Are Filter Bubbles Real?

Axel Bruns

• Publisher: Polity
• Year: 2019
• ISBN:
  9781509536467
• Edition: 1
• Topic: Sprachen & Linguistik
• Subtopic: Journalismus

There has been much concern over the impact of partisan echo chambers and filter bubbles on public debate. Is this concern justified, or is it distracting us from more serious issues?

Axel Bruns argues that the influence of echo chambers and filter bubbles has been severely overstated, and results from a broader moral panic about the role of online and social media in society. Our focus on these concepts, and the widespread tendency to blame platforms and their algorithms for political disruptions, obscure far more serious issues pertaining to the rise of populism and hyperpolarisation in democracies. Evaluating the evidence for and against echo chambers and filter bubbles, Bruns offers a persuasive argument for why we should shift our focus to more important problems.

This timely book is essential reading for students and scholars, as well as anyone concerned about challenges to public debate and the democratic process.

ONE
Introduction: More than a Buzzword?

As they leave the White House, outgoing US Presidents reflect on their experience. In 1961, Dwight D. Eisenhower used his final television address to warn of the ‘grave implications’ of the powerful military-industrial complex that – as a former World War Two general and Supreme Commander of NATO – he had himself helped to create. Only an ‘alert and knowledgeable citizenry’ could act as a corrective and control to the ‘unwarranted influence’ of the military-industrial complex (1961: 2–3); though the Cold War has ended, his warning resonates still.

Barack Obama’s farewell speech in Chicago on 10 January 2017 warned of a different danger, similarly requiring attention from an alert and knowledgeable public: amidst increasingly heated and uncivil political discourse, he suggested, ‘for too many of us it’s become safer to retreat into our own bubbles, whether in our neighbourhoods, or on college campuses, or places of worship, or especially our social media feeds, surrounded by people who look like us and share the same political outlook and never challenge our assumptions’ (Obama 2017: n.p.). Like Eisenhower, Obama should know a good deal about this topic: after all, with his own online community site my.barackobama.com and later with mainstream platforms such as Facebook and Twitter, he is widely recognised as the first US President to successfully use social media in election campaigning.

Obama’s concerns are far from new, however. Under labels such as ‘echo chambers’ and ‘filter bubbles’, they are held responsible for a broad variety of social and political problems. Social media – and the content filtering and recommendation algorithms they rely on – have been identified as one reason that such phenomena have emerged; before them, search engines and the overall multiplication, fragmentation, and personalisation of available media sources in the Internet age were also highlighted as contributing factors. Put simply, the argument goes, if we all disappear into our different information cocoons, where we only ever encounter like-minded others and are served a highly selective media and information diet, then society fragments. Worse still, because it requires a well-informed citizenry and cross-ideological dialogue to reach a broadly supported consensus about the future direction of the state, democracy itself falters.

This book addresses these concerns and reviews the evidence for and against echo chambers and filter bubbles (in search, social media, and beyond). Ultimately, it argues that such fears about the societal effects of disconnected informational spaces, and about the role of new media technologies in their creation, only divert our attention from the much more critical question of what drives the increasing polarisation and hyperpartisanship in many established and emerging democracies.

Obama’s warning followed Donald Trump’s surprise victory in the US presidential election in November 2016; it mirrors similar concerns after the UK’s Brexit referendum in June 2016, and other apparently unexpected wins for populist causes and candidates, despite widespread coverage of their personal and policy flaws. Politicians, journalists, and scholars who support the echo chamber or filter bubble thesis suggest that, with online and social media as the key sources of information for an ever-growing percentage of the public, it became possible for citizens to be locked into ‘ideological filter bubbles that lacked cross-cutting information’ (Groshek and Koc-Michalska 2017: 1390). Brexit supporters might never have seen the warnings of the disastrous economic, social, and political consequences of exiting the European Union; Trump followers might not have encountered reports of the candidate’s personal flaws or the campaign’s collusion with foreign powers.

Away from these momentous occasions, echo chambers and filter bubbles are also held responsible for the emergence of many more communities that espouse contrarian and counterfactual perspectives and ideologies. These include comparatively harmless conspiracy theorists compiling evidence of alien visitations; groups of agitators who steadfastly deny the reality of anthropogenic climate change or the benefits of population-wide vaccination; and extremists preaching racism, homophobia, and religious intolerance. The argument is that echo chambers enable these groups to reinforce their views by connecting with like-minded others, and that filter bubbles shield them from encountering contrary perspectives.

Such disconnection from and ignorance of alternative perspectives is said to result from a combination of individual choice, in selecting the news sources to consult or the social media accounts to follow, and the algorithmic shaping of such choices, as search engines, news portals, and social media platforms highlight and recommend some sources over others. As platform algorithms learn from users’ choices, and users make those choices predominantly from the options promoted by the algorithms, a self-reinforcing feedback loop gradually curtails choice to an increasingly narrow and homogeneous set of options (Bozdag and van den Hoven 2015: 252). Such processes are also susceptible to interference from malicious actors: the operators of the mis- and disinformation campaigns described by the problematic label ‘fake news’ seek to manipulate and exploit the logic of social media platforms and their content recommendation algorithms to give their own, fabricated material more prominence than higherquality content from reputable sources (Woolley and Howard 2017: 6).

Yet even in the absence of deliberate misinformation campaigns, the algorithmic shaping of information streams is problematic. The algorithms follow opaque and non-transparent decision-making processes whose effects are rarely articulated clearly to the user: why does Google Search or Google News present a certain set of results for a given search term, for example? Why does the Facebook newsfeed highlight certain posts from the user’s network and hide others? Why does Twitter recommend specific accounts to follow or trending hashtags to explore? These selections build on the individual user’s and the broader userbase’s actions, but exactly how they are incorporated into the algorithm’s logic remains unknown. At worst, by highlighting the most popular content and thereby making it even more widely visible, the algorithm encourages a form of unconstrained majority rule, where only already popular accounts and content can ever cut through to a wider audience. Amidst a self-perpetuating and even self-amplifying stream of majority voices, minority views are overwhelmed and silenced, and genuinely open public debate between nuanced and diverse perspectives is rendered impossible (Bozdag and van den Hoven 2015: 254).

These, in brief, are the key concerns of those who worry about the impact of echo chambers and filter bubbles; we will explore them in more detail in the following chapters. But are these ‘frequent lamentations’ (Guess 2016: 2) warranted? Is there actual empirical evidence for the existence of such ideological cocoons? Are there other explanations for increasing political fragmentation?

These fears emerged at a time of considerable political upheaval, but also of substantial change in the way news is produced, disseminated, and used. The era of news as ‘a fairly stable product’, delivered via print or broadcast, is almost over (Nielsen 2016: 112). Institutional and technological disruption has always coincided with significant moral panics about what an as yet only imperfectly formed future might hold: the printing press, telegraph, radio, TV, and the early Web each saw similar moral panics about their impact on how citizens might reliably inform themselves about the news (or else be manipulated by populists and demagogues).

This does not mean such fears should be dismissed immediately: the printing press did play an important role in mobilising the public during the American revolution, for example, just as radio did in recruiting popular support for the fascist regimes of the 1930s. Yet these developments did not result inevitably from the technologies themselves, but from how they were utilised by political actors: for good or evil, to reject colonial masters or claim mastery over others, human agency in deploying these technologies channelled their potential power. A moral panic about social media in themselves, then, independent of how and by whom they are used, is no more warranted than one about TV, radio, or the printing press. We would fall for technological determinism: a belief that social media, however platforms might be designed and however citizens might use them, inevitably promote echo chambers and filter bubbles. As we will see, there is no evidence to support such an argument (O’Hara and Stevens 2015: 409). We cannot absolve ourselves from the mess we are in by simply blaming technology.

Because they tap into apparently common-sense understandings of our world, however, moral panics are deeply persuasive, and echo chamber and filter bubble concepts have therefore been accepted widely (often rather unquestioningly) in scholarly and popular literature. In particular, even while ramping up their own social media offerings, mainstream news media have gladly accepted the idea of social media as echo chambers because it enables them to claim that, compared to these new competitors for the attention of news audiences, only the carefully researched and edited news published by established news outlets offers a balanced news diet that penetrates the cocoon. Similarly, for political operatives, the filter bubble has become a handy slur: just as President Trump will dismiss any critical reporting as ‘fake news’, so can politicians now accuse their opponents of ‘living in a filter bubble’, out of touch with ‘ordinary people’.

Such use and abuse of these concepts for political point-scoring is possible because, despite their apparently intuitive meanings, ‘echo chamber’ and ‘filter bubble’ have remained very poorly defined. Even scholarly work tends to use these terms interchangeably (e.g. Graham and Ackland 2017: 190) rather than attempting to distinguish the two. Such conceptual confusion creates considerable variation in how these terms are understood in the research, and frustrates the search for hard evidence of their existence. Even the authors who introduced and popularised these terms, Cass Sunstein (‘echo chambers’) and Eli Pariser (‘filter bubbles’), generally fail to provide clear definitions or to outline concrete methodological approaches for detecting them. As chapter 2 shows in more detail, their early contributions draw largely on hypothetical thought experiments or personal anecdotes, and envisage their deleterious consequences without checking such dark visions against established knowledge about the preferences and behaviours of actual users.

Echo chambers, for instance, would be most effective if their inhabitants use them as the only source of information, but this seems highly improbable in the massively multi-channel environment of the contemporary mediasphere (Dubois and Blank 2018a, 2018b). Filter bubbles, meanwhile, are premised on ‘who one’s friends are . . .: If you have only liberal friends, you’re going to see a dramatic reduction in conservative news’ (Pariser 2015: n.p.). Yet making friends purely on the basis of political leaning seems unlikely for most people. In general, both concepts assign far too much importance to the role of politics in ordinary people’s lives: the selection of friends and followers in social media based on political compatibility is outweighed by friendship, professional, entertainment and other considerations that have little to do with politics in the narrow sense. In discussing ideological cocoons, we must not forget that the vast majority of users are not on social media primarily to talk politics.

The empirical work on echo chambers and filter bubbles that does exist suffers from further limitations, which we will explore in chapters 3 and 4. It is not sufficient to prove that groups on a single social media platform, or even within a specific space within that platform – a Facebook group or Twitter hashtag, for example – appear to constitute an echo chamber or filter bubble, disconnected from the wider world outside. Such studies do not address ‘the actual experiences of individuals’ (Dubois and Blank 2018b: n.p.), because real users do not normally participate only within a single social media space, or engage only with a single platform within the overall mediasphere. A user who participates in an apparent echo chamber on Facebook while consuming a broad diet of mainstream news through other channels does not exist in an information cocoon, however ideologically homogeneous their Facebook community might be.

Indeed, for users who generally maintain a diverse media diet through other channels, the increased personalisation and filtering available in online and social media can be ‘empowering’ (Zuiderveen Borgesius et al. 2016: 5): it provides them with increased autonomy over their news consumption choices. The options available enable different audiences to choose information on the topics that interest them, and to follow the news at a level of detail and complexity that suits their level of information literacy. Participation in specific echo chambers and filter bubbles – if such labels apply to these self-selecting information enclaves at all – could therefore even be beneficial, as long as ‘people have routes out of the chamber’ (O’Hara and Stevens 2...

Table of contents

• Cover
• Front Matter
• 1 Introduction: More than a Buzzword?
• 2 Echo Chambers? Filter Bubbles? What Even Are They?
• 3 Echo Chambers and Filter Bubbles in Action
• 4 Bursting the Bubble
• 5 Conclusion: Polarised but Not Disconnected
• References
• Index
• End User License Agreement