressures to conform and therefore express extreme versions of the dominant disposition to gain a positive reputation within the group. Thirdly, shifts may occur due to a corroboration effect where an individual increases their confidence after seeing their views confirmed and supported by others. All three mechanisms suggest that changes in citizens’ beliefs are epistemically flawed: they result from a biased set of arguments and social cues, rather than from a consideration of diverse reasons and rational persuasion. Although any one like-minded group may, by chance, polarize toward the correct view, these mechanisms suggest that the public will come to hold beliefs with higher confidence than their (subjective) evidence would suggest.

CCT argues that an individual’s assessment of risk and new information is tightly linked to their cultural identity (Kahan 2012; 2016). Individuals do not weigh the available evidence in an impartial manner, but rather through its perceived congruence or conflict with their group identities. Conservatives and liberals, for instance, will evaluate claims concerning climate change in different ways depending on whether they threaten their group membership and preferred social order. ³ CCT therefore points to mechanisms of “identity-protective cognition” where individuals resist information that contradicts their preexisting political views and too easily accept content that reaffirms them (Kahan, Jenkins-Smith, and Braman 2011; Sherman and Cohen 2006). There are two components to identity-protective cognition. The first leads individuals to accept the view of their cultural group to maintain their standing and express their commitment. This is said to be rational given that an individual’s views have little effect on public policy but a significant impact on their in-group relationships. The second leads individuals to see opposing world views as a threat to their standing and therefore to view opposing partisans as either ill-intentioned or unintelligent.

These components suggest that political topics related to prominent group identities—such as climate change or gun rights in the United States—will experience both issue and affective polarization. They also suggest that they will be epistemically dysfunctional. Citizens will tend to value their sense of identity over accuracy and will therefore form unreliable political beliefs. Some empirical studies, for instance, find that people are more likely to accept political information that is concordant with their political partisanship, to question arguments that counter their partisan views while passively accepting arguments that support them, and to resist corrections of their false partisan beliefs (Nyhan and Reifler 2010; Vegetti and Mancosu 2020). ⁴

A PROBLEM OF INDIVIDUAL BELIEFS?

Current discussions of GPT and CCT identify the epistemic problems polarization produces for democracy in its tendency to lead individual citizens to form false or unreliable beliefs, and they point to certain psychological mechanisms to explain these harms. These accounts face two important limitations. Firstly, the psychological mechanisms involved are not necessarily dysfunctional in the ways they influence individuals’ beliefs, and secondly, any epistemic harms they do create at the individual level are often mitigated by other institutions in a democratic system.

GPT, for instance, can point to certain epistemic benefits in terms of promoting new discourses and hermeneutical resources that may otherwise struggle to emerge in a democratic system. Democratic systems can exhibit biases and inequalities that mean that certain political positions or concepts become marginalized or even silenced within public discourse. Homogeneous forms of deliberation and the dynamics of group polarization can therefore work to counteract these systemic biases and inequalities. Dryzek (2012), for instance, argues that the exclusion of certain environmental groups from mainstream German politics allowed them to produce radical environmental discourses not possible within established institutions and dominant narratives around environmental politics. Similarly, Fricker (2007) argues that feminist groups were vital to forming a language of sexual harassment, which was not possible in the broader public sphere dominated by sexist norms.

In these cases, the enclaves and psychological dynamics associated with GPT do not necessarily lead to beliefs that are false or less reliable, but instead work to protect against the wider biases and inequalities that can influence democratic debate. They provide “safe spaces,” which allow for the production of political positions and concepts that may have otherwise struggled to emerge in an imperfect democratic system (Nguyen 2021). Although often absent in more recent accounts (e.g., Talisse 2019), Sunstein (2000, 111) often emphasized the potential of group polarization to produce positive contributions. The point here, however, is that group polarization effects are not necessarily dysfunctional at the level of individual beliefs, and their mere presence is therefore alone insufficient as an explanation of how political polarization produces epistemic problems for democracy.

The same can be said of CCT, as the influence of cultural identity on political beliefs may make one more, rather than less, epistemically reliable. Chambers (2018), for instance, argues that group identity may well be an effective (although imperfect) cognitive heuristic as one’s individual interests often overlap with those of a group. In a world where group identity is important and group members share similar social problems and experiences, one’s group identity is unlikely to be irrelevant to which policy or party to favor. Group identities may therefore allow citizens to more easily act on their interests, providing them with a cognitive heuristic that can be applied in a complex environment.

Lepoutre (2020) goes further in arguing that group identity may even be relevant to the assessment of factual scientific claims, as one’s epistemic standards will in part depend on the costs they associate with a false positive. If an individual comes from a rural community where jobs are scarce and likely made scarcer by increased environmental regulation, then they may reasonably desire a higher epistemic standard for claims concerning climate change than someone who is unaware or unaffected by these costs. At least in principle, then, assessing even scientific claims in relation to the normative concerns linked to one’s group identity is not necessarily an epistemic failure. ⁵ CCT and GPT are not therefore sufficient to explain how polarization produces epistemic problems for democracy at the level of individual beliefs. In fact, these mechanisms may produce epistemic benefits at this level.

To the extent that GPT and CCT do negatively influence individual beliefs, this may also not adversely affect the epistemic quality of a democratic system. A substantial amount of empirical research has long found that citizens generally hold unreliable and incorrect beliefs about politics, and this is true in both polarized and non-polarized conditions (Somin 2016). It is therefore unclear whether these accounts identify epistemic problems specific to political polarization. More importantly, because it is generally the case that citizens possess limited amounts of political knowledge, democracies tend to possess institutional divisions of labor that deal with this fact. As recent work in democratic theory has emphasized, citizen beliefs do not directly inform public policy, but are rather mediated by a range of institutions, which clarify and filter their concerns and allow for the input of more sophisticated political knowledge (Benson 2019b; Christiano 2012; Elliott 2020).

These mediating actors include elected representatives, political parties, civil servants, and expert committees, all of whom play a role in laundering the views of the public and constructing policy packages based on more reliable and specialized knowledge. They also include institutions within what Habermas (2015, 367) calls the “organizational substratum” of the public sphere, such as social movements, charities and nongovernmental organizations, labor unions, and media organizations. Often specializing in particular policy issues, these institutions function to filter and transmit the concerns and views of the public in ways that allow them to influence more informed public policy. While always imperfect and unable to completely eradicate potential problems, it is these mediating institutions that create a division of epistemic labor and allow a democratic system to function effectively even with low levels of voter knowledge, a feature common to democracies in polarized and non-polarized times. The psychological mechanisms highlighted by both CCT and GPT are not therefore necessarily dysfunctional in how they influence individual-level beliefs, and to the extent that they are, this is often offset by other institutions in a democratic system.

THE EPISTEMIC FUNCTIONS OF A DEMOCRATIC SYSTEM

In their current form, neither GPT nor CCT satisfactorily identify and explain the epistemic problems polarization produces for democracy. I therefore aim to offer an alternative account that focuses not on individual-level beliefs, but on system-level diversity. While GPT and CCT can contribute to our understanding of this problem, this is only possible if their psychological mechanisms are combined with a broader account of the political dynamics of polarization. First, however, we need a better understanding of the epistemic functions of a democratic system and why diversity is important to them.

In broad terms, the epistemic task of a democratic system is to address pressing problems of public concern (Anderson 2006). Following Habermas (2015, 307), this task can be separated into two connected functions: justification and discovery. The first refers to the process of making politically binding decisions and involves “justifying the selection of a problem and the choice among competing proposals for solving it” (Habermas 2015). A democratic system needs to determine which social problems should be prioritized given the constraints of time and resources and then which of the available solutions is best placed to ameliorate them. The function of justification, however, is not aimed at “discovering and identifying problems,” nor at developing a sensitivity “to new ways of looking at problems” (Habermas 2015). These tasks instead form part of the function of discovery. This second function is concerned with detecting the many problems facing citizens, identifying their implications for the public interest, and producing and communicating their different interpretations and potential resolutions. The function of justification is therefore always dependent on an effective function of discovery, which identifies the problems that may require regulation through the state and produces the competing proposals to be considered. There will then often be feedback dynamics where processes of justification and decision-making make necessary further processes of discovery.

A range of institutions in a democratic system is necessary to perform these two functions, and some may play a role in both. Habermas (2015, 300) suggested a division of labor between formal political institutions tasked with justifying binding decisions through structured deliberation and a more unstructured public sphere acting as “a far-flung network of sensors that react to the pressures of society-wide problems and stimulate influential opinions.” However, we do not need to endorse such a strict division of labor, nor Habermas’ broader deliberative theory, to recognize the importance of these two functions.

While formal institutions must play a central role in political decision-making, many parts of a democratic system can contribute to processes of justification, including the giving of reasons by supporters and detractors of government policy in the public sphere. Likewise, institutions such as social movements, labor unions, and community associations play a vital role in discovering the problems facing citizens, but formal political institutions can also help detect and understand these problems through official statistical bodies, parliamentary committees, citizen consultation processes, and other initiatives. Similarly, while these two functions are taken from a deliberative theory of democracy, many conceptions of democratic politics will see at least part of its value as coming from its ability to solve problems of public concern. To this extent, they will also require some process of justifying the selection of problems and the adoption of available solutions and some process of discovery that puts these problems and solutions on the table.

What, however, does a democratic system need to discover? I wish to differentiate three main and necessarily connected tasks relevant to the function of discovery. The first is to identify problems of public concern. Social problems and their consequences are not readily apparent to any political authority but must first be discovered through a communicative process. This is why the inclusiveness of a democratic system is not only important for procedural reasons, but also because it allows all those affected by social issues to raise their concerns. Once problem situations are raised, they also need to be understood in terms of whether they amount to issues of public concern, which require regulation through the state. Its possible implications and impacts for members of society therefore need to be uncovered and understood. This then gives rise to two additional tasks: the discovery of information relevant to social problems and the production and engagement with reasons relevant to their conceptualization and resolution.

Relevant information includes what Anderson (2006, 14) calls “situated knowledge,” which derives from citizens’ “experiences of problems and policies of public interest.” Such knowledge is asymmetrically distributed across society and therefore requires an open and inclusive communicative process to uncover. While such knowledge is possessed by citizens, to gain traction in a democratic system, it often requires uptake by institutional actors, such as the media, NGOs, and political parties, who have the means to aggregate and propagate such insights on a larger scale (Benson 2019b, 431–6). Relevant information also includes scientific knowledge that enters the democratic system through organizations such as universities and scientific bodies, as well as official scientific committees and government advisers. When scientific knowledge enters political discourse, however, it will require a broader range of institutions that interpret its implications with respect to normative political judgments.

As this last point suggests, for new information to aid in the understanding and resolution of social problems it requires that it is considered and formulated into reasons. While often making use of pieces of information, political reasons combine them with different values, principles, or interests and can therefore motivate choices over the prioritization of social problems and the selection of potential solutions. A democratic system therefore needs to produce a range of competing reasons relevant to the interpretation and evaluation of problems and their potential resolutions, as well as encourage an engagement with these reasons so they can then be taken up in the process of justification and inform political decision-making. Collections of reasons (and associated facts, values, and principles) may then also be packaged together to form broader political positions or platforms, which refer to many social problems and place emphasis on certain kinds of resolutions.

THE IMPORTANCE OF DIVERSITY

For a democratic system to effectively perform its functions of justification and discovery, certain conditions need to be in place. These include procedural requirements, such as freedom of speech and association, which allow different actors to speak and be heard. Another is perspectival diversity, which I define as the ability of a democratic system to utilize a diversity of perspectives.

Perspectives are particular points of view that emerge from the range of social positions and experiences in a society and influence the ways individuals interpret politics (Bohman 2003; 2006; Young 2002). They emerge from the alternative social roles people inhabit and are therefore shared by some but not all. Given their varied experiences, individuals can possess multiple, and sometimes conflicting, perspectives. Perspectives are not reasons or opinions, but rather practical points of view, which form the background against which reasons are recognized, evaluated, and produced. Reasons are therefore items that are considered in deliberation, while perspectives “inform these reasons and give them their ‘cogency’” (Bohman 2006, 179). Having a different perspective also differs from simply possessing different facts about the world, and they are therefore distinct from Anderson’s situated knowledge. ⁶ Rather, perspectives refer to the background against which people interpret information and recognize it as important. To the extent they involve knowledge, then, it is practical or tacit knowledge, which is exercised through activity (Benson 2019a). People with differing perspectives will therefore interpret politics in different ways, produce different considerations, and see different considerations as important.

Perspectives are also distinct from identities, although closely related. To belong to and identify with a social group is, in part, to hold a certain position in a social structure and to have associated social roles and experiences (Young 2002). Perspectives are similarly social in the sense that they represent practical points of view informed by one’s experiences and roles, which are commonly shared with others in one’s group. Both one’s perspectives and group identification will therefore be formed with respect to often-shared experiences. To offer an (overly) simplified example, a person’s experience growing up in a working-class community can provide them with both a working-class perspective and a working-class identity. Someone’s identity can therefore influence how they evaluate politics, by its connection to perspectives. A perspective does not necessarily come with a strong group attachment, however, and this is important as it allows perspectives to be shared.

Although those directly occupying a social position will likely see best through its associated perspective, this perspective can still be taken up by others to the extent that they understand the social role and related experiences. It is therefore possible for economically privileged individuals to attempt to take up a working-class perspective, or for men to take up a women’s perspective, although this requires a genuine process of learning and reflection on the social position of others. ⁷ While there are good reasons to believe that those inhabiting these social positions will tend to have a better understanding on which to form the relevant perspective, this does not stop others from attempting to take up this perspective based on their (likely more limited) understanding. Without this possibility, the scope for dialog and understanding across perspectives would be greatly constrained. ⁸

The importance of perspectival diversity has been recognized by epistemic democrats and most powerfully defended by Landemore (2013). ⁹ Using the diversity trumps ability theorem (DTA), Landemore argues that the diversity of a group is central to its problem-solving ability and that diverse groups even outperform high-ability groups. ¹⁰ The logic behind the DTA is that high-ability groups think about problems in similar ways and therefore quickly arrive at their highest common local optima. Diverse groups, alternatively, think about problems in very different ways, recognize different kinds of solutions, and can therefore guide each other past their local optima to the global optima. Landemore’s claim that diversity reliably trumps ability has been criticized based on the DTA’s demanding assumptions (Ancell 2017; Brennan 2016). However, my aim here is only to show that diversity is important to democracy’s epistemic quality, not that it outweighs all other variables or makes democracy superior to all other political systems, and the underlying logic of Landemore’s account has been shown to demonstrate diversity’s benefits even under less idealized conditions (Benson 2021).

Even under the best conditions, however, Landemore’s approach offers an incomplete picture of the benefits of perspectival diversity. This is because it primarily applies to the function of justification and to formal and structured sites of deliberation. Landemore (2014, 186–7) uses the DTA to model face-to-face deliberation where individuals can engage in collective problem-solving and assumes that the problem to be solved has already been identified and that relevant knowledge is given in advance. While a useful account of diversities benefits to the function of justification, it therefore leaves out the role of the wider democratic system in addressing problems of public concern and takes for granted that an effective process of discovery has already occurred. Perspectival diversity, however, is also relevant to the function of discovery and can be distributed across a democratic system. Following the discussion above, the discovery function is effectively performed when a democratic system can uncover pressing problems of public concern and produce a broad range of information and reasons.

In terms of problem identification, a situation is only ever perceived as problematic through the lens of one’s perspective. Social problems and the interests of citizens they affect are not given and awaiting representation, but must instead be uncovered and interpreted in a discursive process (Saward 2006). Above bare necessity, what constitutes an individual’s needs is open to public contestation and interpretation and requires an exchange of views to understand (Fraser 1989). Determining which situations constitute a problem requiring regulation through the state therefore similarly needs an exchange of views (Habermas 1985). Understanding whether a problem is of public concern requires the highlighting of alternative need interpretations and the alternative problem perceptions these interpretations create. Given that citizens are differently situated, however, not all will recognize the same problem situations, produce the same need interpretations, or assign these the same level of importance. Having a particular perspective based on one’s position is to be attuned to certain issues and problems, and not others.

Consider again Fricker’s example of sexual harassment. In a public sphere dominated by male perspectives, certain workplace behaviors were not interpreted as problem situations, or to the extent that they were, they were seen as private issues and not therefore problems of public concern. The dominant experiences and social positions that form the male perspective were not attuned to identifying the important interests at stake. It was the greater inclusion of women’s perspectives that then allowed such behaviors to be more fully understood and discussed as potential issues for state regulation. The campaign groups and organizations associated with the feminist movement allowed an alternative set of experiences and social roles to be voiced, and this allowed for new and competing interpretations of women’s workplace needs. As Mansbridge (1999, 647–8) and Young (2002, 140) point out, it was also usually female representatives who saw such issues as important and raised them in formal legislative debates. Similarly, it was the limiting of earlier forms of feminism to white women’s perspectives that restricted early understandings of what counted as harassment, as it again limited the range of experiences and positions through which it could be interpreted. It was therefore the greater inclusion of a range of differing perspectives that allowed this social problem to be better understood over time and to be raised in the democratic system as an issue of public concern.

Having a certain perspective is to view politics through a lens that will be partial with respect to certain interests, some coming easily into focus and others falling outside the field of vision. For a democratic system to have the greatest possibility of identifying the full range of problems confronting citizens, then, it needs to utilize a diversity of perspectives. It is only through having issues viewed through a diversity of social positions and experiences that we can have confidence that the system will not overlook important problem situations or how they impact differently situated citizens. A lack of perspectival diversity, alternatively, risks obscuring pressing problems that cannot easily be identified from certain points of view, just as the problem of sexual harassment was not easily seen from the male perspective.

To aid the function of justification, a democratic system should also aim at producing a broad range of relevant information and reasons, so that political decisions can be informed by all important considerations. We may therefore think that we should directly aim at the diversity of such items, as John Stuart Mill did in aiming at the diversity of opinions. It is perspectives, however, which are more fundamental as they represent the “experiential source” of opinions, information, and reasons (Bohman 2007, 350). A diversity of these items is therefore only achieved through the inclusion and utilization of a diversity of perspectives, which then leads individuals to produce alternative considerations. Anderson’s (2006, 14) experimental model of democracy recognized this with respect to information, arguing that including citizens from different walks of life allows for the maximal inclusion of situated knowledge. Perspectival diversity is not simply about including specific pieces of information, however, but about having political issues evaluated and assessed from alternative social positions. This allows a democratic system to produce a broad range of considerations—knowledge but also reasons and opinions—and allows these considerations to receive attention and uptake.

While perspectives influence the facts and reasons people produce, the mere presence of these items will not lead to their inclusion in political decision-making and processes of justification. It is perspectives, however, which also influence how such items are assigned with importance by others, and so form the background against which they can achieve uptake in a democratic system. Perspectival diversity is therefore required within the institutional structure of the democratic system (e.g., in media, NGOs, and political parties) so that a broad range of considerations will be recognized as important and amplified at scale. Without these institutional actors recognizing their value, information and reasons will not receive attention in political decision-making. For a variety of such items to not only be produced but also receive attention requires that a diversity of perspectives be utilized within the citizenry and the institutional structure of a democratic system.

While the work of epistemic democrats such as Landemore therefore provides resources for understanding how perspectival diversity aids the function of justification, it is also important to show how a democratic system performs the function of discovery. To clarify, this argument need not claim that all perspectives are equally valuable for all social problems. However, we cannot tell in advance which perspectives will be most useful and it is only through an open process where all perspectives are utilized that we can discover their respective values. This argument also does not entail that the truth will always emerge from an unregulated marketplace of ideas. A democratic system will produce many unorganized and erroneous considerations during the process of discovery, and this is one reason why they require many mediating institutions that clarify and filter these considerations for political decision-making. For a democratic system to effectively perform the functions of justification and discovery, however, requires perspectival diversity.

THE POLITICS OF POLARIZATION

The epistemic problems of political polarization are best identified in a system-level reduction in perspectival diversity, rather than in the formation of false or unreliable individual beliefs. Defending this claim, however, requires an understanding of the kind of polarization relevant to perspectival diversity and the associated mechanisms that explain how it comes to reduce it.

The form of polarization important to such a reduction is not issue or affective polarization as much as sorting polarization, which refers to the alignment of social divisions and identities. In this form, polarization involves the increasing alignment of societal divides along a single elevated cleavage. While politics always involves differences and disagreements, sorting polarization occurs when a society’s multiple and cross-cutting “differences become aligned within (normally two) camps with mutually exclusive identities and interests” (McCoy, Rahman, and Somer 2018, 18). While the nature of these camps (religious, cultural, ethnic, economic, ideological, etc.) will differ depending on the context, sorting polarization is defined as the coming together of divisions under an elevated unitary axis often defined in “us vs them” terms. This could be “Republicans vs Democrats” in the United States, “Chavistas vs anti-Chavistas” in Venezuela, or “nationalists vs cosmopolitans” in Hungary. ¹¹

This alignment can produce increased issue polarization, but only to the extent that issue positions are salient to the elevated social divide. In the United States, for instance, ideological, religious, and ethnic divisions have increasingly aligned with the elevated partisan divide, but this has not produced significant issue polarization at the mass level, apart from certain particularly salient issues (Iyengar, Sood, and Lelkes 2012). Affective polarization, alternatively, is more tightly connected as the alignment of social divisions tends to promote a zero-sum perception of politics and intensifies in-group identification and intergroup rivalry (Mason 2015). While often connected to other forms of polarization, it is the alignment of social identities, rather than changing issue positions or affect, which can impact perspective-taking. Understanding how sorting polarization can reduce perspectival diversity, however, requires going beyond GPT and CCT.

While the psychological mechanisms in these accounts have been connected to issue and affective polarization, neither provides a complete account of sorting polarization nor its impact on perspectival diversity. The reason is that while these mechanisms can help crystallize perspectives, they will only do this around identities which are already politically salient in a democratic system. Neither, however, provides an account of why certain identities and perspectives become salient in the first place. CCT, for instance, suggests that people evaluate risk and information with respect to their cultural identities and this can help the formation of perspectives, allowing people’s experiences and social positions to influence how they view politics. The perspectives being formed, however, are only those associated with the identities people already find salient when evaluating political issues. Similarly, if a group shares a common identity and this identity is salient in their political discussions, then group polarization may help to clarify and intensify the associated group perspective. The perspectives formed, however, are again those associated with already salient identities.

Both GPT and CCT therefore simply tend to reflect the identities and perspectives already salient in a democratic system, but neither provides an account of how certain identities and perspectives become salient in the first place. As a result, their impact on perspectival diversity will depend on the nature of the identities and perspectives already prominent in a society. If a variety of identities are politically salient in a democratic system, for instance, then cultural cognition and group polarization may help to crystallize and intensify an associated variety of perspectives, but they will not necessarily lead to a reduction in the diversity of perspectives utilized. In fact, if this involves marginalized groups coming together to voice their otherwise overlooked viewpoint, then they may even help increase perspectival diversity at the system level. On their own then, GPT and CCT are ambivalent with respect to perspectival diversity. What we need is to identify some supplementary mechanisms involved in sorting polarization, which can affect the kind and variety of perspectives that become salient. Doing this, however, requires a greater consideration of the political dynamics of polarization and particularly its discursive dimension.

Comparative studies of sorting polarization show that it is often, in part, a discursive process initiated by political entrepreneurs who aim to activate key societal cleavages for political advantage (Carothers and O’Donohue 2019; McCoy and Somer 2019; Somer 2005). ¹² These actors use polarizing speech and divisive rhetoric to create a perception of politics as based on an “us vs them” conflict and, in doing so, create or undermine political coalitions. Polarizing speech works to elevate and highlight a particular social division and define the relevant in-group and out-groups around certain social identities (partisan, religious, ethnic, etc.). The result is that these groups come to be perceived as mutually exclusive, incompatible, and irreconcilable. Rhetoric such as “Mexicans are rapists” or “Democrats are communists,” for instance, demonizes opposing partisans in ethnic and ideological terms, while claims that “the media is the enemy of the people” undermine the idea of legitimate opposition. Alternatively, statements such as “we are the real Americans” and “make America great again” create an exclusive conception of the legitimate demos defined with respect to a more ethnically and religiously homogeneous past. The result is the alignment of social divisions within mutually exclusive and homogeneous camps and a perception that politics amounts to a “us vs them” conflict.

While political actors may always engage in polarizing speech, it can be incentivized by structural conditions. In the United States, for instance, polarization has been disproportionately driven by the Republican Party, which has taken more extreme issue positions, is less willing to compromise, and engages in greater obstruction and democratic norm-breaking (Hacker and Pierson 2015). The same goes for right-wing media outlets that produce more explicitly partisan content (Benkler, Faris, and Roberts 2018). While political scientists disagree over their relative importance, two factors are often highlighted to explain this increasingly polarizing behavior. These are demographic changes and increasing economic inequality, which both create electoral challenges for Republicans (Grossmann and Hopkins 2016; Hacker and Pierson 2020; McCarty, Poole, and Rosenthal 2016).

Changing demographics present a direct electoral challenge as the relatively homogeneous Republican base is representing a shrinking proportion of the electorate. High levels of inequality, alternatively, create challenges as they cause the economic interests of the general population to increasingly diverge from those of traditional Republican donors who are disproportionately wealthy—a divergence less significant for Democrats who receive less support from this group (Hacker and Pierson 2020). A polarizing discourse that intensifies in-group identity and fosters hostility toward a culturally distinct and threatening out-group is then a strategy for dealing with these structural challenges. It looks to consolidate a homogenous voting bloc in the face of a diversifying public and motivate popular support despite an economic platform that favors an increasingly wealthy minority, by promoting cultural and identity-based issues. ¹³ These structural incentives therefore help explain the increase in polarizing speech over time and are likely one reason why Republican voters have polarized more than Democratic voters (Hacker and Pierson 2015).

Sorting polarization can therefore be understood as involving a discursive process where political actors play an important role in aligning social divisions and creating a politics based on a “us vs them” conflict. While this discursive view is not incompatible with other accounts of polarization and need not be the core driver, it can help to explain how it reduces perspectival diversity. The next section will therefore explore the effect of polarizing discourse on perspectival diversity and how GPT and CCT can be combined with these broader political dynamics.

THE LIMITING OF PERSPECTIVAL DIVERSITY

Once understood, in part at least, as a discourse-driven process where political actors use polarizing speech to create a certain discursive environment, sorting polarization can be seen to involve a strong kind of framing. Framing effects involve a “speaker’s emphasis on a subset of potentially relevant considerations causing individuals to focus on these considerations when constructing their opinions” (Druckman 2001, 1042). By defining the essential problems and factors in any political issue, they (re)organize cognition by altering the weight that individuals give to different considerations, be they facts, values, or reasons (Nelson, Oxley, and Clawson 1997). This works by making certain considerations available in memory, making these accessible at a given time, and making them applicable or central to someone’s evaluation (Chong and Druckman 2007). Based on the inferences that can be drawn from the highlighted considerations, frames also tend to imply certain judgments regarding an issue, such as who is to blame or what is a reasonable solution.

Of course, framing is an unavoidable and not necessarily problematic part of democratic discourse. The effect of frames is not permanent or inaccessible for citizens, and they can form part of a learning process helping citizens organize considerations and views on complex social problems (Nelson and Kinder 1996). In fact, without framing citizens would struggle to form an understanding of an issue or share a common set of references for discussion with others. Political elites therefore play an important role in framing issues and helping to organize the concerns and views of the public (Chambers 2017).

Frames and perspectives are easily confused, with the terms sometimes being used as synonyms for one another (Tversky and Kahneman 1985, 435). As defined here, however, the two concepts are distinct. Discussions of elite framing normally focus on “frames in communication,” which refer to a speaker’s emphasis on certain dimensions of an issue (Chong and Druckman 2007, 105–6). To say that energy policy is “an economic issue,” for instance, is to invoke an economic frame in communication, rather than an environmental one. Frames in communication can then influence a “frame in thought,” which refers to the different dimensions of an issue an individual considers in forming an opinion and the relative weights they assign them (Chong and Druckman 2007). If a person believes that environmental and economic concerns dominate decisions about energy policy, then their frame in thought for this issue includes these two dimensions and their respective weights. Frames in communication therefore refer to the emphasis in speech acts, while frames in thought refer to a set of dimensions that people consider with respect to an issue.

Perspectives, alternatively, refer to practical points of view from which individuals form frames in thought and evaluate frames in communication. To take a certain perspective is to view politics from a certain social position, and as we have seen, certain considerations will appear more relevant and important from this point of view. Those taking a certain perspective will therefore tend to produce a particular frame in thought when confronted with an issue. If someone views the issue of sexual harassment from a woman’s perspective, then certain dimensions of this issue will appear more apparent or weighty, and a certain frame in thought will form. Perspectives are therefore the background position or general orientation from which individuals form their frames in thought with respect to different issues and are one reason differently situated people come to alternative frames in thought. Frames in communication, alternatively, refer to the emphasis placed on certain dimensions in speech and are therefore items offered up and considered in deliberation. Perspectives then represent the social positions from which these items are considered. Empirical work shows that people can reflect on and evaluate competing frames in communication, finding some more appealing than others (Sniderman and Theriault 2004). Perspectives are the general position or orientation from which this is done and are one reason why different people tend to find alternative frames in communication appealing.

Frames in communication (henceforth simply frames) have a two-way relationship with perspectives. While certain frames may appear more appealing from certain perspectives, certain perspectives may also be activated by certain frames. Although work on framing effects has mostly focused on the highlighting of facts, reasons, or values, I wish to suggest that elite framing can also prioritize perspectives. Given citizens’ varied experiences and roles, an individual can possess multiple perspectives and may utilize different perspectives at different times. In the terminology of framing theory, they have multiple perspectives available, but not all will be accessed and applied when considering an issue. Framing effects can then influence which perspective is utilized, by placing emphasis on certain social positions or sets of experiences, making a certain perspective more easily accessible and applicable. The most direct way of doing this is to place emphasis on certain identities, making that social position particularly salient with respect to an issue and therefore activating the associated perspective. Saying that an increase in benefit payments is a “class issue,” for instance, can place emphasis on a working-class identity and encourage the use of a working-class perspective, while saying it is an “act of Christian charity” may emphasize a Christian identity and associated Christian perspective. While individuals may possess both perspectives, these alternative frames make certain identities and social positions more salient, encouraging an audience to consider politics from this point of view.

In this case, the frame makes accessible and applicable a perspective that is already available to the audience. Given that perspectives can be shared, it is also possible to make available perspectives that an audience did not already possess. As noted earlier, taking up a new perspective involves coming to a significant understanding of an alternative social position and associated experiences, and it is therefore a much more difficult task, requiring more intensive communication. In most instances, then, elite framing is most likely to target perspectives already available to an audience, possibly altering how they are applied and understood. What is important, however, is that, unlike GPT and CCT, the mechanisms involved in elite framing can affect the kinds of perspectives utilized by individuals in a democratic system.

While the impact of frames on perspectives is not necessarily problematic (as argued below), the problem with polarizing speech is that it promotes “polarized” frames and these polarized frames become increasingly “dominant” over an ever-larger number of social problems as polarization worsens. ¹⁴ A polarized frame is one that constructs political issues in an oppositional way where the opposition is mutually exclusive and threatening, while a dominant frame is one that excludes alternative ways of viewing an issue. ¹⁵ Sorting polarization can therefore be understood, at least in part, as involving the increasing dominance of a polarized frame. Political actors activate certain social cleavages and in doing so increasingly frame political issues in terms of a single “us vs them” conflict between homogeneous and mutually exclusive groups. The result of such a framing is the elevation of these two social identities as central to politics and therefore the promotion of their associated perspectives.

By framing political issues in terms of an “us vs them” struggle, political actors highlight these two social identities, making accessible and applicable their associated perspectives. Put simply, if politics is represented as a Manichean struggle between an “us” and a “them,” the implication is that everyone must fall into one of these two groups and see politics from their point of view. Those already possessing the relevant identities and perspectives will then be more likely to associate with one side rather than the other. The result is that the perspectives of these groups—whether it be Republican and Democratic perspectives in the United States, Chavista and anti-Chavista in Venezuela, or nationalist and cosmopolitan in Hungary—will increasingly come to represent the only relevant or important perspectives through which to approach politics. Polarized frames therefore work to limit the diversity of perspectives utilized at both the individual and collective levels, as they increasingly establish the two sides of an “us vs them” conflict as the predominant social positions from which to view politics.

There are at least two ways this kind of elite framing reduces perspectival diversity. The first and simplest is that the dominant polarized frame crowds out or pushes out perspectives that do not easily fit with its chosen dimension of conflict. As one social divide gets elevated over others, perspectives directly linked to this divide are promoted, and those not easily connected are marginalized. That is, certain perspectives are made more accessible and applicable, while others are not. Empirical work in the United States, for instance, has found that elite party polarization can lead individuals to prioritize partisan endorsements when evaluating policies, over frames which highlight substantive policy-relevant arguments or reasons (Druckman, Peterson, and Slothuus 2013). Under non-polarized conditions, individuals tend to follow strong arguments over party endorsements, but under polarization they follow the party cue over the argument. One interpretation of these findings is that polarization leads individuals to prioritize the perspective of a Republican or Democrat, and they therefore come to focus on those considerations most important to these partisan perspectives (e.g., party endorsements). This prioritization then comes at the expense of other perspectives, which would have been attentive to other considerations and reasons. Polarized framings therefore promote these partisan perspectives and push out those that do not fit the frame of Republican vs Democrat conflict.

The second way is that the dominant polarized frame works to align perspectives, reducing the variation between them. While multiple identities may be emphasized in polarizing speech, and multiple perspectives made accessible and applicable, this is done in ways that emphasize their connections and therefore create a perception that they are overlapping and mutually reinforcing. The dominant polarized frame does not therefore only elevate the two groups in an “us vs them” conflict but also aligns other identities within these two camps so that each group appears mutually exclusive and internally homogeneous. The effect is that while multiple identities may become salient, they are felt to only reinforce the dominant cleavage. Empirical work, for instance, finds that individuals possessing aligned social identities tend to report stronger partisan identification (Mason and Wronski 2018). White, Christian conservatives therefore identify more strongly as Republicans, while non-white, secular liberals identify more strongly as Democrats.

Mason (2018) has elsewhere referred to this alignment as the formation of “mega-identities,” but the discussion here allows us to also see it as the formation of mega-perspectives. As polarized framings align the perspectives of different social positions, group memberships and experiences that may have previously provided distinct and differing points of view, increasingly converge into a single lens. For example, while Republican and Christian perspectives may have previously been attentive to different social problems and recognized different information and reasons as important, polarization brings these perspectives into greater alignment so that they come to offer a more unified point of view. The Republican and Christian perspectives therefore start to focus on similar issues and produce similar interpretations and reasons. By emphasizing and connecting identities then, polarized frames work to reduce the variation between the most politically salient perspectives. As the Republican perspective becomes whiter, more conservative, and more Christian, the diversity between these perspectives is undermined. The dominant polarized frame therefore reduces diversity not only because certain perspectives are pushed out or marginalized—leading to fewer perspectives being utilized—but also because previously varied and conflicting lenses come to converge within more homogeneous mega-perspectives.

An account of elite framing can therefore point to mechanisms involved in sorting polarization, which help explain how it reduces perspectival diversity. By pushing out certain perspectives and aligning others within more homogeneous mega-perspectives, polarizing frames reduce the range of perspectives utilized at the system level. This is something GPT and CCT struggle to account for, given that they simply tend to reflect the identities and perspectives already salient in the system. This tendency can, however, allow group polarization and cultural cognition to contribute to reductions in perspectival diversity once they are combined with an account of elite framing.

We have seen that CCT may aid in the formation of perspectives as it allows one’s identity to influence their evaluations of politics, but it only does this for identities people already find politically salient. While cultural cognition will not therefore necessarily reduce perspectival diversity in a society with multiple and cross-cutting salient identities, it will help to reduce diversity once a more limited set of mega-identities have started to become prominent. If cultural cognition is therefore combined with a broader political context where elite actors engage in polarized framing and elevate a certain “us vs them” conflict, then the mechanisms of CCT will tend to promote the further formation and crystallization of the associated mega-perspectives. So, while CCT is ambivalent with respect to perspectival diversity on its own, it can be understood to contribute to reducing diversity when supplemented with an account of elite framing and the discursive dynamics of polarization.

Similarly, group polarization can help form and intensify perspectives around the identities already shared and salient within group discussion. While this is ambivalent with respect to diversity in isolation, in a context where elite framing elevates certain mega-identities, group polarization will tend to further crystallize and intensify the associated homogeneous perspectives. Of course, this will only occur if those who share the relevant mega-perspectives also tend to engage in homogeneous deliberation with each other. The kinds of social segregation pointed to by group polarization accounts will therefore likely have an influence on the extent to which polarizing framing reduces diversity.

While polarized frames reduce diversity by making a more limited range of perspectives accessible and applicable, this will be exacerbated if the associated identity groups are also socially segregated and tend to engage in enclave deliberation. Under these conditions, elite framing will combine with group polarization effects, further intensifying a limited range of mega-perspectives. This is likely the case in the United States, which exhibits high levels of social segregation (Mummolo and Nall 2017). Conversely, low levels of social segregation are likely to make polarized frames less effective as people will more often encounter the perspectives of others during political discussion, and diverse deliberation may limit the influence of elite framing (Chong and Druckman 2007). The levels of social segregation and enclave deliberation in a democratic system will therefore likely be important to the potential success of elite polarizing speech.

THE EPISTEMIC PROBLEMS OF POLARIZATION

I have identified the epistemic problems of political polarization in a system-level reduction in perspectival diversity and introduced an account of elite framing to better explain how this reduction comes about. As we saw earlier, perspectival diversity is important to the effective performance of the functions of justification and discovery and therefore the ability of democratic systems to address pressing problems of public concern. It allows for the emergence and understanding of problems and considerations that may be overlooked from any one point of view. In a society undergoing sorting polarization, however, perspectival diversity is compromised as individuals and the institutions they inhabit increasingly view politics from two dominant mega-perspectives. The result is a limiting of the problems identified, and the information and reasons produced and engaged with, to those seen as important from these more homogeneous perspectives. A narrower range of problems will therefore come to be discussed in a democratic system, and those that receive attention will be considered in more limited and binary ways.

Two popular complaints about polarized politics can be recognized in this account. Firstly, it is common to complain that polarization leads to certain problems going unaddressed in favor of a deepening focus on certain divisive issues, such as gun rights and abortion in the United States. If polarization involves the increasing dominance of two perspectives, however, then this is to be expected. Those issues seen as important from these perspectives will receive increased attention in the democratic system at the expense of otherwise pressing issues that may have been recognized by alternative points of view. Secondly, it is also common to complain that issues that are considered are increasingly channeled through polarized conflicts. In the United States, for instance, even highly technical issues such as environmental regulation or the COVID-19 pandemic are increasingly treated as just another partisan debate. This complaint can be understood as a limiting of the kinds of reasons and considerations allowed to inform political decisions. This is again to be expected on the account presented here, as those reasons seen as important to the dominant perspectives become prioritized in political discussion. Of course, polarization is a process and not all societies will be at the extreme. The reduction in perspectival diversity will worsen as sorting polarization progresses, however, so that more polarized societies will experience more of these epistemic problems, ceteris paribus.

It is worth reemphasizing how this view differs from previous accounts, which focused on the formation of false or unreliable individual beliefs, rather than system diversity. While individuals may come to false beliefs while viewing politics from a dominant perspective, this is not required for there to be an epistemic problem on this paper’s account. Similarly, it is not necessarily problematic for any one individual to take on the perspectives highlighted in any polarized conflict, as such perspectives may call attention to important problems and considerations. Instead, epistemic harms only occur when these perspectives become increasingly pervasive so that the diversity of perspectives utilized in the system as a whole is reduced. While any individual distortions may also be mitigated by the broader system, it is systemic distortions that work to undermine the general functions of discovery and justification. My account therefore connects the individual-level impacts of framing effects, GPT, and CCT to a system-level reduction in the range of perspectives utilized, and it is the latter that undermines democracy’s epistemic quality.

Another problem with the previous accounts was that GPT and CCT may provide epistemic benefits in terms of individual beliefs, but something similar may be thought true of elite framing and diversity. Political actors could use frames to help crystallize relevant and possibly marginalized perspectives, encouraging their audience to view issues from alternative points of view and promoting greater perspectival diversity. My account has not taken issue with all forms of elite framing, however, but with the increasing use of polarized frames. Some kinds of elite framing may therefore encourage a variety of perspective-taking, and these can be epistemically productive for democracy. The problem is that polarized frames do the opposite. They look to limit the perspectives utilized to those in a polarized “us vs them” conflict, and while they may help in the formation of these mega-perspectives and the crystallization of their associated concerns, the overall result is a reduction in system-level diversity and therefore democracy’s epistemic quality.

CONCLUSION

This paper has argued that the root epistemic problem that political polarization creates for democracy is best located in a system-level reduction in perspectival diversity, rather than in individual false beliefs, and that appreciating this fact requires an understanding of polarization’s political dynamics. This account’s unique combination of the discursive dimension of polarization and its tendency to undermine diversity also has broader significance for the field.

Firstly, while previous work in democratic theory has focused on psychological accounts of polarization, this paper has analyzed the interaction between psychology and broader political dynamics. Psychology plays an essential role, but polarization is also a political process involving the discursive strategies of political actors, and this will influence how it impacts democratic values. Democratic theorists must therefore incorporate such dynamics if they are to understand the harms of polarization and the best responses to it. While measures aimed at the mechanisms of group polarization and cultural cognition (e.g., increasing intergroup deliberation) remain relevant, this paper suggests that they likely need to be combined with policies targeting the polarizing speech of elite actors. This may include forms of counter-speech and framing, but also measures to address broader structural conditions (e.g., economic inequality), which incentivize the use of polarizing speech.

Secondly, this paper suggests that the recent epistemic turn in democratic theory also needs to pay greater attention to broader political dynamics. Much work in this field focuses on the relationship between democratic inclusion and diversity, rejecting nondemocratic systems based on the epistemic costs of exclusion (e.g., Landemore 2013). However, my discussion of polarization shows that even in a democratic system with a universal franchise, democracy may not achieve the epistemic benefits of diversity if political actors engage in certain discursive strategies. Epistemic democrats should therefore give more consideration to these discursive behaviors as well as the costs of formal exclusion.

Thirdly, while a work in democratic theory, this paper suggests new directions for the empirical study of polarization. Current research has investigated how the alignment of cross-cutting cleavages influences social identities and political attitudes and how forms of polarization may influence framing effects. However, there is limited research on polarization and perspective-taking and that which exists often focuses on how taking the perspective of the other side—or “stepping into their shoes”—may reduce affective polarization (Saveski et al. 2022; Simas, Clifford, and Kirkland 2020). This paper suggests that further research should investigate how polarization and polarized frames influence perspective-taking and its consequences for political behavior and attitudes. I have argued that such influences may have significant implications for democratic values and greater empirical investigation would help broaden our understanding of these threats.