Unbundling the Imaginary Community

How we measure attention incentivises content that erodes our public sphere and polarises communities, including in the Global South. Could better metrics incentivise better content?

  1. Media Metrics Matter

Our ad-based platforms aren’t just fuelling social and political polarisation in post-industrial economies — we know from research that they are having an impact in emerging markets as well. As news emerges of Facebook meeting representatives from the Philippines, Myanmar and Sri Lanka about it’s misinformation problem — it’s become clear that the social platforms which increasingly dominate our source of information, have grave and negative effects. Social media platforms are funded by advertising — indeed, we the users are their products and the advertisers their customers — and to increase their product they have developed very sophisticated mechanisms for showing more advertisements, which equals more of our attention. I argue that changing the model away from more adverts equals more revenue might produce other positive knock-on effects, and that one way to change that model is to change the metrics used to measure what people see.

Colleagues at Caribou Digital have found that the shallow measures of attention that dominate digital advertising incentivise the production of cheap, low value content — what the author describes as unhealthy ‘junk news’ content. Building on this, I argue that that this makes it harder for the media to play a constructive role in upholding the inclusive, public sphere so necessary for inclusive, healthy societies. In work on advertising models and emerging market media this was found to be as true for emerging markets as it is for the US and European countries. As Facebook pushes to increase usage in Africa we should think carefully about the relationship between platforms, advertising based business models and the social value of content.

Of course it’s hardly new to claim there’s a relationship between business models, media content and social outcomes. The anthropologist Benedict Anderson, for example, argued that print capitalism enabled the emergence of a sense of nationhood, as consumers of national newspapers increasingly imagined themselves members of a single community. Communication theorist Jurgen Habermas also viewed information as a critical component of society, but saw the rise of commercial media — and particularly the influence of special interests — as eroding public debate and individual ability to make rational choices. In both cases, the underlying business model played an important role in incentivising the development of particular forms of media, which in turn contributed to social and political outcomes.

Central to the economics of media production has been the way in which attention is measured, and particularly the bargain that publishers make with audiences for their time. Historically publishers paid for newspapers by charging consumers through cover prices, subscription fees or third party patrons. Later, as publishers realised they could make more money and reach larger audiences through advertising, a bargain was struck; instead of paying full production costs for mass media content, audiences traded their attention, accepting the presence of advertising in exchange for access to cheaper news and other content. Publishers started prioritising popular content that was inexpensive to produce but which attracted advertising revenue (such as sports and gossip) bundling it together with content that was more expensive to produce but which reached smaller audiences and possessed higher social value (such as international news and investigative reporting).

Now this bargain has come undone. This is largely because of changes in the way attention is measured and the unbundling of the two types of content, expensive and inexpensive.

In the past the unit of analysis for the channel of communication was the newspaper, radio station or TV channel; and the unit of analysis for attention was total channel sales figures or audience size (often approximated through representative audience panels). Today the measurement of consumption has changed. Instead of stations or publications, consumption analysis tracks individual pieces of content as they are atomised and unbundled from their original channel. Similarly measurement of attention is also far more detailed, with analysis of individual consumers replacing broad audience size and social class breakdowns. Content popularity is measured through blunt metrics such as clicks, likes and shares while metadata and cookies record a wide range of individuals browsing behaviours. The atomization of content also divorces content from brand, and consequently means each piece of content is unable to rely on the brand as a signal to readers about the value/quality/topic of the content, but must survive, in the form of sufficient clicks and likes, on its own. These changes in the measurement of attention and the atomisation content are undoing the bargain made between publishers and audiences.

Commuters on a NYC train (Kubrick)

2. The Unbundling of Content

As this shift in metrics has led to unbundling of content, one consequence is an nutritional inequality with nutritious high-value information on one side and low value ‘junk news’ on the other. The mechanics of monetisation means that this nutritious content is placed behind paywalls where — like organic food — only the wealthy can afford to consume it.

By contrast, ‘junk news’ is paid for through low value advertising, and because revenue is driven by audience volume it is — as my colleague Bryan’s report puts it — ‘scientifically engineered to be cheap, addictive, and marketed to the poor’, just like junk food. At its most benign, this can look like easily-recognisable formats such as generic listicles or clickbait headlines; but more insidious is the incentivisation of content produced solely to attract attention with no concern at all for quality, such as the ‘fake news’ around the 2016 American election, when Macedonian publishers were incentivised to produce anti-Clinton content because it attracted Republican audiences.

The unbundling of content thus leads to an unbundling of the public sphere; the ‘imaginary community’ that Anderson described comes apart as this information divide further separates the rich and the poor. And as we trace this unbundling back to the incentives which generate content to begin with, we start to see that the way in which attention is measured is at the root of the problem.

Commuters on a London Tube (Getty)

3. Media metrics and the public sphere

Advertisers are not concerned with this unbundling of the public sphere; but they are increasingly concerned about the shortcomings of digital advertising. In our interviews with media, advertising and branding representatives, we found widespread recognition that measuring clicks, likes and shares does not reveal very much about audience behaviour before or after engaging with adverts, or the content that ads are linked to. For example in 2017 Vodafone established an agreement with Facebook to advertise only against whitelisted quality publishers; and P&G cut their spend on digital adverts by $100m to prioritise ‘media reach’ through television, audio and ecommerce.

Emerging markets face the same challenges, and in their struggles there are pointers as to how we might mend the metrics that drive cheap content and the subsequent political polarisation. In some countries media companies are fighting back. In Vietnam for example, media companies are asking advertising agencies to train their staff in digital skills to better compete with social media platforms; in Malaysia, the two largest media companies — traditionally competitors — have come together to create a joint market in which to sell their advertising space to clients; and in India, there is a wider push by the government to resist ‘colonisation’ of personal data by Western companies as part of wider efforts to protect their indigenous technology sector.

Producing nutritional content is expensive but important — it’s a vital part of informing citizens so that they can play a constructive role in their society — but in the long term advertising-based business models will not be able to support the quality of media that the public sphere requires. There will need to be efforts in many areas, from finding alternative revenue sources such as micro payments, to building better regulatory frameworks, to promoting digital literacy.

In the short term, efforts to distinguish good from bad content are one approach, such as Santa Clara University and Facebook’s initiative to building trust metrics; but building a more ethical digital media market means that we have to find more insightful metrics, such as the efforts by Chartbeat and Content Insights, who are working to develop metrics that bring the focus back on stories and away from likes and shares. These kinds of services seek to provide publishing platform owners with far more detailed analytics about individual behaviour on their site, allowing publishers to see which content drives engagement. For Chartbeat, that means measurements such as how long someone spends on a page, whether they read to the bottom of the article or watch the entire video. For Content Insights, high value metrics can be achieved through a broader suite of suite of audience behaviours, from time spent on specific content to repeat visits, an indication of loyalty to the brand and content. But both Chartbeat and Content Insights, like most digital analytics, can only measure from the perspective of the specific platform. Cookies and trackers can provide some insight, but the broader picture of what people do after engaging with that content is invisible. For companies, these analytics are unable to describe what individuals did when they visited the site, and what the outcome of their visit was — most notably, whether it translated into sales. Here companies such as Verto Analytics are developing cross-platform insights from panels made up of users who download an app that measures their behaviours. But for those interested in politics and social outcomes of news and entertainment content, these analytics are unable to reveal how people felt about content, who they shared content with and what it means in the context of the rest of their digital life.

Better metrics will make for more nuanced insights into audience behaviour and healthier consumer engagement. In an ideal world, better metrics would reveal that readers spend time on long form, quality content, creating the case to persuade advertisers to fund high nutrition over junk food content. But it’s more likely that such metrics will only confirm our worst fears: that people prefer listicles and sports over long form investigative journalism. However, more detailed metrics might also help us to understand what draws people to quality content, and how they engage with it. We also need metrics that measure quality in other ways, linked to broader behaviours and social outcomes. More detailed insights into individual digital behaviours, and how content fits into their broader social lives, may help provide the basis for metrics that can incentivise and measure good outcomes. Developing better metrics may well help to incentivise the production of content that can strengthen the public sphere, helping build a world where rich and poor alike are able to access nutritious content.