Introduction
The growing electoral success of the far rightFootnote 1 has been accompanied by an increasing presence of far right views and political actors in public life. The entry of far right actors into mainstream politics poses important dilemmas for democratic organisations and institutions, such as mainstream parties (Meguid Reference Meguid2005; Krause, Cohen and Abou-Chadi Reference Krause, Cohen and Abou-Chadi2023; Chou et al. Reference Chou, Dancygier, Egami and Jamal2021) and media outlets. Should they engage with far right actors or avoid providing exposure to their ideas? While there is empirical evidence that accommodation strategies by mainstream political parties contribute to the success of the far right (Abou-Chadi and Krause Reference Abou-Chadi and Krause2020; Krause, Cohen and Abou-Chadi Reference Krause, Cohen and Abou-Chadi2023), there is much less theory and evidence about the role that the media, both traditional media, such as TV channels and newspapers, and online media platforms play in spreading and normalising extreme ideas. Besides media exposure leading to a change in personal attitudes via persuasion (Coppock Reference Coppock2023; Tappin, Berinsky and Rand Reference Tappin, Berinsky and Rand2023), another mechanism through which far right actors and opinions gain influence is via changes in descriptive norms, beliefs about what other people believe (Tankard and Paluck Reference Tankard and Paluck2016; Valentim Reference Valentim2021; Valentim Reference Valentim2024; Bischof and Wagner Reference Bischof and Wagner2019). Research has shown that changes in how descriptive norms are perceived can lead to behavioural change, even in the absence of direct changes in personal attitudes (Tankard and Paluck Reference Tankard and Paluck2016). While a change in descriptive norms can also lead to a change in reported personal attitudes (Valentim Reference Valentim2021), these two processes can operate independently from each other (Tankard and Paluck Reference Tankard and Paluck2016).
While the far right has gained media exposure for decades (DeJonge Reference DeJonge2019; Mudde Reference Mudde2019), the Overton window is currently pushed further to the right: extreme right activists, who are openly racist and sometimes operate outside the democratic constitutional consensus, are increasingly receiving exposure on media platforms. With the ownership of many media outlets concentrated in the hands of wealthy conservative entrepreneursFootnote 2 (Foos and Bischof Reference Foos and Bischof2022; Grossman, Margalit and Mitts Reference Grossman, Margalit and Mitts2022), extreme right activists are increasingly given a platform to air their controversial views. Conservative media platforms (for example, Fox News, GB News) showcase extreme actors, while mainstream networks (for example, CNN and the BBC) claim to invite them to challenge and scrutinise their extreme viewpoints. This includes not only far right politicians such as Marine Le Pen or Donald Trump, but also increasingly more extreme fringe actors with no democratic legitimacy, who are given a platform to comment on events. Similar extreme content is also freely broadcast on the Internet, and accessible on online platforms such as YouTube (Alvares and Dahlgren Reference Alvares and Dahlgren2016; Munger and Phillips Reference Munger and Phillips2020), despite research showing the platform’s algorithmic recommendations contributing little to the popularity of such content (Hosseinmardi et al. Reference Hosseinmardi, Ghasemian, Rivera-Lanas, Ribeiro, West and Watts2024; Liu et al. Reference Liu, Baum, Berinsky, Chaney, de Benedictis-Kessner, Guess, Knox, Lucas, Mariman and Stewart2023), and increasingly on X. Research on social media reveals that de-platforming extreme right activists, such as after the US Capitol storming on January 6, 2021, leads to a reduction of exposure (Rauchfleisch and Kaiser Reference Rauchfleisch and Kaiser2021; Buntain et al. Reference Buntain, Innes, Mitts and Shapiro2023). However, little is known about the attitudinal effects of exposure to such actors on mainstream or online media channels in the first place. While studies show a correlation between social media use and anti-Muslim attitudes (Lajevardi, Oskooii and Walker Reference Lajevardi, Oskooii and Walker2022), the mechanism is unknown.
We contribute to the literature on the far right by exploring a specific mechanism, media interviews with extreme right activists, through which exposure to far right ideas happens in everyday life, and we identify the effects of such exposure on audiences. Research has explored the impact of TV networks like Fox News and tabloids like the British Sun on right-wing attitudes (Arceneaux et al., Reference Arceneaux, Johnson, Lindstädt and Vander Wielen2016; Broockman and Kalla Reference Broockman and Kalla2025; Foos and Bischof Reference Foos and Bischof2022), and political scientists have started to investigate how media channels disseminate right-wing views, for instance via ‘selective exposure’ (Broockman and Kalla Reference Broockman and Kalla2025). By varying both whether viewers are exposed and the strategy that the interviewer pursues when interviewing, we can speak to the conditions under which media affect attitudes. Our findings are consistent with results from other persuasion experiments (Coppock Reference Coppock2023), but we add to this literature by showing that persuasion effects even materialise a) when statements are extreme and b) when the speaker can clearly be identified as far right. We also contribute to the growing literature on changes in social and political norms that are associated with both left-wing (Tankard and Paluck Reference Tankard and Paluck2016) and right-wing political phenomena (Bischof and Wagner Reference Bischof and Wagner2019; Valentim Reference Valentim2021). We show that media exposure to interviews with extreme right actors and exposure to their nativist views, on either TV or online channels, leads to higher agreement with extreme right statements. Attitude change likely happens via persuasion effects, but exposure also affects beliefs about the viability of actors’ views via normalisation effects. Finally, we attempt to make a contribution to the growing literature that looks at the effects of media portrayal of immigrants and refugees on political attitudes (Lo and Lang Reference Lo and Lang2023; Lajevardi Reference Lajevardi2021).
To do that, we draw on two distinct, but similar, real-life interviews with extreme right activists that were broadcast on Sky News in Australia and the UK. Relying on a series of carefully designed audio experiments based on these real-life interviews, we combine increased realism with control over the treatments that study participants are exposed to and the ability to debrief participants after exposure. Using factorial designs, we vary whether participants listen to the audio of the interview, or to neutral content, a weather report. The second experimental factor varies the source cue, whether the platform which broadcasts the interview/forecast is announced as a traditional mainstream TV channel, Sky News,Footnote 3 or an online platform, YouTube. We evaluate three distinct outcome variables: individuals’ agreement with extreme right statements, individuals’ beliefs regarding the prevalence of these statements among others, and individuals’ perceptions about the respectability of the actor interviewed, aiming to distinguish between persuasion and normalisation effects.
We find consistent evidence in line with pre-registeredFootnote 4 expectations across both countries: Unchallenged interviews with extreme right activists do not lead to rejection, but rather fuel extreme right attitudes and foster the belief that a larger share of the public supports the extreme right actors’ statements. The media strategy adopted by the journalist appears crucial, as uncritical platforming leads both to updating on the part of the audience member and to changes in descriptive norms favourable to the extreme right. When we randomly assign the interviewer to robustly challenge the extreme right actor in the British interview, the reputation of the actor decreases, and the effects on attitudes subside. Nevertheless, even critical interviews do not lead participants to move away from the actor’s statements. Moreover, participants continue to update their beliefs about the viability of these views in the population, albeit at a lower rate.
Contrary to our pre-registered expectation, our study also shows that perceived exposure to extreme right actors on the online platform YouTube has a similar effect on attitudes and descriptive norms as platforming extreme right actors on the traditional mainstream TV channel Sky News. This effect is observed in both Australia, where Sky News has a more right-wing tilt, and in the UK, where it does not. Therefore, our results suggest that persuasion and normalisation through media platforming are driven by increased exposure to extreme right actors and their views rather than any added legitimacy conveyed by source cues that would privilege mainstream TV channels over online platforms.
Beyond its theoretical and empirical contributions, this paper has policy implications with regard to the platforming of extreme right actors on various media channels. Across two different contexts, in well-powered experiments, exposing people to extreme right actors’ views did not lead to a decrease in public support for those views. Exposure consistently led people to believe that a larger share of the population agreed with such views. Amidst rampant consumption of extremist content on social media and its rise in traditional media, these findings have significant implications for researchers, politicians, and journalists alike.
Theory
The Role of the Media in Attitudinal and Normative Change
The growing presence of extreme right actors and ideas in mainstream political discourse is often attributed to their increased visibility in the media (Mudde Reference Mudde2019; DeJonge Reference DeJonge2019). Giving a media platform to an extreme right actor has become more commonplace, making extreme right discourse more widespread and available to audiences. There is a large literature on the persuasive effects of media on individuals’ political attitudes. Yet, the effect of the media on the spread of extreme right views remains theoretically unclear. It is commonly agreed that media attention matters in increasing citizens’ interest in a given topic (Ellinas Reference Ellinas2010). The literature on the far right has emphasised the importance of media visibility of far right political actors in their political success. By making issues of immigration and crime salient, the media also create a political environment conducive to the rise of radicalism. Quasi-experiments, which have identified whether media outlets sway citizens’ opinions and if so, in which direction, often treat media (non-) exposure as a black-box (Foos and Bischof Reference Foos and Bischof2022; Grossman, Margalit and Mitts Reference Grossman, Margalit and Mitts2022; DellaVigna and Kaplan Reference DellaVigna and Kaplan2007; Ladd and Lenz Reference Ladd and Lenz2009). Survey experiments that identify the effects of exposure to different arguments find that individuals usually update their attitudes in the direction of the argument they are exposed to (Coppock Reference Coppock2023; Guess and Coppock Reference Guess and Coppock2020). Moreover, experiments that vary whether individuals are exposed to one-sided or two-sided frames show that one-sided frames are more effective at swaying public opinion, while two-sided frames, when they are equally strong, are likely to cancel each other out (Druckman Reference Druckman2004, Reference Druckman2001; Chong and Druckman Reference Chong and Druckman2007).
Yet, one may wonder whether these findings apply equally to all types of messages and messengers, or if there is heterogeneity in how individuals respond based on the type of message and messenger that they encounter. A majority of people may be willing to reject extreme right arguments based on their political predispositions (Zaller Reference Zaller1992). In that sense, when individuals are exposed to extreme right content, those who are sympathetic to extreme right views (that is, authoritarian individuals) would be even more supportive of such views, whereas those who normally reject them (that is, liberal individuals) would oppose these views even more. This directly relates to what scholars call the backlash thesis, which posits that when individuals are exposed to counter-attitudinal evidence, their pre-existing opinions and beliefs are not challenged but strengthened (Kunda Reference Kunda1990; Taber and Lodge Reference Taber and Lodge2006). Despite its theoretical relevance, the backlash hypothesis has lately received little empirical support (Guess and Coppock Reference Guess and Coppock2020; Bishin et al. Reference Bishin, Hayes, Incantalupo and Smith2016; Coppock Reference Coppock2023), even when strong partisan cues are present (Tappin, Berinsky and Rand Reference Tappin, Berinsky and Rand2023).
Another argument for why individuals might reject extreme right messages is that there exist strong social norms that make such views undesirable. Social norms are shared standards of acceptable behaviour which individuals learn over time and in a dynamic fashion (Paluck, Shepherd and Aronow Reference Paluck, Shepherd and Aronow2016). Social norms differ from personal opinions, as the latter relate to agreement with political views, irrespective of individuals’ perceptions of others’ opinions on these views, which are observed in social norms. Persuasion normally refers to a change in personal political views that is distinct from revealing an already held view that was previously subject to self-censorship. Norms against the expression of far right views have been eroding quickly in recent years (Bursztyn, Egorov and Fiorin Reference Bursztyn, Egorov and Fiorin2020).Footnote 5
Key events such as Trump’s 2016 victory in the US presidential election or the entry of extreme right legislators into parliament have changed social norms by signalling that radical right views have become normatively desirable in society (Bursztyn, Egorov and Fiorin Reference Bursztyn, Egorov and Fiorin2020; Valentim Reference Valentim2021; Bischof and Wagner Reference Bischof and Wagner2019). Once extreme right actors gain access to the White House or a national parliament, social norms penalise extreme-right views and actors less. These signals indicate that extreme right attitudes and actors, which used to be socially stigmatised, are now seen as more socially acceptable. Hence, individuals might perceive extreme right views as more popular when the latter are accompanied by a strong signal like the platforming of an actor advocating for those views on media channels. Indeed, media platforming may be directly related to the increasing popularity of extreme right views. The parliamentary entry of extreme right actors is, likely both a function and a cause of increased media coverage, especially since gaining more seats leads to increased media coverage (Dunn and Singh Reference Dunn and Singh2011), as well as more frequent and better access to public broadcasters. Media channels may play a crucial role in the dissemination of far right messages to broader audiences through discursive opportunities (Koopmans and Olzak Reference Koopmans and Olzak2004). Koopmans and Olzak’s (Reference Koopmans and Olzak2004) research highlights the significant impact of media visibility and reported support for right-wing violence on the occurrence of violent acts against ethnic groups in Germany. Thus, we expect that media platforming contributes to normalising extreme right views in society, sending a signal that more citizens than assumed subscribe to these views.
Furthermore, media coverage may not only provide a stage from which to spread and normalise extreme right views, but it may also increase the legitimacy of extreme right actors by portraying them as politically viable and respectable. Through platforming, the media confer ‘legitimacy and authority to political newcomers and (…) dispel voter doubts about their electoral viability’ (Ellinas Reference Ellinas2010). Since extreme right actors are usually marginalised in the political game, offering them a voice gives them the impression that they have a mass following and that they are the voice of the people. Media coverage can also make up for their organisational deficiencies and financial shortages by helping them become known, thus helping leaders of small organisations (Ellinas Reference Ellinas2010). Mainstreaming extreme right discourse is a political resource that can lift marginalised actors from obscurity and push them into the political mainstream. That is why the media can be seen as playing a role in the process of persuasion and normalisation in regard to extreme right views, as well as contributing to the increased respectability of extreme right actors.
Based on these theoretical considerations, we preregistered the following hypotheses:Footnote 6
• Persuasion hypothesis: Subjects who are exposed to an extreme right actor’s interview are more likely to agree with the views expressed in the interview.
• Normalisation hypothesis: Subjects who are exposed to an extreme right actor’s interview are more likely to a) perceive that the rest of the population will agree with these views, and b) perceive that the actor is more respectable.
The Moderating Role of Platform Type and Journalist Media Strategy
We also examine two conditions under which exposure to an extreme right interview can further amplify those attitudinal and normalisation effects: the type of platform (also known as the source cue effect) and the media strategies adopted by the journalists.
First, we expect that extreme right arguments can receive more approval if they are conveyed on a platform which has a credible reputation. This hypothesis directly relates to the large literature on source credibility (Druckman Reference Druckman2001; Chong and Druckman Reference Chong and Druckman2007; Miller and Krosnick Reference Miller and Krosnick2000; Berinsky Reference Berinsky2017). Studies have highlighted the importance that media reputations play in public opinion: trustworthy outlets with well-established reputations and high popularity are more likely to positively influence opinion than outlets viewed as non-mainstream (Druckman Reference Druckman2001; Chong and Druckman Reference Chong and Druckman2007; Miller and Krosnick Reference Miller and Krosnick2000). While issue frames from an untrustworthy news source have been shown to be ineffective, the same frames that are shared by a trustworthy source can affect public opinion (Druckman Reference Druckman2001; Chong and Druckman Reference Chong and Druckman2007). However, recent studies have questioned the size of these source cue effects on trust in news (Blum, Berinsky and Rand Reference Blum, Berinsky and Rand2024).
For the purpose of this study, we define a credible platform as a traditional media outlet that is known to the public and can reach a large audience (for example, Sky News), and distinguish it from an online platform that can be unknown to people (for example, a YouTube Channel). While the public has a good sense of the media’s overall trustworthiness of well-known traditional outlets such as CNN or Fox News in the USA or Sky News in Australia and the UK (our case studies) (DellaVigna and Kaplan Reference DellaVigna and Kaplan2007; Peterson and Allamong Reference Peterson and Allamong2022), it is more difficult to assess the credibility of online outlets given the wide array of options in the digital media landscape (Hindman Reference Hindman2008; Metzger Flanagin and Medders Reference Metzger, Flanagin and Medders2010). One reason for this assumption is that the number of online options individuals can encounter far exceeds the number of sources they are aware of, making it impossible for them to discern which online option is credible. This, of course, does not mean that all online platforms are not credible, but that online platforms do not have clear public reputations (however, see Luca et al., Reference Luca, Munger, Nagler and Tucker2022) who suggest that some online media, even clickbait media, might have higher trust levels among subsections of the population). An interview perceived to be aired on a traditional media platform would thus be assigned more credibility compared to one perceived to feature on an online platform. We therefore preregistered the following hypothesis:
• Platform type hypothesis: Subjects are more likely to a) increase support for the views expressed in the interview, b) perceive that a larger proportion of the population shares the views expressed in the interview, and c) perceive the extreme right actor as respectable if the interview is broadcast by a mainstream platform compared to a non-mainstream platform.
Moreover, the media strategies adopted by journalists could further condition persuasion and normalisation effects. The question of whether it is most effective to ignore, accommodate, or challenge the far right is one that political scientists have been investigating since the far right’s emergence on the political scene (Abou-Chadi and Krause Reference Abou-Chadi and Krause2020; Meguid Reference Meguid2005). Journalists can choose between a variety of strategies when it comes to dealing with extreme right actors, ranging from (1) demarcation to (2) accommodation and (3) confrontation (DeJonge Reference DeJonge2019; Meguid Reference Meguid2005). First, journalists can opt to disengage with the actor by refusing to platform them, which is a way to isolate them. This demarcation strategy has become rare over the past two decades, as there are few cases where far right actors are completely ignored (DeJonge Reference DeJonge2019). The second strategy involves accommodating extreme right actors by offering them a platform to spread their views without directly engaging with them. The journalist gives extreme right actors an implicit endorsement by making issues that are typically pushed by such actors more visible (for example, immigration, nationalism, crime) (Boomgaarden and Vliegenthart Reference Boomgaarden and Vliegenthart2009) or by incorporating some of their rhetoric in their news coverage (for example, by focusing on the ‘silent majority’). This accommodative strategy is likely to amplify the persuasion and normalisation effects of these views.
By contrast, the journalist can adopt a confrontational strategy by being critical towards the extreme right actor or their statements. This confrontational strategy means that the journalist can point to the inaccuracy of the statements made, and/or by raising normative concerns (for example, stressing that the extreme right actor is violating the freedom of religion by adopting anti-Islam positions). We may therefore assume that a journalist who challenges an extreme right actor would undermine the credibility of the actor’s views and image. Discrediting extreme right actors’ statements and image could weaken persuasion and normalisation effects on the public, especially since fact-checking is known to improve factual belief accuracy (Chan et al. Reference Chan, Jones, Jamieson and Albarracin2017; Wood and Porter Reference Wood and Porter2019). We therefore pre-registered the following hypothesis:
• Media strategy hypothesis: Subjects are more likely to a) increase support for the views expressed in the interview, b) perceive that a larger proportion of the population shares the views expressed in the interview, and c) perceive the extreme right actor as respectable if the journalist does not challenge the extreme right actor.
Research Design
Strategies to identify the causal effects of media exposure on political outcomes include field experimental (Gerber, Karlan and Bergan Reference Gerber, Karlan and Bergan2009; Broockman & Kalla 2022) and quasi-experimental designs (DellaVigna and Kaplan Reference DellaVigna and Kaplan2007; Martin and Yurukoglu Reference Martin and Yurukoglu2017; Foos and Bischof Reference Foos and Bischof2022; Grossman, Margalit and Mitts Reference Grossman, Margalit and Mitts2022). Randomised survey experiments have been used to test mechanisms of media influence such as priming and framing (Iyengar and Kinder Reference Iyengar and Kinder1987; Maier and Rittberger Reference Maier and Rittberger2008; De Vreese, Boomgaarden and Semetko Reference De Vreese, Boomgaarden and Semetko2011; Druckman Reference Druckman2001). Survey experiments have the advantage of increasing the control that researchers can exercise over exposure to specific messages. While some of these experiments use student or self-selected samples, others use population-based samples (Mutz Reference Mutz2011). Our study stands in the latter tradition, but increases environmental and external validity by a) exploiting two similar real-life interviews with two extreme right political actors and b) conducting two large population-based survey experiments on representative samples of the Australian and British populations.
Case Studies
In this study, we rely on two separate interviews with extreme right actors that were broadcast on Sky News Australia and Sky News UK in 2018, when both channels were part of Rupert Murdoch’s media empire.Footnote 7 Fielding the same experiment in two countries allows us to address questions of external validity by using two countries with similar extreme right actors, TV channels and treatment conditions. At the same time, these two countries have relatively different shares of conservative media. Although Murdoch’s News Corp empire exists in both countries, it is even more prevalent in Australia than in the UK. Murdoch built his media corporation in Australia and owns 65 per cent of the country’s print media, five popular radio programmes, and a large online news and social media base. By contrast, Murdoch owns 32 per cent of the UK’s newspapers, radio stations, and television channels (MRC 2021). As a result, Sky News Australia is slightly more conservative than Sky News UK, and this is perceived by the respondents in our experiments since respondents perceive Sky News UK to be more ‘mainstream’ than Sky News Australia (3.3 v. 4.0 on a 1–5 scale).Footnote 8
The Australian interview
On August 4, 2018, Sky News Australia aired a ten-minute interviewFootnote 9 with the former United Patriots Front leader, the extreme right activist, Blair Cottrell. He advocated in favour of both skills-based and ‘culture-based’ migration, where Australia should not accept immigrants who were ‘too culturally dissimilar to Australia’. Cottrell also associated the rise of criminality with an increase in ‘African gangs’. He finally promoted his Lads Society, a men-only social club involving regular meetings, and encouraged male viewers to join him. Within hours of the interview going on air and being shared on various Sky News social media platforms, the channel removed the interview from its repeat time slots and online platforms. The broadcast also prompted the interviewer to resign from Sky News, as he adopted an accommodating media approach and refrained from challenging Cottrell’s claims. This led to a public apology from the channel.
The segments that are used in the Australian experiment include Cottrell’s views on skilled and race-based immigration, his association of criminality with ‘African gangs’, and his promotion of his political organisation, the ‘Lads Society’.
The British interview
On September 27, 2018, Sky News UK broadcast segments of an interview with the former co-founder and leader of the English Defence League, Tommy Robinson. The extreme right activist spoke after being released from prison for being found guilty of contempt of court in May 2017 for trying to film and expressing views about suspects in a sexual grooming case in Canterbury. During the one-hour interview,Footnote 10 Robinson also shared his views on typical extreme right themes: Islam, immigration, and terrorism. He notably praised the temporary halt of the construction of mosques and advocated for the introduction of a ‘Trump-style’ travel ban to restrict the number of refugees from ‘failed states’ because he associated refugees with terrorist attacks. Robinson claimed that ‘less people will be murdered’ and ‘less girls would be raped’ if a travel ban would be introduced. These segments are comparable to the extreme right claims Blair Cottrell made during his interview. However, unlike the Australian experiment, Robinson was challenged by the journalist a few times. The journalist pointed towards normative concerns for violating basic principles of democracy, but also raised the inaccuracy of Robinson’s claims. He not only told Robinson that temporarily stopping the construction of mosques is ‘a violation of people’s freedom of religion, of worship’, but he also questioned Robinson multiple times about the credibility of his sources on terrorism and Islam. For instance, the journalist asked Robinson, ‘Where is your source for that?’, after Robinson wrongfully claimed that the majority of rapes are committed by immigrants. The segments used in the British experiment include Robinson’s views on the building of mosques, a Trump-style travel ban, as well as him associating terrorist attacks and rape with refugees.
The Australian and British interviews are similar in terms of audio pitch and, to a lesser extent, content. The activist in both cases avoids irony and metaphors, maintaining a consistent tone that is neither authoritative nor aggressive. This similarity ensures a comparable analysis, minimising the likelihood of the argument’s tone leading to differences in effects. The two interviews also both discuss the traditional topics of the extreme right, namely immigration and crime, making these two cases comparable. However, the outgroups targeted by the extreme right activists are context-dependent. Cottrell targets Black Africans, while Robinson targets Muslims and refugees. Each activist touches upon other topics relevant to the extreme right discourse of their respective countries. Both the comprehensive critical version of the Sky News interview with Robinson and a version that excludes the interviewer’s challenges are used in the experiment to closely resemble the interview style that was used in the Cottrell interview on Sky News Australia. The full transcripts of the interviews and weather reports in Australia and the UK are available in Appendix Section D.
Experimental Design
In our experiments, we use the edited audio recording of these two interviews. One of the main reasons for this choice is that the quality of the videos uploaded on YouTube is not very highFootnote 11 . Relying solely on the audio recording allows us to direct respondents’ focus towards the spoken content and the effect of the explicitly announced media platform, eliminating any potential distractions arising from visual cues of the extreme right actors and interviewers.
In order to minimise potential harms to participants from exposure to the interviews and any questionable statements made during the interviews, we debrief them in full directly after outcome data collection. While this choice prevents us from collecting long-term outcome data, we believe that it is necessary from an ethical point of view to correct any misinformation provided. We discuss our approach to debriefing participants in detail in Ethics Appendix B, where we also provide all information, consent, and debriefing documents. All experiments received full ethics review. The Australian experiments were approved by the LSE Research Ethics Committee under reference 1050 and the British study was approved by the LSE Research Ethics Committee under reference 92361.
Main factors in the experiment
In both experiments, we use a factorial design, a 2 x 2 design in the Australian case, and a 3 x 2 design in the British case. The first factor varies the content that is broadcast. In the Australian case, we vary whether participants are exposed to the unchallenged interview with the extreme right activist or to a weather report, and in the British case, we replicate the first two categories and add a third one: a version of the interview, where the extreme right actor is challenged by the interviewer. This addition allows us to test the effect of the interviewer’s strategy on extreme right beliefs and norms. While we use the exact same segments for the two groups that listen to the unchallenged and challenged interview in the British case, those in the ‘challenged interview’ group also hear the journalist’s responses and criticism of Robinson’s claims. We use a weather report as the placebo condition because we assume the content to be neutral compared to the interview.
The second experimental factor is the same across both experiments: a presenter announces at the start and the end of the segment that the interview is/was either broadcast on Sky News or the presenters’ YouTube channel. An Australian and a British native speaker were recruited to announce the platform on which the interview and weather report were allegedly broadcast at the start and the end of the audio clip. These additions to the clip were meant to ensure that subjects understood the platform on which the interview or weather report was broadcast. Regarding the platform, even though some people may use YouTube as their main source of information, we consider YouTube to be less heavily regulated, and hence more likely to platform extreme content, especially on the extreme right, than mainstream TV channels like Sky News. Indeed, YouTube is a medium where extreme right content is relatively freely accessible, despite algorithms contributing little to its popularity (Hosseinmardi et al. Reference Hosseinmardi, Ghasemian, Rivera-Lanas, Ribeiro, West and Watts2024; Liu et al. Reference Liu, Baum, Berinsky, Chaney, de Benedictis-Kessner, Guess, Knox, Lucas, Mariman and Stewart2023). Empirical work also shows that extreme right content has likely been in decline on YouTube compared to its peak in 2017 (Munger and Phillips Reference Munger and Phillips2020; Hosseinmardi et al. Reference Hosseinmardi, Ghasemian, Clauset, Mobius, Rothschild and Watts2021). At the same time, it remains popular among right-leaning users, and extreme right content receives higher engagement rates than mainstream content (Reeve 2019; Munger and Phillips Reference Munger and Phillips2020).
The Australian recordings last between 1.40 (placebo) to 2.30 minutes (interview), and the British recordings last between 1.30 (placebo) to 2.30 minutes (challenged interview). We display the factorial design of the experiments in Table 1 below.
Table 1. Factorial design
To ensure that respondents complied with the treatments, respondents could not skip through the respective segments. They had to listen to them in full. In both countries, dropout rates are not significantly different across experimental groups. They range from 85 to 126 dropouts in the Australian case, and from 146 to 154 dropouts in the British case.
Data Collection and Outcome Measurement
We fielded the experiments on nationally representative samples of the Australian and British adult populations (N=5062 in Australia and N=5482 in the UK). Both experiments were administered by the high-quality survey firm Survation.Footnote 12 The Australian experiment was administered over two rounds between 1–9 December 2020 and 2–9 March 2022, and the British experiment was fielded from 21 July to 1 August 2022.Footnote 13 This resulted in around 1,250 respondents per group in the Australian case and 950 respondents in the British case.
To test our hypotheses, we include distinct measures aimed at capturing persuasion and normalisation effects. The post-treatment surveys included multiple items measuring extreme right attitudes (four items in the Australian experiment and five items in the British one), extreme right norms (four items in the Australian experiment and five items in the British one), and the respectability of the actors (one item per survey). Concerning extreme right attitudes, we asked respondents how much they agreed with each extreme right statement mentioned by the extreme right actor in the interview. We adapted the statements based on what each actor was talking about. Cottrell talked a lot about immigration and criminality, while Robinson focused more on terrorism and Islam. All items are highly correlated, so we take the mean of the items and reweight the single item on a scale ranging from 0 to 1 (1 means that respondents fully agree with Cottrell’s/Robinson’s views). We present the results of the combined attitudinal scale in the main analysis, but results are comparable for each item, as shown in Tables E.5 and E.7 in the Appendix.
The variable on descriptive norms relates to items that asked respondents to rate the percentage of Australians/British who agree with each of Cottrell’s/Robinson’s statements. This variable refers to descriptive norms that capture people’s perception of how society actually thinks about these statements. This variable, which has been traditionally used to measure social norms in existing studies (Bursztyn, Egorov and Fiorin Reference Bursztyn, Egorov and Fiorin2020; Fieldhouse and Cutts Reference Fieldhouse and Cutts2021), differs from individual attitudes and how an individual thinks about these claims. Since all items are also highly correlated, we take the mean of the items and re-weight the single item on a scale ranging from 0 to 1 (1 means that respondents think 100 per cent of Australians/British agree with Cottrell’s/Robinson’s views). We show the results of the scale in the main body, but findings are similar for each item (see Tables E.5 and E.8 in the Appendix).
Our last outcome variable asked respondents how respectable they think the extreme actor is on a 5-point scale. The question, which is the same in the Australian and the British experiments, is recoded into a categorical variable because there is some differential attrition as a function of the treatment, where respondents are less likely to answer ‘Don’t Know’ in the interview conditions (see Table G.1 in the Appendix). Those who do not find the actor respectable take value 0, those who find him respectable take value 1 and those who don’t know take value 2. We then analyse this question using multinomial logistic regressions. We deviate from the pre-analysis plan as excluding ‘don’t knows’ from this question could introduce bias into our estimates. We discuss this change and other minor changes to the pre-analysis plan in Appendix C.
Recorded pre-treatment covariates include gender, age, region, education, political ideology (authoritarian/libertarian attitudes) and vote in the 2019 general election.
We also added a question in the British survey on whether respondents voted leave or remain in the European Union in the 2016 referendum. Our treatment effects are estimated using OLS regression models with HC2 standard errors when we regress the attitudinal and norms outcomes on the experimental conditions. We deviate from our pre-analysis plan concerning the respectability question and use a multinomial logit regression in this case. We test all predictions in country-specific models without and with covariate adjustment.Footnote 14 Results are consistent throughout and do not depend on model specifications. We present our results with the full sample because our manipulation checks were successful (see Appendix Table F.2). Appendix Tables H1–H3 show that results are conistsent if we restrict the sample to subjects that pass the pre-treatment attention check.
While both survey experiments are very well-powered and similar in substance, we made a few additions in the UK experiment, where we included additional manipulation and attention checks, and randomised the order of the outcome variables. We also added two secondary outcome variables in the British study. For more information on the changes that were made and the power analyses, see our pre-analysis plan of the British experiment. The questionnaires of the Australian and British experiments are displayed in Appendix documents A.5 and A.6.
Results
Interview Effects on Agreement with Extreme Right Statements
Table 2 displays the results of platforming the extreme right actor on individuals’ support for extreme right attitudes in Australia (models 1–4) and the UK (models 5–8). We find that respondents who listen to the unchallenged interview (as opposed to those who listen to the weather report) are more likely to agree with the actors’ extreme right attitudes in both countries, regardless of the platform that is invoked. The effects are highly statistically significant across the board (at p<0.001 significance level), comparable across the two countries and remain similar when we adjust for pre-treatment covariates. According to Cohen’s criteria, this effect is small to medium in terms of size (d=0.16 standard deviations for Australia and d=0.18 for the UK). This effect is nonetheless substantive in this context as it means that interview exposure in the unchallenged interview, compared to listening to the weather report, caused a 3 to 5 percentage points increase in individuals’ support for the actor’s extreme right positions voiced in the interview. This is substantively important given that the actors’ claims are extreme. Nonetheless, a large minority of Australians appear to agree with them. The direction and size of the effects are also comparable to those found in existing field and survey experiments that show the persuasive effects of media (Guess and Coppock Reference Guess and Coppock2020; Guess et al. Reference Guess, Lockett, Lyons, Montgomery, Nyhan and Reifler2020), or campaign messages (Gerber et al. Reference Gerber, Gimpel, Green and Shaw2011; Coppock Reference Coppock2023). Figure 1 displays bar charts including the mean level of agreement with the extreme right actors’ statements (scaled to range from 0 to 1) for each experimental group and 95 per cent confidence intervals surrounding the mean. In line with the ‘persuasion and normalisation’ hypotheses, unchallenged interviews shift participants attitudes further to the right. What is striking is that the base level of agreement with these statements does not seem to matter much. While more Australians agree with the statements in the placebo condition than Brits (around 1/3 of the population), effect sizes are comparable.
Table 2. Effects on agreement with extreme right statements
Heteroskedasticity-robust standard errors in parentheses.
Covariates: age, gender, education, region, vote in the 2019 elections, authoritarian attitudes, and Brexit vote (UK only).
We include a dummy variable in the Australian case to control for the two Australian waves.
$^{***}p \lt 0.001$
;
$^{**}p \lt 0.01$
;
$^*p \lt 0.05$
.
Figure 1. Mean levels of agreement with extreme right statements in Australia and the UK across experimental conditions. 95 per cent confidence intervals.
Table 2 provides further information on the conditions under which respondents may be more or less supportive of the extreme right claims expressed by the actor. First, we do not find that the type of platform conditions the way media interviews affect respondents’ positions on extreme right views. The effect of the type of platform, as well as the interaction between the interview and the platform, are small and non-significant across the board and in the two countries. Listening to the unchallenged interview on the YouTube channel may slightly increase the support towards extreme right attitudes as opposed to those who listen to the interview on Sky News, as Figure 1 suggests, but the difference is minimal and not significant. Although this finding goes against our expectations on the type of platform, it corroborates Peterson and Allamong (Reference Peterson and Allamong2022)’s study that shows that, conditional on exposure, unfamiliar news sources are as effective at shifting public opinion as familiar media with established reputations.
Additionally, we find that if the journalist challenges the extreme right activist, updating effects of the extreme right interview do not materialise. Indeed, Table 2 and Figure 1 show that British respondents who listen to the challenged interview are no more inclined to agree with extreme right positions than those who listen to the weather report, as the effect is non-significant and equal to zero in models 5–8. Moreover, the effect of the challenged interview is significantly different from the effect of the unchallenged interview. This means that the journalist’s critical questioning balances the effect of the extreme right activists’ words. In line with our pre-registered expectations, the effect of the interview on support for extreme right views is larger if the extreme right actor is not challenged by the journalist.
Interview Effects on Descriptive Norms
We now check if exposure to an extreme right interview affects people’s perceptions of descriptive norms as they relate to extreme right statements. Table 3 and Figure 2 replicate the analyses on attitudes, but with a different dependent variable that asks respondents about the proportion of Australians/British they think agree with the statements. We find that respondents who listen to the unchallenged interview are more likely to think that society agrees more with these extreme right views than those who listen to the weather report, regardless of the platform. The effects are statistically significant at the 0.001 level in both countries. Listening to the unchallenged interview increases people’s belief that society has moved in favour of these extreme right views by 2–3 percentage points in Australia, and by 6 percentage points in the UK, as opposed to listening to the weather forecast. These effects translate into small to medium-sized effects in both countries (d=0.16 in Australia and d=0.22 in the UK), and are thus comparable in size to the effects we found on individuals’ approval of extreme right statements. Moreover, in the Australian case, the effect on normalisation is only significantly different from zero on the Sky News platform, although the interaction between the platform and the interview is not statistically significant. There is no evidence of an interaction between platform and interview in the British case. Overall, we do not find that descriptive norms are affected by the platform on which the interview was reported. Listening to the interview where a message is cueing a traditional mainstream platform as opposed to an alternative online platform does not increase people’s beliefs that society agrees with the claims promoted by the extreme right actor.
Table 3. Effects on descriptive norm perceptions
Heteroskedasticity-robust standard errors in parentheses.
Covariates: age, gender, education, region, vote in the 2019 elections, authoritarian attitudes, and Brexit vote (UK only).
We include a dummy variable in the Australian case to control for the two Australian waves.
$^{***}p \lt 0.001$
;
$^{**}p \lt 0.01$
;
$^*p \lt 0.05$
Figure 2. Proportion of others perceived to agree with the actor’s extreme right statements in Australia and the UK across experimental conditions. 95 per cent confidence intervals.
In line with our expectations, listening to the challenged interview attenuates the normalisation effect, but does not entirely reverse it. As shown in Table 2, the normalisation effect declines from 6 percentage points to 2 percentage points in substantive terms, but a 2 percentage point increase is still significantly different from zero at the 0.05 level. Therefore, the unchallenged interview still causes a 2 percentage point increase over a baseline of 34 per cent in people’s belief that society agrees with these extreme right views. This means that being exposed to an interview where the journalist adopts a confrontational stance towards the claims of the extreme right actor (as opposed to listening to a weather report), while not shifting personal attitudes, contributes to the normalisation of extreme right views. This finding suggests that a change in descriptive norms does not always go hand-in-hand with a change in personal attitudes. This evidence is also consistent with an interpretation of the attitudinal change that we are seeing in the unchallenged interview condition as a function of persuasion rather than revealing an already existing attitude. The effect on descriptive norms for those in the challenged interview condition is significantly smaller compared to those who listen to the unchallenged interview, where the interviewer adopts an accommodative media strategy.Footnote 15 Hence, while challenging the extreme right activist’s claims is more effective than not challenging these claims at all, it does not fully reverse the normalisation process.
Interview Effects on the Respectability of the Extreme Right Actor
Finally, we look at how platforming extreme right views affects the perceived level of respectability of the actor who expresses these views. We display the results in Figure 3. The results show some differences by country. While we observe large and statistically significant negative net effects on perceived respectability in the UK on both platforms and both interview conditions, findings are more mixed in Australia. There is both an increase in respondents who see Cottrell as respectable and as unrespectable in the treatment conditions. We also model this answer situation using multinomial logistic regression because ‘Don’t Knows’ are more prevalent in the placebo condition (the weather report), and excluding them would bias our results. The results, which are displayed in Table E.4 in the Appendix, confirm a positive effect on perceived respectability vs ‘Don’t Knows’ and between perceived unrespectability v. ‘Don’t Knows’. The difference between these two options is relatively equal, given that the level of net respectability is already slightly higher in the control group (the weather report). These contrasting findings by country could be explained by the way the two actors are initially perceived in the placebo. The placebo groups in Figure 3 reveal that Robinson was considerably more well-known than Cottrell initially, and a higher number of individuals held negative views of him compared to positive ones. By contrast, Cottrell was an unknown figure. This suggests that giving a platform to unknown extreme right actors might lead to polarised opinions, while it might backfire on those who already have a relatively high level of notoriety.
Figure 3. Level of respectability towards the extreme right actor in Australia (top) and the UK (bottom) across experimental conditions. 95 per cent confidence intervals.
Interestingly, we find some evidence from the Australian case that the platform negatively affects the level of respectability of the actor. We do not find any evidence for such a mechanism in the UK. As shown in column 2 of Table E.4 and in Figure 3, the level of respectability (unrespectability) towards Cottrell is lower (higher) among those who listen to the unchallenged interview on Sky News as opposed to those who listen to the unchallenged interview on the YouTube Channel. This means that the reputation of the Australian extreme right actor is more tarnished when the interview is on a traditional mainstream platform. Yet, while this effect is in the same direction in the UK, it is not statistically significant. We can therefore conclude that there is mixed evidence on whether the type of platform affects the image of the actor.
Last but not least, Table E.4 and Figure 3 show that adopting a confrontational interview strategy negatively affects the image of the extreme right actor. This is consistent with our findings for extreme right attitudes and norms. British respondents who are assigned to the challenged interview are about 9 percentage points less inclined to find Robinson respectable compared to those who listen to the unchallenged interview, which corroborates our media strategy hypothesis. The negative effect of the interview on the respectability of the extreme right actor is larger if the extreme right actor is challenged by the journalist.
Manipulation, Attrition and Attention Checks
We report a series of manipulation, attrition and attention checks in Appendix F, G and H. First, as expressed earlier, the manipulation checks were successful, which means that the cues were well understood by respondents. As Table F.2 shows, a large majority of participants were able to identify the type of content and platform they listened to.
Second, we made sure that our results are not affected by potential differential attrition across experimental conditions. Since respondents had the opportunity to answer ‘Don’t Know’ to any of the questions, we had to check that there was no differential attrition as a function of the experimental condition to which respondents were assigned to. As shown in Table G.1, we found no differential attrition for extreme right attitudes or descriptive norms, but we found some differences across conditions for the level of respectability of the actor. We therefore used multinomial logit models and included ‘Don’t Knows’ as a separate category for this dependent variable.
Third, our findings are robust to the exclusion of respondents who did not pass the pre-treatment attention check (see Tables H.1, H.2, and H.3 in the Appendix). There is one exception in the interaction between the unchallenged interview and the Sky News platform concerning individuals’ support for extreme right attitudes. This result goes in the same direction as in the main analysis but is significant at p<0.05, contrary to the hypothesised direction of the interaction (see model 8 of Table H.1). Since it is only a small minority of respondents who fail the check (12.59 per cent in Australia and 17.86 per cent in the UK), we presented our main findings with all respondents.
Finally, we are confident in our null findings relating to the type of platform because Sky News was ranked as more mainstream and trustworthy than the YouTube Channel by respondents in both countries. Moreover, Sky News was perceived as mainstream and trustworthy as the two other popular channels in each respective country, ABC for Australia and the BBC for the UK.Footnote 16 Additionally, we run the models where we only include respondents who ranked Sky News as a mainstream platform. The models, which are found in Table H.5 in the Appendix, show consistent results with the main models, thereby suggesting that our findings hold for the majority of respondents who consider Sky News to be a mainstream platform.
Mechanisms
While we have established that media platforming of extreme right activists influences individuals’ support for extreme right statements and their perception of others’ endorsement of these statements, we found that the type of platform (YouTube or Sky News) on which the interview was allegedly broadcast does not impact respondents’ attitudes or beliefs. In line with our theory and hypotheses, the most credible explanation for these findings is that respondents update their views in the direction of the information that they receive (Coppock Reference Coppock2023), even if that information is extreme and, at least partially, incorrect.
In Table 4, we provide evidence on potential mechanisms that might explain some of these findings. In line with an explanation based on updating, we find that British respondents are not only more likely to agree with extreme right statements in the unchallenged condition, but they are also more likely to believe that these statements are accurate after exposure. Moreover, once challenged by the interviewer, Table 4 shows that respondents are significantly less likely to rate these statements as accurate, and hence do not approve of them. This finding highlights the role that fact-checking can play in correcting extreme right statements, which is consistent with existing studies (Chan et al. Reference Chan, Jones, Jamieson and Albarracin2017; Wood and Porter Reference Wood and Porter2019). While we cannot statistically identify the causal chain that connects beliefs in the factual accuracy of a statement and agreement with that statement, the evidence is at least consistent with such a mechanism.
Table 4. Perceived accuracy of ER statements and identity of actor
Heteroskedasticity-robust standard errors in parentheses.
Covariates: age, gender, education, region, vote in the 2019 elections, authoritarian attitudes, party-ID and Brexit vote.
$^{***}p \lt 0.001$
;
$^{**}p \lt 0.01$
;
$^*p \lt 0.05$
There are possible alternative interpretations of our findings that we need to address. First and foremost, the interviews might have failed to convey that the activists interviewed are extreme or far right activists, or perhaps this information was only conveyed in the challenged interview condition and not in the unchallenged interview condition. Such an explanation might appear plausible, given the relatively high levels of agreement (1/3 in the UK and just below 50 per cent in Australia) with the actors’ statements in the placebo conditions. We find very strong evidence against an interpretation that questions whether subjects were able to infer the true nature of the actor or his statements. Table 4 clearly shows that an overwhelming majority of subjects are able to correctly identify the interviewee as a far right activist. We find that respondents who listen to the interview are much more inclined to believe that Robinson is a far right political figure, as opposed to those who listen to the weather report. The effect amounts to almost 60 percentage points. The unchallenged interview alone makes his extreme right ideology apparent. In fact, as Table 4 shows, the challenged interview adds little to how respondents categorise the interviewee.
Next, we test whether subjects update their attitudes across the board or if there is significant heterogeneity based on socially conservative pre-treatment attitudes. The latter would suggest that only those already predisposed to conform to extreme right statements would eventually do so. Explanations focused on backlash and polarisation would also suggest that socially liberal individuals should be less likely to agree with the statements made after exposure. However, in our pre-registered test for heterogeneous treatment effects by liberal-authoritarian attitudes, we find only weak evidence in favour of such a mechanism, and no evidence of backlash. Figure 4 displays the Conditional Average Treatment Effects of the (unchallenged) interview, conditional on liberal-authoritarian attitudes (ranging from liberal 0 to 1 authoritarian) that we recorded based on agreement with multiple unrelated statements before the treatment. While the positive effect of the interview appears stronger among those with medium to high authoritarian attitudes, the interaction is not statistically different from zero, and the effect is not negative among liberal respondents. In fact, what is clearly visible from these figures is just how prevalent authoritarian social attitudes are in both countries, although they are even more pronounced in Australia. The absence of heterogeneous effects based on prior liberal-authoritarian beliefs is consistent with attitudinal change coming about via persuasion rather than subjects revealing already-held beliefs if norms are broken. Therefore, changes in attitudes and changes in norms, while sometimes being observed together, are distinct processes, with possibly distinct downstream consequences.
Figure 4. Conditional average treatment effects of unchallenged interview on agreement with extreme right statements, conditional on authoritarian attitudes in Australia (left) and the UK (right): 95 per cent confidence intervals.
In Appendix Table H.5, we showed that respondents rate Sky News as significantly more mainstream and trustworthy than a YouTube Channel. In Figure 5, we display the Conditional Average Treatment Effects of the type of platform by age. It might be possible that older respondents are more likely to perceive platform effects than younger respondents who are more familiar with online media. While we find some evidence in favour of that in Australia, we find no evidence of such an explanation in the UK. Platform effects are zero in the UK across all ages.
Figure 5. Conditional average treatment effects of the platform on agreement with extreme right statements, conditional on age in Australia (left) and the UK (right). 95 per cent confidence intervals.
Conclusion
Does the platforming of extreme right actors contribute to the spread and normalisation of extreme right views, and if so, how? Using two large survey experiments based on real-life interviews with extreme right political activists in Australia and the UK, we consistently find that giving a platform, no matter the type or style of the interview, to an extreme right activist leads to a change in perceived descriptive norms. Our study also suggests that uncritical interviewing leads individuals to agree with the statements made by the actor. Our findings on belief accuracy are consistent with a mechanism suggesting that people become persuaded of the accuracy of their views when they go unchallenged. However, the type of platform on which the extreme right actor diffuses his ideas does not appear to matter per se. Being exposed to an extreme right interview on traditional mainstream TV channels or on the online platform YouTube has the same effect on attitudes and norms. Drawing on the UK experiment, we also show the relevance of a journalist’s critical stance towards the actor and his statements. Individuals are more likely to agree with extreme right statements and think that others agree with these statements if the actor’s extreme and false claims are not challenged by the journalist. Having a journalist challenge the actor’s ideas counteracts the attitudinal effects that we observe in the unchallenged interview and has a large negative effect on how ‘accurate’ people rate his statements. However, we find zero evidence that challenging or adversarial interviewing makes people less likely to agree with extreme right statements, compared to the placebo group. This is an important finding, which speaks directly to the popular claim that (critical) exposure damages extreme right ideas. It seems to have a negative impact on the actor, as respondents exposed to the adverse interview rate him as ‘less credible’. However, it does not appear to defeat extreme right ideas, only fight them to a draw. Moreover, while adverse interviewing attenuates the normalisation of extreme right views in society, these normalisation effects do not entirely disappear, but appear to materialise at a lower rate. Finally, we find that exposure is not costless for extreme right actors, at least not in all contexts. While their name recognition increases overall, the image of the extreme right activist is consistently tarnished in the British experiment, with negative ratings increasing more than positive ratings after exposure. This is not the case in the Australian experiment, where both positive and negative ratings increase in tandem.
This study contributes to the literature on the support for the far right by providing evidence for the attitudinal and normalisation effects of media exposure to extreme right actors and messages via interview formats. Individuals appear to update their attitudes in the ideological direction of the message, which is consistent with more general findings from persuasion experiments conducted in survey environments (Coppock Reference Coppock2023). However, it is striking that updating in the direction of the information still occurs even when extreme statements are made and strong source cues related to extreme right actors are present and perceived by respondents. Clearly, a large majority of participants were able to identify the interviewee as being part of the extreme right using only the statements made, without relying on the challenging questions and contextualisation provided by the interviewer. People are not only more likely to agree with extreme right statements if they are exposed without counter information, but are also more likely to believe that such statements are accurate. This goes against the backlash and polarisation hypotheses, which would predict that people’s attitudes move against the extreme right after exposure or in opposite directions based on their ideological priors, leading to polarisation. This is not the case: exposure to extreme right ideas moves people in the direction of those ideas, not against them. Exposure also does not appear to polarise their views, at least not to a large extent. Where we do observe some polarisation is in the case of the actor’s image, but we cannot predict this polarisation based on prior attitudes.
Our findings also suggest that cueing the type of platform where the interview is broadcast does not influence people’s views or their perceptions of how popular those views are in society at large. While this might contradict some studies that emphasise the importance of source cues in messaging (Druckman Reference Druckman2001; Chong and Druckman Reference Chong and Druckman2007; Miller and Krosnick Reference Miller and Krosnick2000), our findings are consistent with other studies that find that content far outweighs the importance of source cues when it comes to people’s perceptions of information quality or accuracy (Clayton et al. Reference Clayton, Davis, Hinckley and Horiuchi2019; Dias, Pennycook and Rand Reference Dias, Pennycook and Rand2020; Berinsky Reference Berinsky2017). Our results also resonate with Peterson and Allamong’s (Reference Peterson and Allamong2022) recent experimental study, which finds that unfamiliar media sources (that is, with no pre-existing reputation) influence opinion as much as familiar media sources. However, our cue was relatively weak as it only involved a mention at the start and end of the interview by an actor. Therefore, future studies could attempt to test the type of platform hypothesis using additional cues, such as mimicking a high-quality studio set-up or incorporating the platform logo throughout the interview, especially if using videos.
Moreover, by conducting realistic and credible experiments in Australia and the UK, we complement studies of media (non-)exposure that primarily focus on the United States (Broockman and Kalla 2022; Gerber, Karlan and Bergan Reference Gerber, Karlan and Bergan2009). Our experiments combine a high degree of internal validity with environmental and external validity by using real-life interviews and reaching similar conclusions in both Australia and the UK. At the expense of not being able to detect if these attitudinal and normalisation effects are durable, we can show that persuasion and normalisation effects are consistent in two countries with varying shares of conservative media. Interviews with extreme right actors shift people’s views to the extreme right in environments where people are exposed to two-thirds of conservative media (like in Australia), or only one-third (like in the UK) (MRC 2021). If we are willing to assume that short-term effects would translate from the survey environment, in which participants were exposed to the interview recordings, to the real world, then we can estimate, given a daily viewership of around 50,000 of both the Bolt Report, where the Cottrell interview was aired (Guardian 2018) and individual Sky News broadcasts (Press Gazette 2022), that around 2000 people might have shifted their personal opinions to the right after viewing the Cottrell interview. 1000 people in both cases would have seen their descriptive norm perceptions impacted. These are conservative estimates given that copies of the interviews are accessible online.
Our findings are likely to generalise to a large set of extreme right issues, as we observe similar results in both interviews across topics related to immigration, crime, terrorism and Islam. They should also hold across most industrialised countries, where extreme right activists are an increasing presence in the media – whether it’s Alex Jones on InfoWars in the USA or Eric Zemmour in Le Figaro in France. Additionally, as conservative media outlets and platforms such as Fox News or GB News, along with talk radio or podcasts such as the Salem Radio Network or the Daily Wire, have gained popularity, more and more people are likely to be exposed to such interviews. Finally, considering that exposure to extreme right content appears to impact individuals irrespective of their ideological orientation, these persuasion and normalisation effects can occur if mainstream media begins presenting extreme right content, thereby reaching a broader audience.
This study also makes important contributions to the emerging literature on the normalisation of the extreme right (Abou-Chadi and Krause Reference Abou-Chadi and Krause2020; Bischof and Wagner Reference Bischof and Wagner2019; Valentim Reference Valentim2021). While often assumed but rarely tested, we have demonstrated that media platforming of extreme right actors contributes to a change in perceived social norms. We employed descriptive norms as an outcome variable, a rarely utilised approach in political science, to show that exposure to extreme right interviews in the media prompts individuals to revise their perception of others’ agreement with those statements. This finding is significant as others have shown that individuals rely on descriptive norms to shape their own behaviour (Valentim 2023; Tankard and Paluck Reference Tankard and Paluck2017). We also find that changes in descriptive norms do not always go together with changes in attitudes. This finding, combined with the finding that updating of views is happening across the political spectrum and people rate statements as more accurate after one-sided exposure, speaks in favour of persuasion effects on attitudes. Further research should thus examine the downstream consequences of changes in descriptive norms. Just as media outlets respond to perceived demand from viewers for such interviews, it is possible that airing these interviews creates demand via changes in social norms, leading to a positive cycle of airing more extreme content. Additionally, while this study has focused on the extreme right, subsequent studies could investigate the effects of media platforming of views articulated by extreme actors across the political spectrum, including extreme left-wing activists. Based on our results, we have no reason to believe that attitudinal and normative updating would not occur similarly.
Finally, our findings align with recent studies that demonstrate how negative media portrayals of Muslims increase negative sentiment towards them and fuel support for anti-Muslim policies (Lajevardi Reference Lajevardi2021). Moreover, we provide evidence that increased media exposure to derogatory speech drives anti-outgroup attitudes and that these negative effects are comparable for different perceived outgroups, such as Black Africans in Australia and Muslims and refugees in the UK. This is particularly concerning given the increased media coverage of Muslims (Lajevardi Reference Lajevardi2021) and refugee stories over time (Lo and Lang Reference Lo and Lang2023). Most of the extreme right statements we tested are extreme in their nativism, but we also show that effects materialise when discussing skills-based immigration, a less extreme policy also supported by some conservative parties. While we suggest that these persuasion and normalisation effects may also apply to more far right issues, further studies should investigate the scope conditions of our findings more thoroughly.
Our study has significant implications for politicians, policymakers and journalists. First, the finding that platforming unchallenged extreme right content persuades individuals and normalises extreme views in society is sobering, especially for those who expect that extreme right views lose credibility when broadcast.
Second, persuasion and normalisation effects can be significantly alleviated if journalists engage with extreme right actors critically and dare to challenge them robustly. Journalists who fact-check incorrect claims nullify the effectiveness of the conveyed message on attitudes and significantly mitigate, although not nullify, the process of normalising these views in society at large. However, importantly, they do not reverse the process but can hope to counteract the extreme right activist’s influence. As a result, our study suggests that if TV channels decide to platform extreme right activists, journalists who adopt a critical, challenging tone and question the accuracy of false statements can mitigate attitudinal and, to a lesser extent, normative effects. Third, in some contexts, critical engagement appears to negatively affect the actor’s image. This means that exposure does not come without drawbacks for the extreme right activist. Finally, reach matters. Media platforms, whether traditional mainstream TV channels or alternative Internet platforms, can serve as powerful spaces for spreading and normalising extreme right content. The power of traditional mainstream media sources appears to lie more in their ability to capture a larger audience than in any inherent difference in the effectiveness of their messages once they have gained an audience. Therefore media de-platforming still appears as an effective tool to minimise the reach of extreme right actors and hate speech, as shown by recent social media studies (Rauchfleisch and Kaiser Reference Rauchfleisch and Kaiser2021; Buntain et al. Reference Buntain, Innes, Mitts and Shapiro2023). Donovan and Boyd (Reference Donovan and Boyd2021) argue for a new editorial approach, ‘strategic amplification’, which requires news media organisations and platform companies to develop and employ standards for content moderation when producing news content and the algorithmic systems that help spread it. Media companies can enforce standards for content moderation in line with strategic amplification and de-platform individuals who break them. If exposure is necessary or unavoidable, journalists can ask challenging questions, including questioning the accuracy of the statements that extreme right actors routinely make. In times of growing media exposure to extremist actors and content, both media companies and journalists can play their part in countering the empirical pattern that this study documents.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0007123425000195
Data Availability Statement
Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/QF5PAY.
Acknowledgments
We are grateful to Tarik Abou-Chadi, Silja Häusermann and Stefanie Walter for funding and feedback, as well as to the University of Durham and King’s College London for funding our study via Faculty Research Grants. We would like to thank Tarik Abou-Chadi, Daniel Bischof, Francesco Colombo, Andrea de Angelis, Elias Dinas, Sara Hobolt, Rob Johns, Adeline Lo and Pauliina Patana for providing valuable comments on previous versions of the paper. Moreover, we thank the participants of the Social Norms workshop, organised by Daniel Bischof and Elias Dinas in 2019 and 2021, the participants of the 2023 polarisation and political hostility workshop at the LSE, and the participants of three panels at EPSA 2018, EPSA 2019 and EPSA 2020 for valuable comments on research designs and findings. We also thank the seminar participants of the University of Durham, the University of Konstanz, the University of Mannheim and the University of Zurich for their very helpful feedback on earlier versions of this paper. We are particularly grateful to Liz Ralph-Morrow for extended conversations about the initial design of the Australia study.
Financial Support
This work was supported by the Research Priority Programme ‘Equality of Opportunity’ of the University of Zurich, as well as the University of Durham and King’s College London via Faculty Research Grants.
Competing Interests
None.
Ethical standards
The research was conducted in accordance with the protocols approved by the Ethics Committee of the London School of Economics and Political Science.
Open access
