19.1 Introduction

There is a lot happening in platform governance, with multiple legal measures and proposals in many jurisdictions. European Union examples include the Digital Services Act (DSA), the Digital Markets Act, the Strengthened Code of Practice on Disinformation, the Artificial Intelligence Act, and the European Media Freedom Act (EMFA).Footnote ¹ While these measures offer much to analyse and critique,Footnote ² it is worth considering some broader points related to public communication: namely, ways in which infrastructures underlying public communication have changed; how that changes journalism’s role within public discourse; and the regulation of opinion power. While EU developments partially address these changes, it remains unclear how to pursue goals for the sustained plural public speech that has long been sought for democratic governance. A particular challenge is sustaining what journalism did (or was imagined to do) in mass media contexts for public debate in general.

Journalism’s role in opinion formation has reduced, and the role of platforms has grown.Footnote ³ Major platforms ‘shape public communication and distribute attention’ through their ‘architectures, algorithmic amplification, and content moderation’,Footnote ⁴ and they do so with values, scale, and speed quite different from past public speech. Public communication is being ‘privatised according to the logic of an attention-based business model’,Footnote ⁵ and platforms are affecting opinion power in ways that do not appear to support democratic governance.

My focus here links to a long interest in freedom of expression and its relationship with public communication and opinion formation. It is an interest in pluralism, which draws from but also goes beyond historic legal examples. I outline my approach to pluralism briefly in Section 19.2. Section 19.3 sketches certain changes in the infrastructure underlying public speech, particularly the role of journalism within it. Then Section 19.4 considers how a journalistic-editorial role was presumed to exist, and was relied on, in ideas of democratic public speech. Sections 19.5 and 19.6 examine points from the literature about how platforms might serve a speech-related function that would resemble the past role of journalism – even if that past role was idealised. Several challenges are considered, particularly in relation to news recommender systems. Section 19.7 considers coding journalistic norms, which have traditionally focused more on items of speech – a newspaper report, a current affairs segment, and so forth – into automated processes that are applied at huge scale and speed. The challenge is at the least highly demanding. The coding of such norms, if achievable at all, entails a transformation in public speech and a challenge to existing understandings of democratic opinion formation. And the same is true if journalistic norms remain uncoded or unencodable. Public speech is changing.

Much of what is discussed in this chapter could be extended by closely examining recent European developments such as the DSA and the EMFA. The DSA, for example, considers systemic threats to fundamental rights including freedom of expression and media pluralism,Footnote ⁶ while the EMFA contains numerous provisions that have some resemblance to the requirements for democratic free expression and public speech that I will discuss in what follows.Footnote ⁷ These include provisions related to editorial independence, ownership transparency, the adequate funding and independence of public media, ownership concentration, and states’ obligations – including positive obligations – to ‘ensure framework conditions are in place’ to safeguard audience rights to plural, editorially independent content ‘to the benefit of free and democratic discourse’.Footnote ⁸ However, the EMFA is centred on institutional media rather than the wider infrastructure that underlies public speech. There are complex questions about the combined effects of the EU developments on that wider infrastructure, along with many uncertainties about how the multiple measures relevant to democratic freedom of expression will operate.Footnote ⁹ While such issues warrant future attention, my interest here is more conceptual than technical. In part, this is because I suspect that a conceptual analysis may have continuing value as public communication continues to evolve. While the European Acts may well be ‘the beginning of a new era’ of EU regulation,Footnote ¹⁰ my focus here is on what is required for communicatively legitimate democracy in the changed era of communication.

19.2 Sustained Plural Public Speech

I have previously examined aspects of law’s support for pluralism and its links with free expression, in particular links with the freedom’s positive dimensions.Footnote ¹¹ My analysis has mainly concerned mass media contexts, but it need not be limited to that. I labelled what is wanted in an idealised, communicatively legitimate democracy as ‘sustained plural public speech’. It is a pluralism ‘plus’, which extends and modifies common discussions of media pluralism. In shorthand form, it involves media with

• structural diversity of organisations, people, funding, and content; with

• independence from domination by political and economic interests;Footnote ¹² and with

• transparency about ownership, control, and content.

This sort of sustained plural public speech is only ever reached in part, but elements are seen in legal interpretations of expressive freedom in various jurisdictions, including German and French constitutional law and the European Court of Human Rights.Footnote ¹³ They also emerge from decades of political theory, media studies, and legal writing.Footnote ¹⁴ While limited examples exist in law, the rationales underpinning them suggest that there is more to the freedom and that law should go further to support democratic freedom of expression. Structural diversity with real independence and transparency are important parts of what free expression in the service of democracy needs. While representative democracy has different inflections – liberal, republican, agonistic, and so forth – all current forms should have this diversity, independence, and transparency in the infrastructures underlying public communication. An explicitly elite-dominated democracy would be different, but I leave that aside. For present purposes, my interest is that the public speech imagined under this model has been sustained largely by journalism in the past, a point to which I return.

Positive dimensions of communicative freedom, as well as negative ones, are significant if democracy is to be a ‘democracy of expression’, where society has a voice, formulates judgements, and makes effective claims of government.Footnote ¹⁵ This requires ‘equal, substantive communicative freedom’.Footnote ¹⁶ While it has existed as an idealised model more than in substance, what is being sought – sustained plural public speech – remains a democratic goal for public communication. And it can be approached to a greater or lesser extent using freedom of expression principles. That is, the goal is necessary under democratic constitutions, but quite what role free expression plays in it can vary. Freedom of expression can be understood to require some state actions supporting the structural diversity, independence, and transparency noted here; it can be understood to allow such actions, or to prevent substantially such actions – for example, consider some interpretations of the US First Amendment.Footnote ¹⁷ My analysis tends towards the first approach, while work such as Ed Baker’s offers strong arguments for the second, calling for sympathetic judicial assessment of legislation aimed at supporting plural public speech.Footnote ¹⁸ I agree that courts should recognise that such legislation can support expressive freedom, but I suggest that in some instances they should do more to protect the plural public speech that a democratic constitution requires. In any event, questions about which constitutional actors do what to support expressive freedom are separate from questions of what communicative goals should be pursued in the first place.

There are two important points here. First, the two approaches noted here that require or allow state action to support structural diversity, independence, and transparency pursue the same result – something approaching sustained plural public speech – and call for constitutional actors of executives, parliaments, and courts to use a fuller understanding of expressive freedom than one focused mostly on its negative dimensions. Second, both approaches differ from the third one noted here – which substantially prevents such state action. In that approach, freedom of expression is understood as (almost) only a negative freedom against state restrictions on speech. Negative dimensions of this freedom are significant, but free expression also has positive, affirmative, active dimensions (many terms are used in the literature). These dimensions are significant for pluralism, among other things. If sustained plural public speech is a democratic imperative, being concerned only with expressive freedom’s negative dimensions means the freedom actually threatens democratic governance – it becomes a freedom that is rightly concerned with state domination of speech but leaves aside economic domination and the wider requirements of sustained plural public speech.

The first two approaches noted here involve audience interests in communicative freedom, as well as the commonly emphasised interests of speakers. For the audience, the issues are similar whether one concentrates on media freedom or individual freedom of expression. In relation to a freedom of reception,Footnote ¹⁹ there is a common interest in how the freedom supports democratic opinion formation, which links to pluralism. This means that analyses of democratic opinion formation can examine legacy media’s reception by audiences and the role of platforms as they relate to public speech. Platforms have a function that in some ways resembles mass media’s traditional role and, for audience interests in speech and opinion formation, media freedom has large overlaps with free expression. Media freedom is not only about the media as speakers; it is not only about the media’s institutional freedom. It is more than that; it includes a curated pluralism in the audience’s interests. This pluralism would not only support democratic opinion formation – and democracy is a useful focus here – but also could extend beyond news and commentary to encompass matters such as culture and identity.Footnote ²⁰

Overall, democracy calls for sustained plural public speech with structural diversity, independence, and transparency, all aimed at supporting individual and collective opinion formation. I will add to that description later, but it suffices for now. The rationales for democratic free speech go further than many past analyses, and so should law for two related reasons. The first reason to go further concerns the meaning of ‘freedom’ in free expression – if the freedom is to be real or effective, it does not only involve the absence of restrictions. The need for real or effective rights is, for example, a central aspect of the European Court of Human Rights’ approach to positive freedoms.Footnote ²¹ While some readers may think the analysis here places too much within free speech and demands too much of it, I would suggest that negative approaches to freedom of expression place too little. They assume that freedom exists where speech is not restricted. The second reason to go further concerns how legal responses to the effects of platforms on public speech are limited if freedom of expression is only a negative freedom against state restriction. Under European law, the freedom is not only that.Footnote ²²

19.3 A Shift in Journalism

My interest in this chapter concerns the changed role of journalism in supporting opinion formation and the growing position of platforms in shaping public communication. Let me sketch what I mean, focusing on news media for simplicity. In the past, something like the following existed, or was at least presumed to exist.

• A media infrastructure that was largely national and local but with significant transnational content flows, which was comprised of varied organisations, missions, people, and funding.

• Journalism or a journalistic-editorial function for news and commentary in each organisation, with relative independence from political and economic domination and meaningful influence over content.

• Distribution and monetisation that were linked, with the flow of money largely mirroring the distribution of printed, audio, or audiovisual content. This aspect of public communication often disappeared into the background – as infrastructures tend to do – and was not a primary focus of analysis or debate.Footnote ²³

• Audiences that were understood to have a civic disposition, to varying degrees, who accessed and used content to form opinions on varied matters.

Legal support for pluralism came through measures such as regulating mass media structure via ownership limits; regulating for editorial independence; offering direct and indirect state aid to media organisations, especially print; and creating public service media. Such measures are part of many jurisdictions’ traditions of dealing with opinion power, even if challenges to their realisation emerge repeatedly.Footnote ²⁴ While the term ‘opinion power’ appears commonly in German law and scholarship (and analyses influenced by that approach),Footnote ²⁵ the ideas have wider currency.

In addition to laws affecting media structure, many jurisdictions had content-focused laws imposing obligations on broadcasters about local content, children’s programming, election broadcasts, and election advertising. There was comparatively little law about distribution, but issues arose with changes in media infrastructure, such as ‘must carry’ provisions for cable broadcasting or channel placement in electronic programme guides. There were also laws restricting content; some, such as those on defamation and copyright, significantly affecting public speech. Content-related laws often made allowance for the public and democratic role of media through defences or exceptions, or through the ways that general aspects of the law were applied to media workers and similar actors.

Changes from analogue printing to data-driven platforms underline the prescience of Ithiel de Sola Pool’s forty-year-old observation about free speech and communications technology: ‘In ways that cannot yet be precisely identified, the bottleneck for effective [content] monitoring and charging is migrating from reproduction to the continuing service function’; that function is centrally important in ‘the electronic era’.Footnote ²⁶ As those words hinted, intermediaries now affect all the elements of media structure, journalistic-editorial function, distribution and monetisation, and audience actions and dispositions. Intermediaries are a ‘challenge to the positive media order’Footnote ²⁷ that has been imagined and, in part, created in some longstanding democracies.

One notable area of change is distribution and monetisation. News journalism has become partially separated from its distribution and funding in a way in which it was not in the past. Then, concerns were often about keeping a separation between editorial and commercial (or ratings) aspects of media organisations. Those concerns remain but, in addition, media organisations producing news and commentary now do not control many aspects of content distribution as they did before platforms, and the media face substantial challenges to their traditional funding models. This is true for commercial media, with a huge percentage of advertising revenues having moved to platforms and related intermediaries, and for publicly funded media, with increasing pressures on public funding in many longstanding democracies. Both aspects are relevant to media independence, to ‘the ability to determine how news is produced, published, and distributed to the public’.Footnote ²⁸ That is, these changes affect the shape of public communication and the distribution of attention. There is still a configuration of media structure, journalistic-editorial function, distribution, and audience that more or less resembles past formations, even if some of the differences are marked. But platforms, and ways in which media companies are dependent on them, call into question what could be called a presumption of journalism that has long underlaid expressive freedom and debates about it.

19.4 A Presumption of Journalism

Identifying a presumption of journalism is not to idealise some perfect form of journalism. In some contexts, news and commentary have provided real public benefit, and journalism has significantly shaped publics and political processes. There have also been clear weaknesses in journalism practice. The presumption of journalism does not involve some perfect form but refers to a model of content creation, curation, and distribution that, however imperfectly, aligns with public values – and, significantly for my purposes here, it is a model that has been presumed in many analyses of democratic free expression. The model has applied, in theory, across all the varieties of journalism. And there was a wider structure of public communication that allowed the model to be imagined, even if not to be fully realised.

Opinion power is sometimes analysed through a particular element of law or policy, such as media concentration law,Footnote ²⁹ or through the creation, funding, and independence of public service media. Such elements are clearly relevant, but they are insufficient on their own for understanding opinion power. Many analyses have also assumed that a journalistic media underlies news and commentary and substantially shapes public communication. The assumption has been that a journalistic field exists – of the kinds understood to exist during the latter twentieth century in established democracies – and that journalistic-editorial decisions substantially (though not totally) influence content. And there have been important assumptions about audience attitudes and disposition, which are also challenged by platform-distributed speech.Footnote ³⁰ The practices of journalism are often criticised – as being beholden to class interests and elite voices, ridden with discriminations based on race, gender, and more, or dominated by commercial and state powers – but the idea of journalism, of an editorial media, has persisted. Journalism has been assumed in the infrastructure of public debate and, to the degree it has existed, it has helped to make measures such as media concentration laws and public service media useful in controlling opinion power.

I have said that democracy calls for sustained plural public speech with structural diversity, independence, and transparency, to support individual and collective opinion formation. Something can now be added to that description concerning the assumption of journalism. The structural diversity, independence, and transparency that were supported (or should have been supported) by law was understood to result in the distribution of diverse media content that had been substantially shaped by journalistic-editorial decisions. Public communication was, to a large degree, journalistically curated speech. Media content was assumed to be plural across many dimensions and presumed to be received by audiences, partly due to the limited range of content that was then available. Source diversity was thought to lead to exposure diversity. Journalists, editors, and producers continue to exist, even if in many places they are fewer in number and have less funding. Journalistic-editorial decisions continue to shape media content, but there is an overlay of other actors that influence those decisions: intermediaries such as social media and other platforms. Media have far less control of their distribution to significant parts of their audiences, and commercial media are less able to monetise content through advertising.

If you accept that communicatively legitimate democracy needs something like sustained plural public speech – even if pursuing it has challenges – what about now? Public speech was, to a large degree, sustained by journalism. A journalistic-editorial role was assumed to exist and did exist in part. Debates about freedom of expression happened around the base of a journalistic-editorial role. And that base’s presence allowed some analyses to imply that expressive freedom only involved the freedom’s negative dimensions.Footnote ³¹ I would suggest that recent changes make those analyses even less tenable. Media were gatekeepers and framers of issues, focusing attention on particular matters in particular ways. They were doing all that quite imperfectly, and more of the structurally diverse, independent, and transparent media environment mentioned here was needed. Even so, the journalistic-editorial role was important in how public spheres were imagined and how they existed, to the extent and in the styles they did. Journalism was relied on for democratic public speech.

This role still exists, but it has changed along with the context in which it operates. More actors are relevant, and the once relatively independent position of journalists and editors is being constrained. Media organisations are dependent on platforms in multiple ways.Footnote ³² Publishing is becoming ‘platformed’ and ‘some news organisations have almost no control over the distribution of their journalism because they publish primarily to platforms defined by coding technologies, business models, and cultural conventions over which they have little influence’.Footnote ³³ In order to gain audience attention through platforms, media organisations produce somewhat different items, covering different issues with different angles. And journalists and editors are less autonomous than they were. They were always subject to pressures, but now different actors are influencing the content produced, distributed, and received by audiences.

It is not new that a social institution shapes public communication and focuses attention. That role was substantially taken in the past by media, which relieved various other instruments – for present purposes, much media law and policy – of doing more in relation to content. Law shaped media structures and imposed certain positive obligations on content (generally on only some types of organisation or media sectors, or in only some jurisdictions) and certain limitations on expression. Law then left journalism to act relatively autonomously within the broad aim of having an environment of sustained plural public speech. Democratic states were understood in some jurisdictions to have positive obligations to protect pluralism within that infrastructure of public speech. Their obligations should have been understood that way more generally. This is because media had an ‘orienting effect’ on public debate,Footnote ³⁴ an effect that is now joined by, and partially subsumed within, platforms’ focusing of public attention. But the orienting effect of platforms occurs through curation that is subject to very different values from past forms and it operates in very different ways.

19.5 Sustained by Platforms?

Platforms’ effects mean that it is increasingly implausible to assume a relatively independent journalistic-editorial role that has a highly influential effect on public communication. Now platforms have a substantial role in focusing attention on content through amplification and moderation. Media organisations have long done something similar: ‘Whether a media institution amplifies certain messages or sources’ is editorial,Footnote ³⁵ and platforms have some editorial-like roles that are not insignificant. However – and it is an important however – the distribution of content via platforms is data-driven for commercial ends, using data from and about users to generate engagement and more data, in ways that differ fundamentally from the model of institutional media.

Legal reforms such as those at EU level, noted at this chapter’s start, include efforts to support a journalistic-editorial role. That is understandable but also problematic because of differences in platform infrastructures of public speech. And it underlines the value of measures to support institutional media’s pluralism, reach, and independence. It might be thought that platform distribution could happen in some ‘neutral’ fashion, so that content and reach remained media-driven. But social media and search necessarily involve moderation, they necessarily amplify and relegate content – they are not neutral. In addition, more than media content is involved in platform-distributed speech, and platforms treat content systemically more than as individual items. Another possibility could involve creating a journalistic-editorial role within platform distribution, which would be challenging in technical, and perhaps legal, terms. It would fare best legally where there is understanding of freedom of expression’s positive and negative dimensions, and of the freedom’s relevance for audiences as well as speakers. The idea might be pursued by democratic recommender systems or public-media-style intermediaries.Footnote ³⁶ In Martin Moore’s words: ‘Democratic governments need to recognize that dominant technology platforms will never adequately perform certain functions of the public sphere – most notably during election campaigns – and create alternative public service digital public spheres.’Footnote ³⁷ However, states are limited in creating counterpowers to dominant platforms because states are also dependent on platforms for public communication.Footnote ³⁸ In other ways, governments co-opt platform power to their own ends, ‘view[ing] those firms as potential deputies for their own exercises of power’.Footnote ³⁹

As noted in the introduction, the EMFA responds in part to platform effects. Article 18 addresses some aspects of how very large online platforms (VLOPs) treat the content generated by media service providers, seeking to create ‘a privileged status for traditional media’.Footnote ⁴⁰ In summary, the Article provides special protections for media service provider content when a VLOP intends ‘to suspend’ the provision of its service for the content or ‘restrict the visibility’ of the content ‘on the grounds that such content is incompatible with its terms and conditions’.Footnote ⁴¹ Before acting, the VLOP must provide reasons to the media service provider and normally allow twenty-four hours for a reply. Where media service providers believe there have been repeated and unjustified restrictions or suspensions, Article 18(6) provides for ‘meaningful and effective dialogue’ conducted ‘in good faith with a view to finding an amicable solution, within a reasonable timeframe’ to end the restrictions or suspensions and avoid future ones. Actions taken in compliance with DSA Articles 34 and 35 about systemic risks are exempt from EMFA Article 18.Footnote ⁴² These provisions illustrate the flexibility or discretion that the EMFA appears to give to platforms.Footnote ⁴³ Among other things, there are questions around what amounts to suspension or restriction due to incompatibility with terms of service, compared with general moderation practices,Footnote ⁴⁴ as well as doubts about quite what the provisions for dialogue will produce in practice. These provisions illustrate the early commentary, which finds that at least parts of the EMFA may be best understood as containing ‘important and laudable policy goals’ with unclear enforceability: ‘many rules are formulated with wide margins for manoeuvre’ and ‘supervisory and law enforcement instruments are only provided … to a very limited extent’.Footnote ⁴⁵ Overall, this is an attempt to protect journalistic content and its distribution, rather than an attempt to think about moderation as an editorial-like activity with editorial-like effects on public speech.

In some analyses, journalism or editorial ideas about moderation are explicit. For example, a twenty-two-author multidisciplinary academic manifesto in support of diversity in news recommendation argues that because news recommenders extend traditional editorial roles, recommenders should promote public values and have a suitable governance framework.Footnote ⁴⁶ Other research examines news recommenders’ democratic roleFootnote ⁴⁷ – which I read here as their editorial role – or notes how platforms are moving closer to exercising ‘editorial selection’.Footnote ⁴⁸ These ideas are implicit in other proposals,Footnote ⁴⁹ with some focusing on how platforms affect the ability of journalists and editors to exercise editorial judgement as they otherwise would.Footnote ⁵⁰ In short, news recommenders driven by public values would mirror something that, theoretically at least, was previously done by journalism. While some such measures appear possible under current European initiatives, there are real doubts about whether they will emerge, even if they are technically achievable.Footnote ⁵¹

19.6 News Recommenders

News recommenders can be used at two broad levels at least: by media organisations and by platforms. My interest here is in the latter. Or more precisely, it is in the fact that platforms’ automated decision-making is inevitably a form of news (and other content) recommendation – one based on data-driven engagement, which unsurprisingly promotes targeted divisive and socially harmful content. This raises questions about what democratic free expression suggests should be done about platforms’ role in public speech.

A two-part analysis of how to ‘democratise algorithmic news recommenders’ offers some suggestions.Footnote ⁵² Jaron Harambam, Natali Helberger, and Joris van Hoboken consider news recommenders at the media organisation level, aiming to provide choice to individual audience members. As their study notes, wider aspects of recommendation could also be considered; I will come to that soon. The basic idea is that media organisations could deploy multiple ‘algorithmic recommender personae’ with users being able to choose between them and change their choice as they wished. Normative analysis suggested five personae, each name indicating the prioritised content: ‘the Explorer (news from unexplored territory), the Diplomat (news from the ideological other side), the Wizard (surprising news), the Moral Vacationer (guilty pleasures), or the Expert (specialized news based on previous consumption)’.Footnote ⁵³ These personae would ‘offer users an intuitive, one-click option’ to choose different recommendations ‘based on their specific mood, interest and purpose’.Footnote ⁵⁴ In short, users could choose an editorial style.

The example seeks a journalistic-editorial function in algorithmic recommendation. Parallels could be drawn with traditional media outlets, each having different editorial lines focusing attention on different issues or presenting different angles on the same issue. Parallels also arise with different sections of one media outlet – think of weekend supplements or lifestyle pages that commonly remain in online newspapers.

A later experimental project extended the study, developing and evaluating eight personae with a small group of participants acting as illustrative audience members. This led to the eight personae being merged into three – Expert, Challenger, and Unwinder – in order to correspond with audience motivations. However, participants also thought some general news should be offered independently of their own choice of recommender: ‘Some news is simply in the “general interest” and should always be shown.’Footnote ⁵⁵ Thus users would be able to choose one of three editorial styles for their recommendations, with each containing the same general interest news along with its own particular content.

That point mirrors one of seven ‘wish list’ items for democratic platform regulation proposed by Paul Nemitz and Matthias Pfeffer. It would require intermediaries to provide a non-personalised, rather than unmoderated, uniform selection of news:

This suggests an approach that partially mirrors past mass media distribution. Could more be made of these ideas to develop and use editorial personae in distributing news and commentary? The two-part analysis outlined here indicates that personae could be used by media organisations, and that such options would appeal to audiences. But the idea is also interesting at the intermediary level, as Nemitz and Pfeffer’s comments suggest.

In theory, platforms could use one or more recommenders from a range of recommender systems. The systems would have qualities drawn from the analysis and wish list presented here, and they would be validated by bodies independent of political and commercial control, perhaps along the lines of self-regulatory bodies or independent media authorities suggested by Nemitz and Pfeffer.Footnote ⁵⁷ The democratic significance of public speech suggests that platforms be obliged to do this, but that would raise a host of challenges that may make the idea infeasible. Some challenges for platform regulation are widely recognised. These include the relative lack of knowledge and power held by states and the EU compared with platforms; concerns about reinforcing platform power through regulating their role in communication systems; and questions about how far measures would actually limit harms to democratic communication from platform business models. Being free of political and economic domination are widely recognised as central in models of democratic communication, so leaving the business model largely untouched appears particularly problematic for recommender systems’ democratic effectiveness.Footnote ⁵⁸ As Graeme Turner has noted more generally of digital media, ‘Monetisation has been the enemy of democratisation.’Footnote ⁵⁹

19.7 Coding and Delegation

There are also challenges linked to the journalistic-editorial function that this approach would seek to approximate. Here, I note two examples concerning coding and delegation.Footnote ⁶⁰ The first concerns what can be coded. Concepts must be translated into technical terms. Such concepts are inherently contested, and translating them is, to a greater or lesser extent, a transformation. Sometimes called the formalism fallacy, this issue is noted in studies of recommender systems. For example, an analysis of possible news recommenders for use by media organisations proposed five metrics from computer science to evaluate the systems: calibration (the degree to which content reflects user preferences), fragmentation (the degree to which the same stories appear for different users), activation (so users take action), representation (of ideas), and alternative voices (of marginalised groups).Footnote ⁶¹ Some of these metrics require precise identification and classification. Opinions need to be positioned relative to each other. Which opinions support or oppose a particular view, and in what ways? As the authors note, ‘This task is extremely complex, even for humans.’Footnote ⁶² Approximations could be made, such as using known politicians and political parties for the representation of ideas and voices. This resembles some existing regulation of electoral speech that provides broadcast time to political parties. But such approaches have clear limits, and the translation of concepts into code is something that necessarily renders the concepts in a different form. Overall, while the aim is to model theoretical concepts in a form that can be applied at scale and speed, the results are ‘merely approximations and … simplifications of very complex and nuanced concepts that have been contested and debated … for decades’.Footnote ⁶³ In addition, contested and culturally varied journalistic norms have generally been applied to items of content in past media infrastructures. Coding those norms in automated processes that apply at almost unimaginably greater scale and speed is, at the least, a substantial challenge. Journalism involves forming opinions about content, and that is impossible to do in the same way for platforms, given those differences.Footnote ⁶⁴ Public speech will be different under a ‘probabilistic’ model of curation.Footnote ⁶⁵

One suggested approach to coding involves collaboration between journalists and developers to specify editorial values for news recommendation. Jonathan Stray suggests that collaboration could develop four types of technical artifact, involving metrics (what is measured), data sets (used in training or evaluating systems), feedback methods (to systems), and evaluation protocols (for recommender systems).Footnote ⁶⁶ In his analysis: ‘It is necessary to commit to specific definitions, phrased in algorithmic terms, in order to build real recommenders.’Footnote ⁶⁷ Not keeping journalists closely involved would mean the work ‘must be done by technologists, which is closer to delegation than collaboration’.Footnote ⁶⁸ That is, important aspects of public communication would be delegated to those whose understanding and ambitions are to date very different than journalism. Of course, there are clearly ways in which that delegation has already happened. The idea of delegating recommendation, amplification, and moderation to platforms – and in that way delegating important aspects of the infrastructure of public speech – also raises questions about the model’s plausibility for democratic self-government. ‘Relying on computational models that even AI scientists cannot really understand is … in tension with genuine collective self-determination’,Footnote ⁶⁹ and this raises queries about democratic states meeting their obligations to support a system of public speech that serves opinion formation.

That returns us to freedom of expression. The approach to expressive freedom is moving, at least in part, to a systemic level more than focusing on specific items or categories of content. For example, think of the DSA and leave aside doubts and debates about its probable effectiveness.Footnote ⁷⁰ The DSA requires very large platforms and search engines to assess and mitigate systemic risks posed by their systems. This includes risks related to illegal content; fundamental rights, including freedom of expression, media freedom, and pluralism; civic discourse and electoral processes; public security; gender-based violence; public health; minors; and people’s well-being. It is, on paper, an encompassing set of risks in relation to which ‘reasonable, proportionate and effective mitigation’ must occur, giving ‘particular consideration’ to effects on fundamental rights.Footnote ⁷¹ While there is a long tradition of thinking about a system of freedom of expression and the state’s responsibilities towards it,Footnote ⁷² concerns here involve the effects of distribution systems on content where obligations or even liability provisions might apply at the systemic level and not only in relation to content items per se. At the least, there are questions as to how to arrive at, and make operational, values in platform recommenders that could address such issues.

In the past, the actual content produced by a pluralistic mass media system with structural diversity, independence, and transparency was left to a journalistic-editorial function (within general obligations and limitations imposed by content-related laws). The model presumed that decisions were (largely) made at the level of items of content produced by journalists. Now, delegating part of that process to platforms seems implausible given the contested nature of the concepts involved, the scale and speed of moderation that platforms require, and the fundamental aims for pluralism. Even delegating it to multiple algorithms – a version of the editorial personae already discussed – leaves questions about how they would be developed, what values they would prioritise, how those would be operationalised, and how they would be overseen. One can imagine approaches that might work in part, just as one could imagine various models of democratic mass media system. But achieving democratic media systems in the past was highly challenging and only partially successful in any event. Doing something equivalent now, if it is technically possible, could support freedom of expression interests, especially audience interests, but it would need a sophisticated understanding of expressive freedom. EU law has not yet had to develop that; it shows some interesting signs, but more will be needed if the European initiatives are to support that sort of development.Footnote ⁷³

19.8 Conclusion

In the past, media with structural diversity, independence, and transparency was imagined as producing varied, curated content that would reach audiences and support their own opinion formation. Now, media are dependent on distribution through major platforms for important parts of their audiences, and that dependence changes media organisations’ operations. Concurrently, platform infrastructures have changed the circulation of public speech more generally.

In terms of media organisations, having varied institutional forms, staffing, management, regulation, and financing would still help to support sustained plural public speech. Such measures should be pursued. However, even if they were largely achieved, there would be an overlay of decidedly non-journalistic platform distribution that does not show substantive concern for democratic values. In short, platforms do not have an interest in creating a democratically legitimate context for public speech: it is not their aim nor their financial model, and they lack a clear relationship to nation-states.Footnote ⁷⁴ Rather, platforms ultimately moderate ‘in order to sustain themselves’Footnote ⁷⁵ and to sustain user engagement and data extraction,Footnote ⁷⁶ ‘not in order to deal with social problems’.Footnote ⁷⁷ As Julie Cohen has noted, platforms are generally not concerned with opinion formation as such or with achieving some form of neutrality;Footnote ⁷⁸ they are simply ‘giving users what [their] revealed preferences indicate that they want’ in order to sustain engagement and extract more data.Footnote ⁷⁹ And they do that aiming to entrench their own positions.Footnote ⁸⁰ Of course, one could imagine different platform operations, but it seems naive to expect them to arise voluntarily. And that underlines the weakness in a purely negative approach to expressive freedom, which would leave state actions strongly curtailed but platforms free to moderate as they saw fit, as in the current US approach. It means that law applying limitations or obligations to platforms’ moderation would generally be judged as infringing platforms’ own free speech.Footnote ⁸¹

Through these developments, opinion power has changed. There has been ‘a shift over who controls the means to connect with the audience, as well as to define who the audience is and what the audience wants’.Footnote ⁸² Platforms and other intermediaries are increasingly important in terms of their opinion power, even if that has not been their primary aim. Platforms have reconfigured opinion power through overlaying journalism. This has changed who decides about public speech, how they decide, and what they aim for in doing that. The changes make positive dimensions of expressive freedom more important, including state obligations to regulate some actions of non-state entities that affect speech – such as platforms – both through restrictions on speech and through efforts to support an editorial-like role played by those entities. Overall, there is a greater need for states to act to support the conditions for democratic public debate, because a journalistic-editorial role cannot be assumed to do so within the existing infrastructures of public speech. As I have outlined, this poses real challenges, including challenges about the values that can be coded into recommender systems that operate at extremely high scale and speed, the comparative uniformity in recommendations that might result, and the questionable (often unstated) idea that meeting individual content preferences – for example, by giving recipients the ability to choose recommender personae – will meet collective communication needs.

A legal focus on media structure has long been quite common, with editorial control being assumed and relied on within the infrastructures underlying public speech. Now, under the DSA, for example, very large online actors have obligations to assess and mitigate systemic risks, but it is not clear that anything like journalistic-editorial roles will emerge at the systemic level.Footnote ⁸³ It is not even clear that such curation is possible. There will be a different sort of control, supporting a different sort of public speech. It is well recognised that European states have positive obligations in relation to pluralism.Footnote ⁸⁴ I would suggest that the obligation is not just about mass media pluralism but about pluralism of public communication, which now includes pluralism through platform distribution. But how to achieve that remains in doubt.

Of course, the more that regulation affects media content – such as specifying editorial standards in recommender systems – ‘the more strongly it can produce lack of freedoms, and the greater its potential for abuse’.Footnote ⁸⁵ However, not acting can also entail a failure in that effective, democratic freedom of expression remains lacking. And as Cohen states, ‘current patterns of online communication are not inevitable … but rather are the result of infrastructuring work undertaken for particular, self-interested purposes’ of platforms:

That is, one of the first two approaches to expressive freedom outlined in Section 19.2 is what is needed – approaches that require or allow legal measures in support of structural diversity, independence, and transparency for sustained plural public speech. Democratic governments did not fully control the system of public communication in the past, and freedom of expression highlights important ways in which they should not control speech. But for law to frame the system of freedom of expression, in its negative and positive dimensions, is a different matter.

Until recently there was a relatively stable model of mediated public speech in longstanding democracies that, as flawed as it was, provided a form of public debate and individual and collective opinion formation. Its existence involved varied and often substantial action by states – media ownership limits, direct and indirect media subsidies, public media funding, and so forth. Now it is not so clear how democracies should act in a changed system of public communication to support opinion formation and meet their obligations as guarantors of pluralism. But in that context the positive dimensions of expressive freedom have growing importance. Current European measures go only some way there, and European regulators and courts face future challenges in analysing the measures that online actors have (and have not) taken in light of all the dimensions of freedom of expression.