Power asymmetries in the algorithmic formation of public opinion

Structural transformations of the public sphere and algorithmic public opinion

Since the concept of the public sphere gained traction following the 1989 English translation of Jürgen Habermas's habilitation thesis, it has generally been considered in purely normative terms, particularly in critical media and communication studies. It enabled authors to define the core principles of mediated communication that are essential for democratic societies – especially with regard to citizens’ participation in political processes, and thus legitimacy of the political order. Although the lofty ideals of reasonable discussion leading to the formation of an informed public opinion were largely counter-factual, these norms provided an important goal to strive for.

In their article, Gandini et al. (2025) move away from normative interpretations of these concepts, as has often been the case in recent decades. Nevertheless, their description of ‘algorithmic public opinion’ makes a noteworthy contribution to academic literature on these issues. It could be argued that a third structural transformation of the public sphere is taking shape today (Seeliger and Sevignani, 2023: 10–14), and in this process, the gatekeeping role once reserved for institutional (mass) media is increasingly being performed by social media platforms. The authors focus precisely on this role of the platforms, which is central to understanding the public sphere today.

There is little doubt that the introduction of algorithmic curation of information flows signifies a substantial departure from the past. First, and perhaps most obviously, audiences are no longer somewhat passive recipients in these processes. They are both actively involved in producing content and, at least on the surface, can influence the content they receive, since algorithms provide audiences with personalised streams of information based on their past practices.

However, it is noteworthy that users do not necessarily exert influence over this sorting either wittingly or knowingly. Equally important is the fact that these curation practices of platforms-as-corporations are primarily aimed at increasing user engagement and are therefore far from neutral even in this most basic sense. As audiences represent the main commodity sold by these for-profit platforms, simultaneously meaning users are under constant surveillance, the relationship between platforms and audiences becomes highly instrumentalised (Fuchs and Mosco, 2015). Consequently, active user influence over algorithmic decision-making may be, paradoxically, much more illusory than is often tacitly presumed.

(Dis)empowered audiences and gig publics

There are a number of issues at play here, but we can briefly consider some to illustrate the point. For instance, as Gandini et al. (2025) note, the data-driven decision-making of algorithmic systems is derived from measurable and recorded practices. This behaviourist approach ignores ‘beliefs, intentions, and goals’ of audiences. ‘The result is that users’ inferred preferences often fail to align with their personal preferences’ (Gandini et al., 2025), which are never explicitly expressed. This is frequently exemplified by users’ exasperation over the rubbish content they receive on their feeds, and their self-deprecation for even following these platforms in the first place.

In this context, it is vital to take into account the structural pressures of these algorithmic systems in habituating audiences, particularly in terms of creating compulsive patterns of behaviour, a phenomenon that is widely studied in popular psychology today, and to acknowledge the fact that many of the practices on these platforms are now entirely performative in nature. These practices extend well beyond the realm of the formally promotional content, gradually becoming the online norm. Splichal (2025) refers to such phenomena as ‘gig publics’, given they are highly episodic and unable to truly engage in reasonable discourse, which is necessary for the functioning of the public sphere.

These trends further complicate the question of what (kind of) data and user interactions are actually analysed by algorithmic systems, while also blurring the epistemic distinction between appearances and spectacle, on the one hand, and truth and reflection, on the other (Splichal, 2025).

Bearing these issues in mind, it is worth considering whether audiences would explicitly prefer (currently empirically factual) profit-driven algorithmic curation before (normatively oriented) curation aimed at satisfying public interests? (It could be argued that having such hypothetical alternatives is what user influence on algorithmic curation should, in fact, look like).

Despite the perception of voluntariness in these practices of audiences, it is important to recognise that these relationships are inherently asymmetric. They favour a few corporate platforms, with algorithmic habituation subtly exerting influence in accordance with their commercial agenda (cf. Splichal, 2025).

That said, as in the past, audiences undoubtedly play an active role in the formation of public opinion – albeit in a way that is significantly different from how that is typically portrayed – and Gandini et al. (2025) provide a useful overview of several transformations taking place. These include new biases, as well as potential new issues such as hyper-heterogeneity. In normative models, the latter can be considered as especially degenerative for democratic communication, since focused public attention on major societal issues is even more difficult to achieve than in the past (Seeliger and Sevignani, 2023: 11). This again overlaps with Splichal's (2025) description of ‘gig publics’, which are characterised by transient interactions resulting in superficial experiences. Consequently, their attention is fragmented and driven by personal rather than public interests.

Social media platforms as gatekeepers in the public sphere

This brings us to the second significant break that we have already implicitly touched upon, namely, the role of corporate digital platforms in drawing attention to published information. As Gandini et al. (2025) rightly observe, and as has become almost common knowledge, the old gatekeepers – professional journalists and editors in institutional media (Reese, 2021) – have, in many respects, now been supplanted by new gatekeeping actors: the algorithmic systems of the corporate platforms. It is increasingly these automated systems that determine which issues get amplified ‘and which news sources and interpretations are prioritised’ (Gandini et al., 2025).

The authors (Gandini et al., 2025) distinguish between direct and indirect ways in which platforms perform their gatekeeping function. Direct gatekeeping, among other things, involves prioritising content, particularly with regard to the aforementioned aim of user engagement. However, it is the indirect forms of gatekeeping that seem especially relevant. These lie at the intersection of how platforms act, and how external actors try to operate on these platforms. They include attempts of actors to optimise content with the purpose of promoting its visibility, algorithmic gaming, and algorithmic activism. Distinctions between these approaches are, to a certain extent, arbitrary. Nonetheless, ‘they all share the ultimate goal of influencing online visibility and shaping public opinion’ (Gandini et al., 2025).

Arguably, it is precisely these indirect forms of gatekeeping that further exacerbate power imbalances. Contrary to the illusion that anyone can now influence public opinion simply by creating an online account, asymmetries appear as stark as ever. Winner-takes-all logic (i.e. popularity bias) has long been a feature of the online environment, but even more importantly, it has become far easier for ‘promotional intermediaries’ (Davis, 2013) to translate their financial might into communicative power.

Power asymmetries and journalism in the (global) flows of information

The advent of public relations industries in the early 20th century, followed by the emergence of think tanks and political spin in subsequent decades, has resulted in the pervasive influence of moneyed interests on the formation of public opinion via the mass media (Golding and Murdock, 2022: 39–40; Prodnik and Vobič, 2024). In their capacity as the ‘old gatekeepers’, journalists and editors have historically depended on these actors to provide them with ‘information subsidies’, to use Gandy's (1982) term. As news production has always been expensive, these subsidies have significantly reduced the cost of journalistic labour. The outsized influence of powerful external actors has therefore always been an issue, but today, the distinction between publicly relevant information, unfiltered news, promotional content, advertisements, public relations material, astroturfing, political propaganda and advertorials has become utterly blurred on social media platforms.

Although the ‘wall of separation between church and state’ in journalism – which stipulated that the business side of the media should never interfere with the news department – was always porous, it has at least been an explicit normative goal and a mark of the profession's credibility. Such norms have never really existed on social media platforms, since they never claimed to serve the public interest. Worryingly, the collapse of the financial models of news media that relied on advertising revenue, which is now monopolised by the largest platforms (Pickard, 2020: Chapter 3), has led to the further breakdown of these norms in journalism as well. Given the increasing number of cases of journalistic institutions on the verge of collapse, it seems inevitable that these boundaries will crumble further. Producers of ‘information subsidies’ therefore have increasing free rein to flood the public sphere with information they choose. Meanwhile, there are also few guarantees that digital platforms will not exert direct control over information curation at whim and start enabling authoritarian politicians (cf. Akbari, 2025). As always, there are opportunities to counter such narratives, but these fights are increasingly being fought on unequal grounds.

Conclusion: Are we missing something here?

This brings us to the final point. It is my unnerving feeling that media and communication studies have largely failed to properly integrate how digital platforms are reshaping social relations into their analyses. Most of the studies remain abstract, as if these transformations are innocuous, as if information flows frictionlessly across the digital realm, and as if algorithms are in fact acting as neutral arbiters in the public sphere.

Platforms have indeed become central nodes in the formation of public opinion, but what exactly does this mean? Does it make sense to treat this as a game of ‘equal partners’, in which audiences (and many other actors) are almost on the same footing as the platforms? Even if that was the case at some point in time, it has long since passed. Shoemaker (2020), for instance, refers to the digital platforms as ‘supra-gatekeepers’. She is not alone in contending that these titans wield more power than any other comparable individual actor did in the past. Consequently, they are increasingly the ones that define our social reality. In these communicative relations, even traditional news publishers – in many cases still large corporate actors – are wholly dependent on platforms for journalistic sourcing and, in particular, disseminating news in the attention economy (Prodnik and Vobič, 2024). Meanwhile, platforms now have the power to destroy large parts of an entire industry with the click of a button (or, more literally, by adjusting their algorithms to downgrade the visibility of news). It matters little that these (news) industries are still credited with being essential to democratic order (Reese, 2021).

Still, even these observations merely touch upon the extensive range of issues that would need to be incorporated within a truly holistic framework. While they have indeed become key players in the public sphere, these platforms are not merely media entities in the conventional sense. Perhaps media and communication studies are struggling because their influence is so pervasive. For starters, they are influential in political communication, yet they have also become political actors themselves: they act as vast lobbying entities. Their power is superior to that of nation states and they are aware of it. Their owners have increasingly explicit political interests and either actively use their platforms to change the political landscape or see their interests progressively overlap with those of dominant states (cf. Akbari, 2025). They are vast economic entities in their own right; not only are they the largest media businesses in history, they are also among the world's biggest corporations (Akbari, 2025). Furthermore, they are not only important adopters of technological innovation; they are driving these changes by pouring trillions of dollars into the development of generative AI models (Morozov, 2025) (paradoxically, it is precisely the extraction of ‘the public opinion’ that is an essential raw material of this AI race). As a result, these actors have inevitably found themselves embroiled in political and even geopolitical struggles while participating in an increasing number of military projects. It is perplexing to read in the news that the EU is continually losing its ability to regulate these platforms, while media scholars debate the influence of audiences. In an era where the power of Big Tech seems almost unchecked, are we still referring to the same actors?

Although these matters extend well beyond issues directly connected to the formation of public opinion, media and communication scholars cannot simply ignore them. They influence whether and how these actors are able to exert their communicative power. This should not be viewed as a veiled critique of the excellent paper by Gandini et al. (2025), but rather as an observation of a broader challenge facing the field. After all, combining the often abstract theoretical models with existing material reality is a daunting task, particularly given that these platforms are at the forefront of the development of digital capitalism as a whole. A genuinely critical framework would encompass all the aforementioned aspects, neither underestimating their importance for the public sphere nor overestimating their role in a fatalistic and conspiratorial manner.