Media Pluralism and Online News: The Consequences of Automated Curation for Society

The book arises from an international research project that explores the future of media pluralism policies for online news. It investigates the latest European policies and techniques for regulatory intervention, and examines the consequences of innovative news practices asking, ‘How will automation of news affect public opinion in the age of social media platforms, and what are the consequences?’

In Media Pluralism and Online News the authors make the argument that there is an urgent need for revitalised thinking for a media policy agenda to deal with the trends to platform power and concentrated media power, which is an ongoing global risk to public interest journalism.

In the transition to a media landscape increasingly dominated by broadband internet distribution and the dominance of US-centric new media behemoths Google, Facebook, Apple, Amazon and Netflix the book investigates measures that can be taken to reduce this ongoing march of concentration and the attenuation of media voices.

Securing the public interest in a vibrant and sustainable news media sector will require that merger decisions assess whether there is a ‘reduction in diversity’ -- calling for a new public interest test and a more expansive policy focus than in the past. This would include consideration of the sustainability of local businesses; the encouragement of original and local news content; quality of content, in terms of the promotion of news standards; and new modes of delivery and consumption, including the ‘automated curation’ of news content by digital platforms. 

• Language: English
• Publisher: Intellect Books
• Release date: Aug 23, 2023
• ISBN: 9781789388503

Book preview

1

Introduction

Tim Dwyer and Derek Wilding

This book represents our research and critical reflections from a four-year Australian Research Council-funded ‘Discovery Project’ entitled ‘Media Pluralism and Online News’.

The broader purpose of the project was to track the dynamic developments in the way news is produced, distributed and consumed online and to account for this in public policy designed to promote media pluralism (Media Pluralism Project 2019). The researchers were also seeking to transition the understanding of media pluralism by studying contemporary European policy approaches and a series of innovative news practices, including through making use of a big data approach to collecting media content. To that end, the project reviewed the best international policy practices, and we conducted interviews with policymakers, regulators and academic experts in Europe and Asia, and we used computational data collection for empirical analysis of news articles on a variety of platforms.

One of the principal aims of the ‘Media Pluralism and Online News’ project was to discover methods of computationally evaluating media pluralism that are relevant for assessments of multi-platform news ecosystems. We investigated how the concept of ‘public affairs’ allows us to separate out and identify the categories of news content that contribute to media pluralism (Media Pluralism Project Dashboard 2020). Our view is that this term is useful in showing the news content that should be valued; that is, content which is constitutive of a news public good, and, for which, the business model is in a state of crisis.

In this context, an understanding of major trends in dominant media-tech, including platformisation, data surveillance flows, and broader system-wide innovation in connectivity and distribution informs the research; these developments all portend radical implications for news consumption. Algorithmic visibility of news has emerged as a core concept for those researching questions of media pluralism. News content distribution is embedded in platform power and regulatory debates in relation to how access is shaped by automated machine-learning processes. Research indicates that the way audiences become exposed to news content will increasingly be a matter of these automated, platform-based mechanisms (Wilding et al. 2018; Helberger 2018; Helberger et al. 2018; Helberger 2021).

As Wilding et al. (2018) note, in the context of a discussion regarding the importance of obtaining a diverse news diet for quality and informed citizenship, ‘Algorithmic filtering methods, however, run the risk of constraining diversity, which may cause information blindness for consumers’ (p. 58). They anticipate that a reverse situation is also possible: that algorithmic or automated news content may lead to more diverse content when that objective is ‘programmed in’ to a platform's algorithms. To develop this point a step further, we see that in either scenario, whether or not there is likely to be more or less diversity in content, which is always an important consideration arising in online news contexts, platforms are curating such provision in novel computational ways. This departs from the predominately human curation of news content that operated in the pre-digital world. Such filtering processes are, of course, the source of ongoing debates regarding their intrinsic merit or otherwise.

Bodo's discussion of algorithmic agency and platforms in News, Algorithms and Automation provides an account that places the consequences for news distribution into stark relief. His research distinguishes the use of algorithmic personalization by digital news organizations from that used by platform intermediaries. In the latter context, he argues:

Platforms upset the logics of news production, delivery and consumption in more than one way: they compete for advertising revenues with news organizations; through their control of access to audiences, they play an increasingly important role in news delivery; and through their own personalization efforts, they are able to set their own agendas, at the expense of news media.

(Bodo 2019: 1056)

This power to curate the news, then, and to indirectly ‘set their own agendas’, lies at the core of their societal influence. Virality, popularity, misinformation and disinformation together with the personalized delivery to audiences is a demonstrable shift accompanying the rise of tech-media platform intermediaries. For Bodo, these developments are linked to ‘economic, technological and sociocultural conditions of the production, distribution, and consumption of news’ and the ‘algorithmic control of information flows and the customization of the information environment around each of us constitute the latest development in that process’ (Bodo 2019: 1070). The platform logic of personalization is characterized by abundant personal usage data (and datafication more broadly), vast amounts of user and publisher content, an aggressively ad-based business model, ‘almost limitless technological and financial resources’ and a more diffuse approach to editorial or curatorial oversight (Bodo 2019: 1070). Quite rightly, too, Bodo's view is that the role of these algorithmically personalized controls is ultimately a matter for how societies wish to interpret their impact on the prevailing forms of governance.

Helberger's research points to the increasingly important role that personalized recommender systems now play in news distribution, arguing that recommender systems can be theorized in terms of four main varieties of algorithmic recommendation. First, she argues that liberal recommenders inform within limits of what people want/expect. Second, participatory recommenders map the diversity of ideas and opinions in society, responding to differences in information needs, styles and preferences. Third, deliberative recommenders nudge people to encounter different perspectives, serendipity, as well as activating people to comment, share and the like. Finally, constructivist recommenders can nudge people to encounter and acknowledge minority opinions but also support seeking out the like-minded (Helberger 2021).

While these varieties of priorities for algorithmic recommender systems are particularly relevant for the responsibilities of public service media (PSM) sites and platforms, they can be applied more broadly in relation to certain categories of commercialized personalized media delivery. Public service media have a recent history of blending machine-determined automation alongside human-driven curation to effectively operate in expanding automatic media content environments. For example, YLE, the Finnish public service media broadcaster, has long given its journalists the ability to target their content via an insights dashboard that presents them with audience performance data (Hutchinson and Sørensen 2021). The next step for public media operators is to enable a form of exposure mechanism, similar to the Helberger (2021) model, to determine where the targeted content will appear in these kinds of news content delivery mechanisms. Given the historical trajectory of PSM, it seems logical to develop the application of this kind of pathway not only for traditional public service media but more broadly to help safeguard democracy.

Just as all media content is not targeted for all audiences on social media, algorithmic media can target diverse audiences for news content. This algorithmic thinking can be described as a form of ‘hidden innovation’, exemplifying research and development that begins in public media, and which is then adopted by the broader commercial media sector (Cunningham 2013). The idea of aligning content production with audience analytics is nothing new of course – but how it is applied to satisfy aspects of exposure mechanics within automated media has emerged as the frontier of news and algorithmic studies. Helberger's recommender model (discussed in more detail below) is well suited to media produced by organizations with a legislated social cohesion purpose; in other words, public service media (2021). Indeed the concept of targeted exposure, based on algorithmic alignment, is also related to the content exposure management techniques employed by certain social media influencers. Bishop (2019) notes that this targeted approach towards specific exposure visibility is typically a strategy of content producers seeking maximum visibility within algorithmically determined environments – something news distribution is yet to fully address. However, we would make the observation that the application of this model to more traditional commercial media contexts where editorial and business priorities can innately conflict, renders some of the more interventionist varieties of recommender models a less plausible proposition.

The need to take new platform distribution processes into account in assessments of media pluralism has been recognized by EU Member states in EU Council deliberations. In the Council's discussions regarding findability and discoverability of pluralistic content, it has foreshadowed that future media policy needs to consider how algorithmic manipulation affects users’ exposure to informational content and news sources as well as their overall consumption. Significantly, though, the Council has gone a step further by recommending that Member States’ media policies should consider ‘must show’ rules and criteria for discoverability and findability to favour media pluralism and cultural diversity. The Council specifically invites the European Commission to:

continue and further develop research to identify potential risks to media pluralism and further understand the changed position of editorial media outlets in relation to social media, search engines, video-sharing platforms and other media platforms; concepts of information science should explicitly be taken into account.

(Council of the EU 2020, paragraph 25)

Although the EU is the leading development of this approach, there are indications that it could feature in policy or regulatory arrangements in other jurisdictions. In Australia, as explained further below, the recent legislation that prompted the dominant digital platforms Google and Meta to make payments to news organizations includes a requirement that the platform develop a proposal to ‘recognise original […] news content when it makes available and distributes that content’, along with an obligation to inform news organisatons of certain algorithmic changes that might have a significant effect on their referral traffic. However, as Lee and Molitorisz (2021) point out, these aspects are not in operation as the legislative scheme itself is designed to come into effect only if the two platforms make insufficient progress in their own commercial agreements with news providers; these agreements need not contain provisions dealing with algorithmic governance.

The main reason behind the call for greater transparency is that there is now a body of evidence that has shown us how social media news sharing on platforms, driven by algorithms, shapes how news stories are selected, distributed, discussed, and valued (Martin and Dwyer 2019; Flew et al. 2021). The overall impacts of digital intermediaries (social media, search, online news and aggregation platforms) on news indicate important trends in the way audiences now get access to journalistically produced news content. Transformations wrought by the rise of algorithms and artificial intelligence in online intermediaries, including in social media feeds and search and online news sites themselves mean that our exposure to news and information products is shaped by our own usage (Pasquale 2015: 79; 2017). It is not an overstatement to recognize that these digital modes of distributing and discovering news are baked into a new logic of the media–industrial complex. In other words, news as a public good has been recast by private interests in platform capitalism (Srnicek 2017). In turn, this gives heightened meaning to the resurgence of interest in news as a public good – evident for example, in the work of the Australian Competition and Consumer Commission (ACCC 2019) and the OECD (2021) – and emphasizes the importance for research on journalism and critical algorithmic studies to engage with rapidly evolving policy.

Algorithmic curation of news articles in tech-media spaces now relies on the routines and logics of automated plaformization, shaping how news content becomes visible for audiences. The influence of platformization on journalism, including the use of new metrics and performance priorities, has emerged as a key part of the online news ecosystem, shifting the foundations of how audiences will discover sources of news and information. These processes are the terrain of this volume.

What is at stake for online news audiences?

The subtitle for our book, The Consequences of Automated Curation for Society, focuses on what we understand to be the high stakes arising from the unprecedented rise of tech-media platforms and news distribution shaped by algorithmic manipulation.

In 2020, the University of Canberra's Digital News Report found that around 52 per cent of Australians get news via social media, and the number is growing (Park et al. 2020: 50). Facebook also boasts of its investments in news via deals with publishers and new products such as Facebook News. The Reuters Institute for the Study of Journalism in a five-nation survey found that the primary source of news ranged from 31 per cent of people in Germany using social media as their main source of news to 45 per cent in the United States to 66 per cent in Brazil (Newman et al. 2019 in Flew et al. 2021).

The consequences of being beholden to the algorithmic provision of news by platforms was made shockingly clear the day Facebook decided to shut down the visibility of news and information sites on the platform for 17 million Australians in early 2021. The move was widely anticipated in the context of the Australian Government's decision to legislate for a mandatory bargaining code for news, in response to recommendations by the competition regulator, the ACCC (2019) in its Digital Platforms Inquiry (DPI) report, two years previously.

In a move that attracted a great deal of international attention and indeed scrutiny, the Australian government enacted a law in early 2021 that puts in place a mandatory code to help support the sustainability of the Australian news media sector. The key rationale for the new law is to address bargaining power imbalances between digital platforms and Australian news businesses. One of the main recommendations of the ACCC's (2019) DPI was to establish a ‘bargaining’ code (which was initially to be voluntary, until the government decided that to have any impact it needed to be mandatory). The code provides for a system of ‘final offer’ arbitration between news media businesses registered with the Australian Communications and Media Authority (ACMA) and designated ‘responsible digital platform corporations’. The term ‘designated’ has emerged as a very ‘over-determined’ one; to be designated requires a failure on the part of certain platforms to enter into sufficient commercial arrangements with news content businesses.

However, platform corporations have not been put onto the designated list and Meta and Google have already entered into contracted commercial arrangements with news media businesses that former ACCC Chair Rod Sims has estimated exceed AUD 200 million annually (Butler 2021). This pre-emptive move was widely interpreted as an attempt to avoid being caught in a regulatory net that would not be on the most favourable terms for the platforms (Dwyer 2021).

The Mandatory News Media Bargaining Code has a requirement to provide registered news businesses with advance notification of planned changes to an algorithm that will significantly affect ‘covered’ news, to provide information about the collection and availability of user data; and to develop a proposal to recognize original news (Senate 2021). This is intended to capture algorithm changes which involve an active decision to modify how content will be distributed on a designated digital platform service. The Code provisions are not intended to apply in situations where the algorithmic changes are mostly automated or a result of machine learning. The policy intent is that changes are likely to be considered significant if they are likely to result in an approximately 20 per cent or greater change in referral traffic to registered news businesses as a whole (Senate 2021: 1.126). The key point here is that the enforcement processes, including very significant fines, will not be available because of the lack of designation under the legislation. Against this background, the decision of Meta (then known as Facebook) to shut down news in the days before the legislation was passed to give effect to this scheme was widely interpreted as a form of blackmail and an attempt to prevent the government from proceeding with the new law.

In Australia, the ABC's online headline was: ‘Facebook just restricted access to news in Australia: Here's what that means for you’ (ABC Online 2021). The impact was well reported around the world. A sample of international news headlines caught the key aspects of story:

‘Australia's Scott Morrison slams Facebook's move to block news as arrogant.’

(South China Morning Post)

‘Facebook ban on news in Australia provokes fierce backlash.’

(Financial Times)

‘Australia Reacts to a Facebook Without News.’

(The New York Times)

‘Facebook's Australia news ban: what is the social media giant up to and how will you be affected?’

(The Guardian)

When Australian users (including those living abroad) attempted to share a news story on Facebook they received the following pop-up message: ‘This post can't be shared: In response to Australia's proposed new Media Bargaining law, Facebook will restrict publishers and people in Australia from sharing or viewing Australian and international news content.’

Richard Glover, a popular Australian radio presenter on the public service broadcaster, writing in the Washington Post posed the question:

People and governments everywhere have wondered whether the world has allowed American tech giants to become too powerful and too central to the way our societies work. What would happen if they suddenly chose to use that enormous power to win an argument with a democratically elected government?

(Glover 2021: n.pag.)

Glover was not alone in thinking through the stark consequences of these events in terms of Facebook as a news source: ‘The problem is that many Australians have developed the habit of using Facebook as their access point to news and information.’ Facebook pages run by Government departments responsible for fire and emergency services, hospitals, charities, domestic violence support groups all went dark.

Publishers’ Facebook pages were replaced with an information page with contact details only. Net data analytics group Chartbeat noted that from 5.30 a.m. on the morning of 18 February 2021 (the morning after), the traffic sent to Australian and international publishers fell off a cliff, going down to a trickle except for a few anomalous publisher sites that were still able to link to their news on the social media platform (Benton 2021a).

This was not the first time that that Meta had threatened to metaphorically ‘shoot the hostages’ to achieve its own political or comercial ends (Benton 2021b). Another occasion when Meta used its market clout was in 2017 when it undertook live experiments to the newsfeed by introducing a new ‘Explore Feed’ product in Slovakia, Sri Lanka, Serbia, Bolivia, Guatemala and Cambodia, moving news stories to the new feed. It meant that users needed to click on the new product to access their favourite news sites, and it resulted in the biggest drop in organic reach seen to that point in time (Struharik 2017).

A further example is when Facebook shut down completely in August 2018 for 45 minutes, providing a glimpse into the world without Facebook news (Whittaker et al. 2018). The result confirmed that without a newsfeed, users go straight to publishers’ mobile apps and news sites. Chartbeat data indicates that built in mobile aggregators emerged as important referrers, driving ‘significant traffic to publishers’ (Schwartz 2018: n.pag.). These aggregators include Apple News, Google Chrome Suggestions, news aggregator Flipboard and the Google News app which is pre-installed on Android devices. Publisher owned apps also gain when the Facebook news feed is not available and these went up by as much as 40 per cent in 2017 during the ‘Explore Feed’ incident (Struharik 2017). But perhaps the most heartening message for news publishers in the 2018 dark mode event was that news users seek out alternatives and can change their practices in a matter of seconds.

Similarly, platforms have changed content curation algorithms without warning, reducing traffic to news websites, as occurred with Facebook in 2017 and 2018 (Flew et al. 2021). There have also been very well-documented cases of audience manipulation by Facebook (Martin and Dwyer 2019). In early 2018, Facebook with no warning decided to flick the algorithmic switch to prioritize posts made by friends and family over public content. In this action, Facebook was actively deprioritizing professionally made news journalism, and in effect allowing a lesser quality information to dominate news feeds. The other less well-publicised instance of algorithmic change was when in March 2018, Facebook tweaked the algorithm to favour local news. They first rolled out the changes in the US market, claiming that this was to ensure that readers were being exposed to content that was more meaningful in their community contexts.

Such instances are all clear evidence of the power of the platform to control the delivery of news content. The actions of the platforms to curate content and to shape what becomes more, less or even invisible demonstrate their power. However, some research is beginning to emerge indicating that the impact of these kinds of algorithmic changes on the businesses models of different categories of news institutions (digital native, print, PSM, social news and television) in terms of referral traffic, can be quite variable. For example, Bailo et al. note that referral traffic to public service media organizations in Australia has proven to be quite resilient to the 2018 prioritization of ‘family and friends’ changes in Facebook's algorithms, whereas social news was detrimentally impacted (Bailo et al. 2021).

Online news and platform power

For Helberger, the ability of platforms to make algorithmic changes of this kind is an example of their communicative power and indeed the main problem with solving questions of diminishing structural diversity in the media. She makes the point that platforms are not concerned with the provision of diverse forms of news and information; instead, they seek to control exposure to available information. Platforms, she argues, ‘stage encounters with media content, affect the findability of content, order and prioritise existing content, manage and direct user attention as a scarce resource, and influence the choices consumers make’ (2018: 162). Therefore, platforms’ roles in curating opinion power should not be underestimated. In their recent work colleagues at the University of Amsterdam note that their research, ‘aims not only at better understanding shifts of opinion power, but also at developing ways and informing policy debates and regulatory thinking on how to deal with concentrated power in the media’ (Seipp et al. 2023: 2).

Schlesinger locates power in relation to a ‘regulatory field’, adapting Bourdieu's formulation, and suggests that the ‘upsurge of British activism’ by relevant agencies is related to the way ‘they exercise power over cultural production, circulation and consumption’ (Schlesinger 2020). He maps a litany of what are, at their core, content concerns on platforms and which herald significant consequences for society: ‘fake news, exposure to harmful or illegal content, anti-competitive behaviour, misleading political advertising, the uses of consumer data, expressions of violence and terrorism, online indecency and interference by foreign governments in the domestic electoral process’ (Schlesinger 2020).

Underlying many of these issues, the development of algorithmic patterns in online news distribution and consumption through platformization dynamics has inevitably given rise to a call for a renovated set of questions and re-framings for policymakers in nation states (Wilding et al. 2018; Flew and Wilding 2020: 6). In effect, these changed dynamics require new orthodoxies for expanding conceptions of voice pluralism based largely on media ownership; they require rethinking and adapting the long-standing policy objectives for media plurality. Fenton et al. (2020) argue that in the context of increasing media concentration, there is a need for counter narratives and mechanisms ‘to promote meaningful plurality’ (2020: 104). Therefore, these authors argue that the performance of news algorithms should be scrutinized so that they do not ‘unduly favour particular types of news providers and voices over others’ (2020: 105). We agree that this is the reality of online news distribution over intermediary platforms where the algorithms have a direct role in the selection of news articles surfacing on platforms, at both aggregate and individual levels. However, in this context, we would distinguish ‘meaningful plurality’ from the conceptualisation of ‘sufficient plurality’ – critiqued by Gibbons (2015: 22–27) among others – used in UK law. The latter is concerned with the hard political decision limits or thresholds in law and regulation which impact on the number and ‘sufficiency’ of voices, while the former focuses on expanding formulations that apply in new media ecologies. As Valcke et al. (2016: 2) have observed, ‘the concentration of where the audience goes – in terms of aggregators and sites – is every bit as damaging to pluralism as limitations on spectrum and concentration of ownership’.

The preparedness to completely turn off the availability of news in these ecologies represents a phase-shift in platform power. Writing in The Conversation in relation to the social value of news distributed by platforms, Dwyer noted that executives from Google and Meta told an Australian Senate committee that they were prepared to take drastic action if Australia's news media bargaining code, which, as noted above, would force the internet giants to pay news publishers for linking to their sites, came into force (Dwyer 2021).

People realize news is important – that it shapes their interactions with the world and provides meaning and helps them navigate their lives. So the threat to withdraw services from the Australian media landscape was all about wanting to convey a sense of power to Australian lawmakers. At the same time, the rhetorical flourish from the platform behemoths was also about how news meant little to them in financial terms, a claim that was at odds with research indicating that 52 per cent of Australians were currently getting their news via social media platforms (although, news access on social media varies with age and platform usage) (Park et al. 2020, 2021).

Australia's Treasurer, Josh Frydenberg described Meta's action to turn off news in people's Facebook news feeds as ‘heavy handed’ and ‘confirming the immense market power of these digital giants’ (Snape 2021). Predictably, Meta argued that poorly drafted legislation was to blame. When Meta's Simon Milner appeared to give evidence at a public hearing of the Senate Economic Legislation Committee, inquiring into the Treasury Laws Amendment (News Media and Digital Platforms Mandatory Bargaining Code) Bill, he was questioned about the intentions of Meta in sending Facebook news to black in Australia for a week. The senior spin doctor was questioned in relation to the social responsibility of ‘turning off’ a swathe of government and community health information sites that were caught up in the Facebook dragnet (Taylor 2021). In part, Milner's justification for the social media platform's action was that the definition of news in the new law, as argued in the corporation's written submission, was ‘so broad and vague that they will inevitably result in differences in interpretation and result in large, overreaching ambit claims from news publishers’ (Facebook 2021: 27). The argument was made that ‘core news’ and ‘covered news’ – the legislative categories of news that help establish who has access to regulated bargaining, mediation and arbitration over remuneration for news content – could include any content that ‘reports, investigates or explains events of public significance at a local level’. The submission also questions the use of the phrase ‘of interest to Australians’ in the new law. This lack of specificity in the definition of news, Meta argued in its submission, would result in ‘overreaching ambit claims for news publishers’ (Facebook 2021: 27).

As we explain in Chapter 2, this issue of the definition of ‘news’ was one that researchers in the Media Pluralism Project were frequently reminded of during the many hours spent in classifying news stories scraped from online brands. Was a particular story news – even though it could also be classified as ‘sport’, or ‘health information’, or ‘celebrity’? We noted that the varieties of news would often depend on where one's gaze was directed: online news sites, blogs, entertainment news sites all push out different contents that could be interpreted as examples of a range of media forms, genres and topics.

Meta's decision to block access to Facebook news for Australians had quite an adverse impact on health (including vaccination and COVID-19) information, weather and other unexpected community pages hosted on the social media platform. The ‘collateral damage’ sites affected by the decision included the mental health information line, 1800 Respect, the WA Department of Fire and Emergency Services, the Hobart Women's Shelter and the Bureau of Meteorology, among many others (Smyth et al. 2021; Snape 2021).

Joel Nothman, the lead data scientist on our project, suggested that there would have been some weeks of data analytics work in creating the blacklist of sites that Meta prevented from surfacing in Australians’ newsfeeds. It is possible to surmise, then, that the humans in that curation process made some bizarre and yet also quite telling decisions about what constitutes a source of ‘news’ from Meta's perspective.

Despite these concerns over news content definitions, the Treasury Laws Amendment (News Media and Digital Platforms Mandatory Bargaining Code) Act 2021 was passed by parliament on 25 February 2021 and commenced on 2 March 2021. The efficacy