AbstractPolitical polarization poses major societal risks, yet a globally relevant, interdisciplinary understanding remains lacking, as nearly half of all studies focus on the United States. We argue for renewed effort to bring global equity to polarization research, highlighting insights robust across countries, those unique to specific contexts, and key literature gaps. Closing these gaps means overcoming cultural and systemic barriers, including data-access restrictions and misaligned incentives. It also demands interdisciplinarity bridging traditional approaches and those addressing the polarizing role of the internet, social media, and AI. Otherwise, efforts to counter polarization, and its democratic harms, risk resting on unsuitable evidence.IntroductionWith a folk understanding that political polarization refers to socio-political divisions within a society, many have proclaimed that we are more divided than ever1,2,3. In this account, polarization has been blamed for populism4, the erosion of social cohesion and rise of incivility5,6,7,8,9, the loss of trust in the institutions of democracy10,11, legislative dysfunction11,12,13,14, and the collective failure to address existential risks such as Covid-19 or climate change15,16,17.But political polarization is far from a new problem–it has been a key focus of research since the birth of the social sciences–and it is reasonable to argue that, when viewed in a historical context, current societal divisions are minor compared to those which led to the major global conflicts of the 20th century18. What has fundamentally changed is the emergence of the information age (the internet, social media, AI) and its integration into, and domination over, our contemporary political discourse19. This shift, and the election of Donald Trump to the U.S. Presidency in 2016 on the back of a campaign that heavily leveraged social media, has driven a renewed interest in understanding the causes and consequences of political polarization. However, while polarization research has blossomed in the U.S., a comparable evidence-base for understanding polarization in a global context remains limited (Fig. 1; see Supplementary Note 1 for data collection details), especially outside Europe and other high-income democracies. If insights from the U.S. were to generalize to other regions, the lack of localized evidence could, perhaps, be tolerated. However, as we will outline, there are numerous results from the U.S. which are outliers when viewed in a global context, where there are substantial conceptual challenges which threaten our normative understanding of what polarization is and how it should be addressed (Box 1; see ref. 20).Fig. 1: The unequal distribution of polarization research.Full size imageAcademic papers on political polarization which refer to a specific country overwhelmingly reference the United States. In total, 5965 papers use the term “political polarization”, “ideological polarization” or “affective polarization” (or equivalent spelling in British English) in their title or abstract of which 2243 refer to at least 1 country. The United States is mentioned in 967 papers (43.1% of papers with a country mention). Many papers which do not mention a country are U.S. focused, but do not state so explicitly. Contextual factors which contribute to the unequal distribution of polarization research include, but are not limited to, legitimate global interest in the polarization of U.S. politics given its disproportionate influence on global affairs, differences in academic research funding across countries, the challenges of conducting studies in countries with “hard to reach populations”, and the perception that studying polarization in countries where polarization is particularly high, or damaging, is more important than in less polarized countries.Ensuring a globally relevant literature on political polarization is, or should be, a fundamental goal for the academic community (as several have argued previously1,7,8,21,22). But there are substantial systemic and cultural barriers to achieving this goal. First, how polarization is studied is often influenced by norms established in U.S.-focused research (Box 1). Second, prominent reviews6,23,24,25, and research articles26,27,28,29, published in top interdisciplinary journals are often limited to the U.S., which may in turn shape academic incentives. Finally, when research does focus on non-U.S. contexts, the approach employed is usually confined to limited disciplinary silos, either employing traditional social science research methods (often using survey data), or applying data- and computational social science approaches to study polarization in our contemporary information ecosystem (e.g., on social media). Rarely are these streams of research synthesized into a single coherent narrative and understanding which is able to capture the complex societal feedback loops across the different domains (political institutions, the economy, legacy and social media) which contribute to driving and sustaining polarization. Quantitative polarization research is contingent on high-quality data. But in the case of digital media research, access to this data has become increasingly difficult (an issue which the European Union’s Digital Services Act aims to fix30,31,32), and is sometimes only available to U.S.-based researchers (for example, in the case of Meta’s U.S. election studies27,28,33,34,35).Seeking to address these issues, this article offers a broad overview of polarization research globally, focusing on key trends in ideological and affective polarization (the most widely studied polarization sub-types), the influence of socio-economic factors, the impact of the traditional news media and social media, and the increasingly important role of recommendation systems and artificial intelligence (AI)36,37,38. Drawing from a growing set of examples, we discuss key results in the literature that are country-specific and those that, broadly speaking, generalize across a wide range of nations. We also highlight key areas where globally relevant literature is lacking and provide concrete suggestions for advancing towards a more globally equitable understanding of polarization.Changes to information production, dissemination, and consumption have changed political discourse at a global scale, and an understanding and analysis of these factors, and their global variability, must be integrated into any coherent understanding and analysis of the causes and consequences of polarization in the future. A failure to address these gaps is not only potentially dangerous, risking inefficient or counterproductive polarization interventions, but is intellectually complacent, further deepening existing inequities in our global understanding of the phenomenon.Our review focuses on cross-national patterns in ideological and affective polarization, on the role of socio-economic and institutional factors, and on the impact of the news media, social media, and AI; many adjacent topics necessarily fall outside its scope. Parts of the literature we discuss will seem disconnected, and some may argue that it would be conceptually clearer to present these streams of research in separate articles. However, it is our view that the tendency to study aspects of polarization in a siloed manner only perpetuates this disconnect. For this reason, our review retains a broad scope in the topics covered, pointing to additional resources where relevant, with the overarching aim of fostering greater interdisciplinary cohesion in future polarization research.Box 1 The normative U.S. framing of polarizationIn selecting “polarization” as their word of 2024, Merriam-Webster defined polarization as “division into two sharply distinct opposites; especially: a state in which the opinions, beliefs, or interests of a group or society no longer range along a continuum but become concentrated at opposing extremes”197. While such a framing may be appropriate for the public’s understanding of the U.S.’s unusual Manichean two-party democracy, there are many cases where such a framing may not be appropriate20. For instance, it is far from clear how polarization–conceptualised in a variety of ways (ideological, affective), and differently by different research fields (e.g., distributions of opinions in surveys, politically homophilous interactions on social media9)–should be conceptualised20, and measured, in multi-party democracies.Researchers have adapted conventional two-party measures of polarization for use in multi-party contexts7,20,70,74,78, but, to some extent, it remains an open question as to whether key concepts in the study of polarization change when extended from a one-dimensional political spectrum to a multi-dimensional context. Key areas of dispute include the importance of in-group identification in defining affective polarization78, the distinction between animosity aimed at political elites56, partisans, and voters more generally, the precise nature of political animosity20, and whether we should expect multiple parties to fall into two, or more, opposed camps20,78.Similarly, there is substantial work discussing how polarization can be formulated without explicit reference to the extremes, e.g., by referencing the bimodality or the spread of the left-right distribution, or without reference to political ideology, changing how the concept is interpreted44. While such issues can seem abstract, they can directly affect the global context for understanding polarization when, for example, cross-country political values surveys ask individuals to place themselves ideologically on a one-dimensional left-right scale, but not on other axes of political belief which may be equally important for understanding political polarization (e.g., the establishment/anti-establishment dimension15,198,199).Global trends in polarizationUntil the early 21st century, scholars across many world regions focused largely on the study of ideological polarization, typically understood as the gap between liberals and conservatives, and by extension between the political left and right39,40,41,42,43. Most commonly, ideological polarization has been quantified at the level of political elites or the mass public according to the dispersion of perceived (in the case of elites) or self-reported (mass public) left-right scores using surveys44. Although not aligned a priori, these ideological labels are often considered indicative of specific positions or attitudes on key political issues (e.g., in relation to the economy, immigration and social issues). However, since the 2010s, evidence from the U.S. has called into question this association, noting that the animosity between partisan groups has grown disproportionately relative to policy differences. This has led researchers to shift focus away from ideological polarization and towards affective polarization6,23,45,46, the phenomenon whereby an individual has negative feelings towards their political out-group (and/or positive feelings towards their in-group). Although approaches vary, affective polarization is most commonly measured by employing “feeling thermometers” which gauge partisans’ warmth (or animosity) towards their political opponents6.Ideological polarizationAmongst political elites, comparative data from 21 high-income countries reveal a small but significant increase in ideological polarization since the 1990s. In absolute terms, Canada and Australia are cited as countries with low ideological polarization, in contrast to Spain, Italy and Sweden where ideological polarization is particularly high47. Similarly, across 20 democracies in Western Europe, research finds that the dispersion of political parties along the left-right ideological axis has increased between 1945 and 2021, with researchers suggesting that this has contributed to a fragmented and volatile political context, at the risk of democracy48. But such trends in the polarization of political elites do not necessarily reflect growing ideological polarization in the general population.At a global scale, ideological polarization has not increased across the mass public49, which is in clear contrast to the trend observed in the United States, where growing ideological polarization (both mass and elite) is now well documented13,50, albeit this increase is smaller than the variability observed across countries6,45,46.However, the appropriateness of the left-right ideological scale for assessing political beliefs, and therefore the appropriateness of the concept for measuring polarization, has been disputed in low- and middle-income countries. Here, research suggests that ideological labels do not translate coherently from country to country51. This lack of coherence also applies in authoritarian contexts, which pose a unique challenge to studying polarization (Box 2). For example, researchers have suggested that the traditional left-right divide, commonly observed in Western political systems, is reversed in China52. Here, the lack of a pluralist party system means that ideological divides cannot be mapped onto partisan divides. Researchers have also shown that political attitudes do not map consistently to pro- and anti-regime groups53. These ambiguities challenge how to understand ideological polarization in different national contexts, and especially in authoritarian states where the concept remains understudied53.Box 2 Challenges in conceptualising and studying polarization in authoritarian statesDespite its normative framing as a democratic phenomenon, polarization is an important concept for authoritarian states52,53,200,201,202. As in democracies, individuals in authoritarian societies have diverse ideological views and may have positive or negative affect towards various groups (e.g., towards supporters of the regime in power). But studying polarization in these contexts involves unique challenges (see below). Beyond its inherent value, the study of polarization in these states can offer valuable insights for new democracies which retain a legacy of authoritarianism (e.g., in Kenya203 and Liberia204) and can guide scholars interested in polarization’s role in eroding democracy5,60.Particularly in authoritarian states, polarization can be exploited as a tool to cement power and discredit opponents205. Researchers can play an important role in understanding political dynamics in these contexts. For example, research has shown how the increased polarization of online interaction patterns preceded periods of violence during the Arab Spring129, how the Chinese government produced up to half a billion social media posts each year to distract citizens from political topics206, and how politicians in Zimbabwe deliberately exploited polarization to reshape boundaries of conflict and mobilize constituencies201.Crucially, given the risks posed to research subjects in authoritarian states–including severe physical repression202, or exile207–researchers must address a number of practical and ethical challenges to studying polarization:Respondents in authoritarian contexts are “hard to survey”208, meaning that studies suffer from numerous biases, including the censorship of data (e.g., in China179), preference falsification in surveys, and coordinated inauthentic activity. These biases may be mitigated by using list experiments or conjoint experiments209. Additionally, preference falsification may be counteracted using “anonymity sampling”210 which improves survey accuracy while protecting respondent anonymity.Scientists are increasingly emphasizing the need for publicly available data for study replication. However, in authoritarian contexts, such data may put survey respondents at risk if not carefully anonymised. Researchers must take extra care when providing public replication datasets–including the IDs of social media posts–that contain politically sensitive content. The use of secure data archives may offer a path for data sharing without putting research subjects at risk211.Even if data are anonymised, scholars should be conscious that their work may be used as “opposition research”, and indirectly benefit the regime in power. For example, papers with real-time monitoring of opposition groups could be used as justification for cracking down on protesters. An example which highlighted these issues is the ENCORE study, in which internet censorship was measured at the user level across a number of countries–including China, Iran, Pakistan, and Saudi Arabia–but without individual consent212.Affective polarizationIn the United States, the state of affective polarization has been described as “political sectarianism”23, drawing parallels to religious and ethnic sectarianism, and having often been linked to the rise of populism (although there is substantial variability in the ideological positions of populist parties globally, see ref. 54). Referencing data from 12 OECD countries, Finkel et al. argue that the state of U.S. affective polarization is exceptional, with the U.S. having suffered the largest increase over the last 40 years55. However, estimates of the absolute levels of affective polarization from 53 countries point towards Türkiye, Hungary, and Kenya as examples of high affective polarization, and the Philippines, the Netherlands, Belgium, and Japan as examples of countries with low levels of affective polarization, with the U.S. falling in the middle of the pack5 (although this varies depending on whether the focus is on the out-party in general, or on political elites; animosity towards political elites is disproportionately high in the U.S., see ref. 56).For many countries (including authoritarian states where measuring polarization is particularly challenging) quantitative estimates of affective polarization are still unavailable5. In these contexts, where populations are hard to survey (see Box 2), researchers have proposed using expert estimates of polarization, and other markers of democracy, as an alternative to conventional survey studies. The Varieties of Democracy (V-Dem) collaboration57 is probably the best-known example of such an international effort. However, it is important to note that expert estimates remain subjective and are open to criticism. In particular, scholars have noted that V-Dem indicators often fail to capture evolving trends, with changes in V-Dem indicators often lagging behind those derived from social surveys58. Additionally, the subjectivity of evaluations may result in confirmation bias amongst expert annotators59.Using V-Dem data, researchers have defined the concept of pernicious polarization as quantifying the extent to which a country is divided into mutually antagonistic camps60. This measure points towards Brazil, Mexico, the United States, and India as examples of democracies with particularly high levels of pernicious polarization, and reveals that every world region other than Oceania has suffered an increase in pernicious polarization since 2005, albeit levels were much higher in Europe and the Americas during the first half of the twentieth century. However, pernicious polarization is only one of many proposed polarization measures. It is not always clear how different types, or definitions, of polarization are related, in which contexts it is better to study one polarization type over another, and whether polarization, especially ideological polarization, is even undesirable61.How are ideological and affective polarization related?In the United States, the relationship between ideological and affective polarization is debated, rooted in the question of whether policy differences drive partisan animosity, the key component of affective polarization. Some scholars have argued that ideological and affective polarization are largely distinct45,46, while others have identified ideological polarization as a key driver of affective polarization and partisan animosity62,63.At the heart of this debate is the distinction between two forms of sorting into partisan groups: ideological sorting–the growing alignment of specific issue-positions, and issue-extremity, to partisan affiliation (e.g., Republican or Democrat)62–and social sorting–the alignment between individuals’ partisan affiliation and various social identities including race, ethnicity, religion, class, socioeconomic status, location46,64,65, and ideological identity (e.g., liberal or conservative)6. In the former case, scholars argue that affective polarization is driven by opposed politicians adopting increasingly divergent issue-positions62,66, with the pairwise alignment of different issues becoming stronger over time (or, alternatively, the diversity of aligned issues broadening; see the “oil-spill” model of polarization67). In the latter case, researchers argue that social sorting has resulted in fewer social identities that cut across party lines. Over time, this has resulted in partisans seeing themselves as having less and less in common with their political opponents, which in turn has strengthened the impact of stereotypes and “othering”, driving affective polarization66. In both cases, researchers have argued that the news media and social media play a key role in facilitating growing issue or identity alignment68. Here, the crucial and often overlooked observation is that individuals can label themselves as belonging to a specific ideological or partisan group, even if their issue-specific beliefs and attitudes do not appear consistent with those labels (“ideological incongruence”; see ref. 69).The framing of partisanship in terms of social identity is a crucial change in the conceptual framing of polarization–a feature of politics also observed outside the U.S., including in the UK, Sweden, and the Netherlands70. The concept builds on an existing literature from social identity theory which discusses the relationship between group identity and out-group animosity71, and the role that social identity plays in shaping attitudes during times of political flux (e.g., in Hong Kong in the 1990s72). But the challenge here is that the growing importance of social identity is not easy to disentangle from growing ideological sorting (in the U.S., both have risen in the last 40 years). The consequence is that scholars in the U.S. continue to debate the relationship between ideological and affective polarization due to the lack of a clear causal relationship between the two66.This debate draws parallels globally, where the evidence for the relationship between ideology and affect displays “relatively weak and inconsistent” trends73. Survey evidence suggests that ideological and affective polarization are broadly correlated74, but some countries display significant ideological polarization without high out-party animosity (e.g., the Netherlands), while others are afflicted with severe affective polarization despite the principal parties being largely moderate (e.g., Montenegro and Serbia). Importantly, survey evidence from 53 countries suggests that high levels of affective polarization are correlated with democratic backsliding, but high levels of ideological polarization are not5 (some previously debated whether ideological polarization is even democratically problematic, since, they argued, some ideological diversity is essential for preventing radicalism75). While this comparative evidence is not conclusive, specific examples such as the rise of affective polarization in the UK around Brexit, and associated issues such as immigration, have led scholars to introduce the cross-cutting concept of “issue-based affective polarization”76. Scholars have also noted that, in contrast to the conventional framing of affective polarization between partisan groups (e.g., Republicans vs. Democrats), that affective hostility is common across factions within a political party77 (e.g., between the progressive and centrist wings of the Labour party in the UK–Corbyn vs. Starmer–or the Democrats in the USA–Sanders vs. Biden). In some instances, the measured level of within-party affective polarization may exceed inter-party polarization77, challenging the primacy of partisanship as a political identity.The reasons for country-level differences in the relationship between ideological and affective polarization are complex and disputed (see20): The U.S.-centric literature has made a two-party “us-vs-them” framing of polarization the norm (Box 1), but in multi-party states it is not uncommon for individuals to have positive (or negative) feelings towards multiple parties (e.g., in Finland or Sweden7). This seriously complicates how to measure polarization. For instance, even when affective polarization is measured across multiple parties7, the measurement continues to frame the concept in terms of politics “divided into two distinct camps”, a choice criticised by some scholars20. Brazil is a good example of a country which is not well-suited to this formulation of polarization, with politics being less institutionalised and partisan identities being less evident78, resulting in affective polarization with low in-group identification. Additionally, framing polarization in a multi-party context allows for new political dynamics which can have both positive and negative democratic impacts, for instance, related to the sustainability and effectiveness of coalition governments74. These distinct phenomena are important, since they will contribute to differences in the complex societal feedback loops driving and sustaining polarization.Differences in the institutional framework of countries are key to understanding cross-national differences in affective polarization and its relationship to ideological polarization. Research suggests that affective polarization is higher in majoritarian political systems73, where political power is concentrated in the hands of a single party (e.g., France, UK) and which feature single-member electoral districts73, than in systems based on consensual governance (e.g., Switzerland, the Netherlands). However, Gidron et al.73 caution against concluding that majoritarian political systems cause affective polarization. In contrast, the ideological polarization of political parties appears lower in more majoritarian electoral systems47. As highlighted by ref. 47, one of the best examples of how the political system affects ideological polarization is the example of New Zealand: When the electoral system changed to proportional representation in 1993, the measured level of ideological polarization doubled79.Focusing explicitly on institutions and the intersection between affective polarization and ideological extremity (those individuals who describe themselves as being on the far-left or far-right), evidence from 45 countries shows that the most ideologically extreme individuals are usually those who display the highest out-group affect, but that this relationship is stronger in older democracies, in countries with parliamentary systems, and where the party system of a country is less fragmented80.Beyond institutions, socio-economic factors and income inequality are often implicated as drivers of polarization47,73,81,82,83,84,85,86. Evidence from 20 high-income democracies (Europe, USA, Canada, New Zealand, Australia) shows that affective polarization is worse when levels of income inequality and unemployment are high73. But, again, countries with high income inequality show lower levels of ideological polarization amongst the mass public (e.g., the U.S.) than lower income inequality countries (e.g., Sweden81,86). Other important institutional factors may alter how polarization manifests such as cross-national differences in political engagement73 and the role of supra-national organisations (e.g., the European Union, or NATO87). However, it is often hard to establish causal evidence of a relationship. The difficulty of acquiring causal evidence for what is, and is not, polarizing is one of the major challenges facing researchers. And arguably, this issue is worsened by studying polarization in domain-silos, since such research cannot address the complex societal feedback loops involved in driving and sustaining the phenomenon.Finally, it is often assumed that polarization has roots in ethnic or religious differences (Box 3). However, while these factors may correlate with observed differences in polarization, and may be used as the basis of polarizing strategies by political leaders, multiple comparative studies have shown that social cleavages do not, inherently, predict damaging forms of polarization60,88. For further detail on the drivers of affective polarization and how it manifests across comparative contexts (including in the Global North, Latin America, and Asia) we point the reader to the recently published Handbook of Affective Polarization89.Box 3 Do ethnic and religious differences lie at the root of polarization?It is often assumed that polarization has roots in ethnic or religious differences. However, while these factors may correlate with observed differences in polarization49, multiple comparative studies have shown that social cleavages do not, inherently, predict damaging forms of polarization21,60,88,213.Ethnic, religious or social class divisions become politically relevant cleavages when people develop shared identities and values, and organizations like parties or unions represent them214,215. And these cleavages become polarizing when political actors make them the basis of polarizing strategies, for instance, through the mobilization of anti-immigration sentiment in Europe and the U.S.216. But as Benson217 points out, while the social psychology of intergroup conflict may explain the intensification of the in-group identification and out-group animosity of affective polarization, it does not explain how those identities become politically salient in the first place. For this, scholars turn to elite agency and the divisive rhetoric of polarizing politics that politicizes certain social cleavages, grievances and political identities201,218.For example, ethnic and religious divides were politicized in India and Türkiye as the secular solution of the founders gave way to Hindu nationalism under Modi’s BJP in India and the Islamist movement and Kurdish rift under Erdogan’s AKP in Türkiye219,220. On the other hand, some countries with significant ethnic divides have not descended into an ethnicity-defined political rift221. A comparative study of four Asian countries finds that those with clearly identifiable social cleavages, Taiwan and Indonesia, suffered only short episodes of damaging polarization, while those with less identifiable social cleavages, Thailand and the Philippines, suffered more destructive consequences, including democratic collapse in the Thai case222.Finally, certain cleavages may be more prone to damaging polarization, such as the unresolved, emotionally-charged historic debates around citizenship, national identities, and founding myths that Somer and McCoy223 call “formative rifts”. Examples include the claim that political legitimacy is to be conferred only on those with national liberation war experience in Zimbabwe224, or the legacy of unequal citizenship rights and political exclusion around race, gender and religion in the U.S.225,226.Polarization in the information ageThe role of the news mediaPartisan news media are often assumed to be polarizing, but concrete evidence detailing the polarizing role of the news media is mixed and beset by technical issues and geographic limitations. Again, the available literature is heavily U.S. and Europe-centric90 and the role of the news media in developing countries or in autocracies is rarely considered (a discussion of why this research is challenging is provided in Box 2).Most studies on the polarizing role of the news media concern either the supply side of the news–how the news media ecosystem is aligned with specific political ideologies–or the reception side–namely what news individuals consume. Addressing the former, Western news media systems have been categorized according to “political parallelism”, referring to the one-to-one correspondence between news outlets and political parties91,92. Using this framework, countries were classified as having low levels of political parallelism (e.g., the U.S., UK, Ireland), high levels but lowering over time (e.g., Austria, Belgium, Finland, Germany), or sustained high levels (e.g., Italy, France, Spain, Greece), revealing substantial differences in the politicization of the news media across developed countries, indicative of differences in ideological polarization.With regard to the consumption side, the proliferation of the internet has enabled easy access to increasingly diverse online content, some of which is highly partisan or radical36,93,94. Nevertheless, survey self-reports in 12 developed countries find little ideological bias in news consumption patterns22 with most individuals favouring centrist and moderate news sources, both on- and offline. This is despite fears that social media algorithms and “filter bubbles” will drive users to more radical content which in turn may drive increasing affective polarization (note that evidence for the existence of filter bubbles is very limited95,96,97,98). The U.S. has the most partisan news consumption, and individual communities do have an ideologically skewed news media diet (more so in the U.S. than other countries28,99), but absolute levels of partisan news consumption remain low. How to compare these results to countries where the politicization of the news media is more explicit is unclear, especially in authoritarian states where news outlets are forced to support the regime in power or risk political backlash100.But questions of news supply and demand do not answer whether the news media impact population-level polarization across the general population–do we expect changes in the news media to affect the polarization trends measured by traditional survey studies? Limited experimental evidence from the U.S. and Israel suggests that exposure to partisan news outlets and their messages reinforces prior political attitudes and affective polarization94,101,102,103. However, given the challenges in measuring exposure to (partisan) media using self-reports and in ascertaining their effects outside experimental settings, this research may not portray the role of the news media in the real world accurately. In this context, a major challenge is acquiring high-quality causal evidence104. This is because standard research designs are ill equipped to disentangle the effects of media exposure on the pre-existing beliefs and behaviours of experimental subjects, from the effect of experimental media treatments105. Additionally, establishing a population-level baseline for the content individuals are exposed to in their day-to-day lives is far from trivial, and may be conflated by self-selection biases in how individuals report their media consumption habits106.Without high-quality causal evidence, disentangling whether partisan media is a reflection of underlying population-level polarization due to self-selection biases, or whether partisan media is the cause of polarization, remains difficult105. Partly addressing this challenge, a 2020 study incentivised regular Fox News viewers in the U.S. to watch CNN for a month, finding that this switch had a moderating effect on their political attitudes107, thereby implying that continued exposure to partisan media is polarizing. Yet, other work finds null effects from exposure to politically aligned news media: Online behavioral data shows that partisan news consumption has little impact on attitudes or affective polarization in the U.S.33,108,109, but how these results translate to other world regions is unclear.Finally, the global loss of trust in the news, and very low engagement with news media content in general, mean that some researchers (but certainly not all) are increasingly questioning the polarizing role of the news media110,111,112,113. Self-surveys carried out across 46 countries (in Europe, the Americas, Asia, and some in Africa) show that interest in the news has fallen in the last decade19. No country has seen a growth in interest, and few countries have maintained stable levels–notably Finland, the Netherlands, and South Korea. Providing direct evidence of this, studies using web browsing data show that only 3.4% of all URLs visited by large samples in the U.S., Poland, and the Netherlands are news, and, even when individuals visit partisan domains, more than half of the content they consume is apolitical, such as sports or food recipes108; see also ref. 114.On social media–which has become a leading gateway for accessing news content19–news makes up only a small part of a user’s feed (less than 3% for U.S. Facebook users115). This lack of interest in the news and political content may partly explain diverging trends in mass and elite polarization across countries. However, a competing explanation suggests that the lack of engagement with highly politicized hot-button issues is not the key for understanding increasing affective polarization. Rather, a broader set of beliefs and preferences have become captured in a network of ideologically aligned belief packages (what DellaPosta refers to as an “oil spill”67), capturing issues which are seemingly apolitical (e.g., liberals’ preference for Lattes116), such that over time fewer issue-preferences remain which cut across partisan lines. Empirical research suggests that digital media play a key role in facilitating this process, whereby “identities, beliefs, and cultural preferences become drawn into an all-encompassing societal division”68, but evidence from outside the U.S. is limited.Rather than reducing exposure to political content, some argue that increasing public exposure to high-quality news may have a depolarizing effect. This would pull more moderate citizens into the democratic process, increase public resilience to misinformation117,118,119, and decrease their susceptibility to populist or hyper-partisan rhetoric which is believed to play a key role in driving affective polarization36. However, the widespread decline in trust in the news (16 out of 19 surveyed countries19) casts a shadow on this approach (see also112 for a discussion on the distinction between a lack of trust, and mistrust, of the media). Strong public service broadcasters may play a key role in maintaining trust, but in many markets politicians and other public figures are actively undermining trust through their continued criticism of the news media, e.g., in Greece, Peru, the U.S., and the UK19, or have deliberately defunded public service broadcasting.The role of the internet and social mediaThe internet and social media are playing an increasingly dominant role in how citizens (and politicians) produce and consume content, but this fundamental change to our lives is too often ignored in traditional polarization research. This should not be understood to imply that the internet and social media are ultimately responsible for recent trends in polarization–claims of social media’s negative impacts are often anecdotal, or based on correlational evidence120, and there are substantial differences across countries in the extent to which social media is seen as democratically harmful (Americans see social media as far more damaging than other surveyed populations121)–but it is critical to understand the role that this new information ecosystem may play in driving polarization.With a focus on user-user interactions, one perspective argues that social media drives polarization (both ideological and affective) by enabling politically engaged users to primarily interact with other ideologically aligned individuals. This phenomenon is sometimes referred to as interactional polarization9,122, and is observed across many platforms (albeit to varying extents123,124,125) and countries (e.g., in the U.S.126,127, Brazil128, Egypt129, Israel122, Venezuela130, Pakistan131, several Western countries9, and Japan132). These patterns are considered to be indicative of an epistemic bubble, or echo chamber123,133,134, driving greater ideological isolation, limiting out-group exposure, and thereby generating greater out-group affect135.However, another perspective argues that these homophilous interaction patterns do not constitute ideological isolation. First, social media interactions in most countries are not primarily with political actors or content9,127,136,137, and the interaction patterns of politically engaged users differ to those who are less politically engaged9. Second, when individuals interact with news content rather than individuals, their news diet is often politically moderate22,99,109. Third, while certain interaction types (like reposts) give the appearance of isolated communities, other interaction types (mentions or follows) reveal that out-group interaction is not so rare9,134,138,139, albeit these interactions are often toxic attacks (a feature common across many countries9,140). These differences across interaction types also translate to differences across user groups, with more active users less likely to view social media content as toxic and unconstructive, when compared to passive users (“lurkers”)141. Fourth, empirical evidence suggests that, contrary to the echo chamber hypothesis, ideologically isolated interactions on social media may increase the plurality of cross-cutting beliefs across user groups, whereas non-local interactions across partisan lines are precisely those which drive the alignment of ideology and identity68. Finally, even if online interactions are ideologically homogeneous, individuals may in fact be more isolated in their offline environments82,142,143, a factor likely to be highly country- and context-dependent. For these reasons, some scholars have suggested that research should look beyond the mere act of interaction and instead focus on interaction quality as well as on promoting prosocial behavior144,145,146,147.Despite these ongoing debates, we must not overlook the fact that social media has become a cornerstone of contemporary political campaigns, and many are convinced that effective social media campaigns have the power to swing political elections148,149, with the 2016 election of Donald Trump to the U.S. presidency being one key example. Evidence of substantial user heterogeneity in the engagement with, and sharing of, political content on social media150 suggests that there are likely substantial differences in the polarizing effect of social media across user groups, possibly related to factors such as online literacy (which we note is hard to measure accurately151). However, given that elections are often won on fine margins, the persuasive effects of social media on small subpopulations may still have substantial effects across broader society.These biases in exposure also apply to political elites. When strategists for the UK Labour Party were asked in 2022 how to improve the party’s electability, they cited MPs' use of Twitter as one of the biggest impediments to winning: “For members of parliament, Twitter has become a constant focus group on what (allegedly) matters”152. This is crucial–we must study the polarizing role of social media not because it is necessarily representative of the wider population, but precisely because its unrepresentativeness may skew perceptions of public and elite opinion.A core question in the study of digital media and polarization is whether biased information exposure leads groups to adopt more extreme views, driving changes in the level of polarization measured in political values surveys. The largest Facebook field experiment to date–a collaboration with Meta which was limited to U.S. researchers35–finds that reducing users’ exposure to content from like-minded sources does not measurably affect affective polarization or ideological extremity27. However, similar research (with smaller study sizes) suggests that such results are likely to be country-dependent: A U.S. study found that deactivating Facebook reduced both ideological and affective polarization153, while a replication in Bosnia-Herzegovina found an increase in polarization154, with evidence to suggest that this unexpected increase may be related to individuals’ lack of offline exposure to other ethnicities (although this result did not replicate in another ethnically polarized setting, Cyprus155).A key area of differentiation across countries is the different social platforms which dominate political messaging. In the last decade, Twitter, now X, and Facebook have been the dominant platforms for political communication in high-income democracies. But in countries such as Brazil, the Philippines, and India, WhatsApp is a substantially more important channel than X or Facebook for the dissemination of political content from political parties to voters156. In Eastern European states, Telegram plays a similar role in facilitating political communication. These comparative differences are important because platforms based around private peer-to-peer, or group, messaging pose a major barrier to data acquisition for academic research.The risks of AI and bots exacerbating polarizationThere are broad concerns that the polarizing effect of social media may be exacerbated by recommendation algorithms, AI generated content, and bots36,157,158,159, especially given the opaqueness of how these systems operate, and the scales at which they can affect the information that individuals are exposed to.Since the quantity of online content vastly exceeds what any individual can consume, social media companies must–by necessity–curate the content which their users are exposed to. For example, on Youtube recommended content makes up over 70% of the videos watched160. However, with AI-based recommendation systems optimizing for user engagement161, there are fears of a self-reinforcing cycle in which consumers are exposed to increasingly homogeneous content which–if the content is political in nature–may drive polarization (although this is not inevitable; AI could also be used to counter polarization, see below).Unfortunately, the commercial interests of tech companies mean that independently auditing this polarizing role is difficult. From the few studies available, U.S. data on Instagram and Facebook suggest that various changes to their recommendation algorithms do not impact ideological or affective polarization. These changes include removing re-shared content33, removing all content rankings34, and down-ranking content from ideologically aligned sources27. In contrast, recent work, again U.S. focused, shows that changes to algorithmic ranking on X can both raise or lower affective polarization depending on how content expressing partisan animosity, or antidemocratic attitudes, was ranked29. On Youtube, data suggests that recommendations are largely reflective of individual preferences, and that the algorithm only has a mild role in driving users towards more moderate political content37.Similar fears are raised concerning how AI generated content (often shared by bots) alters social media. Many studies have shown that bots amplify political content157,159,162,163, and a growing body of evidence demonstrates how large language models (LLMs) are often ideologically biased164, and can be tuned to promote specific political positions (e.g., xAI’s Grok, which was intended to counteract the “woke AI” bias of other models165). These observations have driven concerns that swarms of AI bots may, for example, produce a fabricated grassroots consensus, a fragmented shared reality, suppress or mobilise voters at the micro level, and erode institutional trust166.A global survey covering the elections of 2024 suggests that initial fears about the negative democratic impacts of generative AI were overblown167, but there is still cause for concern. One major study shows that LLMs are highly effective when deployed for political persuasion168, especially if they have been fine-tuned for persuasiveness. However, this capacity for persuasion can also be leveraged for good. Multiple studies demonstrate how LLMs can be used for democratic deliberation, to correct misconceptions, and to durably reduce conspiracy beliefs146,169,170,171. Crucially, whether AI will drive polarization, or contribute to mitigating its damaging consequences, is fundamentally a question of how AI is used and the different cultural contexts in which it is used, not an intrinsic property of the technology itself.Understanding the polarizing role of AI in non-experimental settings requires the ability to identify AI and bot generated content using appropriate detection tools. But the efficacy of these tools is context, and therefore country, dependent. For example, recent research shows that Google’s toxicity detection tool “Perspective” is far more likely to label content in German as toxic, than content in other languages172. These challenges may be exacerbated as generative AI further improves and it becomes harder to distinguish between machine- and user-generated content173. Improved detection tools will be required to distinguish between the true population-level impact of AI from a mirage of false polarization which reflects the actions of bots rather than human actors166. Such tools are especially necessary for low-resource languages. However, researchers should be conscious that the development of such tools can be a double-edged sword if used for malicious purposes (e.g., in authoritarian contexts; see Box 2). For a comprehensive discussion on the possible polarizing role of AI, especially if exploited by bad actors, see ref. 174.Limiting the internet’s polarizing roleIf we are to prevent digital media from driving polarization, we must consider how social media and AI companies moderate their platforms, and the country-specific regulatory environments in which they operate175,176. On social media, state regulators can issue takedown requests to remove content they deem to be problematic at a national level (see for example, https://help.x.com/en/rules-and-policies/post-withheld-by-country). However, in most cases regulators cannot remove this content without using the–typically U.S.-based–social media companies as an intermediary. Furthermore, there have been numerous cases where takedown requests have been rejected, e.g., in India177. This state of affairs puts an outsized emphasis on U.S. regulators, U.S. political interests, and a U.S. philosophy on platform regulation.This outsized emphasis on U.S. interests has directly impacted the quality of polarization research. Notably, Meta’s U.S. election studies–the gold standard for acquiring causal evidence into the online drivers of polarization27,33,34–are part of a recent academic-industry collaboration. But such collaborations are currently dependent on the goodwill of, and the parameters set by, social media companies. These companies have incentives to prioritize U.S. research, which they pursue by granting researchers some “independence by permission”35. Consequently, the Meta studies only provide evidence for the U.S., and only involve U.S.-based researchers. Other scientists may use more creative methods to causally estimate the role of social media in driving polarization37,178, but without the ability to alter a social platform’s algorithms, and with no passive consumption data, the questions which such studies can answer are limited.These systemic biases are a barrier to diversifying polarization research leading some to call for social media research to be democratized30,31. The European Union’s new Digital Services Act (DSA) has the potential to transform access to online platform data, enabling a new phase of in-depth research, but only if the data access provided is sufficient for reliable and consistent measurements of online social phenomena (social media studies are especially prone to sampling bias137). Early evidence suggests poor compliance with the DSA by tech companies32, and for researchers outside Europe and the U.S. barriers to data access will remain. Similarly, the EU’s AI Act aims to establish a common regulatory and legal framework for the use of AI, and in doing so aims to limit AI’s potentially harmful impacts, but whether this framework will be effective is unclear (in July 2025, Meta stated that it will not sign the EU’s AI code of practice).When considering digital media regulation at a global scale, it is important to emphasise that regulators (and governments) may make decisions in the interest of preventing online harms, but they could equally decide to manipulate online discourses in order to support the political regime in power179,180,181. Either way, these choices are fundamentally political and rest on value judgments which differ between countries. Given the outsized role of U.S. platforms and regulators, there has been a growing desire for digital sovereignty182, especially with regards to the development of sovereign AI183 and the regulation of online harms (e.g., Australia’s social media ban for minors). Academics should be conscious of how diverging regulatory environments may bias future polarization research, and should consider how these environments may impact evolving polarization trends.Challenges moving forward: countering polarization globallyPreventing the potential damaging societal impacts of political polarization requires effective and long-lasting interventions. However, the current dominance of U.S. research has put an outsized emphasis on developing interventions for affective polarization between Democrats and Republicans in the U.S.26, with few interventions having been tested elsewhere. These include measures promoting intergroup contact184, emphasizing a shared identity between partisans185, meaningful, non-political conversations across party lines186,187, and inoculating against polarizing content188,189. However, existing approaches often suffer from not being both durable (lasting for extended periods) and scalable (easily and effectively distributed)25, and in some cases interventions counter-intuitively increase polarization, or result in other negative side-effects (see Box 1 in ref. 25; also154,190,191). Relying on such narrowly designed measures risks introducing inappropriate or even counterproductive solutions in diverse societal and cultural contexts.Fortunately, scholars are increasingly aware of the need to study interventions globally, with several projects already underway (for example globalsocialmediastudy.com/ and templetonworldcharity.org/LLPW). However, the impact of these studies risks remaining limited without systemic and cultural changes in how polarization research is produced and communicated.Researchers working on understudied regions often face career disadvantages due to the perceived lack of visibility, relevance, and therefore impact, of their work192,193. Academic institutions, publishers, and policymakers must actively promote and reward work that brings global balance to polarization research. This is not a call for less research on the U.S., which will likely remain prolific even if this research has been heavily impacted by recent funding cuts in the U.S., but rather to increase recognition and support for research that expands our understanding to include diverse global perspectives.A first step to improving the current state of affairs, given the constraints on academic resources, is to prioritize comparative research that identifies both shared and unique aspects of polarization across different countries. As part of this, it is crucial that country-specific research avoids over-generalizing their results. For example, journals could mandate that studies focusing solely on the U.S. explicitly state this in their title or abstract (many currently don’t), as is commonplace for research focusing on almost any other country. Additionally, particular efforts are needed to address the conceptual challenges of studying polarization in authoritarian states, and in low-income democracies where key political concepts, used widely in the Global North, may not apply (e.g., the framing of ideological polarization along a left-right axis15,51).To prevent further global disparities in polarization research, we argue that researchers must push back against growing restrictions on data access now. Since 2023, several social media companies (including Meta, X, Reddit) have shuttered academic access to their platforms’ data, and new opportunities for data access such as the European Union’s Digital Services Act remain largely unproven30,31,32. If effective, these data provisions may enable continued observational research, but this will not enable the high-quality experimental research which is necessary to establish clear causal relationships between various aspects of polarization and digital media use37,178. Such research will likely only be possible across countries, and at scale, with the buy-in of social media companies, or with substantially greater cooperation across the research community. Interdisciplinary, collaborative research is necessary to disentangle the complex web of factors driving different forms of polarization, and greater effort is required to establish the aspects of polarization on which there is academic consensus. One option here would be for interdisciplinary researchers to collaborate on the production of polarization “consensus statements”, as is common in medical research.Building a globally equitable picture of political polarization, and its complex manifestations and drivers across different publics and media modalities, requires integrating research methods from the social sciences, with the advanced data science approaches which are necessary for understanding the structure of, and spread of information through, our digital world, and how this content impacts the political beliefs of individuals. Examples of promising methods which bridge disciplinary divides include study designs leveraging matched survey and digital trace data58, including data acquired through data donation frameworks59, and novel computational methods for understanding the interdependent structure of political attitudes using networks194,195. However, for researchers newly entering the study of political polarization, there is a burden of responsibility to ensure that proposed technical solutions to key research challenges remain conscious of polarization’s conceptual nuances. Complex feedback loops between different aspects of polarization cannot be understood if this nuance is lost.Finally, as our understanding of polarization evolves and new interventions are developed, both academics and policymakers must account for the unique regulatory and practical challenges of each country176,196. Interventions must be both ethically and practically tailored to the local context. We cannot assume that societies outside the U.S.–where empirical evidence is comparably limited–face the same problems which can be solved with the same solutions. Academics play a critical role in gathering evidence to better understand how polarization manifests across countries, but the implementation of interventions is necessarily political, and often controversial. These decisions must rest with local stakeholders and decision makers. However, without high-quality evidence which is contextually appropriate, even the best-intentioned decision makers may draw inappropriate conclusions from flawed evidence, risking the further erosion of democratic institutions.ReferencesMcCoy, J., Rahman, T. & Somer, M. “Polarization and the global crisis of democracy: common patterns, dynamics, and pernicious consequences for democratic polities”. Am. Behav. Sci. 62, 16–42 (2018).Article Google Scholar World Economic Forum, “The Global Risks Report 2025.” https://www.weforum.org/publications/global-risks-report-2025/in-full/ (2025).IPSOS, “Global Trends about Polarization: A Divided World,” Global Trends about Polarization: A Divided World. https://www.ipsos.com/sites/default/files/2023-Ipsos-Global-Trends-Report.pdf (2023).Norris, P. & Inglehart, R. Cultural backlash: Trump, Brexit, and authoritarian populism. Cambridge University Press, (2019).Orhan, Y. E. The relationship between affective polarization and democratic backsliding: comparative evidence. Democratization 29, 714–735 (2022).Article Google Scholar Iyengar, S., Lelkes, Y., Levendusky, M., Malhotra, N. & Westwood, S. J. The origins and consequences of affective polarization in the United States. Annu. Rev. Polit. Sci. 22, 129–146 (2019).Article Google Scholar Wagner, M. Affective polarization in multiparty systems. Elect. Stud. 69, 102199 (2021).Article Google Scholar Gidron, N., Adams, J. & Horne, W. American Affective Polarization In Comparative Perspective. Cambridge University Press (2020).Falkenberg, M., Zollo, F., Quattrociocchi, W., Pfeffer, J. & Baronchelli, A. Patterns of partisan toxicity and engagement reveal the common structure of online political communication across countries. Nat. Commun. 15, 9560 (2024).Article ADS CAS PubMed PubMed Central Google Scholar Citrin, J. & Stoker, L. Political trust in a cynical age. Annu. Rev. Polit. Sci. 21, 49–70 (2018).Article Google Scholar Hetherington M. J. & Rudolph, T. J. Why Washington Won’t Work: Polarization, Political Trust, and the Governing Crisis, Illustrated edition. University of Chicago Press, (2015).Pew Research Center. As Partisan Hostility Grows, Signs of Frustration With the Two-Party System. https://www.pewresearch.org/wp-content/uploads/sites/20/2022/08/PP_2022.09.08_partisan-hostility_REPORT.pdf (2022).Pew Research Center, “The polarization in today’s Congress has roots that go back decades,”. https://www.pewresearch.org/short-reads/2022/03/10/the-polarization-in-todays-congress-has-roots-that-go-back-decades/ (2024).Kalmoe, N. P. & Mason, L. Radical American partisanship: mapping violent hostility, its causes, and the consequences for democracy. in Chicago studies in American politics. The University of Chicago Press, (2022).Nielsen, L. H. & Petersen, M. B. Beyond left and right: the role of system trust in COVID-19 attitudes and behaviors across eight western countries. Eur. J. Polit. Res. 64, 1233–1256 (2025).Falkenberg, M. et al. Growing polarization around climate change on social media. Nat. Clim. Change 12, 1114–1121 (2022).Article ADS Google Scholar Caldwell, D., Cohen, G. & Vivyan, N. Long-run trends in partisan polarization of climate policy-relevant attitudes across countries. Environ. Polit. https://doi.org/10.1080/09644016.2024.2403957 (2024).McCoy, J., Press, B., Somer, M., & Tuncel, O. “Reducing Pernicious Polarization: A Comparative Historical Analysis of Depolarization,” Carnegie Endowment for International Peace, https://carnegie-production-assets.s3.amazonaws.com/static/files/McCoy_et_al_-_Polarization_final_3.pdf (2022).Newman, N., Fletcher, R., Eddy, K., Robertson, C. & Nielsen, R. “Digital news report 2023,” Reuters Institute for the Study of Journalism, (2023).Röllicke, L. Polarisation, identity and affect - conceptualising affective polarisation in multi-party systems. Elect. Stud. 85, 102655 (2023).Article Google Scholar Carothers, T. & O’Donoghue, A. “Introduction,” in Democracies Divided, Carothers, T. & O’Donoghue, A., Eds., in The Global Challenge of Political Polarization, Brookings Institution Press, (2019).Fletcher, R., Cornia, A. & Nielsen, R. K. How polarized are online and offline news audiences? a comparative analysis of twelve countries. Int. J. Press. 25, 169–195 (2020).Article Google Scholar Finkel, E. J. et al. Political sectarianism in America. Science 370, 533–536 (2020).Article CAS PubMed Google Scholar Jost, J. T., Baldassarri, D. S. & Druckman, J. N. Cognitive–motivational mechanisms of political polarization in social-communicative contexts. Nat. Rev. Psychol. 1, 560–576 (2022).Article PubMed PubMed Central Google Scholar Hartman, R. et al. Interventions to reduce partisan animosity. Nat. Hum. Behav. 6, 1194–1205 (2022).Article PubMed Google Scholar Voelkel, J. G. et al. Megastudy testing 25 treatments to reduce antidemocratic attitudes and partisan animosity. Science 386, eadh4764 (2024).Article CAS PubMed Google Scholar Nyhan, B. et al. Like-minded sources on Facebook are prevalent but not polarizing. Nature 620, 137–144 (2023).Article ADS CAS PubMed PubMed Central Google Scholar González-Bailón, S. et al. Asymmetric ideological segregation in exposure to political news on Facebook. Science 381, 392–398 (2023).Article ADS PubMed Google Scholar Piccardi, T. et al. Reranking partisan animosity in algorithmic social media feeds alters affective polarization. Science 390, eadu5584 (2025).Article CAS PubMed Google Scholar Roozenbeek, J. & Zollo, F. Democratize social-media research—with access and funding. Nature 612, 404–404 (2022).Article CAS PubMed Google Scholar Rathje, S. To tackle social-media harms, mandate data access for researchers. Nature 633, 36–36 (2024).Article ADS CAS PubMed Google Scholar Jaursch, J., Ohme, J. & Klinger, U. “Enabling research with publicly accessible platform data: early DSA compliance issues and suggestions for improvement,” https://doi.org/10.34669/WI.WPP/9 (2024).Guess, A. M. et al. Reshares on social media amplify political news but do not detectably affect beliefs or opinions. Science 381, 404–408 (2023).Article ADS CAS PubMed Google Scholar Guess, A. M. et al. How do social media feed algorithms affect attitudes and behavior in an election campaign? Science 381, 398–404 (2023).Article ADS CAS PubMed Google Scholar Wagner, M. W. Independence by permission. Science 381, 388–391 (2023).Article ADS CAS PubMed Google Scholar Haroon, M. et al. Auditing YouTube’s recommendation system for ideologically congenial, extreme, and problematic recommendations. Proc. Natl. Acad. Sci. 120, e2213020120 (2023).Article CAS PubMed PubMed Central Google Scholar Hosseinmardi, H. et al. Causally estimating the effect of YouTube’s recommender system using counterfactual bots. Proc. Natl. Acad. Sci. 121, e2313377121 (2024).Article CAS PubMed PubMed Central Google Scholar Santos, F. P., Lelkes, Y. & Levin, S. A. Link recommendation algorithms and dynamics of polarization in online social networks. Proc. Natl. Acad. Sci. 118, e2102141118 (2021).Article MathSciNet CAS PubMed PubMed Central Google Scholar Abramowitz, A. I. & Saunders, K. L. Is polarization a myth? J. Polit. 70, 542–555 (2008).Article Google Scholar Fiorina, M. P. & Abrams, S. J. Political polarization in the American public. Annu. Rev. Polit. Sci. 11, 563–588 (2008).Article Google Scholar DiMaggio, P., Evans, J. & Bryson, B. Have American’s social attitudes become more polarized? Am. J. Sociol. 102, 690–755 (1996).Article Google Scholar Sanders, D. et al. Citizens and the European Polity: Mass Attitudes Towards The European And National Polities, First Edition. In Intune Series. Oxford: Oxford University Press, (2012).Ford, R. & Jennings, W. The changing cleavage politics of Western Europe. Annu. Rev. Polit. Sci. 23, 295–314 (2020).Article Google Scholar Bramson, A. et al. Understanding polarization: meanings, measures, and model evaluation. Philos. Sci. 84, 115–159 (2017).Article MathSciNet Google Scholar Iyengar, S., Sood, G. & Lelkes, Y. Affect, not ideology. Public Opin. Q. 76, 405–431 (2012).Article Google Scholar Mason, L. et al. I disrespectfully agree’: the differential effects of partisan sorting on social and issue polarization: partisan sorting and polarization. Am. J. Polit. Sci. 59, 128–145 (2015).Article Google Scholar Dalton, R. J. Modeling ideological polarization in democratic party systems. Elect. Stud. 72, 102346 (2021).Article Google Scholar Emanuele, V. & Marino, B. Party system ideological polarization in Western Europe: data, trends, drivers, and links with other key party system properties (1945–2021). Polit. Res. Exch. 6, 2399095 (2024).Google Scholar Young, D. J. et al. A new measure of issue polarization using k -means clustering: US trends 1988–2024 and predictors of polarization across the world. R. Soc. Open Sci. 13, 251428 (2026).Article ADS Google Scholar Pew Research Center, “Political Polarization in the American Public,” https://www.pewresearch.org/wp-content/uploads/sites/4/2014/06/6-12-2014-Political-Polarization-Release.pdf (2014).Geser, H. The limits of ideological globalization. Current patterns of ‘left and right’ in different geographical regions. https://doi.org/10.5167/UZH-48431 (2008).Beattie, P., Chen, R. & Bettache, K. When left is right and right is left: the psychological correlates of political ideology in China. Polit. Psychol. 43, 457–488 (2022).Article Google Scholar Pan, J. & Xu, Y. China’s Ideological Spectrum. J. Polit. 80, 254–273 (2018).Article Google Scholar Norris, P. et al. Measuring populism worldwide. Party Polit 26, 697–717 (2020).Article Google Scholar Boxell, L., Gentzkow, M. & Shapiro, J. M. “Cross-country trends in affective polarization.” Rev. Econ. Stat. https://doi.org/10.1162/rest_a_01160 (2022).Reiljan, A., Garzia, D., Ferreira Da Silva, F. & Trechsel, A. H. Patterns of affective polarization toward parties and leaders across the democratic world. Am. Polit. Sci. Rev. 118, 654–670 (2024).Article Google Scholar Coppedge, M. et al. Varieties of Democracy: Measuring Two Centuries of Political Change. (Cambridge University Press, 2020).Stier, S., Breuer, J., Siegers, P. & Thorson, K. Integrating survey data and digital trace data: key issues in developing an emerging field. Soc. Sci. Comput. Rev. 38, 503–516 (2020).Article Google Scholar Kmetty, Z. et al. Determinants of willingness to donate data from social media platforms. Inf. Commun. Soc. 28, 1324–1349 (2025).Article Google Scholar McCoy, J. & Somer, M. Toward a theory of pernicious polarization and how it harms democracies: comparative evidence and possible remedies. Ann. Am. Acad. Pol. Soc. Sci. 681, 234–271 (2019).Article Google Scholar Kreiss, D. & McGregor, S. C. A review and provocation: on polarization and platforms. New Media Soc 26, 556–579 (2024).Article Google Scholar Rogowski, J. C. & Sutherland, J. L. How Ideology Fuels Affective Polarization. Polit. Behav. 38, 485–508 (2016).Article Google Scholar Webster, S. W. & Abramowitz, A. I. The ideological foundations of affective polarization in the U.S. electorate. Am. Polit. Res. 45, 621–647 (2017).Article Google Scholar Mason, L. et al. Uncivil agreement: How Politics Became Our Identity. (University of Chicago Press, 2018).Brown, J. R. & Enos, R. D. The measurement of partisan sorting for 180 million voters. Nat. Hum. Behav. 5, 998–1008 (2021).Article PubMed Google Scholar Lelkes, Y. et al. Affective polarization and ideological sorting: a reciprocal, albeit weak, relationship. Forum (Genova) 16, 67–79 (2018).Article Google Scholar DellaPosta, D. et al. Pluralistic collapse: the ‘oil spill’ model of mass opinion polarization. Am. Sociol. Rev. 85, 507–536 (2020).Article Google Scholar Törnberg, P. et al. How digital media drive affective polarization through partisan sorting. Proc. Natl. Acad. Sci. 119, e2207159119 (2022).Article PubMed PubMed Central Google Scholar Schul, Y. & Ganzach, Y. Divergent views of party positions: How ideology and own issue position shape party perception through convergence and divergence processes. Group Process. Intergroup Relat. 28, 256–276 (2025).Article Google Scholar Bankert, A., Huddy, L. & Rosema, M. Measuring partisanship as a social identity in multi-party systems. Polit. Behav. 39, 103–132 (2017).Article Google Scholar Brewer, M. B. & Pierce, K. P. Social identity complexity and outgroup tolerance. Pers. Soc. Psychol. Bull. 31, 428–437 (2005).Article PubMed Google Scholar Brewer, M. B. Multiple identities and identity transition: implications for Hong Kong. Int. J. Intercult. Relat. 23, 187–197 (1999).Article Google Scholar Gidron, N., Adams, J. & Horne, W. “How ideology, economics and institutions shape affective polarization in democratic polities,” in Annual conference of the American Political Science Association, (2018).Reiljan, A. et al. Fear and loathing across party lines’ (also) in Europe: affective polarisation in European party systems. Eur. J. Polit. Res. 59, 376–396 (2020).Article Google Scholar The American Political Science Review, “Part III. The Prospect for Action,” Am. Polit. Sci. Rev. 44, 85, https://doi.org/10.2307/1951001 (1950).Hobolt S. B. & Tilley, J. Affective Polarization Around Issues. in Handbook of Affective Polarization, (eds. Torcal, M.& Harteveld, E.) https://doi.org/10.4337/9781035310609.00025 (Edward Elgar Publishing, 2025).Young, D. J. & de-Wit, L. H. Affective polarization within parties. Polit. Psychol. https://doi.org/10.1111/pops.12973. (2024).Areal, J. Them’ without ‘us’: negative identities and affective polarization in Brazil. Polit. Res. Exch. 4, 2117635 (2022).Google Scholar Vowles, J., Banducci, S. A. & Karp, J. A. Forecasting and evaluating the consequences of electoral change in New Zealand. Acta Polit 41, 267–284 (2006).Article Google Scholar Riera, P. & Madariaga, A. G. Overlapping polarization: on the contextual determinants of the interplay between ideological and affective polarization. Elect. Stud. 84, 102628 (2023).Article Google Scholar Iversen, T. & Soskice, D. Information, inequality, and mass polarization: ideology in advanced democracies. Comp. Polit. Stud. 48, 1781–1813 (2015).Article Google Scholar Monti, C., D’Ignazi, J., Starnini, M. & De Francisci Morales, G. “Evidence of Demographic rather than Ideological Segregation in News Discussion on Reddit,” in Proceedings of the ACM Web Conference 2023, https://doi.org/10.1145/3543507.3583468 (2023).Stewart, A. J., McCarty, N. & Bryson, J. J. Polarization under rising inequality and economic decline. Sci. Adv. 6, eabd4201 (2020).Article ADS PubMed PubMed Central Google Scholar McCarty, N. M. et al. Polarisation: What Everyone Needs To Know. (University Press, 2019).N. M. McCarty, K. T. Poole, and H. Rosenthal, Polarized America: the dance of ideology and unequal riches, Second edition. in The Walras-Pareto Lectures. The MIT Press (2016).Hübscher, E., Sattler, T. & Wagner, M. Does austerity cause polarization? Br. J. Polit. Sci. 53, 1170–1188 (2023).Article Google Scholar Xia, Y. et al. “The Russian invasion of Ukraine selectively depolarized the Finnish NATO discussion,” Preprint arXiv: http://arxiv.org/abs/2212.07861 (2023).Somer, M. & McCoy, J. Déjà vu? polarization and endangered democracies in the 21st century. Am. Behav. Sci. 62, 3–15 (2018).Article Google Scholar M. Torcal and E. Harteveld, Eds., Handbook of affective polarization. in Elgar handbooks in political science. Cheltenham, UK; Northampton, MA, USA: Edward Elgar Publishing, (2025).Kubin, E. & Von Sikorski, C. The role of (social) media in political polarization: a systematic review. Ann. Int. Commun. Assoc. 45, 188–206 (2021).Google Scholar D. C. Hallin and P. Mancini, Comparing Media Systems: Three Models Of Media And Politics. Cambridge University Press, (2004).Brüggemann, M., Engesser, S., Büchel, F., Humprecht, E. & Castro, L. Hallin and Mancini revisited: four empirical types of western media systems: hallin and mancini revisited. J. Commun. 64, 1037–1065 (2014).Article Google Scholar Van Aelst, P. et al. Political communication in a high-choice media environment: a challenge for democracy? Ann. Int. Commun. Assoc. 41, 3–27 (2017).Google Scholar Stroud, N. J. et al. Polarization and partisan selective exposure. J. Commun. 60, 556–576 (2010).Article Google Scholar Bruns, A. et al. Are filter bubbles real? in Digital futures. (Cambridge, UK; Medford, MA: Polity Press, 2019).Ross Arguedas, A., Robertson, C. T., Fletcher, R. & Nielsen, R. K. “Echo chambers, filter bubbles, and polarisation: a literature review,” Reuters Institute for the Study of Journalism. https://doi.org/10.60625/RISJ-ETXJ-7K60 (2022).Cardenal, A. S., Aguilar-Paredes, C., Galais, C. & Pérez-Montoro, M. Digital technologies and selective exposure: how choice and filter bubbles shape news media exposure. Int. J. Press. 24, 465–486 (2019).Article Google Scholar Robertson, R. E. et al. Users choose to engage with more partisan news than they are exposed to on Google search. Nature 618, 342–348 (2023).Article ADS CAS PubMed Google Scholar Mangold, F., Schoch, D. & Stier, S. Ideological self-selection in online news exposure: evidence from Europe and the US. Sci. Adv. 10, eadg9287 (2024).Article PubMed PubMed Central Google Scholar Enikolopov R. & Petrova, M. Media Capture: Empirical Evidence. in Handbook of Media Economics, vol. 1, 687–700 (Elsevier, 2015).Garrett, R. K. et al. Implications of pro- and counterattitudinal information exposure for affective polarization: partisan media exposure and affective polarization. Hum. Commun. Res. 40, 309–332 (2014).Article Google Scholar Garrett, R. K., Long, J. A. & Jeong, M. S. From partisan media to misperception: affective polarization as mediator. J. Commun. 69, 490–512 (2019).Article Google Scholar Levendusky, M. S. Why do partisan media polarize viewers? Am. J. Polit. Sci. 57, 611–623 (2013).Article Google Scholar Prior, M. et al. Media and political polarization. Annu. Rev. Polit. Sci. 16, 101–127 (2013).Article Google Scholar J. De Benedictis-Kessner, M. A. Baum, and A. J. Berinsky, “Polarization and Media Usage: Disentangling Causality,” in The Oxford Handbook of Electoral Persuasion, E. Suhay, B. Grofman, and A. H. Trechsel, Eds., Oxford University Press, 553–571 https://doi.org/10.1093/oxfordhb/9780190860806.013.22 (2020).Arendt, F. Media stereotypes, prejudice, and preference-based reinforcement: toward the dynamic of self-reinforcing effects by integrating audience selectivity. J. Commun. 73, 463–475 (2023).Article Google Scholar Broockman, D. E. & Kalla, J. L. Consuming cross-cutting media causes learning and moderates attitudes: A field experiment with Fox News viewers. J. Polit. 87, 246–261 (2025).Wojcieszak, M. et al. “No polarization from partisan news: over-time evidence from trace data,”. Int. J. Press. 28, 601–626 (2023).Article Google Scholar Guess, A. M. et al. (Almost) everything in moderation: new evidence on Americans’ online media diets. Am. J. Polit. Sci. 65, 1007–1022 (2021).Article Google Scholar Strömbäck, J. et al. News media trust and its impact on media use: toward a framework for future research. Ann. Int. Commun. Assoc. 44, 139–156 (2020).Google Scholar Fletcher, R. et al. The link between changing news use and trust: longitudinal analysis of 46 countries. J. Commun. 75, 1–15 (2025).Article Google Scholar Park, S. et al. The relationship between news trust, mistrust and audience disengagement. Journalism 26, 2285–2304 (2025).Article Google Scholar Guess, A. M., Barberá, P., Munzert, S. & Yang, J. The consequences of online partisan media. Proc. Natl. Acad. Sci. 118, e2013464118 (2021).Article CAS PubMed PubMed Central Google Scholar Allen, J., Howland, B., Mobius, M., Rothschild, D. & Watts, D. J. Evaluating the fake news problem at the scale of the information ecosystem. Sci. Adv. 6, eaay3539 (2020).Article ADS PubMed PubMed Central Google Scholar Eisenach, J. A. et al. Meta and the News: Assessing the Value of the Bargain. NERA Economic Consulting, https://about.fb.com/wp-content/uploads/2023/03/NERA-Global-White-Paper-FINAL-March-2023.pdf (2023).DellaPosta, D., Shi, Y. & Macy, M. Why Do Liberals Drink Lattes? Am. J. Sociol. 120, 1473–1511 (2015).Article Google Scholar Arechar, A. A. et al. Understanding and combatting misinformation across 16 countries on six continents. Nat. Hum. Behav. 7, 1502–1513 (2023).Article PubMed Google Scholar Del Vicario, M. et al. The spreading of misinformation online. Proc. Natl. Acad. Sci. 113, 554–559 (2016).Article ADS PubMed PubMed Central Google Scholar Van Der Linden, S. Misinformation: susceptibility, spread, and interventions to immunize the public. Nat. Med. 28, 460–467 (2022).Article ADS PubMed Google Scholar Lorenz-Spreen, P., Oswald, L., Lewandowsky, S. & Hertwig, R. A systematic review of worldwide causal and correlational evidence on digital media and democracy. Nat. Hum. Behav. 7, 74–101 (2022).Article PubMed PubMed Central Google Scholar Pew Research Center, “Social media seen as mostly good for democracy across many nations, but U.S. is a major outlier,” (2022).Yarchi, M., Baden, C. & Kligler-Vilenchik, N. Political polarization on the digital sphere: a cross-platform, over-time analysis of interactional, positional, and affective polarization on social media. Polit. Commun. 38, 98–139 (2021).Article Google Scholar Cinelli, M., De Francisci Morales, G., Galeazzi, A., Quattrociocchi, W. & Starnini, M. The echo chamber effect on social media. Proc. Natl. Acad. Sci. 118, e2023301118 (2021).Article CAS PubMed PubMed Central Google Scholar Bessi, A. et al. Users Polarization on Facebook and Youtube. PLoS ONE 11, e0159641 (2016).Article PubMed PubMed Central Google Scholar Waller, I. & Anderson, A. Quantifying social organization and political polarization in online platforms. Nature 600, 264–268 (2021).Article ADS CAS PubMed Google Scholar Barberá, P. et al. Birds of the same feather tweet together: Bayesian ideal point estimation using Twitter data. Anal. 23, 76–91 (2015).Google Scholar Wojcieszak, M., Casas, A., Yu, X., Nagler, J. & Tucker, J. A. Most users do not follow political elites on Twitter; those who do show overwhelming preferences for ideological congruity. Sci. Adv. 8, eabn9418 (2022).Article PubMed PubMed Central Google Scholar Recuero, R., Soares, F. B. & Gruzd, A. Hyperpartisanship, disinformation and political conversations on twitter: the Brazilian presidential election of 2018. Proc. Int. AAAI Conf. Web Soc. Media 14, 569–578 (2020).Article Google Scholar I. Weber, V. R. K. Garimella, and A. Batayneh, “Secular vs. Islamist polarization in Egypt on Twitter,” in Proceedings of the 2013 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, Niagara Ontario Canada: ACM, https://doi.org/10.1145/2492517.2492557 (2013).Morales, A. J., Borondo, J., Losada, J. C. & Benito, R. M. Measuring political polarization: Twitter shows the two sides of Venezuela. Chaos Interdiscip. J. Nonlinear Sci. 25, 033114 (2015).Article CAS Google Scholar Baqir, A., Galeazzi, A., Drocco, A. & Zollo, F. “Social media polarization reflects shifting political alliances in Pakistan,” 2023. Preprint at http://arxiv.org/abs/2309.08075 (2024).Takikawa, H. & Nagayoshi, K. Political polarization in social media: Analysis of the ‘Twitter political field’ in Japan. in 2017 IEEE International Conference on Big Data (Big Data), Boston, MA: IEEE, 3143–3150 https://doi.org/10.1109/BigData.2017.8258291 (2017).Nguyen, C. T. et al. Echo chambers and epistemic bubbles. Episteme 17, 141–161 (2020).Article Google Scholar Terren, L. et al. Open University of Catalonia, R. Borge-Bravo, and Open University of Catalonia, “Echo Chambers on Social Media: A Systematic Review of the literature. Rev. Commun. Res. 9, 99–118 (2021).Google Scholar Sunstein, C. R. The Law of Group Polarization. J. Polit. Philos. 10, 175–195 (2002).Article Google Scholar Eady, G., Nagler, J., Guess, A., Zilinsky, J. & Tucker, J. A. How many people live in political bubbles on social media? evidence from linked survey and twitter data. SAGE Open 9, 215824401983270 (2019).Article Google Scholar Di Bona, G. et al., “Sampled social media data risk biased and inconsistent estimates of online social phenomena,” Preprint at SocArXiv. https://doi.org/10.31235/osf.io/37jz9_v1 (2026).Williams, H. T. P., McMurray, J. R., Kurz, T. & Hugo Lambert, F. Network analysis reveals open forums and echo chambers in social media discussions of climate change. Glob. Environ. Change 32, 126–138 (2015).Article Google Scholar Gruzd, A. & Roy, J. Investigating political polarization on twitter: a Canadian Perspective: investigating political polarization on twitter. Policy Internet 6, 28–45 (2014).Article Google Scholar Mekacher, A., Falkenberg, M. & Baronchelli, A. The systemic impact of deplatforming on social media. PNAS Nexus 2, pgad346 (2023).Article PubMed PubMed Central Google Scholar Oswald, L., Schulz, W. S. & Lorenz-Spreen, P. Disentangling participation in online political discussions with a collective field experiment. Sci. Adv. 11, eady8022 (2025).Article PubMed Central Google Scholar Niezink, N. M. D. “Offline Consequences of Echo Chambers,” in Proceedings of the Fourth International Workshop on Social Sensing, Montreal QC Canada: ACM, 5–5 https://doi.org/10.1145/3313294.3313381 (2019).Guess, A., Nyhan, B., Lyons, B. & Reifler, J. “Avoiding the echo chamber about echo chambers. Kn. Found. 2, 1–25 (2018).Google Scholar Bail, C. et al. Breaking the Social Media Prism: How to Make Our Platforms Less Polarizing. (Princeton, NJ: Princeton University Press, 2022).Klein, M., Gruzd, A. & Lannigan, J. Using deliberation-centric social network analysis to measure balkanization. SSRN Electron. J. https://doi.org/10.2139/ssrn.2914554 (2017).Article Google Scholar Argyle, L. P. et al. Leveraging AI for democratic discourse: chat interventions can improve online political conversations at scale. Proc. Natl. Acad. Sci. 120, e2311627120 (2023).Article CAS PubMed PubMed Central Google Scholar González-Bailón, S. & Lelkes, Y. Do social media undermine social cohesion? a critical review. Soc. Issues Policy Rev. 17, 155–180 (2023).Article Google Scholar Tapsell, R. et al. Social media and elections in southeast asia: the emergence of subversive, underground campaigning. Asian Stud. Rev. 45, 117–134 (2021).Article Google Scholar Oh, C. & Kumar, S. How Trump won: the role of social media sentiment in political elections. PACIS 2017 Proceedings, 48, (2017).Mosleh, M., Yang, Q., Zaman, T., Pennycook, G. & Rand, D. G. Differences in misinformation sharing can lead to politically asymmetric sanctions. Nature 634, 609–616 (2024).Article ADS CAS PubMed PubMed Central Google Scholar Guess, A. M. & Munger, K. Digital literacy and online political behavior. Polit. Sci. Res. Methods 11, 110–128 (2023).Article Google Scholar S. Payne, “Twitter’s bubbles are a blight on British politics,” The Financial Times, https://www.ft.com/content/66df41ad-5835-4f8a-b9f8-7e819722d0c7 (2025).Allcott, H., Braghieri, L., Eichmeyer, S. & Gentzkow, M. The welfare effects of social media. Am. Econ. Rev. 110, 629–676 (2020).Article Google Scholar Asimovic, N., Nagler, J., Bonneau, R. & Tucker, J. A. Testing the effects of Facebook usage in an ethnically polarized setting. Proc. Natl. Acad. Sci. 118, e2022819118 (2021).Article CAS PubMed PubMed Central Google Scholar Asimovic, N., Nagler, J. & Tucker, J. A. Replicating the effects of Facebook deactivation in an ethnically polarized setting. Res. Polit. 10, 20531680231205157 (2023).Article Google Scholar Chauchard, S. & Garimella, K. What circulates on partisan WhatsApp in India? insights from an unusual dataset. J. Quant. Descr. Digit. Media, https://doi.org/10.51685/jqd.2022.006 (2022).Shao, C. et al. The spread of low-credibility content by social bots. Nat. Commun. 9, 4787 (2018).Article ADS PubMed PubMed Central Google Scholar Gruzd, A., Mai, P. & Soares, F. B. How coordinated link sharing behavior and partisans’ narrative framing fan the spread of COVID-19 misinformation and conspiracy theories. Soc. Netw. Anal. Min. 12, 118 (2022).Article PubMed PubMed Central Google Scholar Yang, K. et al. Arming the public with artificial intelligence to counter social bots. Hum. Behav. Emerg. Technol. 1, 48–61 (2019).Article ADS Google Scholar J. E. Solsman, “YouTube’s AI is the puppet master over most of what you watch,” CNET, https://www.cnet.com/tech/services-and-software/youtube-ces-2018-neal-mohan/ (2018).Milano, S., Taddeo, M. & Floridi, L. Recommender systems and their ethical challenges. AI Soc 35, 957–967 (2020).Article Google Scholar Hagen, L., Neely, S., Keller, T. E., Scharf, R. & Vasquez, F. E. Rise of the machines? examining the influence of social bots on a political discussion network. Soc. Sci. Comput. Rev. 40, 264–287 (2022).Article Google Scholar Stella, M., Ferrara, E. & De Domenico, M. Bots increase exposure to negative and inflammatory content in online social systems. Proc. Natl. Acad. Sci. 115, 12435–12440 (2018).Article ADS CAS PubMed PubMed Central Google Scholar Rettenberger, L., Reischl, M. & Schutera, M. Assessing political bias in large language models. J. Comput. Soc. Sci. https://doi.org/10.1007/s42001-025-00376-w (2025).Humphreys, D. “AI’s epistemic harm: reinforcement learning, collective bias, and the new AI Culture war,” Philos. Technol. https://doi.org/10.1007/s13347-025-00928-y (2025).Schroeder, D. T. et al. How malicious AI swarms can threaten democracy. Science 391, 354–357 (2026).Simon F. M. & Altay, S. “Don’t Panic (Yet): assessing the evidence and discourse around generative AI and elections,” KNIGHT FIRST AMENDMENT INSTITUTE, (2025).Hackenburg, K. et al. The levers of political persuasion with conversational AI. Science 390, eaea3884 (2025).Zhou, X., Sharma, A., Zhang, A. X. & Althoff, T. Correcting misinformation on social media with a large language model. arXiv preprint arXiv:2403.11169, https://doi.org/10.48550/arXiv.2403.11169 (2024).Costello, T. H., Pennycook, G. & Rand, D. G. Durably reducing conspiracy beliefs through dialogues with AI. Science 385, eadq1814 (2024).Article ADS CAS PubMed Google Scholar Tessler, M. H. et al. AI can help humans find common ground in democratic deliberation. Science 386, eadq2852 (2024).Article CAS PubMed Google Scholar Nogara, G. et al. Toxic Bias: Perspective API Misreads German as More Toxic https://doi.org/10.48550/ARXIV.2312.12651 (2023).Article Google Scholar Baronchelli, A. Shaping new norms for AI. Philos. Trans. R. Soc. B Biol. Sci. 379, 20230028 (2024).Article Google Scholar Burton, J. Algorithmic extremism? The securitization of artificial intelligence (AI) and its impact on radicalism, polarization and political violence. Technol. Soc. 75, 102262 (2023).Article Google Scholar Persily, N. & Tucker, J. A. Eds., Social Media and Democracy: The State of the Field, Prospects for Reform, 1st ed. Cambridge University Press, https://doi.org/10.1017/9781108890960 (2020).Fukuyama F. & Grotto A. Comparative Media Regulation in the United States and Europe. in Social Media and Democracy: The State of the Field, Prospects for Reform, N. Persily and J. A. E. Tucker, Eds., in SSRC Anxieties of Democracy, (Cambridge University Press, 2020).M. Singh, “India court rejects Twitter’s lawsuit against gov’t challenging block orders,” TechCrunch, 2023 https://techcrunch.com/2023/06/29/court-dismisses-twitter-plea-against-india-government/?guccounter=1 (2024).Mosleh, M., Martel, C., Eckles, D. & Rand, D. G. Shared partisanship dramatically increases social tie formation in a Twitter field experiment. Proc. Natl. Acad. Sci. 118, e2022761118 (2021).Article CAS PubMed PubMed Central Google Scholar Roberts, M. et al. Censored: Distraction and Diversion Inside China’s Great Firewall. Princeton: Princeton University Press, https://doi.org/10.23943/9781400890057 (2018).Farzam, A., Moradi, P., Mohammadi, S., Padar, Z. & Siegel, A. A. Opinion manipulation on Farsi Twitter. Sci. Rep. 13, 333 (2023).Article ADS CAS PubMed PubMed Central Google Scholar Jones, M. O. Digital authoritarianism in the Middle East: deception, disinformation and social media. London: Hurst & Company, (2022).Pohle, J. & Thiel, T. “Digital sovereignty,” Internet Policy Rev. https://doi.org/10.14763/2020.4.1532 (2020).Mügge, D. EU AI sovereignty: for whom, to what end, and to whose benefit? J. Eur. Public Policy 31, 2200–2225 (2024).Article Google Scholar Wojcieszak, M. & Warner, B. R. Can interparty contact reduce affective polarization? a systematic test of different forms of intergroup contact. Polit. Commun. 37, 789–811 (2020).Article Google Scholar Levendusky, M. S. Americans, not partisans: can priming American national identity reduce affective polarization? J. Polit. 80, 59–70 (2018).Article Google Scholar Balietti, S., Getoor, L., Goldstein, D. G. & Watts, D. J. Reducing opinion polarization: effects of exposure to similar people with differing political views. Proc. Natl. Acad. Sci. 118, e2112552118 (2021).Article CAS PubMed PubMed Central Google Scholar Santoro, E. & Broockman, D. E. The promise and pitfalls of cross-partisan conversations for reducing affective polarization: evidence from randomized experiments. Sci. Adv. 8, eabn5515 (2022).Article PubMed PubMed Central Google Scholar Harrop, I., Roozenbeek, J., Madsen, J. & van der Linden, S. Inoculation can reduce the perceived reliability of polarizing social media content. Int. J. Commun., 17, 5291–5315 (2023).Garibay, I., Mantzaris, A. V., Rajabi, A. & Taylor, C. E. Polarization in social media assists influencers to become more influential: analysis and two inoculation strategies. Sci. Rep. 9, 18592 (2019).Article ADS CAS PubMed PubMed Central Google Scholar Bail, C. A. et al. Exposure to opposing views on social media can increase political polarization. Proc. Natl. Acad. Sci. 115, 9216–9221 (2018).Article ADS CAS PubMed PubMed Central Google Scholar Wojcieszak, M. & Garrett, R. K. Social identity, selective exposure, and affective polarization: how priming national identity shapes attitudes toward immigrants via news selection. Hum. Commun. Res. 44, 247–273 (2018).Article Google Scholar Gomez, C. J., Herman, A. C. & Parigi, P. Leading countries in global science increasingly receive more citations than other countries doing similar research. Nat. Hum. Behav. 6, 919–929 (2022).Article PubMed PubMed Central Google Scholar Kahalon, R., Klein, V., Ksenofontov, I., Ullrich, J. & Wright, S. C. Mentioning the sample’s country in the article’s title leads to bias in research evaluation. Soc. Psychol. Personal. Sci. 13, 352–361 (2022).Article Google Scholar Carpentras, D., Lueders, A. & Quayle, M. Response Item Network (ResIN): a network-based approach to explore attitude systems. Humanit. Soc. Sci. Commun. 11, 589 (2024).Article Google Scholar M. Quayle, “Social identity networks: People holding attitudes are a collective social identity information system and bipartite networks are a useful way to represent them,” Eur. Rev. Soc. Psychol. https://doi.org/10.1080/10463283.2025.2514433 (2025).Napoli, P. et al. Social Media and the Public Interest: Media Regulation in the Disinformation Age. Columbia University Press, https://doi.org/10.7312/napo18454 (2019).Merriam-Webster, “Polarization.,” Merriam-Webster.com Dictionary. https://www.merriam-webster.com/dictionary/polarization (2026).Uscinski, J. E. et al. American politics in two dimensions: partisan and ideological identities versus anti-establishment orientations. Am. J. Polit. Sci. 65, 877–895 (2021).Article Google Scholar Petersen, M. B., Osmundsen, M. & Arceneaux, K. The ‘need for chaos’ and motivations to share hostile political rumors. Am. Polit. Sci. Rev. 117, 1486–1505 (2023).Article Google Scholar Tucker, J. et al. Social media, political polarization, and political disinformation: a review of the scientific literature. SSRN Electron. J. https://doi.org/10.2139/ssrn.3144139 (2018).Article Google Scholar LeBas, A. Polarization as craft: party formation and state violence in Zimbabwe. Comp. Polit. 38, 419 (2006).Article Google Scholar Pan, J. & Siegel, A. A. How saudi crackdowns fail to silence online dissent. Am. Polit. Sci. Rev. 114, 109–125 (2020).Article Google Scholar LeBas, A. & Gray, K. Authoritarian experience and electoral success: the fate of authoritarian diasporans in Kenya. Democratization 28, 502–520 (2021).Article Google Scholar Bowles, J., Larreguy, H. & Liu, S. How weakly institutionalized parties monitor brokers in developing democracies: evidence from postconflict Liberia. Am. J. Polit. Sci. 64, 952–967 (2020).Article Google Scholar Arriola, L. R., Devaro, J. & Meng, A. Democratic subversion: elite cooptation and opposition fragmentation. Am. Polit. Sci. Rev. 115, 1358–1372 (2021).Article Google Scholar King, G., Pan, J. & Roberts, M. E. How the Chinese government fabricates social media posts for strategic distraction, not engaged argument. Am. Polit. Sci. Rev. 111, 484–501 (2017).Article Google Scholar Esberg, J. & Siegel, A. A. How exile shapes online opposition: evidence from Venezuela. Am. Polit. Sci. Rev. 117, 1361–1378 (2023).Article Google Scholar Tourangeau, R., Edwards, B., Johnson, T. P., Wolter, K. M., & Bates, N. Eds., Hard-to-Survey Populations, 1st ed. Cambridge University Press, https://doi.org/10.1017/CBO9781139381635. (2014).Horiuchi, Y., Markovich, Z. & Yamamoto, T. Does conjoint analysis mitigate social desirability bias? Polit. Anal. 30, 535–549 (2022).Article Google Scholar Maleki, A. et al. “How do leading methods mislead?: Measuring public opinions in authoritarian contexts,” https://wc2021.ipsa.org/wc/paper/how-do-leading-methods-mislead-measuring-unmeasurable-public-opinions-authoritarian (2021).Lukito, J. et al. “The State of Digital Media Data Research, 2023,https://doi.org/10.26153/TSW/46177 (2023).Narayanan, A. & Zevenbergen, B. No encore for encore? ethical questions for web-based censorship measurement. SSRN Electron. J. https://doi.org/10.2139/ssrn.2665148 (2015).Article Google Scholar Haggard, S. & Kaufman, R. Backsliding: Democratic Regress in the Contemporary World, 1st ed. Cambridge University Press,https://doi.org/10.1017/9781108957809 (2021).Bartolini, S. and Mair, P. Identity, competition and electoral availability: the stabilisation of European electorates 1885–1985, 1. publ. Cambridge: Cambridge University Press, (1990).Mair, P. et al. Cleavages. in Handbook of Party Politics, SAGE Publications Ltd, 371 (2006).Alba, R. & Foner, N. Immigration and the geography of polarization. City Community 16, 239–243 (2017).Article Google Scholar Benson, J. et al. Democracy and the epistemic problems of political polarization. Am. Polit. Sci. Rev. https://doi.org/10.1017/S0003055423001089. (2023).Enyedi, Z. et al. The role of agency in cleavage formation. Eur. J. Polit. Res. 44, 697–720 (2005).Article Google Scholar Vaishnav, M. et al. “India,” US AID, Contract No. GS-10F-0033M / Order No. 7200AA18M00016, Tasking N068, https://pdf.usaid.gov/pdf_docs/PA0215TX.pdf (2024).Somer, M. et al. Turkey: the slippery slope from reformist to revolutionary polarization and democratic breakdown. Ann. Am. Acad. Pol. Soc. Sci. 681, 42–61 (2019).Article ADS Google Scholar Basedau, M., Erdmann, G., Lay, J. & Stroh, A. Ethnicity and party preference in sub-Saharan Africa. Democratization 18, 462–489 (2011).Article Google Scholar Slater, D. & Arugay, A. A. Polarizing figures: executive power and institutional conflict in asian democracies. Am. Behav. Sci. 62, 92–106 (2018).Article Google Scholar Somer, M. & McCoy, J. Transformations through polarizations and global threats to democracy. Ann. Am. Acad. Pol. Soc. Sci. 681, 8–22 (2019).Article Google Scholar LeBas, A. & Munemo, N. Elite conflict, compromise, and enduring authoritarianism: polarization in Zimbabwe, 1980–2008. Ann. Am. Acad. Pol. Soc. Sci. 681, 209–226 (2019).Article Google Scholar Mettler, S. & Lieberman, R. C. Four Threats: The Recurring Crises of American Democracy, First edition. New York: St. Martin’s Press, (2020).Abramowitz, A. & McCoy, J. United States: racial resentment, negative partisanship, and polarization in Trump’s America. Ann. Am. Acad. Pol. Soc. Sci. 681, 137–156 (2019).Article Google Scholar Download referencesFundingMax Falkenberg acknowledges funding from the European Union through the MSCA Actions Postdoctoral Fellowship “POLARCLIMATE”, grant number 101148808.Author informationAuthors and AffiliationsKing’s College London, London, UKMax FalkenbergCentral European University, Vienna, AustriaMax FalkenbergSapienza University of Rome, Rome, ItalyMatteo Cinelli, Antonio Scala & Walter QuattrociocchiUniversity of Padova, Padova, ItalyAlessandro GaleazziDuke University, Durham, NC, USAChristopher A. BailUniversidad Politécnica de Madrid, Madrid, SpainRosa M. BenitoQueensland University of Technology, Brisbane, AustraliaAxel BrunsToronto Metropolitan University, Toronto, CanadaAnatoliy GruzdNortheastern University, Boston, MA, USADavid LazerHarvard University, Cambridge, MA, USADavid LazerSungkyunkwan University, Seoul, South KoreaJae Kook LeeGeorgia State University, Atlanta, GA, USAJennifer McCoyUniversity of Tokyo, Tokyo, JapanKikuko NagayoshiCornell University, Ithaca, NY, USADavid G. RandUniversity of Colorado Boulder, Boulder, CO, USAAlexandra A. SiegelUniversity of Cambridge, Cambridge, UKSander van der LindenSabanci University, Istanbul, TürkiyeOnur VarolSaarland University, Saarbrücken, GermanyIngmar WeberUniversity of Warsaw, Warsaw, PolandMagdalena WojcieszakUniversity of California Davis, Davis, CA, USAMagdalena WojcieszakCa’ Foscari University of Venice, Venice, ItalyFabiana ZolloCity St. George’s University of London, London, UKAndrea BaronchelliAuthorsMax FalkenbergView author publicationsSearch author on:PubMed Google ScholarMatteo CinelliView author publicationsSearch author on:PubMed Google ScholarAlessandro GaleazziView author publicationsSearch author on:PubMed Google ScholarChristopher A. BailView author publicationsSearch author on:PubMed Google ScholarRosa M. BenitoView author publicationsSearch author on:PubMed Google ScholarAxel BrunsView author publicationsSearch author on:PubMed Google ScholarAnatoliy GruzdView author publicationsSearch author on:PubMed Google ScholarDavid LazerView author publicationsSearch author on:PubMed Google ScholarJae Kook LeeView author publicationsSearch author on:PubMed Google ScholarJennifer McCoyView author publicationsSearch author on:PubMed Google ScholarKikuko NagayoshiView author publicationsSearch author on:PubMed Google ScholarDavid G. RandView author publicationsSearch author on:PubMed Google ScholarAntonio ScalaView author publicationsSearch author on:PubMed Google ScholarAlexandra A. SiegelView author publicationsSearch author on:PubMed Google ScholarSander van der LindenView author publicationsSearch author on:PubMed Google ScholarOnur VarolView author publicationsSearch author on:PubMed Google ScholarIngmar WeberView author publicationsSearch author on:PubMed Google ScholarMagdalena WojcieszakView author publicationsSearch author on:PubMed Google ScholarFabiana ZolloView author publicationsSearch author on:PubMed Google ScholarAndrea BaronchelliView author publicationsSearch author on:PubMed Google ScholarWalter QuattrociocchiView author publicationsSearch author on:PubMed Google ScholarContributionsM.F., M.C., A.G., F.Z., A.B. and W.Q. coordinated the collaboration and collated author contributions. M.F., M.C., A.Ga., C.A.B., R.M.B., A.Br., A.Gr., D.L., J.K.L., J.M., K.N., D.G.R., A.S., A.A.S., S.v.d.L., O.V., I.W., M.W., F.Z., A.Ba., W.Q. all contributed to the writing of the manuscript and approved the final version. A.B. and W.Q. supervised the project.Corresponding authorsCorrespondence to Andrea Baronchelli or Walter Quattrociocchi.Ethics declarationsCompeting interestsThe Authors declare no competing interests.Peer reviewPeer review informationNature Communications thanks Naijia Liu, Anna Petherick, Markus Wagner and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary informationRights and permissionsOpen Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.Reprints and permissionsAbout this article