The Rise and Fall of the Filter Bubble

The term "filter bubble" was brought to prominence by Eli Pariser in his 2011 book, The Filter Bubble: What the Internet is Hiding From You. Ironically for a book that anthropomorphised the Internet, imagining it as a malevolent entity that sought to confound us, the subject was personalisation, but in this case the attempt to produce more "relevant" content for users through the analysis of their networks, past choices and demographic characteristics. Pariser's theory was that personalisation in social media was simply reinforcing existing biases by excluding conflicting viewpoints. Google's search results and Facebook's news feed were being subtly censored in a way that simultaneously increased plurality (there was now an infinite number of bubbles rather than the single narrative of the totalitarian dystopia) while reducing individual cognitive diversity (you were isolated in your own, singular bubble). This wasn't a new idea - it's just a version of the echo chamber, after all - but Pariser caught the zeitgeist by suggesting that the tech titans were at best indifferent to social capital and at worst malign. Despite ample evidence that default personalisation produced little variation in results, the filter bubble became an established fact in the discourse on the social impact of the Internet.

One reason for this was that the idea chimed with the narrative of increased political polarisation, particularly in the USA. The Internet was thought to be driving people into more entrenched camps, producing greater partisanship and "alternative facts". Inevitably, the filter bubble has been blamed for Brexit and the election of Donald Trump. This distracts attention from the more overt bias and influence on popular discourse exercised by newspapers and television, which may be getting worse as the global news market takes a more nationalistic turn. Insofar as polarisation is an actual social phenomenon, rather than just a reflection of changes in the dynamics of party politics (e.g. the declining significance of race as a bipartisan issue, both in progressive and conservative registers, in the US after the 1960s), it appears to be most acute among older voters with a greater reliance on traditional media. As one American academic study noted, "We find that the growth in polarization in recent years is largest for the demographic groups least likely to use the internet and social media".

In practice, a greater exposure to content of any sort will lead to greater diversity. Even if you only augment The Sun with The Daily Mail you will increase the diversity of views, simply because the overlap between the two is not perfect. The net effect on diversity of opinion will be positive, even if slight (it can't be negative). If you use social media, the odds are that you will be exposed to a broad variety of viewpoints simply because your own family and friends will likely include people with different opinions and affinities. They may be utterly wrong about everything, but you're still going to be exposed to their perspective. Unless you rigorously curate your preferences, and so take control of your personal filter (in which case the idea of an imposed filter bubble does not apply), you are probably exposed to far more diversity of opinion today than you were twenty years ago. Some of that additional opinion will be nonsense or propaganda, but there is no good reason to believe that the relative proportions of "good" and "bad" in new media will be any different to traditional media.

Despite the flaws in the filter bubble theory, its proponents continue to believe its fundamental truth even in the face of contrary evidence. For example, Rachel Botsman (a former director of the William J Clinton Foundation) writing in The Guardian concedes that social media actually increase diversity but remains determined to salvage the filter bubble: "A recent Reuters Institute Digital News Report found that 44% of people in the US who use social media for news end up seeing sources from both the left and the right, at more than twice the rate of people who don’t use social media. However, that’s not to say they necessarily pay attention to any contrary views. When Facebook rolled out its 'related articles' feature last year, users continued to ignore information that undermined their favoured narrative." If you can be bothered to click on the link in Botsman's article you will discover that she has misrepresented the Facebook change, which actually led to lower rates of sharing of contentious articles by providing counter-narratives.

Botsman is typical of the neoliberal establishment in simultaneously extolling freedom and insisting on the need for restraint: "I’m a huge advocate of free speech, open democracy and online dialogue that mirrors the exchange of opinions that happens offline – at work, at home and even in classrooms. The issue is that it has become a free-for-all, a corruptible beast that we can’t, or haven’t, yet learned to control". It should be obvious here that the "beast" is not the Internet but that more traditional animal, the bestial mob. Social media has become a symbolic substitute for society, allowing establishment figures like Botsman to not just attack the common herd for its lack of manners and good sense but to revel in their own contempt. To this end, the inflation of the troll from a social nuisance to a state actor with malign intent (those Russian bots) is a distraction from what would otherwise be a whine against the impropriety and disrespect that characterises the everyday exchanges that Botsman claims to value. The hyping of the threat to democracy is an anti-democratic manoeuvre.

That the filter bubble has actually come to prominence during a period of increased cognitive diversity should lead us to wonder what purpose the concept serves. One explanation can be found in the space where political centrism intersects with the anxieties of traditional media outlets over falling advertising revenues. Centrism by definition deplores partisanship and polarisation, but it also believes that regulation is necessary in the public sphere to ensure a level playing field. This obviously conflicts with the traditional liberal exaltation of the free press, requiring the hyperbole of extreme threats, from antisocial corporations to state actors, to excuse intervention. For traditional outlets like newspapers, the existential crisis presented by the changing advertising landscape is enough to justify a state of exception, but they have no better idea how to tackle the problem of increased diversity than the centrists. The consequence has been a steady inflation in the scare stories and an increasing note of hysteria among traditional media commentators facing falling sales and thus the prospect of their own eventual redundancy.

The limited attempts to address the circulation of "fake news" via social media have centred on the independent verification of news sources. These have tended towards ranking, i.e. classifying content as more or less reliable based on the reputation of the source. This means providing a weighting of content providers that can be incorporated into algorithms, which is little advance on Google's Page Rank. The obvious aim is to reconcile the different interests by giving traditional outlets a privileged position as sources, but this simply substitutes credibility for traffic and is therefore economically unsustainable. The fight is ultimately over money, not truth. Platforms like Facebook are comfortable with this approach not only because it doesn't threaten their revenues but because it allows them to avoid the responsibility of editorial control. To rub salt into the wounds of traditional outlets, this can as easily be done by outsourcing judgement to unpaid volunteers (community self-policing) or not-for-profits as to established media brands. As Mark Zuckerberg disingenuously explained it: "We do not want to be arbiters of truth ourselves, but instead rely on our community and trusted third parties".

The filter bubble, as a controlled environment in which the consumer is passive, has gradually given way to the idea of an active network in which people share junk news (i.e. deliberately false and extremely partisan). From slaves of the machine they have become active amplifiers of each other's prejudice (the way these networks are described often employs tropes familiar from reports of Islamic radicalisation). This evolving idea has been politicised by the evidence that such networks are more prevalent on the political right than the left, leading to the idea of "network contagion" being associated not just with polarisation but with political populism. A recent study by the Oxford Internet Institute (OII) has been widely cited by liberal media in this regard, despite its methodology being quite hilariously flawed. Its sample of junk news is biased towards the right (so comparisons with junk news circulation on the left are impossible) and no rationale is provided for its segmentation of consumers into value-laden groups such as "the resistance". Many of the junk news sources turn out to be vanilla conservative outlets, like National Review.

The stylistic characterisation of junk news in the OII study includes this classic line: "These outlets use emotionally driven language with emotive expressions, hyperbole, ad hominem attacks, misleading headlines, excessive capitalization, unsafe generalizations and fallacies". This description would obviously apply to many mainstream newspapers. The real reason for the rightward bias in junk news is simple economics: there is a large market for rightwing junk and a willingness to spend disposable income on value-reinforcement (the gammon pound, in British terms). This isn't just a structural reality, reflecting the correlation of age with both increased conservatism and financial comfort, but the result of a deliberate cultivation of this market segment through traditional media since the 1970s, notably cable TV, shock-jock radio and partisan newspapers. Though exploited mainly by conservatives, the key enablers of this development have been successive neoliberal reforms that simultaneously removed the social obligations that had been placed on traditional media and insisted that editorial strategy should be subservient to market forces. If you want to understand junk news, start with the deregulation of US TV news in the mid-90s (enacted under a certain William J Clinton).

Having started out as a vision of passive isolation, the filter bubble has matured into the trope of an active cult. The next stage in its evolution points towards irrelevance as more apocalyptic visions of media and content-generation emerge, such as the idea that bad actors will soon be able to fake video so well that we will struggle to believe the evidence of our own eyes (it is worth remembering that cinema has been simulating contemporary reality since Sergei Eisenstein restaged the storming of the Winter Palace, while the tradition of impersonating public figures on the telephone is as old as, well, the telephone). What is telling about this imaginative hellscape is less the implausibility (the technology capable of determining a simulation will advance in parallel with the technology for creating one) but the assumption that we will have once more been reduced to passive idiots, which tells you a lot about the media editors who commission such fantasies. The filter bubble is dead. Long live the filter bubble.