Four Kinds of Polarization and Their Opposites

To assess the democratic implications of polarization, we must make clear how we understand the phenomenon—or phenomena. It is part of my aim to clarify and compare the different ways in which “polarization” is used and to argue that different kinds of polarization should be evaluated differently. This section’s contribution is to provide a synthesis of different strands of research on polarization, sharpen the conceptual distinctions between different kinds of polarization, add a fourth kind of polarization (“intransigent polarization”) and some subtypes to existing typologies, and offer an integrated and comprehensive framework that more clearly connects the different kinds of polarization to democratic theory. Thus, in what follows I distinguish and conceptualize four different kinds of polarization and identify their opposites in a manner that is instructive for democratic assessment.

The classical understanding of political polarization in political science regards it as a matter of “ideological distance (in contra-distinction to ideological proximity)” (Sartori Reference Sartori2005, 120). When people have much dispersed positions on ideology and policy, we are witnessing ideological polarization. The further apart people are on the political spectrum, the more ideological polarization there is among them. While scholars working with this understanding of polarization tend to speak of polarization as a “process where extreme views on some matter of public policy have become more common over time” (McCarty Reference McCarty2019, 9), they also tend to speak of polarization in terms of “bimodality.” Polarization as bimodality refers to the situation where the political opinions and votes of politicians or the public are clustered at different poles of the political spectrum and there is little or no overlap between them (McCarty Reference McCarty2019, 9; Fiorina and Abrams Reference Fiorina and Abrams2008, 566). However, dispersion is not the same as bimodality. The two poles of a bimodal distribution of political attitudes need not be far apart and there can be high dispersion of ideological positions without bimodality (DiMaggio, Evans, and Bryson Reference DiMaggio, Evans and Bryson1996, 696). Thus, there are two subspecies of ideological polarization, where one is about moving to the ideological extremes, and the other is about opinions clustering around opposing ideological poles (where the poles can be more or less ideologically extreme). Thus, we should distinguish between polarization as ideological extremism and polarization as binary ideological clustering.

Whereas the first kind of polarization is about what people believe, the second kind of polarization is about people’s attitudes to their opinions or how strongly they hold their beliefs. In other words, what I call intransigent polarization is not about the content of people’s opinions, as ideological polarization is, but concerns how people relate to their beliefs—how insistently they stand on them. Both ideological polarization and intransigent polarization might be understood as about going to extremes, but they refer to two different understandings of “extremism” (Talisse Reference Talisse. Robert2019, 106-110), and thus to different phenomena. To clarify the distinction, consider the difference between the following two cases. In the first, a person goes from believing in “lower taxes” to believing in “no taxes.” This is a case of changing the content of one’s opinion and becoming more ideologically extreme. In the second, a person goes from believing that “low taxes” is a somewhat attractive position to holding that “low taxes” is an extremely attractive and nonnegotiable position. This is a case of changing one’s attitude to one’s belief and of extremism as intransigence. To be intransigent means to hold one’s opinion with such confidence and rigidity that one is unwilling to change it. Political polarization is about the relations between political opponents, and intransigent polarization is about being unwilling to learn from, accommodate, or agree to anything proposed by opponents.

Today, affective polarization has become an important competing understanding of the partisan divisions in politics and society. Affective polarization was first introduced as an alternative and “more diagnostic” indicator of mass polarization than ideology and policy preferences (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012, 406). Affective polarization is identity based and expressed in feelings of dislike of and anger at the other party and its adherents. According to its proponents, affective polarization is not a mere spillover from ideological polarization, and extremity in ideology is not a necessary condition for the existence of animosity toward opponents (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012, 424; Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019, 131). Thus, the claim is that the public can be deeply divided and hate their political opponents without this being related to strong political disagreement over policy (Mason Reference Mason2015; Reference Mason2018). The theory is that affective polarization is driven by “the power of partisanship as social identity,” rather than ideological differences (Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019, 130). When the public is in the grip of affective polarization, citizens will see their opponents as hypocritical, selfish, and closeminded, and they will avoid social interaction with them. In short, affective polarization is characterized by negative feelings of animosity, anger, and distrust of people associated with the other party and positive feelings for and trust of one’s own party.

Some researchers also speak of partisan sorting as a form of polarization. Generally speaking, partisan sorting entails that political parties become more internally homogenous. However, it is important to distinguish between two different processes of homogenization: ideological sorting and social sorting (Mason Reference Mason2018). Ideological partisan sorting refers to the process where ideology and partisan identity become increasingly matched. An example is the process in the United States, beginning in the 1960s, where liberals moved to the Democratic party and conservatives to the Republican party (Hare and Poole Reference Hare and Poole2014, 415-418; McCarty Reference McCarty2019, 14-16). Notice that partisan sorting is different from people having become more ideologically extreme (Fiorina and Adams Reference Fiorina and Abrams2008) but contributes to ideological clustering. Social partisan sorting goes beyond ideology and includes the alignment of different group identities or cleavages. Parties are socially sorted when partisan identities fall into alignment with several different social identities (Mason Reference Mason2018, 61-63). Thus, social sorting means that partisan identity becomes connected to particular social identities, defined in terms of, for example, ethnicity, race, religion, and place. Moreover, social sorting entails a process in which multiple cleavages are aligned under one single difference of “Us versus Them” (McCoy, Rahman, and Somer Reference McCoy, Rahman and Somer2018). Thus, the key expression of polarization as partisan sorting is that people perceive the parties as entirely distinct and with no overlap in terms of ideology or social identities.

It is instructive for the explication and subsequent evaluation of the four different kinds of polarization to identify their respective opposites. By identifying the opposites of polarization, we can ask whether those are what democracies should aim for.

To identify the respective opposites of ideological polarization and intransigent polarization, it is necessary to distinguish more clearly between consensus and compromise than political scientists often do.Footnote ² As I suggest we use the terms, “consensus” refers to ideological convergence or absence of disagreement, and “compromise” refers to a kind of agreement that is motivated by or at least made necessary by ideological divergence or disagreement (Gutmann and Thompson Reference Gutmann and Thompson2012, 10-12; Rostbøll Reference Rostbøll2017, 621-622). Whereas consensus ends disagreement or ideological polarization, compromise doesn’t do away with disagreements between the involved parties but embodies them. Thus, the opposite of ideological polarization is ideological consensus, while the opposite of intransigent polarization is compromise-willingness. Importantly, the disposition to compromise is needed only in cases where there is some ideological polarization (or some other kind of disagreement).

It is difficult to identify one opposite of affective polarization, because the concept refers to two ideas: that politics has become more affective and that the affects characterizing polarization are negative ones. I propose that we regard the general opposite of affective polarization as the composite notion “absence of negative affects.” This opposite can be divided into a) a politics characterized by absence of affects (apathy), and b) a politics defined by positive feelings toward opponents. What we regard as the most important opposite of affective polarization depends on whether we focus on its affectivity or its negativity.

The overall opposite of partisan sorting is heterogeneous or internally diverse parties. The two kinds of partisan sorting have each their opposite. The opposite of parties that are internally ideologically homogenous is parties that are internally ideologically diverse, and the opposite of socially sorted parties are parties with cross-cutting cleavages or identities. Thus, when there is no or less ideological partisan sorting, there will be greater overlap between the ideological views of opposed partisans and the parties will have less clear ideological identities. And when there is no or less social partisan sorting, partisans will share some social identities or group memberships with their political opponents.

Table 1 summarizes the definitions of the four kinds of polarization and their opposites.

Table 1 Four kinds of polarization and their opposites

Criteria for the Quality of Democracy

In order to evaluate whether the different kinds of polarization are good or bad for democracy, we need democratic criteria that can be used as standards to judge the quality of democracy. I shall not provide a full list of necessary and sufficient conditions for a democratic system but highlight the five criteria that are most relevant in relation to the normative assessment of polarization in existing democracies:

1. C1 Inclusive participation. Those subject to collective decisions should be equally included as participants in the making of those decisions (Dahl Reference Dahl1989, 119-131; Habermas Reference Habermas and Rehg1996, 104-111; Warren Reference Warren2017, 44; Young Reference Young2000, 23). Since we are concerned with the quality of existing democracies, the question is not only whether those subjected have political rights of participation but also the actual level of political inclusion and participation.

2. C2 Contestation. It must be possible to oppose the government, to offer alternative points of view, and there should be a clear range of alternatives for voters to choose between (Dahl Reference Dahl1971, 4; Mouffe Reference Mouffe2018, 17). This is required both for government to be responsive to citizens’ preferences and for holding government accountable. Again, the issue is not just formal opportunities for contestation but also the actual level and quality of contestation.

3. C3 Mutual respect. Political opponents are respected as legitimate adversaries rather than as enemies to be destroyed (Levitsky and Ziblatt Reference Levitsky and Ziblatt2018, 102-106; Mouffe Reference Mouffe2009, 101-105). Mutual respect is compatible with strong disagreements and even dislike of one’s opponents, as long as one respects their equal standing as fellow citizens (Rostbøll Reference Rostbøll2023).

4. C4 Enlightened understanding. The electorate should, in the light of sufficient information and good reasons, be able to develop an adequate understanding of political alternatives (Dahl Reference Dahl1989, 111-112; Habermas Reference Habermas and Rehg1996, 315-316; Mansbridge et al. Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012, 11). Only if citizens know something about their own and others’ interests as well as the expected consequences of different political programs and policies, can they through the political process get what they want and find right.

5. C5 Collective decision-making capacity. Citizens must as a collective have the capacity to impose binding decisions on themselves (Warren Reference Warren2017, 44). Only if citizens or their representatives have this ability of “getting things done” can the people be said to rule themselves. Democracy is not just a matter of inclusion, contestation, and learning, it is a form of governing.

To further appreciate the significance of these criteria, it is helpful to identify the main threats to each, which I have done in table 2.

Table 2 Criteria for assessing democratic quality and threats to these

Degree versus Kind

While polarization is commonly seen as detrimental to the quality of democracy and sometimes as a threat to the very survival of democracy, more sophisticated accounts emphasize that polarization can have both good and bad effects on democracy. Writers who mention the potential beneficial effects of polarization often caution that they materialize only under a moderate level of polarization and argue that if the level of polarization becomes severe, the negative effects of polarization will predominate. To improve or save democracy, the task, these writers suggest, is reducing polarization (Levitsky and Ziblatt Reference Levitsky and Ziblatt2018, 115, 223). Thus, on this view, the problem of polarization is a question of degree and for committed democrats the aim should be finding “the Goldilocks point” of the right level of polarization (McCarty Reference McCarty2019, 20). This section questions the suggestion that “the problem of polarization” is mainly a matter of degree.

One disadvantage of speaking of the problem of polarization as one of degree is that it is unclear exactly what it would mean to reduce it or find its right level. If we speak of “political polarization” in general terms, we risk obscuring the fact that the term can refer to different phenomena. Considering our four different understandings of polarization, the suggestion could be finding the right level of ideological polarization, intransigence, affective polarization, or partisan sorting. One complication is that the four need not go together. For example, there can be much affective polarization without strong ideological disagreement (Mason Reference Mason2018), and there can be partisan sorting without anyone having become more ideologically extreme (Fiorina and Adams Reference Fiorina and Abrams2008). So, what counts as reducing or finding the right level of “polarization”?

While there can be a conceptual difficulty of specifying what it would mean to reduce the diverse phenomena called “polarization,” the main danger is communicative and normative. Indicating that there is a “right level” of polarization moves one into normative terrain and commits one to a conservative view of what (a good) democracy is and requires. If researchers and other commentators indiscriminately speak of severe polarization as bad for democracy, this seems to imply that polarization of all kinds should be reduced. However, from the perspective of normative theory that is a questionable position. In general, there is a risk that political scientists become servants of the existing order when they warn against high degrees of polarization. If we are not clearer regarding the exact forms of polarization that threaten democracy or on what kind of relations democracy requires, there is the danger that any form of dissent is delegitimized, that the status quo ante is glorified as “good democracy,” and that any fundamental change is depicted as fatal for democracy (Mouffe Reference Mouffe2018, 10, 17, 22; Stavrakakis Reference Stavrakakis2018).

Even taking our four kinds separately, it is not evident what reducing polarization means and if it is necessarily democratically desirable. The meaning of reducing ideological polarization might seem clear enough: less ideological distance between parties or voters. However, there is a difference between reduction of the distance between the most extreme ideological positions and reduction of clustering around two opposing poles. Ideological extremism and ideological clustering create different kinds of democratic challenges and should be evaluated differently. Whether ideological extremism is a problem or not depends on the content of the dispersed ideological positions more than on the distance between them. In other words, it might be more important what the parties (and voters) disagree about than how much they disagree. Particularly, disagreements over fundamental democratic values and the rules of the game create greater challenges for democracy than disagreements about policy. Hence, it is not clear that all ideological disagreements are of the same kind and that they all should be reduced. Ideological clustering creates a different kind of problem for democracy. Clustering is less a problem of extremism than that the dispersion of ideological views becomes concentrated around two opposing poles with no overlap between them.

Perhaps part of the appeal of the idea that there is a Goldilocks point of the right level of polarization relies on an aversion to extremism and a belief in the value of moderation (Graham and Svolik Reference Graham and Svolik2020). However, it is important not to confuse the different meanings of “extremism” involved in ideological polarization and intransigent polarization, respectively. From a normative perspective, we should clarify whether “the problem of extremism” is that the ideological content of people’s views is extremely divergent (for example, far left versus far right), or that partisans hold their views with such confidence and inflexibility that they are unwilling to listen to or accommodate anything proposed by political opponents (Almagro Reference Almagro2023). In political negotiations as well as voting behavior, the degree of confidence in beliefs among the parties might be just as important as the content of their beliefs. For example, an extreme degree of confidence in the rightness of one’s beliefs might make one less likely to compromise and less tolerant of political opponents. Thus, what destroys “the compromising mindset” might not be extremity in ideological views (belief content) but extreme confidence in one’s views (Gutmann and Thompson Reference Gutmann and Thompson2012; Muirhead Reference Muirhead2014, 105-110). Moreover, extreme confidence in (and commitment to) one’s views can also beset ideological centrists.

As the opposite of extremism, “moderation” can also refer to two different phenomena: ideological centrism and compromise-willingness. It is possible to have radical ideological views and nevertheless be willing to compromise with political opponents. And one can be an ideological centrist who is unwilling to cooperate and compromise with “extremists” (Craiutu Reference Craiutu2017, 6, 22-23, 32, 237-238, 242; Gutmann and Thompson Reference Gutmann and Thompson2010, 1134-1138; May Reference May, Rostbøll and Scavenius2018: 37; Rostbøll Reference Rostbøll2021; Theriault Reference Theriault, Hopkins and Sides2015). What is most important for meeting the criteria of both mutual respect (C3) and collective decision-making capacity (C5) is not that the parties become more moderate in the sense of moving to the ideological center, but that they become more moderate in the sense of being willing to compromise.

The last point is important to keep in mind when considering issues of institutional design and reform proposals for countering polarization. Political scientists have, for example, discussed the effects on polarization of electoral systems, campaign finance, and political party reform, but they have done so without clearly distinguishing between the aim of creating more centrist parties and providing incentives for compromise (Carothers and O’Donohue Reference Carothers, O’Donohue, Carothers and O’Donohue2019b, 279-281; Levitsky and Ziblatt Reference Levitsky and Ziblatt2018, 202-206; McCarty Reference McCarty2019, 119-132). It goes beyond the scope of this article to go into the question of which institutions provide incentives for compromise. The issue is complicated by the suggestion that sometimes majoritarian institutions may be better at generating compromise than so-called consensus institutions (Schwartzberg Reference Schwartzberg and Knight2018). Moreover, we should note that similar patterns of polarization exist across societies with very different institutions (McCoy et al. Reference McCoy, Rahman and Somer2018, 19).

Affective polarization tends to be deemed always bad for democracy and something that should be reduced or eradicated for the sake of democracy’s quality or survival. However, speaking of reducing affective polarization might not be the most enlightening way of thinking about the challenges it poses to democracy. Here we should remember that affective polarization is a composite phenomenon including both a dimension of negativity and of affectivity. It might make sense to say that there is a golden mean for the level of affectivity in politics, where too much affectivity crowds out any form of deliberation and too little entails apathy or indifference. However, I want to argue that the main issue for the quality of democracy is the kind of feelings partisans have for their political opponents. Here we should speak in more qualitative terms. Democracy depends on mutual respect (C3), which is a kind of feeling. However, it is important to note that respect is neither the opposite of hatred, nor some golden mean between hatred and love. Respect is something qualitatively different than these other feelings. Mutual respect is based on a regulative (moral) principle that tells us how to treat others, irrespective of whether we hate or love them (Kant Reference Kant and Gregor1996, 54-56). Thus, the negativity of affective polarization becomes a problem for democracy only in the absence of a feeling of respect that overrides the impulses connected with hatred of political opponents.

When it comes to partisan sorting, most political scientists think that democracy works best when parties provide voters with clear ideological alternatives or distinct policy programs (APSA 1950; McCarty Reference McCarty2019, 19-20; McCoy, Rahman, and Somer Reference McCoy, Rahman and Somer2018, 17-18; Muirhead Reference Muirhead2010, 132). Political contestation (C2) requires that voters can distinguish between the parties. However, this benefit of partisan sorting seems only to be connected to ideological partisan sorting and not to social partisan sorting. What is detrimental to democracy is when partisan sorting goes beyond what the vote should be about—ideological and policy disagreements—to being only a matter of identity. When parties become socially sorted, the danger is that what determines the vote is less which program and policies voters find good or bad, but mainly a question of which group they identify with (Mason Reference Mason2018; Vachudova Reference Vachudova2019). Social partisan sorting also harms the development of enlightened understanding (C4), because party politics becomes a question of who you are and with whom you belong rather than which policies are best. Thus, the issue is not having the right level of sorting but having the right kind of sorting.

It might be objected that it is unrealistic that people should vote according to their policy preferences or ideology rather than their identity (Achen and Bartels Reference Achen and Bartels2016, 18). However, while it is true that normative theory must be consistent with human abilities (“ought implies can”), we should not tailor our democratic norms to how citizens behave under present circumstances (“ought cannot be derived from is,” especially not “is now”). Perhaps people would act differently with different institutional incentives or under other socio-economic conditions (Landemore Reference Landemore2020, 44-47). Normative theory should speak to potential human competencies rather than only the realized ones. Moreover, realism requires that we examine variation across time and space and note that group behavior and polarization do not always take the same form (Chambers Reference Chambers2018). It is likely that identity or group attachment will always play a role in political behavior, but as I shall argue later, this need not be seen in contrast to acting for reasons, as Achen and Bartels (Reference Achen and Bartels2016, 215, 228) do.

The democratic assessment of the different kinds of polarization is complicated by possible interaction effects among them. Affective polarization, for example, is often seen as causally connected to social partisan sorting (Mason Reference Mason2015), and intransigence can be seen as closely connected to affective polarization. While it is important to understand these connections, I have emphasized that the different kinds of polarization do not always go together and that it is equally important to understand the distinct democratic implications of the different kinds of polarization taken separately. It provides for a clearer analysis to keep apart the question of whether a kind of political polarization has bad democratic effects on its own and the question of whether it spills over into other kinds of polarization. But, of course, in the end we should bring all the issues together in an all things considered assessment.

Ideological versus Affective Polarization

The conclusion of the previous section is that the problem of polarization is more a question of its kind than of its degree. This section continues the analysis by providing some reasons why from a normative democratic perspective it is important to distinguish between the implications of ideological and affective polarization. While there are several reasons to regard ideological polarization as less of a democratic problem than affective polarization, this section and the next argue that it is premature to conclude that the threats of polarization all lie with affective polarization and that affective polarization is only detrimental to democracy, as one might think from reading the literature based on social psychology.

I begin with the reasons for regarding the distinction between ideological and affective polarization as normatively significant, and for regarding the latter as especially problematic for democracy. First, most democratic theorists agree that ideological disagreement is a natural and legitimate aspect of a free society. Insofar as we can speak of ideological disagreement as a fundamental “circumstance of politics” and respect for disagreement as a fundamental democratic norm (Waldron Reference Waldron1999), ideological polarization as a form of dispersion of political views cannot be regarded as a democratic wrong or something that should be eliminated. Nothing similar can be said of affective polarization if we understand the latter as a matter of animosity toward political opponents. It would be better if no one hated their opponents. Only by distinguishing between ideological polarization and affective polarization can we appreciate the democratically important distinction between legitimate disagreement and destructive hatred, between opponents with whom one disagrees and enemies who one wants to eliminate, as required by the criterion of mutual respect (C3).

Second, ideological polarization among ideologically sorted political parties is a necessary condition for providing citizens clear and meaningful alternatives to choose between, as required by C2 (Levendusky Reference Levendusky2010; Lupu Reference Lupu2015). Nothing like that can be said about affective polarization. Indeed, affective polarization means that partisans do not just regard the other party as mistaken but as an illegitimate choice or an enemy (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012, 428; Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019, 143). Since affective polarization means that partisans “loose perspective on the difference between opponents and enemies” (Mason Reference Mason2018, 6), it is a threat not only to mutual respect (C3) but also to contestation and enlightened understanding (C2 and C4). In an environment of affective polarization, a governing party will not seek to promote clear ideological choices that citizens can freely choose between and have equal opportunities to attain an understanding of. Rather than improving the democratic process of free and enlightened choice, affective polarization undermines it. Thus, unless we distinguish clearly between ideological and affective polarization, we have the paradoxical conclusion that “polarization” is both the promotion and the limitation of dissent, contestation, and electoral competition.

Third, it is a common idea that diversity of opinions has positive effects on learning and thus some kind of polarization is epistemically beneficial (C4). At the same time, the literature on affective polarization shows that the latter has bad epistemic effects such as bias in belief formation, simplification of the issues, and prejudices against the opposition (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Mason Reference Mason2015; McCoy, Rahman, and Somer Reference McCoy, Rahman and Somer2018). Here it is crucial not to confuse the effects of the two kinds of polarization. Indeed, the epistemic value of ideological polarization may depend on the absence of affective polarization. Thus, John Dewey (Reference Dewey, Morris and Shapiro1993, 243) argues that the epistemic value of disagreement depends on “the habit of amicable cooperation—which may include … rivalry and competition,” but also requires that we “treat those who disagree—even profoundly—with us as those from whom we may learn, and in so far, as friends.” Amicable cooperation is exactly what affective polarization erodes. However, it is essential to distinguish this issue from the existence of ideological disagreement. Failing to differentiate between ideological polarization and affective polarization combined with the typical negative evaluation of “polarization” has the danger of delegitimizing expressions of dissent and strong disagreement.

While there are strong reasons to highlight the divergent democratic implications of ideological and affective polarization, as well as to note that some of the positive effects of ideological polarization require absence of affective polarization, we should be careful not to assume that ideological polarization is always good and affective polarization always bad. Most important, there is the obvious point that not all ideological positions accept democratic norms. While democracy allows for and even profits from much ideological dispersion or disagreement, it also depends on consensus on some basic democratic norms and institutions. Without an ideological commitment to basic norms of freedom, equality, and respect for opponents (C1–3), as well as the institutional rules of the game, we have a kind of disagreement or ideological polarization that may undermine democracy (Levitsky and Ziblatt Reference Levitsky and Ziblatt2018). In a democracy, there cannot be disagreement about everything. If you disagree with the notion that disagreement should be respected, you are not respecting disagreement but denying its legitimacy. Even democratic theorists who have done most to emphasize its “agonistic” or conflictual character agree that democracy requires “certain forms of consensus” regarding basic norms and institutions (Mouffe Reference Mouffe2018, 93). In short, if ideological polarization includes undemocratic positions, this kind of polarization can also be detrimental to democracy.

Reasonable Affective Polarization

The claim that affective polarization is not always or only bad for democracy is controversial and will require more argument. Notice that my contention is not that affective polarization should be part of our ideal conception of democracy. My argument belongs to non-ideal theory in its technical definition, as a theory that applies to conditions in which not everyone follows ideal principles (Rawls Reference Rawls1999, 8, 216). If everyone were in full compliance with the norms of democracy (the definition of ideal theory), there clearly would be no reasons for hating opponents, or regarding them as corrupt and self-serving. My claim is that under non-ideal conditions—which are the conditions that characterize actual democracies—there are sometimes good reasons for these affects.Footnote ³

As said, affective polarization is defined by partisans hating and being angry at people from other parties, seeing them as “hypocritical, selfish, and close-minded” (Iyengar Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019 et al., 130). Under non-ideal conditions, we cannot exclude the possibility that some people deserve to be viewed in that way. A non-ideal theory of democracy assumes that some people do not comply with democratic principles, which could include forms of corruption and lack of concern for the common good. My contention is that the types of feelings that constitute affective polarization in principle could be reasonable. In other words, sometimes there might be good reasons for being angry at and distrusting political opponents, or regarding them as hypocritical and selfish. When it comes to normatively evaluating affective polarization, it makes a difference whether the involved parties have good reasons to feel the way they do. Indeed, there might be such a thing as reasonable affective polarization, which consists in emotionally reacting to and distancing oneself from others’ wrongdoing.Footnote ⁴

The question of whether people have good reasons for their negative feelings toward political opponents is not raised in the existing literature on affective polarization. The latter literature is typically inspired by Social Identity Theory, which suggests that positive feelings for the in-group and negative feelings for the out-group are consequences of categorization and identification (Tajfel Reference Tajfel1970; Tajfel and Turner Reference Tajfel, Turner, Austin and Worchel1979). Social Identity Theory regards affects as unconsciously formed and discriminatory behavior toward out-groups occurring for “no ‘reasons’” (Tajfel and Turner Reference Tajfel, Turner, Austin and Worchel1979, 47). On this basis, the literature does not regard affective polarization as based on choice and reflection but as something that is caused by unconscious and automatic psychological mechanisms (Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019, 130; Mason Reference Mason2015, 130). This theory of affective polarization excludes the very possibility that people could have good reasons for distrusting and hating political opponents, as well as for regarding them as self-serving, hypocritical, and narrow-minded. Thus, any discussion of the potential reasonableness of affective polarization is blocked from the start.

We should be clearer than the existing literature is on whether “affect” in affective polarization is posited as the cause or the matter of polarization—or both. It is one thing to say that political polarization has an affective cause—and thereby imply that people’s emotions rather than their cognitive faculties or reason drive it. It is another thing to say that political polarization is about the affects that people who identify with different parties have toward one another. Hence, we should distinguish between the notions that people’s relation to their party is caused by affect and that people exhibit certain affects. We should not assume that just because affective polarization involves strong emotions, it is necessarily caused exclusively by emotions. Emotions need not be blind, and they can have reasons as well as causes.Footnote ⁵ If we regard the affective aspect of polarization only in terms of or as necessarily caused by unconscious and automatic psychological mechanisms, we cannot distinguish between cases in which people have good reasons for their negative feelings toward their opponents and those where they do not. Making this distinction is crucial for the democratic assessment of affective polarization.

My proposal is that we examine whether or to what extent the negative feelings toward political opponents described in the literature on affective polarization could be understood as what we might call principle-dependent feelings, rather than assuming that they always are automatic or noncognitive reactions to external stimuli. John Rawls (Reference Rawls2000, 45-48, 148-152) describes a principle-dependent feeling as a feeling we can only explain with reference to the violation of a principle. They are indirect or reflective feelings, which only persons who have principles and are endowed with reason can have. For example, the feelings of resentment, indignation, and anger at the corrupt or selfish behavior of others can only be explained with reference to the violation of principles of honesty and the common good. A person can simply not have those kinds of feelings without understanding and applying principles.

If we analyze the affects involved in polarization as principle-dependent, we can ask whether these affects could be explained by the violation of a principle and thus whether there are good reasons for them. Of course, the fact that a feeling can be associated with a principle is not sufficient for it to be reasonable, since the principle might be invalid; anti-democrats, for example, may feel principled resentment at political equality. Moreover, a person might be mistaken in their belief that a principle has been violated. Still, the very possibility that negative affects toward political opponents can be motivated by principles and reasons opens a way for a discussion that has been blocked by seeing them as necessarily irrational, as is the case in the social psychology-inspired literature on affective polarization.Footnote ⁶ Noting the cognitive dimension of, for example, anger enable us to consider whether this emotion is “more or less correctly directed, or fitting” (Lepoutre Reference Lepoutre2018, 401).

The feelings associated with affective polarization are sometimes feelings of resentment and indignation of the fact that the other party does not treat you with proper regard or respect. If you feel that members of the other party are close-minded, corrupt, untrustworthy, or selfish, these could all be feelings that they do not treat you as an equal or respect your dignity (violating C1 and C3). Populist parties are often described as the most polarizing parties and as the most in the grip of affective polarization (Cohen Reference Cohen2019; Müller Reference Müller2016b, 4; Pappas Reference Pappas2019, 212; Roberts Reference Roberts2022; Urbinati Reference Urbinati2019, 50-51, 70, 74). Moreover, there is a large literature that explains populism with reference to feelings of lack of respect and recognition (Cramer Reference Cramer2016; Gidron and Hall Reference Gidron and Hall2020; Zurn Reference Zurn, Hirvonen and Koskinen2022). Some people might see these developments as detrimental to democracy because they distract from what politics should really be about, namely policy discussion and choices (Mason Reference Mason2015, 142; Müller Reference Müller2016a). However, I would argue that sometimes the kind of resentment expressed in affective polarization is a democratic sentiment. Democracy is a form of government in which citizens owe one another inclusion and mutual respect (C1 and C3). If resentment and hatred of political opponents can be explained by people being ignored, marginalized, or oppressed, there is nothing democratically wrong about these feelings. They might be cries for being seen and heard, included and respected as equals in the political process (Cohen Reference Cohen2019, 26; Cramer Reference Cramer2016, 52, 66, 105; Mouffe Reference Mouffe2018, 22-23). Since democracy requires that everyone be seen, heard, and treated as equals, this is a reasonable demand (C1 and C3). If certain groups are not included as equals in the political process, this is the primary democratic problem, rather than the fact that they feel resentment, anger, or distrust of those who exclude them.Footnote ⁷

To be sure, very few actual cases of affective polarization may be products of violations of democratic principles. Some cases of affective polarization are based on nondemocratic principles, and some people see violations where there are none. Sometimes, elites for strategic purposes speak to potential supporters as if the latter were not heard or in other ways are disadvantaged, even though this rhetoric has no factual basis. And sometimes, people feel resentment at losing privileges that cannot be defended with reference to egalitarian democratic principles (Fukuyama Reference Fukuyama2018, 22; Mutz Reference Mutz2018). My point is simply that to evaluate actual cases of affective polarization, we must also look at the reasons people (could) have for feeling as they do.

I should emphasize that my contention is not that it is in any way ideal when people have the negative feelings characteristic of affective polarization. As mentioned, I am making this argument as part of a non-ideal theory of democracy. It is only when some people do not comply with democratic principles that there can be good reasons for resentment, anger, and distrust. Moreover, I do not deny that affective polarization often has detrimental consequences for the democratic system and that it often expresses itself in violation of democratic principles.Footnote ⁸ One of the problems with affective polarization is that it can create a vicious circle of mutual hostility. When the dynamic of affective polarization has taken hold, everyone might be said to have good reasons to feel hostility toward their opponents. Thus, all things considered, even reasonable affective polarization can have negative effects on democracy.

Still, we must be careful not to place the blame in the wrong place. To secure this, we should not only look at the psychological causes and external effects of affective polarization, but we should also examine to what extent people have good reasons for their negative feelings toward political opponents.

I have argued both that it often is analytically and normatively important to distinguish between ideological and affective polarization, and that we should avoid assuming that all the problems lie with affective polarization. The more fundamental question concerns the norms and principles to which the partisans are committed. From a democratic perspective, we cannot blame people for taking up diverse ideological positions as long as these do not violate basic democratic norms and accept democratic institutions. Regarding affective polarization, we should not blame people for their negative feelings toward political opponents if—and this is an important qualification—these can be explained by actual violations of democratic principles. However, the question of affective polarization is complex. For while we should be careful not to blame people for feelings that they have democratic reasons to have—feelings of resentment and anger at being ignored or oppressed, for example—we might still blame people for acting on these feelings in ways that are destructive of democracy.

I don’t suggest that all or even most actual cases of affective polarization have any principled democratic basis. What I call for is a more nuanced evaluation that at least considers the possibility that some cases of affective polarization are reasonable reactions to deeper problems. A more nuanced analysis and discussion of affective polarization involves three stages. First, assuming that people’s feelings toward their own party and political opponents depend on their experiences and principles, we—scholars and fellow citizens—should try to identify what these experiences and principles are. That is, we should ask which reasons people (could) have for feeling as they do. Second, recognizing that having reasons is not the same as having good reasons, we should discuss the validity of different interpretations of people’s experiences and the principles involved in evaluating them. Do people have good empirical and normative reasons to feel about different parties as they do? Are their emotions fitting? Third, we should assess the political effects of affective polarization at the mass level. The existing literature tends only to consider the third stage and thus risks dealing only with a symptom and ignoring people’s potential reasons for having polarizing affects. By including the first two stages in the analysis, we recognize that there is something in relation to affective polarization not only to condemn but also to discuss and reason about as fellow citizens, namely the validity of our interpretations of reality and our principles.

Invoking the idea of “good reasons” for affective polarization is admittedly not without challenges. The idea is easily twisted and everyone in emotionally charged political struggles might think they have good reasons for their anger and hostility. And in so far as there is disagreement on what the proper democratic principles are and whether they have been violated in particular cases, it seems that we need some way or someone to adjudicate on these issues. However, I do not think this is a unique challenge to the approach I suggest in this article. Political scientists in all subfields routinely judge whether certain attitudes and actions are positive or pernicious for democracy. It might be more straightforward to judge all cases of affective polarization detrimental to democracy than differentiating between the validity of the reasons for different instances of it, but that does not make the former less problematic. If I am right that people can have better and worse reasons for their affects toward political opponents, and if we want to assess their democratic pedigree, there is no way around making judgments about what are valid democratic principles and whether they have been violated in particular cases. Of course, no such judgments are infallible, and they should be seen as part of a continuing dialogue rather than final verdicts. The role of normative democratic theory is to provoke self-reflection and common deliberation among citizens, not to substitute for the latter.

Group Polarization and Opinion Formation

I now turn to the process of belief formation, which is the focus of the literature on “group polarization.” The latter is not yet another kind of polarization, but an explanation for how and why polarization occurs. Rather than measuring distance between different groups or their feelings toward each other, group polarization concerns the group internal process that makes its members “go to extremes” (Sunstein Reference Sunstein2009; Talisse Reference Talisse. Robert2019, 96-98). More specifically, group polarization refers to the phenomenon where people after within group interaction move toward more extreme positions than they were originally inclined (Sunstein Reference Sunstein2017, 68). In short, the dynamic of group polarization makes people more extreme versions of themselves.

Group polarization can contribute to all four kinds of polarization: It makes people more ideologically extreme, it creates intransigence, it fosters negative feelings toward the outgroup, and it increases within group homogeneity (Sunstein Reference Sunstein2009, 3; Reference Sunstein2017, 68-69; Talisse Reference Talisse. Robert2019, 96-106). Group polarization is a consequence of “enclave deliberation” among insulated, likeminded groups, who are “mostly hearing more and louder echoes of their own voices” (Sunstein Reference Sunstein2017, 66). There are different explanations of group polarization, including that the groups are exposed to limited information and argument pool, that the members are concerned with their within-group reputation, and that they become more confident of their views from the corroboration by their group (Sunstein Reference Sunstein2017, 71-75; Talisse Reference Talisse. Robert2019, 110-115).

This section discusses which democratic concerns are relevant in relation to group polarization. I argue that the appropriate democratic assessment is a procedural one, that is, it should focus on the quality of the process of opinion formation rather than the substantive content of people’s beliefs and opinions.

The notion and reality of group polarization require us to evaluate not only the state of affairs of polarized ideological positions, intransigence, hatred of opponents, or partisan sorting but also the process that creates these kinds of polarization. From the perspective of normative democratic theory, the phenomenon of group polarization raises the issue of the quality of processes of opinion formation in society. Especially deliberative democrats have emphasized the level of autonomous and rational opinion formation as an essential indicator of the quality of democracy (Habermas Reference Habermas and Rehg1996, 360-366). The idea is that the quality of democracy depends not only on how people’s preferences are aggregated in the democratic process, but also that background conditions enable people to form their political opinions in free and rational ways. However, one does not need to accept the high standards of deliberative democracy to agree that processes of opinion formation that are unfree or grossly irrational pose a problem for the quality of democracy (cf. C4).

The problem of group polarization is often regarded as an epistemic one. The warning is, and empirical findings show, that when people engage in “enclave deliberation”—discussion among like-minded people in insulated groups—the process will not be one of learning or approaching true beliefs (Sunstein Reference Sunstein2017, 86-87). Another concern has less to do with the substantive merit of polarized beliefs than with the fact that group polarization is a process that changes people’s beliefs regardless of the evidence (Talisse Reference Talisse. Robert2019, 106). Why should we regard these epistemic shortcomings as democratic shortcomings? My suggestion is that we should do so because they threaten a fundamental democratic concern with freedom and autonomy. If people form their political opinions in irrational ways, their freedom and autonomy are diminished (Rostbøll Reference Rostbøll2008, 135-141; Sunstein Reference Sunstein2017, 11). This is the case because people who lack understanding of their own interests and political affairs cannot rule themselves in the sense of giving laws that correspond with their interests and solve their problems.

According to the procedural view, the task of democratic theory is neither to tell people which ideological opinions are right nor what their true interests are. No one has privileged insight into these questions, and democracy requires that citizens treat each other as authorities regarding their own interests and that everyone is free to form their own opinions (C3). However, this does not mean that democratic theory should not be concerned with the quality of the background conditions and the processes in which people form their beliefs and political opinions. It would be a strange democratic theory that had no concern for the social and institutional conditions that shape how citizens interact and form their opinions. Democratic processes are not merely processes where each has an opportunity to have an equal impact on political outcomes but just as importantly are processes in which people interact and influence the opinion formation of one another (Dworkin Reference Dworkin2000, 191). Some forms of political interaction, some institutions and social conditions, are more conducive to the free and rational development of political opinions than are others (C4). Democratic theory should say something about the rationality and autonomy of citizen interactions in different political processes. There is no point outside social conditions and political processes where true interests or rational opinions can be found. People’s political opinions and conceptions of themselves are shaped by the social conditions and political institutions under which they live (Rawls Reference Rawls1999, 229). The task of democratic theory, then, is to analyze what free and rational democratic processes look like and to discuss which social conditions, political institutions, as well as individual behaviors undermine such processes.

Earlier I warned against the danger that the predominantly negative evaluation of political polarization among political scientists and other commentators risks delegitimizing dissent and ideologically radical views. In this connection, the advantage of the procedural approach suggested here is that it focuses on the processes of opinion formation rather than the content of people’s beliefs. In other words, it does not evaluate people’s opinions per se or as a state of affairs (how divergent or ideologically extreme they are) but concentrates on the interactions and institutions that contribute to people forming their opinions as they do. If we can find fault with these interactions for being coercive, manipulative, or otherwise violating democratic criteria of freedom and enlightened understanding (C1–4), we can provide a democratic critique of them. This would not be a critique of people for holding certain views but of processes that fail to allow people to form their opinions freely and rationally. Such a critique, focusing on the process of opinion formation, might be just as critical of centrist as of extreme views, since what matters from this perspective is not the content of people’s views but the quality of the process of opinion formation (Rostbøll Reference Rostbøll2008, 20-24, 135-141).

Sites of Polarization

Until now, I have argued that democratic assessment should discriminate more clearly between different kinds of polarization as well as include considerations of whether people have good reasons for their polarized opinions, affects, and behavior. This section suggests another way in which we should be more discriminating in our democratic evaluation of polarization. I argue that whether (different kinds of) polarization benefits or harms democracy also depends on the site at which it occurs. Politics takes place at different sites (in civil society, in election campaigns, in legislative assemblies, and so on), and the democratic process can be divided into different stages (opinion formation, campaigning, voting, and governing). There is a division of political work among different actors and sites that complement one another only by serving different purposes (Goodin Reference Goodin2005; Habermas Reference Habermas and Rehg1996; Mansbridge et al. Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012). Because different political arenas have different purposes their democratic quality should not be judged by the same standards. To assess the democratic implications of polarization, therefore, we should avoid assuming that all parts of the democratic system must live up to the exact same ideal. Some forms of polarization might be beneficial in some parts of the democratic system, while they are detrimental in others.

The main harm caused by polarization is often said to be that it inhibits compromise and the solving of societal problems, while the main benefit is said to be that it provides clear electoral alternatives (Lieberman, Mettler, and Roberts Reference Lieberman, Mettler, Roberts, Lieberman, Mettler and Roberts2022, 3; McCarty Reference McCarty2019, 19-20; McCoy, Rahman, and Somer Reference McCoy, Rahman and Somer2018, 17-18). From the perspective of the more differentiated and systemic approach suggested here, it is important to emphasize that lack of compromise-willingness is mainly a problem in legislative assemblies, while the benefit of clear alternatives for voters to choose between pertains to elections. Moreover, the main harm of polarization refers to collective decision-making capacity (C5), while the main benefit refers to contestation (C2). Some forms of ideological polarization, partisan sorting, and even affective polarization can be valuable during elections because they motivate participation and include otherwise marginalized people (C1) (McCoy, Rahman, and Somer Reference McCoy, Rahman and Somer2018, 17-18; Roberts Reference Roberts2022). However, the kinds of polarization that benefit participation and inclusion during the campaign stage are detrimental to the deliberation and compromise required in legislative assemblies (C5) (Gutmann and Thompson Reference Gutmann and Thompson2012). This is just one example of how our democratic assessment of polarization should be sensitive to the question of in which part of the democratic system it occurs, and that sometimes different democratic criteria should be applied to different sites.

The empirical literature on polarization tends to have a narrow focus on the formal part of the democratic system, as does the division between governing and campaigning in the previous paragraph. While my approach agrees that it is important to distinguish between the democratic implications of (different kinds of) polarization in legislative assemblies and in electoral campaigns, it holds a more expansive view of the democratic system by including the informal arena of civil-society politics. Thus, we can divide the democratic system into an informal part consisting of civil society with its associations, movements, and media, and a formal part consisting of elections and the legislative assembly (Cohen and Arato Reference Cohen and Arato1992; Habermas Reference Habermas and Rehg1996, 304-387). Civil society has a different political purpose than elections and formal decision-making bodies. Rather than making binding decisions or providing electoral alternatives, civil society with its public sphere is the place where new problems are discovered, experiences are articulated, and opinions are formed. Table 3 lists the main functions of the three principal arenas of the democratic system. Of course, what we regard as the purposes of the different parts of the democratic system depend on our normative model of democracy and the following fits a more deliberative conception of democracy.

Table 3 Democratic arenas and their functions

The most important difference between the public sphere of civil society and a legislative assembly is that the former is not able or required to make binding political decisions and, thus, do not need to live up to the criterion of collective decision-making capacity (C5). The most important difference between civil-society politics and elections is that the former does not have the same representative functions and requirements of equal power as the latter. As part of the democratic system, the purpose of civil society is to detect new problems, give voice to marginalized groups, protest existing decisions, learn about one’s own and others’ interests and their connection to politics, and form a public opinion that can influence elections and legislative assemblies. According to Habermas (Reference Habermas and Rehg1996, 186, 307), these purposes require that civil society remains a “wild” complex with an “anarchic structure.”

While it is crucial for the ability of civil society to serve its democratic purposes that it does not become structured along the same lines as the formal part of the democratic system and that it need not fulfill the exact same criteria, this does not mean that there are no democratic criteria for civil-society action. Indeed, the quality of civil society should be judged with the first four criteria (C1–4). However, the type of inclusion, contestation, respect, and enlightened understanding that we should expect and hope for in civil society differ from what is required in the formal political system, especially legislatures. Here I limit myself to indicating what this means in relation to the criteria of mutual respect (C3) and enlightened understanding (C4).

Consider first mutual respect (C3). In a legislative assembly, this norm is connected to a willingness to listen to others, cooperate with them, and to make the concessions needed for compromise (Rostbøll Reference Rostbøll2017; Reference Rostbøll, Rostbøll and Scavenius2018).Footnote ⁹ Civil-society groups do not need to show these virtues, because the functions they have within the democratic system often can be realized only through more confrontational forms of action. Social movements are not always respectful of political opponents in the strong sense required for cooperation in legislative assemblies, but they further other democratic desiderata by bringing new issues to the agenda, including otherwise marginalized voices, and motivating participation (C1 and C2) (Mansbridge et al. Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012, 7, 19; Young Reference Young2001). The actions of social movements are often polarizing, both ideologically and affectively, and movements are often engaged in enclave deliberation that entails group polarization. In evaluating the polarizing and disrespectful actions of social movements, it is crucial to remember in which arena they act and not subject them to the same criteria as, for example, political parties engaged in governing.Footnote ¹⁰ Thus, insofar as affective polarization motivates political participation and insofar as group polarization helps to generate new ideas, these might have a place in civil-society action, even if they are detrimental in legislative assemblies.

However, we should avoid a risk in the systemic approach, as this is programmatically laid out in Mansbridge et al. (Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012), which is to justify all kinds of actions within the parts if they contribute positively to the democratic system as a whole. If civil-society action has the effect of including new voices or motivating participation by violating fundamental democratic norms such as mutual respect (for example, through racist speech or violence), it is normatively dubious to deem it “democratic” (Owen and Smith Reference Owen and Smith2015). To judge actions only in terms of their effects on the democratic system, rather than on their intrinsic qualities, would be to treat some persons as means for the realization of some aggregate goal or state of affairs, which violates fundamental democratic norms of equality and mutual respect (C3). For this reason, I propose that while the different parts of the democratic system do not need to fulfill the democratic criteria in the same way, all the parts must satisfy democratic criteria such as mutual respect in some minimal sense. For example, insofar as affective polarization entails regarding political opponents as enemies to be eradicated, this cannot be regarded as “democratic” no matter its consequences in terms of mobilization or contestation. This is also because even if such forms of affective polarization might mobilize and include some groups, it is not itself based on a commitment to respect for norms of inclusion and contestation (C1 and C2) (Rostbøll Reference Rostbøll2023, 205-210).

Consider next enlightened understanding (C4). From the literature, we know that both affective polarization and group polarization can have negative epistemic effects (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Sunstein Reference Sunstein2017). However, it has also been argued that there can be epistemic advantages to people clustering in groups of likeminded others (Mansbridge et al. Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012, 24; Sunstein Reference Sunstein2017, 85-97). When people meet in likeminded groups, they are more likely to share their experiences and opinions and to get a hearing than in large diverse settings. By promoting otherwise invisible or silenced forms of knowledge, unearthing new information, and generating new types of arguments, enclave deliberation can counter “epistemic injustice” (Fricker Reference Fricker2007). Thus, when people deliberate among likeminded people in diverse groups, this can create a form of second-order epistemic diversity within the wider society (Sunstein Reference Sunstein2017, 85-97).

To understand the value of second-order diversity requires us to take a systemic approach to democracy. It is the political role of civil society to form groups in which people can exchange experiences, become aware of their problems, and develop their political opinions. This form of learning is an essential part of developing enlightened understanding (C4). However, the challenge is that a society cannot benefit from second-order diversity if the different groups are isolated from one another and do not listen to one another. For the viewpoints of different civil-society groups to have a beneficial effect on societal learning, there needs to be an audience that is not polarized in the same way as the civil-society groups are (Mansbridge et al. Reference Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, Warren, Parkinson and Mansbridge2012, 24). When enclave deliberation also leads to hatred and distrust of other groups (affective polarization) there will be no persuasion across group or party lines (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012, 428). The upshot of this brief analysis is that the kind of process that takes place in individual groups and the ones taking place in the mass public or among legislators need not be the same and need not be held to the same criteria. Different kinds of learning can take place at different sites of the democratic system, which means that polarization does not create the same problems everywhere.

I have argued that democracy should be seen as a complex system with different parts that partly divide democratic functions between them. This means that we cannot use the exact same standards to assess the democratic implications of (different kinds of) polarization in all political arenas. Commentators who speak of negative and positive effects of polarization in general terms fail to see how political work is and should be divided within the democratic system. I would finally like to argue that the way in which and the reason why polarization plagues many contemporary societies is connected to the fact that it does not respect the division of labor between different sites of the democratic system. Logics of conflict that belong to civil society are allowed to intrude on the formal part of the political system, and logics that are necessary and beneficial in electoral campaigns intrude on processes of legislation and governing. This problem cannot be understood by applying one universal standard to democratic politics as such, nor by indiscriminately speaking of too much polarization or the right kind of polarization. What level and kind of polarization is good or bad for democracy also depends on where it takes place. The problem for democracy is less the level of polarization or specific kinds of polarization, but that the same kinds of polarization are allowed to colonize all arenas. For democracy to work well, it is imperative that different arenas follow different logics and operate in different ways, which means (among other things) that they can tolerate and even benefit from different kinds and levels of polarization.

The breakdown of the proper division of political work between the different sites of the democratic system is undoubtedly connected to the dysfunction or decline of political parties (Caramani Reference Caramani2017; Mair Reference Mair2013; Pildes Reference Pildes, Hopkins and Sides2015). The traditional understanding of the function of political parties is to see them as linking civil society and the state, citizens, and government. However, it is important that this linking function does not involve a collapse of the separation of the informal and formal parts of the democratic system. For political parties to work well, they must be intermediary bodies that uphold the differentiation between civil-society (or movements) logics and political society (or party) logics (Arato and Cohen Reference Arato and Cohen2022, 53-106). A well-functioning party system institutionalizes pluralism, recruits and educates candidates, and ensures a form of regulated rivalry (Muirhead and Rosenblum Reference Muirhead and Rosenblum2020). Today, political parties in many places clearly fail to live up to these ideals. Political theorists have long neglected the democratic value of political parties but over the last two decades they have picked up the task of theorizing the proper place and form of political parties and partisanship in democratic systems (Muirhead and Rosenblum Reference Muirhead and Rosenblum2020). It takes us too far afield to consider the many interesting proposals in the normative literature on parties and partisanship, but I want to emphasize that we should not merely lament the decline of parties and party democracy nor try to find ways to return to some illusory golden age. Rather, we should think of reform proposals that respond to and fit present circumstances and challenges. My hope is that the conceptual and normative framework developed in this article—connecting different kinds of polarization, democratic criteria, and the diverse sites of political work—can guide further inquiry into the proper role and form of political parties in a well-functioning and differentiated democratic system.