The “neo-intermediation” of large on-line platforms: Perspectives of analysis of the “state of health” of the digital information ecosystem

1 Introduction

The renewed global debate over the organization of the communication order has necessarily extended media analysis to what may be labeled the neo-intermediation of information flows in the context of what is variously referred to as global “platform capitalism” (Srnicek, 2017), “informational capitalism” (Cohen, 2019), or “big data capitalism” (Fuchs, 2019). The digital metamorphosis of the structural power of the new systemic constraints requires a review of the current tools for mapping and investigating the digital information ecosystem. By highlighting the political nature of such an analysis, relations of authority are reconstructed, new political entities are created, and novel interpretative frameworks are established. The enormous volume of online activity occurring through “big tech,” and the opaque way in which such content is managed, curated, and distributed to multiple users, has made evident the “de facto” outsourcing of the monitoring of online public debate. This includes a small number of private media corporations that support and govern the overall global space that people use to communicate (Balkin, 2018; Langvardt, 2019). As a result of the growing power of platforms – not only as major information intermediaries on a global scale, but also as providers of infrastructure technology (Webb, 2019) – legacy media are also in a worrying state of dependence. Digital platforms are able to redefine the conditions under which legacy media operate by directing and shaping the ways in which political power is exercised, whether public or private (Helberger, 2020). Indeed, it is now clear that there is a need to address the metamorphosis of power relations between the social players behind the “hybrid forms of control” (Cobbe, 2020) of content production, and access to digital infrastructure.

Emergency scenarios such as the cross-border impact of information wars on global geopolitics, from the pandemic information crisis to the Russo-Ukrainian war, have forced policy makers to focus on platforms supplying disinformation in the short term. As a result, the push toward automated content moderation has been reinforced (Meyer and Hanot, 2020). Yet, an almost fully automated system of self-policing carries with it the danger of hiding the inner workings of platforms as “governors” of public speech in which the fundamentally political nature of content moderation is executed by algorithms (Gillespie, 2018; Gorwa et al., 2020).

Thus, a cross-disciplinary approach is necessary to develop adequate regulatory proposals and draw the attention of researchers to the ethical challenges underlying the functioning of datafication, commodification, and selection algorithms (Burrell and Fourcade, 2021; Kitchin, 2017). Indeed, these are analytical prisms that help us understand the way in which the ecosystem modifies power relations: There is no doubt that the analysis of the forms and algorithmic dissemination strategies of public discourse, under the new systemic constraints, sheds light on the nature and kinds of responsibilities that we must attribute to them in addressing rights and public values (and the surrounding debates).

Starting from a comparative analysis of the growing international “regulatory field” of the subject of platform governance (Schlesinger and Kretschmer, 2020), this essay will focus on the ongoing European Commission strategy to build a European digital sovereignty: In particular, the European regulatory reshaping of platforms’ editorial accountability will be under scrutiny.

The purpose is to address the following questions:

1. What are the limits and potentialities of the new European digital governance model in dealing with such a disruptive phenomenon as algorithmic neo-intermediation of information?

2. What might be the potential effects of the new legislation’s enforcement on the health of the digital information ecosystem, and, thus, on the resilience of an autonomous and authentically plural public sphere?

This contribution is organized in two sections: The first section provides a brief synoptic framework of the main models of platform governance that emerge from a review of the literature, in order to frame the ongoing European digital strategy from a theoretical point of view. The second section centers on the close analysis of some crucial provisions of the Digital Services Act (DSA regulation 2022/2065) which dictates the new rules for hosting providers by reshaping their liability regime. It will make clear that the limitations of this regulatory framework are inherent to both the type of governance under which Regulation 2022/2065 falls, and the special features of neo-intermediation. Therefore, by focusing on the regulatory challenges that have arisen from algorithmic moderation systems (the so-called Recommender systems [RS]) – which represent one of the main breaking points compared to traditional forms of information broadcasting – we will seek to predict their likely impact on the cognitive pluralism of the “neo-intermediated” public information space. Regulatory processes, in fact, co-produce the algorithms that regulate (Campo et al., 2018; Cheney-Lippold, 2018). Thus, recognizing these as socio-technical constructs (Beer, 2017; Bucher, 2018; Airoldi, 2021; Hildebrandt, 2022) shall be the first step in constructing an analytical direction that will open up the possibilities of advancing the cognitive link in which the elements in question are resolved. Insofar, it will enable a multilevel analysis capable of contributing to the lines of development of an integrated regulatory framework at the supranational level.

As for methodology, the comparative analysis of the main models of platform governance, which have been elaborated on in the doctrine, will take Drahos’ (2017) broad theory of regulation as a parameter. In this multidimensional view, the state is seen as part of a network in which the tasks of regulation are redistributed to actors within it in various ways. Next, we will seek to shed light on the possible consequences for the health of the digital information ecosystem of what at an abstract level is a shift toward a Limited Governance Regulation model (LGR) resulting from the introduction of new legislation. In other words, we will try to understand if the (re)positioning of public authority will result in an effective rebalancing of the power of actors involved in governance in favor of users’ freedom of expression, a prerequisite that is essential to ensure the resilience of an authentically democratic digital public sphere.

In a subsequent phase we analyze reasons why the enforcement of public intervention in the form of LGR fails to counterbalance the “systemic opinion power” (Helberger, 2020) of platforms, confirming the limitations of the model already identified in the doctrine at a general level.

Therefore a solution will be proposed that can obviate the criticalities peculiar to the LGR, that is, a multisourcing model of regulating, which can determine a redistribution of both editorial accountability, and regulatory production, in line with the “tripartism regulatory model” described in 1992 by Ayres and Braithwaite (1992), arisen within the theory of “responsive regulation” (Drahos and Krygier, 2017). Finally, the need for “critical data literacy” (Nichols and Smith, 2021) will be emphasized both as a prerequisite for making the proposed model of accountability feasible, which we call “neo-editorial,” and to obviate the principal limitation of LGR, that is, the depoliticization of the ideological assumptions that underly informational capitalism.

2 Preliminary questions. Platform accountability: Beyond the opinion power

Content moderation and the prescriptive regulation of media accountability have a long history. The “opinion power” assigned to the media, and, thus, the ability to affect social and political institutions, is the main reason they have always been subject to a different regime of liability and transparency, compared to private companies operating in different sectors (Garton Ash, 2016).

Media accountability can be defined as those “voluntary or involuntary processes by which media are directly or indirectly accountable to their society for the quality and/or consequences of publication” (McQuail, 2005, p. 207). As such, it is part of inter-institutional accountability, a dimension that influences the quality of democracy (Morlino, 2012). At the same time, it involves the relationship between the media and their audiences, as the latter can hold the media directly accountable without the intermediation of public institutions (McQuail, 2005).

However, any statement about the principles that should inform the regulatory policies of platforms inevitably raises the complex question of what a media company is, and whether digital platform companies can be considered as such (van Drunen, 2020). Precisely because of their “amphibious” nature, digital platforms have rarely been discussed from the perspective of opinion power (Helberger, 2020). Since they do not fall under the concept of media in the traditional sense, insofar as they do not publish their own content, they have been traditionally considered neutral hosting providers, falling within the area of the so-called “safe harbor” principle. On the basis of this principle, the E-commerce Directive 2000/31 has granted them a regime of exemption from liability for illegal or harmful third-party content. Thus, the first issue that the current regulating approaches to platform accountability need to address is a reframing of the meaning of “media companies,” by including global-scale digital information gatekeepers (Gillespie, 2014). However, this is only a preliminary issue.

The neutral appearance of platforms’ activities is the major feature of digital content moderation based on algorithmic data mining. Thereby, the main challenge for current regulatory attempts, aimed at introducing new liability regimes, is the extent and pervasiveness through which digital platforms can algorithmically control, shape, and personalize the global information flow, through predictive analysis based on behavioral data in the full quantification of human experience (Napoli, 2019; Hildebrandt, 2022). Beyond explicit exercises of power, the ability to shape and direct the flow of political information and communication, as well as the power to amplify or downgrade particular voices, can also be as much a consequence of algorithmic ordering (which is never neutral) as of conscious design. As a result of this discontinuity with traditional content intermediation historically carried out by media companies, we prefer to speak of “neo-intermediation” instead of “reintermediation” (Giacomini, 2018) to better define the activity carried out by platforms. Although the gatekeeping process has been extensively studied by multiple disciplines, in the digital ecosystem some important changes have occurred: (a) the editorial role delegated to the algorithms, (b) the growing role of audiences as secondary gatekeepers for which users co-determine what makes the news, and (c) the change in the position of the journalist from gatekeeper to gate watcher.

In particular, from the moment it is born to when it reaches its widest audience, information is modeled, filtered, and hidden within a dense mixture of elements that come together in the algorithmic infrastructure of social media and digital platforms (Moeller and Helberger, 2018).

Thus, the concept of neo-intermediation is intended to focus on the central role of recommendation and personalization algorithms, which are new gatekeeping infrastructures, in combination with the role played by third-party mediators, also known as data brokers. From a theoretical point of view, the identification of the phenomenon of neo-intermediation and the limits of the concept of disintermediation, allows us to focus our attention on the processes of selection and dissemination of information content, so much so that we can speak of “algorithmic publishing” (platform press) and the social and political responsibility of online platforms towards the public sphere. The gatekeeping process, which drives how certain events are deemed more newsworthy than others, involves both how institutions or influential individuals determine what information to convey to recipients, and the moral perspective with its own stereotypes and biases. Indeed, the notion of “algorithmic public opinion” (Airoldi and Gambetta, 2018) emerges, which is inevitably influenced by the governance policies of online platforms (Friedman et al., 2013). As suggested by Natalie Helberger (2020) the “systemic opinion-power” hidden under the algorithmic neo-intermediation phenomena based on datafication, commodification, and selection “is not only the power to influence political processes (such as democratic will formation) but it is political power in its own right.”

In this scenario, the EU strategy to build a so-called “European Digital Sovereignty” is the turning point toward the “new internet era” (Floridi, 2021) in the European context. The current Digital Services Act is mainly aimed at trying to counterbalance the infrastructural platform power, in order to ensure a safer transnational digital space, in which fundamental rights are protected. The European model represents an attempt to democratize and “constitutionalize” the internet (De Gregorio, 2020; Santaniello, 2021). While promoting the values and principles of liberal democracies, it opposes isolationist instances with a polyphonic approach, which is also open to input from civil society (Santaniello, 2022).

Nevertheless, a fundamental premise of such a sovereignty is to ensure the conditions for thinking and reshaping a space, that would be capable of being defined as a European digital public sphere, or rather, a transnational public sphere (Schlesinger, 2020). This would be a precondition for legitimizing the sovereignty of supranational institutions in a post-Westphalian context of supposedly sovereign states. The metaphorical space of the public sphere is indeed a crucial hermeneutical grid for thinking about tactics of media governance; thereby the possibility of resilience of its own autonomy will have to remain the counterfactual parameter used to evaluate their results, being media, in turn, the primary means to create a public sphere and to practice self-government (Garton Ash, 2016).

3 Comparative analysis: Platform accountability models review

The aim of this section is to give a synoptic framework of the main models of platform governance up to now elaborated in the doctrine, and to show how there has been a shift, along an ideal continuum that takes European regulation from Industry Self Regulation to a second model: the Limited Government Regulation (LGR), as mentioned above. Indeed, the EU claims its role as rule maker, dictating a coercive ex ante discipline aimed at regulating, among other things, the private activity of information intermediation. The aim of this intervention is to ensure the safeguarding of fundamental rights related to freedom of expression and information, which need to be carefully balanced with interrelated rights such as privacy, dignity, equality, and non-discrimination.

We define regulation – as suggested by Drahos and Krygier (2017, p. 4) – as “a dynamic multilevel process” in which “the state becomes part of a network and the tasks of regulation are redistributed in various ways among actors within the network.” This includes non-legal forms of norm making, which opens up a number of possible entry points for empirical study. The concept describes a system capable of generating regulation from many actors, at different levels, and using a variety of instruments to communicate and enforce their chosen norm (Drahos, 2017, p. 764).

Recent studies, conducted over the period from 2017 to 2020, have mapped the international regulatory landscape through quantitative processing techniques, allowing interventions and relationships to be categorized into three macro areas according to the homogeneity of drivers (Flew and Gillett, 2021): competition, moderation, and rights. On the basis of these studies, we will go on to enucleate the operations and relationships of the agencies in charge of prescriptive regulation of cultural and informational content, as they exercise power over the production, circulation, and consumption of the communication stream.

With specific reference to the theme of the Social Platform Accountability, the most recent literature (De Blasio and Selva, 2021) has identified four ideal regulation paradigms from empirical evidence at the European level: (a) accountability set by law, (b) co-decided accountability, (c) regulated self-regulation, and (d) pure self-regulation.

They are based on three elements: (a) principles of media legislation; (b) main actors, identifying the role played by public institutions, media companies, professionals and audiences; and (c) tools used to ensure media accountability (Eberwein et al., 2018; Fengler et al., 2014), as well as “any non-state means to empower the media to the public” (Bertrand, 2000, p. 108).

Nevertheless, in order to frame the current European Strategy from a political point of view, we will follow an approach which, while encompassing the described models, is mainly based on the criterion of overall expansiveness of governance tactics (Rochefort, 2020; Gorwa, 2019; Rahman, 2018). It allows us to screen the purposes and effectiveness of the interventions by assessing their impact on informational capitalism constraints as a counterfactual perspective.

The main models of platform governance that have emerged on the basis of this criterion are (a) industry self-regulation, (b) limited government regulation, and (c) comprehensive government regulation.

4 From Industry Self-Regulation to Limited Government Regulation: The end of an era

5 Concluding remarks

Through the concept of discursive closure, we can see how particular values and (infra)structures are naturalized, neutralized, and legitimated, with the active marginalization of likely alternatives, even on an imaginal level. Approaches based on technological solutionism (Dencik and Hintz, 2017), like the Limited Government Regulation, which inspires the ongoing “European digital package”, would not address regulatory issues inherent in fundamental problems such as the systemic opinion power of big platforms. None of the provisions of the ongoing regulation are aimed at creating countervailing powers, even though this is essential to preventing certain social media platforms from becoming “quasi-governments” of online speech (Cohen, 2019; Helberger, 2020). Their vocabulary, in turn, enforces the discursive depoliticization of structural phenomena such as “Dataism”, the gradual normalization of datafication, as a new paradigm in science and society (Van Dijck, 2014; Dencik, 2018). In so doing, it seems to lack a crucial requirement for considering the democratic tactics of governance in the digital ecosystem: the restitution of the intellectual privacy of individuals (Eskens, 2020), understood as control of both “incoming” and “outgoing” information flow (Rodotà, 2014). This is a primary condition for ensuring the resilience of autonomy and cognitive self-sovereignty, and, therefore, for reflecting on the possible resilience of an autonomous public sphere.

While we are aware that a sectoral approach cannot be a definitive answer to the problem, we have described a shared model of accountability that seeks to ensure that the empowerment of platforms does not result in a right to censorship or go on to increase discretion, if not arbitrariness, in defining the standards allowed in public discourse.

Furthermore, empowering “prosumers” means that they at the time of identification choice, should be empowered to assess the potential harmfulness of their own content through adequate disclosure and media literacy. The enforcement of the latter should be a prerequisite for the feasibility of such a model as outlined in this paper. For the implementation of such media literacy, it is possible to envisage triangulation: bodies specialized in specific issues, such as independent intermediaries between regulators and platforms. They would aim to assist the producer, at their request, in the assessment of the compliance or non-compliance of their content. Depending on the nature of the filtering powers attributed to them on a case-by-case basis, these intermediaries could, if authorized, be given responsibility for publications. These actors, therefore, would be both regulators and regulated. This would be a further application of the model of “regulatory tripartism” (Ayres and Braithwaite, 1992), which, recognizing that society cannot rely exclusively on law and its agencies of implementation, thinks about models of regulation that draw third parties into the circle of regulation. (Drahos and Krygier, 2017). This also requires that public interest groups receive the information available to a regulator and the opportunity to participate in regulatory negotiation. They would then have a role not only in enforcement and compliance, but also in increasing the regulatory capacity of the network society.

In our case, these intermediate bodies would go on to constitute a third neutral element, which both through the production of rules “from below” and through the enforcement of users’ critical media literacy (Nichols and Smith, 2021; Markham, 2019) could help self-immunize the health of the digital ecosystem by creating a survival habitat for a democratic public sphere.

We argue that the main objective to ensure the effectiveness of any tactic of governance should be, first and foremost, to find tools capable of unveiling the new forms of structural oppression exercised by the cultural hegemony. Unlike in the past, this could be done in an invisible, and therefore more pervasive and insidious manner by the seemingly neutral algorithms that neo-intermediate public discourse (Kramsch, 2020; Amoore, 2020). This also means challenging the “normalizing” perceptual paradigms (Mertala, 2020) imposed by the dominant political discourse both by private corporations and by government institutions. Last, it should be added that carrying on research in this field also means unveiling such ideological normalizations, which in turn justify the construction of “data relations” (Couldry and Meijas, 2019), ensuring the “natural” conversion of daily life into a data stream, without which neo-intermediation could not exist in its current form.